2023-04-27 12:19:53,968 INFO [train.py:976] (1/8) Training started 2023-04-27 12:19:53,968 INFO [train.py:986] (1/8) Device: cuda:1 2023-04-27 12:19:53,970 INFO [train.py:995] (1/8) {'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 12:19:53,970 INFO [train.py:997] (1/8) About to create model 2023-04-27 12:19:54,681 INFO [zipformer.py:178] (1/8) 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 12:19:54,697 INFO [train.py:1001] (1/8) Number of model parameters: 70369391 2023-04-27 12:19:57,229 INFO [train.py:1016] (1/8) Using DDP 2023-04-27 12:19:58,265 INFO [multidataset.py:46] (1/8) About to get multidataset train cuts 2023-04-27 12:19:58,265 INFO [multidataset.py:49] (1/8) Loading LibriSpeech in lazy mode 2023-04-27 12:19:58,283 INFO [multidataset.py:65] (1/8) Loading GigaSpeech 1998 splits in lazy mode 2023-04-27 12:20:00,752 INFO [multidataset.py:72] (1/8) Loading CommonVoice in lazy mode 2023-04-27 12:20:00,755 INFO [asr_datamodule.py:230] (1/8) Enable MUSAN 2023-04-27 12:20:00,755 INFO [asr_datamodule.py:231] (1/8) About to get Musan cuts 2023-04-27 12:20:02,950 INFO [asr_datamodule.py:255] (1/8) Enable SpecAugment 2023-04-27 12:20:02,950 INFO [asr_datamodule.py:256] (1/8) Time warp factor: 80 2023-04-27 12:20:02,950 INFO [asr_datamodule.py:266] (1/8) Num frame mask: 10 2023-04-27 12:20:02,950 INFO [asr_datamodule.py:279] (1/8) About to create train dataset 2023-04-27 12:20:02,951 INFO [asr_datamodule.py:306] (1/8) Using DynamicBucketingSampler. 2023-04-27 12:20:07,437 INFO [asr_datamodule.py:321] (1/8) About to create train dataloader 2023-04-27 12:20:07,438 INFO [asr_datamodule.py:435] (1/8) About to get dev-clean cuts 2023-04-27 12:20:07,439 INFO [asr_datamodule.py:442] (1/8) About to get dev-other cuts 2023-04-27 12:20:07,440 INFO [asr_datamodule.py:352] (1/8) About to create dev dataset 2023-04-27 12:20:07,675 INFO [asr_datamodule.py:369] (1/8) About to create dev dataloader 2023-04-27 12:20:25,619 INFO [train.py:904] (1/8) Epoch 1, batch 0, loss[loss=7.439, simple_loss=6.748, pruned_loss=6.901, over 16452.00 frames. ], tot_loss[loss=7.439, simple_loss=6.748, pruned_loss=6.901, over 16452.00 frames. ], batch size: 75, lr: 2.50e-02, grad_scale: 2.0 2023-04-27 12:20:25,620 INFO [train.py:929] (1/8) Computing validation loss 2023-04-27 12:20:32,880 INFO [train.py:938] (1/8) Epoch 1, validation: loss=6.911, simple_loss=6.238, pruned_loss=6.721, over 944034.00 frames. 2023-04-27 12:20:32,880 INFO [train.py:939] (1/8) Maximum memory allocated so far is 11798MB 2023-04-27 12:20:36,371 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=5.0, num_to_drop=2, layers_to_drop={0, 3} 2023-04-27 12:20:51,539 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=23.0, num_to_drop=1, layers_to_drop={0} 2023-04-27 12:21:11,866 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.2634, 4.2354, 4.2594, 4.2504, 4.2712, 4.2520, 4.2276, 4.2280], device='cuda:1'), covar=tensor([0.0090, 0.0054, 0.0037, 0.0039, 0.0039, 0.0042, 0.0022, 0.0052], device='cuda:1'), in_proj_covar=tensor([0.0009, 0.0009, 0.0009, 0.0009, 0.0009, 0.0009, 0.0009, 0.0009], device='cuda:1'), out_proj_covar=tensor([9.0285e-06, 9.1438e-06, 9.0451e-06, 9.0237e-06, 9.3466e-06, 9.1170e-06, 9.1485e-06, 9.2227e-06], device='cuda:1') 2023-04-27 12:21:15,139 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([5.9422, 5.9435, 5.9434, 5.9445, 5.9438, 5.9386, 5.9140, 5.9444], device='cuda:1'), covar=tensor([0.0007, 0.0007, 0.0011, 0.0009, 0.0008, 0.0012, 0.0009, 0.0009], device='cuda:1'), in_proj_covar=tensor([0.0009, 0.0009, 0.0009, 0.0009, 0.0009, 0.0009, 0.0009, 0.0009], device='cuda:1'), out_proj_covar=tensor([8.9173e-06, 9.0445e-06, 8.8401e-06, 9.0573e-06, 8.8582e-06, 8.9895e-06, 8.9730e-06, 9.0361e-06], device='cuda:1') 2023-04-27 12:21:17,132 INFO [train.py:904] (1/8) Epoch 1, batch 50, loss[loss=1.356, simple_loss=1.194, pruned_loss=1.436, over 17044.00 frames. ], tot_loss[loss=2.166, simple_loss=1.962, pruned_loss=1.954, over 751980.91 frames. ], batch size: 50, lr: 2.75e-02, grad_scale: 2.0 2023-04-27 12:21:20,833 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=5.42 vs. limit=2.0 2023-04-27 12:21:37,039 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=7.02 vs. limit=2.0 2023-04-27 12:21:46,521 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=83.0, num_to_drop=1, layers_to_drop={1} 2023-04-27 12:21:59,097 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=21.51 vs. limit=2.0 2023-04-27 12:22:01,739 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=28.03 vs. limit=2.0 2023-04-27 12:22:02,791 WARNING [train.py:894] (1/8) Grad scale is small: 0.001953125 2023-04-27 12:22:02,791 INFO [train.py:904] (1/8) Epoch 1, batch 100, loss[loss=1.164, simple_loss=0.9982, pruned_loss=1.314, over 16203.00 frames. ], tot_loss[loss=1.633, simple_loss=1.453, pruned_loss=1.609, over 1325599.98 frames. ], batch size: 164, lr: 3.00e-02, grad_scale: 0.00390625 2023-04-27 12:22:03,297 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=4.62 vs. limit=2.0 2023-04-27 12:22:13,669 INFO [optim.py:368] (1/8) 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 12:22:20,964 WARNING [optim.py:388] (1/8) Scaling gradients by 0.0112030990421772, model_norm_threshold=1019.0284423828125 2023-04-27 12:22:21,071 INFO [optim.py:450] (1/8) 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 12:22:43,778 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=144.0, num_to_drop=2, layers_to_drop={2, 3} 2023-04-27 12:22:46,126 WARNING [optim.py:388] (1/8) Scaling gradients by 0.0022801109589636326, model_norm_threshold=1019.0284423828125 2023-04-27 12:22:46,233 INFO [optim.py:450] (1/8) 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 12:22:49,649 WARNING [optim.py:388] (1/8) Scaling gradients by 0.04246773198246956, model_norm_threshold=1019.0284423828125 2023-04-27 12:22:49,754 INFO [optim.py:450] (1/8) 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 12:22:51,348 WARNING [optim.py:388] (1/8) Scaling gradients by 0.000716241542249918, model_norm_threshold=1019.0284423828125 2023-04-27 12:22:51,452 INFO [optim.py:450] (1/8) 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] (1/8) Epoch 1, batch 150, loss[loss=0.9981, simple_loss=0.8383, pruned_loss=1.14, over 17210.00 frames. ], tot_loss[loss=1.396, simple_loss=1.224, pruned_loss=1.443, over 1766384.49 frames. ], batch size: 44, lr: 3.25e-02, grad_scale: 0.00390625 2023-04-27 12:22:53,729 WARNING [optim.py:388] (1/8) Scaling gradients by 0.049951765686273575, model_norm_threshold=1019.0284423828125 2023-04-27 12:22:53,835 INFO [optim.py:450] (1/8) 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 12:22:58,541 WARNING [optim.py:388] (1/8) Scaling gradients by 0.00609818659722805, model_norm_threshold=1019.0284423828125 2023-04-27 12:22:58,647 INFO [optim.py:450] (1/8) 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 12:22:59,158 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=3.97 vs. limit=2.0 2023-04-27 12:23:02,550 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=7.28 vs. limit=2.0 2023-04-27 12:23:16,873 WARNING [optim.py:388] (1/8) Scaling gradients by 0.059935860335826874, model_norm_threshold=1019.0284423828125 2023-04-27 12:23:16,977 INFO [optim.py:450] (1/8) 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 12:23:17,638 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=86.72 vs. limit=5.0 2023-04-27 12:23:28,341 WARNING [optim.py:388] (1/8) Scaling gradients by 0.060559310019016266, model_norm_threshold=1019.0284423828125 2023-04-27 12:23:28,447 INFO [optim.py:450] (1/8) 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 12:23:32,763 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.6643, 3.3505, 2.8731, 2.9155, 2.9898, 3.4126, 3.0091, 2.6855], device='cuda:1'), covar=tensor([0.3545, 0.1382, 0.1606, 0.1671, 0.1517, 0.0505, 0.3126, 0.1145], device='cuda:1'), in_proj_covar=tensor([0.0009, 0.0009, 0.0010, 0.0009, 0.0010, 0.0009, 0.0010, 0.0009], device='cuda:1'), out_proj_covar=tensor([9.4388e-06, 9.2503e-06, 9.4726e-06, 9.5782e-06, 9.6212e-06, 9.2900e-06, 9.6331e-06, 9.4307e-06], device='cuda:1') 2023-04-27 12:23:42,816 WARNING [train.py:894] (1/8) Grad scale is small: 0.00390625 2023-04-27 12:23:42,817 INFO [train.py:904] (1/8) Epoch 1, batch 200, loss[loss=0.9311, simple_loss=0.7806, pruned_loss=0.9965, over 16869.00 frames. ], tot_loss[loss=1.254, simple_loss=1.088, pruned_loss=1.311, over 2108564.24 frames. ], batch size: 42, lr: 3.50e-02, grad_scale: 0.0078125 2023-04-27 12:23:51,003 INFO [optim.py:368] (1/8) 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] (1/8) Scaling gradients by 0.002041660714894533, model_norm_threshold=541.4743041992188 2023-04-27 12:23:51,109 INFO [optim.py:450] (1/8) 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 12:24:00,575 WARNING [optim.py:388] (1/8) Scaling gradients by 0.02974529005587101, model_norm_threshold=541.4743041992188 2023-04-27 12:24:00,682 INFO [optim.py:450] (1/8) 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] (1/8) Scaling gradients by 0.01955481991171837, model_norm_threshold=541.4743041992188 2023-04-27 12:24:01,587 INFO [optim.py:450] (1/8) 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 12:24:02,165 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=59.21 vs. limit=5.0 2023-04-27 12:24:18,146 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=5.23 vs. limit=2.0 2023-04-27 12:24:29,663 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=59.41 vs. limit=5.0 2023-04-27 12:24:30,813 INFO [train.py:904] (1/8) Epoch 1, batch 250, loss[loss=0.9034, simple_loss=0.769, pruned_loss=0.8577, over 12275.00 frames. ], tot_loss[loss=1.157, simple_loss=0.9965, pruned_loss=1.202, over 2370580.05 frames. ], batch size: 248, lr: 3.75e-02, grad_scale: 0.0078125 2023-04-27 12:24:31,288 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=19.70 vs. limit=2.0 2023-04-27 12:24:33,600 WARNING [optim.py:388] (1/8) Scaling gradients by 0.057925041764974594, model_norm_threshold=541.4743041992188 2023-04-27 12:24:33,709 INFO [optim.py:450] (1/8) 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 12:24:37,994 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=5.01 vs. limit=2.0 2023-04-27 12:25:05,024 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([5.8516, 5.8823, 5.4260, 5.8850, 5.6731, 5.8716, 5.8777, 5.7071], device='cuda:1'), covar=tensor([0.0050, 0.0029, 0.1181, 0.0033, 0.1244, 0.0065, 0.0082, 0.0859], device='cuda:1'), in_proj_covar=tensor([0.0009, 0.0009, 0.0009, 0.0009, 0.0009, 0.0009, 0.0009, 0.0009], device='cuda:1'), out_proj_covar=tensor([9.2797e-06, 9.1038e-06, 9.2423e-06, 8.9527e-06, 9.5141e-06, 9.4213e-06, 9.0398e-06, 9.4243e-06], device='cuda:1') 2023-04-27 12:25:16,071 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=296.0, num_to_drop=1, layers_to_drop={0} 2023-04-27 12:25:20,576 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=300.0, num_to_drop=2, layers_to_drop={2, 3} 2023-04-27 12:25:21,112 WARNING [train.py:894] (1/8) Grad scale is small: 0.0078125 2023-04-27 12:25:21,112 INFO [train.py:904] (1/8) Epoch 1, batch 300, loss[loss=0.9989, simple_loss=0.8315, pruned_loss=0.977, over 17259.00 frames. ], tot_loss[loss=1.093, simple_loss=0.9348, pruned_loss=1.12, over 2584644.43 frames. ], batch size: 52, lr: 4.00e-02, grad_scale: 0.015625 2023-04-27 12:25:21,636 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=4.70 vs. limit=2.0 2023-04-27 12:25:28,256 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=4.61 vs. limit=2.0 2023-04-27 12:25:30,081 INFO [optim.py:368] (1/8) 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 12:25:45,099 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=11.76 vs. limit=2.0 2023-04-27 12:26:12,602 INFO [train.py:904] (1/8) Epoch 1, batch 350, loss[loss=0.8328, simple_loss=0.6912, pruned_loss=0.7839, over 16752.00 frames. ], tot_loss[loss=1.043, simple_loss=0.8864, pruned_loss=1.053, over 2747293.52 frames. ], batch size: 89, lr: 4.25e-02, grad_scale: 0.015625 2023-04-27 12:26:14,093 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=45.20 vs. limit=5.0 2023-04-27 12:26:18,642 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=357.0, num_to_drop=2, layers_to_drop={2, 3} 2023-04-27 12:26:26,571 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=3.05 vs. limit=2.0 2023-04-27 12:26:52,611 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=387.0, num_to_drop=1, layers_to_drop={1} 2023-04-27 12:27:06,004 INFO [train.py:904] (1/8) Epoch 1, batch 400, loss[loss=0.9469, simple_loss=0.7745, pruned_loss=0.8917, over 17189.00 frames. ], tot_loss[loss=1.007, simple_loss=0.8498, pruned_loss=1.001, over 2881317.36 frames. ], batch size: 46, lr: 4.50e-02, grad_scale: 0.03125 2023-04-27 12:27:17,879 INFO [optim.py:368] (1/8) 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 12:27:46,053 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=439.0, num_to_drop=2, layers_to_drop={2, 3} 2023-04-27 12:27:56,277 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=448.0, num_to_drop=2, layers_to_drop={1, 2} 2023-04-27 12:27:57,949 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=28.51 vs. limit=5.0 2023-04-27 12:27:59,545 INFO [train.py:904] (1/8) Epoch 1, batch 450, loss[loss=0.9565, simple_loss=0.7735, pruned_loss=0.8917, over 17205.00 frames. ], tot_loss[loss=0.9804, simple_loss=0.8213, pruned_loss=0.9584, over 2986051.05 frames. ], batch size: 46, lr: 4.75e-02, grad_scale: 0.03125 2023-04-27 12:28:37,439 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=10.35 vs. limit=2.0 2023-04-27 12:28:51,183 INFO [train.py:904] (1/8) Epoch 1, batch 500, loss[loss=0.843, simple_loss=0.6853, pruned_loss=0.7482, over 16467.00 frames. ], tot_loss[loss=0.9646, simple_loss=0.802, pruned_loss=0.9277, over 3067613.70 frames. ], batch size: 146, lr: 4.99e-02, grad_scale: 0.0625 2023-04-27 12:29:01,368 INFO [optim.py:368] (1/8) 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 12:29:03,892 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.99 vs. limit=2.0 2023-04-27 12:29:38,125 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=4.47 vs. limit=2.0 2023-04-27 12:29:44,925 INFO [train.py:904] (1/8) Epoch 1, batch 550, loss[loss=0.8696, simple_loss=0.6998, pruned_loss=0.7652, over 16857.00 frames. ], tot_loss[loss=0.9527, simple_loss=0.7867, pruned_loss=0.9007, over 3128543.60 frames. ], batch size: 96, lr: 4.98e-02, grad_scale: 0.0625 2023-04-27 12:29:46,907 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=3.01 vs. limit=2.0 2023-04-27 12:29:56,818 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=21.34 vs. limit=5.0 2023-04-27 12:29:58,101 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=562.0, num_to_drop=1, layers_to_drop={1} 2023-04-27 12:30:03,848 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=568.0, num_to_drop=1, layers_to_drop={0} 2023-04-27 12:30:07,312 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=12.05 vs. limit=5.0 2023-04-27 12:30:26,878 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=590.0, num_to_drop=1, layers_to_drop={1} 2023-04-27 12:30:38,276 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=600.0, num_to_drop=2, layers_to_drop={0, 1} 2023-04-27 12:30:38,745 INFO [train.py:904] (1/8) Epoch 1, batch 600, loss[loss=0.8705, simple_loss=0.6957, pruned_loss=0.754, over 16846.00 frames. ], tot_loss[loss=0.9409, simple_loss=0.7724, pruned_loss=0.873, over 3174268.01 frames. ], batch size: 39, lr: 4.98e-02, grad_scale: 0.125 2023-04-27 12:30:45,001 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=2.22 vs. limit=2.0 2023-04-27 12:30:48,385 INFO [optim.py:368] (1/8) 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 12:31:02,523 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=623.0, num_to_drop=2, layers_to_drop={0, 1} 2023-04-27 12:31:07,588 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=629.0, num_to_drop=2, layers_to_drop={2, 3} 2023-04-27 12:31:24,888 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=6.07 vs. limit=5.0 2023-04-27 12:31:28,004 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=648.0, num_to_drop=1, layers_to_drop={1} 2023-04-27 12:31:31,002 INFO [train.py:904] (1/8) Epoch 1, batch 650, loss[loss=0.8864, simple_loss=0.7097, pruned_loss=0.7436, over 16473.00 frames. ], tot_loss[loss=0.9278, simple_loss=0.7584, pruned_loss=0.843, over 3211602.17 frames. ], batch size: 75, lr: 4.98e-02, grad_scale: 0.125 2023-04-27 12:31:31,351 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=651.0, num_to_drop=2, layers_to_drop={1, 3} 2023-04-27 12:31:32,143 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=652.0, num_to_drop=2, layers_to_drop={2, 3} 2023-04-27 12:31:40,310 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=6.03 vs. limit=5.0 2023-04-27 12:31:54,794 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=2.99 vs. limit=2.0 2023-04-27 12:32:22,394 INFO [train.py:904] (1/8) Epoch 1, batch 700, loss[loss=0.8336, simple_loss=0.6645, pruned_loss=0.6874, over 16754.00 frames. ], tot_loss[loss=0.9149, simple_loss=0.7467, pruned_loss=0.811, over 3233340.15 frames. ], batch size: 102, lr: 4.98e-02, grad_scale: 0.25 2023-04-27 12:32:31,774 INFO [optim.py:368] (1/8) 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 12:33:00,825 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.6180, 3.2791, 3.4853, 3.8905, 3.2502, 3.4155, 3.2090, 3.3455], device='cuda:1'), covar=tensor([0.6776, 0.8415, 0.9227, 0.6435, 1.0898, 0.6250, 0.7263, 1.1335], device='cuda:1'), in_proj_covar=tensor([0.0049, 0.0048, 0.0054, 0.0041, 0.0053, 0.0043, 0.0051, 0.0049], device='cuda:1'), out_proj_covar=tensor([4.1403e-05, 4.5042e-05, 5.1100e-05, 3.6571e-05, 4.5586e-05, 3.9843e-05, 4.6218e-05, 4.5838e-05], device='cuda:1') 2023-04-27 12:33:00,831 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=739.0, num_to_drop=2, layers_to_drop={0, 1} 2023-04-27 12:33:05,887 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=743.0, num_to_drop=2, layers_to_drop={1, 3} 2023-04-27 12:33:14,014 INFO [train.py:904] (1/8) Epoch 1, batch 750, loss[loss=0.8008, simple_loss=0.6618, pruned_loss=0.6012, over 16775.00 frames. ], tot_loss[loss=0.8992, simple_loss=0.7348, pruned_loss=0.7744, over 3251547.59 frames. ], batch size: 124, lr: 4.97e-02, grad_scale: 0.25 2023-04-27 12:33:51,295 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=787.0, num_to_drop=1, layers_to_drop={1} 2023-04-27 12:34:06,024 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.96 vs. limit=2.0 2023-04-27 12:34:06,256 INFO [train.py:904] (1/8) Epoch 1, batch 800, loss[loss=0.6903, simple_loss=0.5848, pruned_loss=0.4834, over 16826.00 frames. ], tot_loss[loss=0.8714, simple_loss=0.7156, pruned_loss=0.7264, over 3268745.32 frames. ], batch size: 39, lr: 4.97e-02, grad_scale: 0.5 2023-04-27 12:34:17,530 INFO [optim.py:368] (1/8) Clipping_scale=2.0, grad-norm quartiles 1.716e+02 2.855e+02 4.131e+02 5.648e+02 8.889e+02, threshold=8.261e+02, percent-clipped=19.0 2023-04-27 12:34:54,107 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 1, batch 850, loss[loss=0.7068, simple_loss=0.6071, pruned_loss=0.4738, over 16635.00 frames. ], tot_loss[loss=0.8397, simple_loss=0.6942, pruned_loss=0.6765, over 3280685.35 frames. ], batch size: 62, lr: 4.96e-02, grad_scale: 0.5 2023-04-27 12:35:47,096 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=8.14 vs. limit=2.0 2023-04-27 12:35:50,488 INFO [train.py:904] (1/8) Epoch 1, batch 900, loss[loss=0.6619, simple_loss=0.5637, pruned_loss=0.4456, over 16785.00 frames. ], tot_loss[loss=0.8044, simple_loss=0.6706, pruned_loss=0.6257, over 3287408.66 frames. ], batch size: 83, lr: 4.96e-02, grad_scale: 1.0 2023-04-27 12:35:56,985 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=908.0, num_to_drop=2, layers_to_drop={1, 3} 2023-04-27 12:36:00,422 INFO [optim.py:368] (1/8) 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 12:36:08,838 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=918.0, num_to_drop=2, layers_to_drop={0, 2} 2023-04-27 12:36:14,425 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=924.0, num_to_drop=2, layers_to_drop={0, 2} 2023-04-27 12:36:37,787 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=946.0, num_to_drop=2, layers_to_drop={0, 2} 2023-04-27 12:36:42,348 INFO [train.py:904] (1/8) Epoch 1, batch 950, loss[loss=0.6442, simple_loss=0.5716, pruned_loss=0.3949, over 17217.00 frames. ], tot_loss[loss=0.773, simple_loss=0.6504, pruned_loss=0.5801, over 3302224.61 frames. ], batch size: 46, lr: 4.96e-02, grad_scale: 1.0 2023-04-27 12:36:44,141 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=952.0, num_to_drop=2, layers_to_drop={0, 2} 2023-04-27 12:36:51,595 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=5.34 vs. limit=5.0 2023-04-27 12:37:11,200 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.98 vs. limit=2.0 2023-04-27 12:37:35,472 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=1000.0, num_to_drop=1, layers_to_drop={1} 2023-04-27 12:37:35,504 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([5.7459, 5.6702, 5.7420, 5.6705, 5.6649, 5.7747, 5.7622, 5.2991], device='cuda:1'), covar=tensor([0.0524, 0.1072, 0.0537, 0.1127, 0.0783, 0.0573, 0.0672, 0.4008], device='cuda:1'), in_proj_covar=tensor([0.0058, 0.0072, 0.0059, 0.0071, 0.0062, 0.0060, 0.0062, 0.0090], device='cuda:1'), out_proj_covar=tensor([5.1688e-05, 6.2321e-05, 5.3138e-05, 5.6678e-05, 5.8478e-05, 5.3275e-05, 5.6921e-05, 7.4848e-05], device='cuda:1') 2023-04-27 12:37:35,994 INFO [train.py:904] (1/8) Epoch 1, batch 1000, loss[loss=0.6094, simple_loss=0.529, pruned_loss=0.3866, over 16826.00 frames. ], tot_loss[loss=0.7393, simple_loss=0.6276, pruned_loss=0.5364, over 3309661.73 frames. ], batch size: 102, lr: 4.95e-02, grad_scale: 1.0 2023-04-27 12:37:46,288 INFO [optim.py:368] (1/8) 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 12:38:05,256 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=5.16 vs. limit=5.0 2023-04-27 12:38:21,603 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=1043.0, num_to_drop=2, layers_to_drop={0, 2} 2023-04-27 12:38:29,481 INFO [train.py:904] (1/8) Epoch 1, batch 1050, loss[loss=0.5767, simple_loss=0.5218, pruned_loss=0.335, over 16756.00 frames. ], tot_loss[loss=0.7085, simple_loss=0.6074, pruned_loss=0.4971, over 3317540.12 frames. ], batch size: 76, lr: 4.95e-02, grad_scale: 1.0 2023-04-27 12:39:12,051 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=1091.0, num_to_drop=1, layers_to_drop={1} 2023-04-27 12:39:21,405 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=5.48 vs. limit=5.0 2023-04-27 12:39:21,796 INFO [train.py:904] (1/8) Epoch 1, batch 1100, loss[loss=0.5764, simple_loss=0.5111, pruned_loss=0.3459, over 16887.00 frames. ], tot_loss[loss=0.6801, simple_loss=0.5887, pruned_loss=0.4618, over 3322327.48 frames. ], batch size: 109, lr: 4.94e-02, grad_scale: 1.0 2023-04-27 12:39:33,204 INFO [optim.py:368] (1/8) 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] (1/8) Epoch 1, batch 1150, loss[loss=0.5863, simple_loss=0.5279, pruned_loss=0.3404, over 16389.00 frames. ], tot_loss[loss=0.6545, simple_loss=0.5723, pruned_loss=0.4308, over 3316389.57 frames. ], batch size: 146, lr: 4.94e-02, grad_scale: 1.0 2023-04-27 12:40:19,571 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=1153.0, num_to_drop=1, layers_to_drop={0} 2023-04-27 12:40:41,029 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=2.00 vs. limit=2.0 2023-04-27 12:41:07,799 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=5.26 vs. limit=5.0 2023-04-27 12:41:10,401 INFO [train.py:904] (1/8) Epoch 1, batch 1200, loss[loss=0.5652, simple_loss=0.5015, pruned_loss=0.3345, over 16770.00 frames. ], tot_loss[loss=0.6308, simple_loss=0.5574, pruned_loss=0.4026, over 3319281.70 frames. ], batch size: 83, lr: 4.93e-02, grad_scale: 2.0 2023-04-27 12:41:12,653 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=1203.0, num_to_drop=2, layers_to_drop={0, 1} 2023-04-27 12:41:21,557 INFO [optim.py:368] (1/8) 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 12:41:25,698 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=1214.0, num_to_drop=2, layers_to_drop={0, 3} 2023-04-27 12:41:29,940 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=1218.0, num_to_drop=2, layers_to_drop={0, 2} 2023-04-27 12:41:34,827 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=1224.0, num_to_drop=1, layers_to_drop={1} 2023-04-27 12:41:57,147 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=1246.0, num_to_drop=1, layers_to_drop={1} 2023-04-27 12:42:03,009 INFO [train.py:904] (1/8) Epoch 1, batch 1250, loss[loss=0.506, simple_loss=0.4629, pruned_loss=0.2834, over 16264.00 frames. ], tot_loss[loss=0.6116, simple_loss=0.5447, pruned_loss=0.3804, over 3319205.87 frames. ], batch size: 165, lr: 4.92e-02, grad_scale: 2.0 2023-04-27 12:42:17,897 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=1266.0, num_to_drop=1, layers_to_drop={0} 2023-04-27 12:42:24,722 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=1272.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 12:42:49,319 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=1294.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 12:42:56,580 INFO [train.py:904] (1/8) Epoch 1, batch 1300, loss[loss=0.5571, simple_loss=0.5162, pruned_loss=0.3046, over 16869.00 frames. ], tot_loss[loss=0.5935, simple_loss=0.5328, pruned_loss=0.3604, over 3323784.40 frames. ], batch size: 42, lr: 4.92e-02, grad_scale: 2.0 2023-04-27 12:43:07,776 INFO [optim.py:368] (1/8) 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] (1/8) Epoch 1, batch 1350, loss[loss=0.4483, simple_loss=0.4374, pruned_loss=0.2241, over 17241.00 frames. ], tot_loss[loss=0.5767, simple_loss=0.5227, pruned_loss=0.3415, over 3324580.88 frames. ], batch size: 45, lr: 4.91e-02, grad_scale: 2.0 2023-04-27 12:44:49,663 INFO [train.py:904] (1/8) Epoch 1, batch 1400, loss[loss=0.55, simple_loss=0.5032, pruned_loss=0.3054, over 12433.00 frames. ], tot_loss[loss=0.5619, simple_loss=0.5133, pruned_loss=0.3259, over 3313659.93 frames. ], batch size: 247, lr: 4.91e-02, grad_scale: 2.0 2023-04-27 12:45:00,101 INFO [optim.py:368] (1/8) 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 12:45:27,157 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 1, batch 1450, loss[loss=0.538, simple_loss=0.5097, pruned_loss=0.2831, over 17059.00 frames. ], tot_loss[loss=0.5494, simple_loss=0.5057, pruned_loss=0.3126, over 3315418.06 frames. ], batch size: 55, lr: 4.90e-02, grad_scale: 2.0 2023-04-27 12:46:14,034 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.4210, 4.3173, 4.3436, 4.4657, 4.1189, 4.4851, 4.3914, 4.0331], device='cuda:1'), covar=tensor([0.0453, 0.0278, 0.0474, 0.0270, 0.0360, 0.0300, 0.0284, 0.0378], device='cuda:1'), in_proj_covar=tensor([0.0067, 0.0059, 0.0074, 0.0061, 0.0066, 0.0066, 0.0059, 0.0063], device='cuda:1'), out_proj_covar=tensor([6.1651e-05, 4.9631e-05, 7.0212e-05, 5.7294e-05, 6.1972e-05, 5.9197e-05, 5.4222e-05, 5.6622e-05], device='cuda:1') 2023-04-27 12:46:34,497 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=1495.0, num_to_drop=2, layers_to_drop={0, 3} 2023-04-27 12:46:41,351 INFO [train.py:904] (1/8) Epoch 1, batch 1500, loss[loss=0.5809, simple_loss=0.5216, pruned_loss=0.329, over 12294.00 frames. ], tot_loss[loss=0.5381, simple_loss=0.4983, pruned_loss=0.3015, over 3301374.61 frames. ], batch size: 247, lr: 4.89e-02, grad_scale: 2.0 2023-04-27 12:46:44,272 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=1503.0, num_to_drop=1, layers_to_drop={0} 2023-04-27 12:46:51,080 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=1509.0, num_to_drop=2, layers_to_drop={1, 3} 2023-04-27 12:46:52,703 INFO [optim.py:368] (1/8) Clipping_scale=2.0, grad-norm quartiles 2.422e+02 5.181e+02 6.478e+02 8.944e+02 1.260e+03, threshold=1.296e+03, percent-clipped=1.0 2023-04-27 12:47:38,770 INFO [train.py:904] (1/8) Epoch 1, batch 1550, loss[loss=0.4754, simple_loss=0.486, pruned_loss=0.2228, over 17018.00 frames. ], tot_loss[loss=0.5279, simple_loss=0.492, pruned_loss=0.2915, over 3304880.51 frames. ], batch size: 50, lr: 4.89e-02, grad_scale: 2.0 2023-04-27 12:47:39,031 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=1551.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 12:47:56,078 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=1566.0, num_to_drop=1, layers_to_drop={0} 2023-04-27 12:48:22,117 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.7913, 4.7099, 4.9585, 4.9216, 4.6307, 4.8159, 4.3979, 4.8769], device='cuda:1'), covar=tensor([0.1614, 0.1287, 0.1049, 0.1395, 0.2389, 0.1079, 0.1911, 0.1897], device='cuda:1'), in_proj_covar=tensor([0.0032, 0.0033, 0.0034, 0.0026, 0.0029, 0.0032, 0.0027, 0.0026], device='cuda:1'), out_proj_covar=tensor([2.3312e-05, 2.3992e-05, 2.5469e-05, 1.9997e-05, 2.2481e-05, 2.3097e-05, 2.1425e-05, 2.1681e-05], device='cuda:1') 2023-04-27 12:48:35,335 INFO [train.py:904] (1/8) Epoch 1, batch 1600, loss[loss=0.5346, simple_loss=0.5142, pruned_loss=0.2757, over 16592.00 frames. ], tot_loss[loss=0.5212, simple_loss=0.4894, pruned_loss=0.2834, over 3310329.00 frames. ], batch size: 62, lr: 4.88e-02, grad_scale: 4.0 2023-04-27 12:48:39,700 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.6372, 4.4819, 4.4758, 4.5372, 4.0904, 4.3665, 4.5477, 4.0650], device='cuda:1'), covar=tensor([0.0468, 0.0328, 0.0408, 0.0245, 0.0564, 0.0547, 0.0337, 0.0465], device='cuda:1'), in_proj_covar=tensor([0.0063, 0.0058, 0.0072, 0.0058, 0.0064, 0.0065, 0.0056, 0.0060], device='cuda:1'), out_proj_covar=tensor([5.9560e-05, 4.7531e-05, 6.9671e-05, 5.4423e-05, 5.9566e-05, 5.7446e-05, 5.0928e-05, 5.3834e-05], device='cuda:1') 2023-04-27 12:48:47,163 INFO [optim.py:368] (1/8) 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 12:48:47,478 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=1611.0, num_to_drop=1, layers_to_drop={1} 2023-04-27 12:49:05,692 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=1627.0, num_to_drop=2, layers_to_drop={0, 2} 2023-04-27 12:49:32,287 INFO [train.py:904] (1/8) Epoch 1, batch 1650, loss[loss=0.5124, simple_loss=0.5046, pruned_loss=0.2563, over 16679.00 frames. ], tot_loss[loss=0.5161, simple_loss=0.4874, pruned_loss=0.2775, over 3312056.58 frames. ], batch size: 57, lr: 4.87e-02, grad_scale: 4.0 2023-04-27 12:49:42,355 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.8809, 3.9003, 3.6287, 3.4908, 3.9722, 3.4311, 3.7073, 3.6871], device='cuda:1'), covar=tensor([0.0461, 0.0353, 0.0609, 0.0486, 0.0423, 0.0723, 0.0490, 0.0492], device='cuda:1'), in_proj_covar=tensor([0.0070, 0.0064, 0.0063, 0.0063, 0.0069, 0.0073, 0.0072, 0.0067], device='cuda:1'), out_proj_covar=tensor([6.6447e-05, 5.8202e-05, 6.2281e-05, 6.0298e-05, 6.4334e-05, 7.0081e-05, 6.5216e-05, 6.1940e-05], device='cuda:1') 2023-04-27 12:49:57,824 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=1672.0, num_to_drop=2, layers_to_drop={2, 3} 2023-04-27 12:50:29,818 INFO [train.py:904] (1/8) Epoch 1, batch 1700, loss[loss=0.4821, simple_loss=0.4877, pruned_loss=0.2332, over 17035.00 frames. ], tot_loss[loss=0.5085, simple_loss=0.4845, pruned_loss=0.2695, over 3304902.84 frames. ], batch size: 53, lr: 4.86e-02, grad_scale: 4.0 2023-04-27 12:50:34,235 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=2.24 vs. limit=2.0 2023-04-27 12:50:41,826 INFO [optim.py:368] (1/8) 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 12:50:45,771 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.3649, 4.1614, 4.4529, 4.5046, 4.7371, 4.2999, 4.1332, 4.6625], device='cuda:1'), covar=tensor([0.0441, 0.0518, 0.0645, 0.0459, 0.0368, 0.0524, 0.0626, 0.0310], device='cuda:1'), in_proj_covar=tensor([0.0092, 0.0095, 0.0107, 0.0105, 0.0088, 0.0096, 0.0106, 0.0088], device='cuda:1'), out_proj_covar=tensor([7.3855e-05, 9.0759e-05, 1.1068e-04, 9.1086e-05, 8.7085e-05, 8.7208e-05, 9.3540e-05, 7.8500e-05], device='cuda:1') 2023-04-27 12:50:56,376 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.7668, 3.6263, 3.4215, 4.1701, 3.5164, 3.9053, 3.6679, 3.5713], device='cuda:1'), covar=tensor([0.1532, 0.1321, 0.1297, 0.0544, 0.1395, 0.0733, 0.0939, 0.1213], device='cuda:1'), in_proj_covar=tensor([0.0085, 0.0077, 0.0093, 0.0075, 0.0087, 0.0079, 0.0084, 0.0083], device='cuda:1'), out_proj_covar=tensor([7.4555e-05, 6.5368e-05, 7.8601e-05, 6.2765e-05, 7.6431e-05, 6.5826e-05, 7.6113e-05, 7.4716e-05], device='cuda:1') 2023-04-27 12:51:28,763 INFO [train.py:904] (1/8) Epoch 1, batch 1750, loss[loss=0.5071, simple_loss=0.4938, pruned_loss=0.2584, over 16132.00 frames. ], tot_loss[loss=0.4982, simple_loss=0.4795, pruned_loss=0.2602, over 3312907.70 frames. ], batch size: 164, lr: 4.86e-02, grad_scale: 4.0 2023-04-27 12:52:15,501 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=1790.0, num_to_drop=1, layers_to_drop={2} 2023-04-27 12:52:27,926 INFO [train.py:904] (1/8) Epoch 1, batch 1800, loss[loss=0.4855, simple_loss=0.4764, pruned_loss=0.2456, over 16240.00 frames. ], tot_loss[loss=0.4906, simple_loss=0.4761, pruned_loss=0.2532, over 3321715.21 frames. ], batch size: 165, lr: 4.85e-02, grad_scale: 4.0 2023-04-27 12:52:37,425 INFO [zipformer.py:625] (1/8) 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] (1/8) 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 12:53:10,944 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 1, batch 1850, loss[loss=0.4314, simple_loss=0.4448, pruned_loss=0.2062, over 16812.00 frames. ], tot_loss[loss=0.4828, simple_loss=0.4723, pruned_loss=0.2467, over 3327468.84 frames. ], batch size: 42, lr: 4.84e-02, grad_scale: 4.0 2023-04-27 12:53:32,838 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=1857.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 12:53:32,945 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=1857.0, num_to_drop=1, layers_to_drop={1} 2023-04-27 12:53:48,685 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=1870.0, num_to_drop=1, layers_to_drop={1} 2023-04-27 12:53:50,656 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.6555, 4.6752, 4.1916, 4.1528, 4.7179, 4.7811, 4.0560, 4.7040], device='cuda:1'), covar=tensor([0.0160, 0.0208, 0.0288, 0.0347, 0.0106, 0.0138, 0.0241, 0.0172], device='cuda:1'), in_proj_covar=tensor([0.0051, 0.0056, 0.0068, 0.0064, 0.0052, 0.0056, 0.0067, 0.0064], device='cuda:1'), out_proj_covar=tensor([4.3820e-05, 5.2598e-05, 6.7858e-05, 5.8687e-05, 4.4851e-05, 5.1636e-05, 6.4725e-05, 6.0091e-05], device='cuda:1') 2023-04-27 12:54:21,400 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=1899.0, num_to_drop=2, layers_to_drop={1, 3} 2023-04-27 12:54:23,725 INFO [train.py:904] (1/8) Epoch 1, batch 1900, loss[loss=0.3974, simple_loss=0.4207, pruned_loss=0.1848, over 17232.00 frames. ], tot_loss[loss=0.4717, simple_loss=0.4662, pruned_loss=0.2381, over 3324754.38 frames. ], batch size: 45, lr: 4.83e-02, grad_scale: 4.0 2023-04-27 12:54:36,242 INFO [optim.py:368] (1/8) 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 12:54:43,885 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=1918.0, num_to_drop=2, layers_to_drop={2, 3} 2023-04-27 12:54:49,110 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=1922.0, num_to_drop=1, layers_to_drop={3} 2023-04-27 12:55:00,132 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=1931.0, num_to_drop=2, layers_to_drop={0, 2} 2023-04-27 12:55:22,814 INFO [train.py:904] (1/8) Epoch 1, batch 1950, loss[loss=0.4472, simple_loss=0.4531, pruned_loss=0.2199, over 16861.00 frames. ], tot_loss[loss=0.4629, simple_loss=0.4622, pruned_loss=0.2311, over 3324240.35 frames. ], batch size: 116, lr: 4.83e-02, grad_scale: 4.0 2023-04-27 12:55:42,108 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=1967.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 12:56:23,561 INFO [train.py:904] (1/8) Epoch 1, batch 2000, loss[loss=0.3888, simple_loss=0.4126, pruned_loss=0.1825, over 16869.00 frames. ], tot_loss[loss=0.4548, simple_loss=0.4573, pruned_loss=0.2255, over 3331336.27 frames. ], batch size: 42, lr: 4.82e-02, grad_scale: 8.0 2023-04-27 12:56:36,802 INFO [optim.py:368] (1/8) Clipping_scale=2.0, grad-norm quartiles 3.491e+02 5.646e+02 7.242e+02 8.659e+02 1.834e+03, threshold=1.448e+03, percent-clipped=2.0 2023-04-27 12:57:21,188 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.7688, 4.0984, 3.8983, 4.3393, 4.1595, 1.7517, 4.2022, 4.1834], device='cuda:1'), covar=tensor([0.3023, 0.1242, 0.2093, 0.0580, 0.2380, 0.8004, 0.0662, 0.0388], device='cuda:1'), in_proj_covar=tensor([0.0047, 0.0031, 0.0048, 0.0035, 0.0028, 0.0061, 0.0033, 0.0023], device='cuda:1'), out_proj_covar=tensor([4.4353e-05, 2.8237e-05, 4.3037e-05, 2.6833e-05, 2.7034e-05, 5.3311e-05, 2.6143e-05, 2.2071e-05], device='cuda:1') 2023-04-27 12:57:27,604 INFO [train.py:904] (1/8) Epoch 1, batch 2050, loss[loss=0.3955, simple_loss=0.4274, pruned_loss=0.1818, over 17077.00 frames. ], tot_loss[loss=0.4448, simple_loss=0.4515, pruned_loss=0.2186, over 3335120.87 frames. ], batch size: 53, lr: 4.81e-02, grad_scale: 8.0 2023-04-27 12:58:09,485 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.97 vs. limit=2.0 2023-04-27 12:58:18,989 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=2090.0, num_to_drop=2, layers_to_drop={1, 2} 2023-04-27 12:58:26,084 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=4.43 vs. limit=5.0 2023-04-27 12:58:32,223 INFO [train.py:904] (1/8) Epoch 1, batch 2100, loss[loss=0.4725, simple_loss=0.4698, pruned_loss=0.2376, over 16421.00 frames. ], tot_loss[loss=0.4357, simple_loss=0.4473, pruned_loss=0.2117, over 3338324.96 frames. ], batch size: 146, lr: 4.80e-02, grad_scale: 16.0 2023-04-27 12:58:36,882 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=4.26 vs. limit=5.0 2023-04-27 12:58:45,932 INFO [optim.py:368] (1/8) 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 12:59:19,749 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 1, batch 2150, loss[loss=0.3342, simple_loss=0.3771, pruned_loss=0.1457, over 16767.00 frames. ], tot_loss[loss=0.4321, simple_loss=0.4455, pruned_loss=0.2091, over 3333990.49 frames. ], batch size: 39, lr: 4.79e-02, grad_scale: 16.0 2023-04-27 13:00:20,750 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.82 vs. limit=2.0 2023-04-27 13:00:32,100 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=2194.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 13:00:40,251 INFO [train.py:904] (1/8) Epoch 1, batch 2200, loss[loss=0.368, simple_loss=0.4187, pruned_loss=0.1587, over 17124.00 frames. ], tot_loss[loss=0.4265, simple_loss=0.4427, pruned_loss=0.2049, over 3327067.35 frames. ], batch size: 48, lr: 4.78e-02, grad_scale: 16.0 2023-04-27 13:00:53,195 INFO [optim.py:368] (1/8) 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 13:00:56,680 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=2213.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 13:01:07,609 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=2222.0, num_to_drop=2, layers_to_drop={1, 2} 2023-04-27 13:01:11,971 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 1, batch 2250, loss[loss=0.3816, simple_loss=0.4339, pruned_loss=0.1647, over 17264.00 frames. ], tot_loss[loss=0.4204, simple_loss=0.4392, pruned_loss=0.2006, over 3322867.40 frames. ], batch size: 52, lr: 4.77e-02, grad_scale: 16.0 2023-04-27 13:02:05,704 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=2267.0, num_to_drop=1, layers_to_drop={2} 2023-04-27 13:02:08,814 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 1, batch 2300, loss[loss=0.3739, simple_loss=0.4297, pruned_loss=0.159, over 17110.00 frames. ], tot_loss[loss=0.4134, simple_loss=0.4357, pruned_loss=0.1954, over 3329638.22 frames. ], batch size: 49, lr: 4.77e-02, grad_scale: 16.0 2023-04-27 13:02:57,985 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.0270, 3.9835, 3.7951, 4.3462, 3.5521, 4.1639, 3.1411, 4.4538], device='cuda:1'), covar=tensor([0.0714, 0.0845, 0.0889, 0.0810, 0.3813, 0.0856, 0.2061, 0.1151], device='cuda:1'), in_proj_covar=tensor([0.0042, 0.0045, 0.0042, 0.0037, 0.0068, 0.0043, 0.0049, 0.0037], device='cuda:1'), out_proj_covar=tensor([3.4073e-05, 3.6782e-05, 3.2325e-05, 3.5658e-05, 6.1314e-05, 3.4872e-05, 4.3192e-05, 3.6867e-05], device='cuda:1') 2023-04-27 13:03:01,864 INFO [optim.py:368] (1/8) 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 13:03:07,886 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=2315.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 13:03:09,093 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.9535, 4.0416, 3.6304, 3.5045, 3.9423, 3.7087, 3.7634, 3.9252], device='cuda:1'), covar=tensor([0.0239, 0.0273, 0.0388, 0.0353, 0.0301, 0.0327, 0.0468, 0.0270], device='cuda:1'), in_proj_covar=tensor([0.0050, 0.0050, 0.0044, 0.0048, 0.0050, 0.0052, 0.0056, 0.0050], device='cuda:1'), out_proj_covar=tensor([4.7305e-05, 4.8652e-05, 4.2863e-05, 4.6538e-05, 4.4769e-05, 5.2325e-05, 5.5773e-05, 4.7977e-05], device='cuda:1') 2023-04-27 13:03:39,688 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=2340.0, num_to_drop=1, layers_to_drop={1} 2023-04-27 13:03:53,136 INFO [train.py:904] (1/8) Epoch 1, batch 2350, loss[loss=0.4106, simple_loss=0.4175, pruned_loss=0.2018, over 16863.00 frames. ], tot_loss[loss=0.4102, simple_loss=0.4335, pruned_loss=0.1933, over 3327642.76 frames. ], batch size: 96, lr: 4.76e-02, grad_scale: 16.0 2023-04-27 13:04:21,773 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.9020, 4.9675, 5.0765, 5.2046, 5.4432, 4.9823, 4.8922, 5.3052], device='cuda:1'), covar=tensor([0.0244, 0.0235, 0.0413, 0.0311, 0.0245, 0.0257, 0.0332, 0.0156], device='cuda:1'), in_proj_covar=tensor([0.0108, 0.0097, 0.0114, 0.0114, 0.0108, 0.0104, 0.0108, 0.0093], device='cuda:1'), out_proj_covar=tensor([9.9248e-05, 1.0919e-04, 1.2225e-04, 1.1191e-04, 1.1613e-04, 1.0713e-04, 1.0726e-04, 8.7909e-05], device='cuda:1') 2023-04-27 13:04:54,412 INFO [train.py:904] (1/8) Epoch 1, batch 2400, loss[loss=0.3145, simple_loss=0.3624, pruned_loss=0.1333, over 16939.00 frames. ], tot_loss[loss=0.4076, simple_loss=0.4328, pruned_loss=0.1911, over 3326960.84 frames. ], batch size: 41, lr: 4.75e-02, grad_scale: 16.0 2023-04-27 13:04:55,616 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=2401.0, num_to_drop=2, layers_to_drop={1, 2} 2023-04-27 13:05:07,295 INFO [optim.py:368] (1/8) 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 13:05:55,899 INFO [train.py:904] (1/8) Epoch 1, batch 2450, loss[loss=0.3973, simple_loss=0.4422, pruned_loss=0.1762, over 17291.00 frames. ], tot_loss[loss=0.4021, simple_loss=0.4298, pruned_loss=0.1871, over 3335432.37 frames. ], batch size: 52, lr: 4.74e-02, grad_scale: 16.0 2023-04-27 13:06:51,030 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=2494.0, num_to_drop=1, layers_to_drop={2} 2023-04-27 13:06:56,508 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=4.79 vs. limit=5.0 2023-04-27 13:06:58,990 INFO [train.py:904] (1/8) Epoch 1, batch 2500, loss[loss=0.439, simple_loss=0.4596, pruned_loss=0.2092, over 15642.00 frames. ], tot_loss[loss=0.3972, simple_loss=0.4262, pruned_loss=0.184, over 3334263.89 frames. ], batch size: 191, lr: 4.73e-02, grad_scale: 16.0 2023-04-27 13:07:11,499 INFO [optim.py:368] (1/8) 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 13:07:14,714 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=2513.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 13:07:29,244 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=2526.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 13:07:50,934 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=2542.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 13:08:01,506 INFO [train.py:904] (1/8) Epoch 1, batch 2550, loss[loss=0.4147, simple_loss=0.4306, pruned_loss=0.1994, over 16733.00 frames. ], tot_loss[loss=0.3953, simple_loss=0.4255, pruned_loss=0.1825, over 3332569.75 frames. ], batch size: 124, lr: 4.72e-02, grad_scale: 16.0 2023-04-27 13:08:15,979 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=2561.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 13:08:32,315 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 1, batch 2600, loss[loss=0.3418, simple_loss=0.3943, pruned_loss=0.1446, over 16440.00 frames. ], tot_loss[loss=0.3924, simple_loss=0.4241, pruned_loss=0.1803, over 3326745.61 frames. ], batch size: 68, lr: 4.71e-02, grad_scale: 16.0 2023-04-27 13:09:20,088 INFO [optim.py:368] (1/8) 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 13:09:36,893 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([5.8495, 6.1471, 5.6659, 6.1901, 5.6139, 5.6932, 5.8689, 6.2720], device='cuda:1'), covar=tensor([0.0300, 0.0403, 0.0385, 0.0204, 0.0360, 0.0227, 0.0228, 0.0149], device='cuda:1'), in_proj_covar=tensor([0.0119, 0.0132, 0.0115, 0.0091, 0.0119, 0.0104, 0.0109, 0.0089], device='cuda:1'), out_proj_covar=tensor([1.0270e-04, 1.1809e-04, 9.7404e-05, 6.7818e-05, 9.9228e-05, 8.4570e-05, 9.2033e-05, 7.8546e-05], device='cuda:1') 2023-04-27 13:10:11,887 INFO [train.py:904] (1/8) Epoch 1, batch 2650, loss[loss=0.3354, simple_loss=0.3937, pruned_loss=0.1385, over 17280.00 frames. ], tot_loss[loss=0.3876, simple_loss=0.4216, pruned_loss=0.1768, over 3331941.00 frames. ], batch size: 52, lr: 4.70e-02, grad_scale: 16.0 2023-04-27 13:10:55,330 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=2685.0, num_to_drop=1, layers_to_drop={1} 2023-04-27 13:11:09,984 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=2696.0, num_to_drop=1, layers_to_drop={2} 2023-04-27 13:11:12,606 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=2.03 vs. limit=2.0 2023-04-27 13:11:15,038 INFO [train.py:904] (1/8) Epoch 1, batch 2700, loss[loss=0.3338, simple_loss=0.3889, pruned_loss=0.1393, over 16833.00 frames. ], tot_loss[loss=0.3834, simple_loss=0.4199, pruned_loss=0.1735, over 3325576.78 frames. ], batch size: 39, lr: 4.69e-02, grad_scale: 16.0 2023-04-27 13:11:29,641 INFO [optim.py:368] (1/8) Clipping_scale=2.0, grad-norm quartiles 2.771e+02 4.333e+02 5.264e+02 6.238e+02 1.242e+03, threshold=1.053e+03, percent-clipped=1.0 2023-04-27 13:12:13,544 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=2746.0, num_to_drop=2, layers_to_drop={2, 3} 2023-04-27 13:12:20,236 INFO [train.py:904] (1/8) Epoch 1, batch 2750, loss[loss=0.3862, simple_loss=0.4423, pruned_loss=0.165, over 17241.00 frames. ], tot_loss[loss=0.3788, simple_loss=0.4173, pruned_loss=0.1702, over 3336044.63 frames. ], batch size: 52, lr: 4.68e-02, grad_scale: 16.0 2023-04-27 13:13:23,780 INFO [train.py:904] (1/8) Epoch 1, batch 2800, loss[loss=0.3922, simple_loss=0.44, pruned_loss=0.1722, over 17038.00 frames. ], tot_loss[loss=0.3786, simple_loss=0.4167, pruned_loss=0.1702, over 3331601.55 frames. ], batch size: 55, lr: 4.67e-02, grad_scale: 16.0 2023-04-27 13:13:35,359 INFO [optim.py:368] (1/8) 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 13:13:54,353 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=2824.0, num_to_drop=1, layers_to_drop={1} 2023-04-27 13:14:26,001 INFO [train.py:904] (1/8) Epoch 1, batch 2850, loss[loss=0.3466, simple_loss=0.3789, pruned_loss=0.1572, over 16855.00 frames. ], tot_loss[loss=0.3742, simple_loss=0.4129, pruned_loss=0.1678, over 3332856.21 frames. ], batch size: 90, lr: 4.66e-02, grad_scale: 16.0 2023-04-27 13:14:52,612 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.2690, 3.3065, 3.2033, 2.9076, 3.2486, 3.0284, 3.2467, 3.3052], device='cuda:1'), covar=tensor([0.0178, 0.0195, 0.0228, 0.0248, 0.0222, 0.0349, 0.0273, 0.0211], device='cuda:1'), in_proj_covar=tensor([0.0046, 0.0039, 0.0039, 0.0044, 0.0043, 0.0045, 0.0049, 0.0043], device='cuda:1'), out_proj_covar=tensor([5.0097e-05, 4.5074e-05, 4.4352e-05, 4.7853e-05, 4.6489e-05, 5.7456e-05, 5.5329e-05, 4.7238e-05], device='cuda:1') 2023-04-27 13:15:05,456 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([5.4464, 5.0171, 5.3573, 5.3920, 4.7979, 5.1430, 5.2630, 4.8542], device='cuda:1'), covar=tensor([0.0186, 0.0200, 0.0169, 0.0095, 0.0713, 0.0226, 0.0162, 0.0216], device='cuda:1'), in_proj_covar=tensor([0.0088, 0.0073, 0.0111, 0.0081, 0.0131, 0.0093, 0.0082, 0.0088], device='cuda:1'), out_proj_covar=tensor([1.0966e-04, 8.1205e-05, 1.3408e-04, 9.2466e-05, 1.6555e-04, 1.1769e-04, 9.9316e-05, 1.1458e-04], device='cuda:1') 2023-04-27 13:15:09,574 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=2885.0, num_to_drop=2, layers_to_drop={1, 3} 2023-04-27 13:15:27,286 INFO [train.py:904] (1/8) Epoch 1, batch 2900, loss[loss=0.4407, simple_loss=0.4367, pruned_loss=0.2224, over 16149.00 frames. ], tot_loss[loss=0.3756, simple_loss=0.4116, pruned_loss=0.1698, over 3325519.87 frames. ], batch size: 164, lr: 4.65e-02, grad_scale: 16.0 2023-04-27 13:15:40,740 INFO [optim.py:368] (1/8) 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 13:16:13,947 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=2937.0, num_to_drop=1, layers_to_drop={0} 2023-04-27 13:16:32,274 INFO [train.py:904] (1/8) Epoch 1, batch 2950, loss[loss=0.3658, simple_loss=0.3928, pruned_loss=0.1694, over 16725.00 frames. ], tot_loss[loss=0.3726, simple_loss=0.4082, pruned_loss=0.1685, over 3326753.15 frames. ], batch size: 89, lr: 4.64e-02, grad_scale: 16.0 2023-04-27 13:17:29,885 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=2996.0, num_to_drop=1, layers_to_drop={0} 2023-04-27 13:17:32,171 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=2998.0, num_to_drop=2, layers_to_drop={0, 1} 2023-04-27 13:17:35,317 INFO [train.py:904] (1/8) Epoch 1, batch 3000, loss[loss=0.3781, simple_loss=0.4251, pruned_loss=0.1655, over 16720.00 frames. ], tot_loss[loss=0.3708, simple_loss=0.407, pruned_loss=0.1673, over 3329836.33 frames. ], batch size: 62, lr: 4.63e-02, grad_scale: 16.0 2023-04-27 13:17:35,317 INFO [train.py:929] (1/8) Computing validation loss 2023-04-27 13:17:45,053 INFO [train.py:938] (1/8) Epoch 1, validation: loss=0.2847, simple_loss=0.3895, pruned_loss=0.08992, over 944034.00 frames. 2023-04-27 13:17:45,054 INFO [train.py:939] (1/8) Maximum memory allocated so far is 15930MB 2023-04-27 13:17:59,838 INFO [optim.py:368] (1/8) 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 13:18:07,459 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.79 vs. limit=2.0 2023-04-27 13:18:09,277 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=3019.0, num_to_drop=1, layers_to_drop={1} 2023-04-27 13:18:36,923 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=3041.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 13:18:41,577 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 1, batch 3050, loss[loss=0.4721, simple_loss=0.463, pruned_loss=0.2406, over 12325.00 frames. ], tot_loss[loss=0.3683, simple_loss=0.4053, pruned_loss=0.1656, over 3329927.90 frames. ], batch size: 246, lr: 4.62e-02, grad_scale: 16.0 2023-04-27 13:19:05,993 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=7.56 vs. limit=5.0 2023-04-27 13:19:27,676 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=3080.0, num_to_drop=2, layers_to_drop={2, 3} 2023-04-27 13:19:33,330 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.5001, 2.8533, 3.0258, 1.5685, 2.9918, 3.0280, 3.1245, 2.4473], device='cuda:1'), covar=tensor([0.0292, 0.0230, 0.0200, 0.0534, 0.0226, 0.0102, 0.0131, 0.0345], device='cuda:1'), in_proj_covar=tensor([0.0029, 0.0033, 0.0035, 0.0036, 0.0032, 0.0034, 0.0035, 0.0033], device='cuda:1'), out_proj_covar=tensor([3.7128e-05, 3.6832e-05, 4.0065e-05, 3.9103e-05, 3.6010e-05, 3.6926e-05, 3.5999e-05, 3.8045e-05], device='cuda:1') 2023-04-27 13:19:53,976 INFO [train.py:904] (1/8) Epoch 1, batch 3100, loss[loss=0.3336, simple_loss=0.3884, pruned_loss=0.1394, over 17236.00 frames. ], tot_loss[loss=0.3662, simple_loss=0.4038, pruned_loss=0.1643, over 3326160.61 frames. ], batch size: 45, lr: 4.61e-02, grad_scale: 16.0 2023-04-27 13:20:07,651 INFO [optim.py:368] (1/8) 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 13:20:50,387 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=4.93 vs. limit=5.0 2023-04-27 13:21:00,201 INFO [train.py:904] (1/8) Epoch 1, batch 3150, loss[loss=0.355, simple_loss=0.4144, pruned_loss=0.1478, over 17254.00 frames. ], tot_loss[loss=0.3633, simple_loss=0.4017, pruned_loss=0.1625, over 3325386.26 frames. ], batch size: 52, lr: 4.60e-02, grad_scale: 16.0 2023-04-27 13:21:38,516 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 1, batch 3200, loss[loss=0.3553, simple_loss=0.4035, pruned_loss=0.1536, over 16431.00 frames. ], tot_loss[loss=0.3593, simple_loss=0.4, pruned_loss=0.1593, over 3335208.94 frames. ], batch size: 68, lr: 4.59e-02, grad_scale: 16.0 2023-04-27 13:22:12,211 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=4.97 vs. limit=5.0 2023-04-27 13:22:17,347 INFO [optim.py:368] (1/8) Clipping_scale=2.0, grad-norm quartiles 3.273e+02 4.612e+02 6.158e+02 7.733e+02 1.150e+03, threshold=1.232e+03, percent-clipped=3.0 2023-04-27 13:23:07,319 INFO [train.py:904] (1/8) Epoch 1, batch 3250, loss[loss=0.2864, simple_loss=0.3424, pruned_loss=0.1152, over 16144.00 frames. ], tot_loss[loss=0.3559, simple_loss=0.3978, pruned_loss=0.157, over 3336648.25 frames. ], batch size: 36, lr: 4.58e-02, grad_scale: 16.0 2023-04-27 13:23:56,333 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=3288.0, num_to_drop=1, layers_to_drop={1} 2023-04-27 13:24:01,881 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 1, batch 3300, loss[loss=0.3473, simple_loss=0.3854, pruned_loss=0.1547, over 16710.00 frames. ], tot_loss[loss=0.3531, simple_loss=0.3965, pruned_loss=0.1549, over 3342498.07 frames. ], batch size: 89, lr: 4.57e-02, grad_scale: 16.0 2023-04-27 13:24:25,122 INFO [optim.py:368] (1/8) 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 13:24:25,568 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.2440, 4.3307, 4.0511, 2.5733, 3.6673, 4.0403, 3.4734, 4.2528], device='cuda:1'), covar=tensor([0.0137, 0.0158, 0.0202, 0.1232, 0.0522, 0.0170, 0.0282, 0.0148], device='cuda:1'), in_proj_covar=tensor([0.0038, 0.0040, 0.0045, 0.0072, 0.0041, 0.0039, 0.0039, 0.0043], device='cuda:1'), out_proj_covar=tensor([4.2601e-05, 4.4550e-05, 5.0768e-05, 7.8411e-05, 5.2735e-05, 4.3852e-05, 5.0356e-05, 4.7183e-05], device='cuda:1') 2023-04-27 13:25:03,543 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=3341.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 13:25:09,034 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.5003, 4.4667, 3.8870, 3.9404, 4.4940, 4.4043, 4.1779, 4.3709], device='cuda:1'), covar=tensor([0.0223, 0.0151, 0.0192, 0.0334, 0.0092, 0.0224, 0.0134, 0.0201], device='cuda:1'), in_proj_covar=tensor([0.0054, 0.0049, 0.0068, 0.0070, 0.0049, 0.0057, 0.0064, 0.0061], device='cuda:1'), out_proj_covar=tensor([7.3670e-05, 6.3192e-05, 1.0567e-04, 9.5873e-05, 5.7421e-05, 7.1992e-05, 9.2634e-05, 8.7368e-05], device='cuda:1') 2023-04-27 13:25:16,106 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=3349.0, num_to_drop=2, layers_to_drop={0, 1} 2023-04-27 13:25:17,759 INFO [train.py:904] (1/8) Epoch 1, batch 3350, loss[loss=0.4031, simple_loss=0.4218, pruned_loss=0.1922, over 16934.00 frames. ], tot_loss[loss=0.3495, simple_loss=0.394, pruned_loss=0.1524, over 3340520.90 frames. ], batch size: 109, lr: 4.56e-02, grad_scale: 16.0 2023-04-27 13:25:49,836 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=3375.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 13:26:09,262 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 1, batch 3400, loss[loss=0.4456, simple_loss=0.4584, pruned_loss=0.2164, over 12323.00 frames. ], tot_loss[loss=0.3495, simple_loss=0.3938, pruned_loss=0.1526, over 3331710.07 frames. ], batch size: 246, lr: 4.55e-02, grad_scale: 16.0 2023-04-27 13:26:39,219 INFO [optim.py:368] (1/8) Clipping_scale=2.0, grad-norm quartiles 2.681e+02 4.195e+02 5.225e+02 6.801e+02 1.040e+03, threshold=1.045e+03, percent-clipped=0.0 2023-04-27 13:26:51,865 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.91 vs. limit=2.0 2023-04-27 13:27:00,372 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.77 vs. limit=2.0 2023-04-27 13:27:11,642 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=3.39 vs. limit=5.0 2023-04-27 13:27:32,313 INFO [train.py:904] (1/8) Epoch 1, batch 3450, loss[loss=0.3439, simple_loss=0.3825, pruned_loss=0.1526, over 16703.00 frames. ], tot_loss[loss=0.3463, simple_loss=0.3915, pruned_loss=0.1506, over 3334907.52 frames. ], batch size: 134, lr: 4.54e-02, grad_scale: 16.0 2023-04-27 13:28:12,015 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=3480.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 13:28:14,690 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.8198, 4.8280, 5.1323, 5.2143, 5.5152, 5.1526, 4.9928, 5.2180], device='cuda:1'), covar=tensor([0.0306, 0.0237, 0.0598, 0.0520, 0.0277, 0.0279, 0.0459, 0.0198], device='cuda:1'), in_proj_covar=tensor([0.0122, 0.0105, 0.0136, 0.0135, 0.0138, 0.0117, 0.0127, 0.0104], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0001], device='cuda:1') 2023-04-27 13:28:15,057 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.71 vs. limit=2.0 2023-04-27 13:28:29,373 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.0710, 3.9218, 3.5630, 3.6661, 4.0711, 3.9567, 3.6406, 3.9026], device='cuda:1'), covar=tensor([0.0153, 0.0155, 0.0159, 0.0303, 0.0072, 0.0163, 0.0135, 0.0148], device='cuda:1'), in_proj_covar=tensor([0.0053, 0.0048, 0.0067, 0.0070, 0.0048, 0.0056, 0.0063, 0.0061], device='cuda:1'), out_proj_covar=tensor([7.4883e-05, 6.4169e-05, 1.0913e-04, 1.0012e-04, 5.9328e-05, 7.5254e-05, 9.5929e-05, 9.3613e-05], device='cuda:1') 2023-04-27 13:28:39,711 INFO [train.py:904] (1/8) Epoch 1, batch 3500, loss[loss=0.4256, simple_loss=0.4268, pruned_loss=0.2122, over 11811.00 frames. ], tot_loss[loss=0.3432, simple_loss=0.3887, pruned_loss=0.1488, over 3329871.51 frames. ], batch size: 248, lr: 4.53e-02, grad_scale: 16.0 2023-04-27 13:28:53,813 INFO [optim.py:368] (1/8) 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 13:29:16,288 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=3528.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 13:29:40,067 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=3546.0, num_to_drop=1, layers_to_drop={0} 2023-04-27 13:29:46,547 INFO [train.py:904] (1/8) Epoch 1, batch 3550, loss[loss=0.3647, simple_loss=0.3975, pruned_loss=0.1659, over 16690.00 frames. ], tot_loss[loss=0.3435, simple_loss=0.3882, pruned_loss=0.1494, over 3318173.40 frames. ], batch size: 134, lr: 4.51e-02, grad_scale: 16.0 2023-04-27 13:30:44,871 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=3593.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 13:30:54,497 INFO [train.py:904] (1/8) Epoch 1, batch 3600, loss[loss=0.3426, simple_loss=0.3722, pruned_loss=0.1565, over 16882.00 frames. ], tot_loss[loss=0.3402, simple_loss=0.3854, pruned_loss=0.1475, over 3312184.39 frames. ], batch size: 109, lr: 4.50e-02, grad_scale: 16.0 2023-04-27 13:31:04,155 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=3607.0, num_to_drop=2, layers_to_drop={1, 3} 2023-04-27 13:31:08,701 INFO [optim.py:368] (1/8) 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] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=3641.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 13:31:52,079 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.0419, 4.1063, 3.8825, 2.2844, 3.6892, 3.8956, 3.5982, 3.9135], device='cuda:1'), covar=tensor([0.0117, 0.0109, 0.0183, 0.1433, 0.0290, 0.0096, 0.0167, 0.0180], device='cuda:1'), in_proj_covar=tensor([0.0042, 0.0044, 0.0050, 0.0086, 0.0043, 0.0044, 0.0042, 0.0050], device='cuda:1'), out_proj_covar=tensor([4.9169e-05, 5.2808e-05, 6.1520e-05, 1.0114e-04, 5.9561e-05, 5.4504e-05, 5.7681e-05, 5.6485e-05], device='cuda:1') 2023-04-27 13:31:56,476 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=3644.0, num_to_drop=1, layers_to_drop={1} 2023-04-27 13:32:06,054 INFO [train.py:904] (1/8) Epoch 1, batch 3650, loss[loss=0.3221, simple_loss=0.3589, pruned_loss=0.1426, over 16864.00 frames. ], tot_loss[loss=0.3371, simple_loss=0.3825, pruned_loss=0.1459, over 3311669.73 frames. ], batch size: 116, lr: 4.49e-02, grad_scale: 16.0 2023-04-27 13:32:42,420 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=3675.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 13:32:43,501 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=3676.0, num_to_drop=1, layers_to_drop={1} 2023-04-27 13:33:20,376 INFO [train.py:904] (1/8) Epoch 1, batch 3700, loss[loss=0.32, simple_loss=0.3611, pruned_loss=0.1395, over 16462.00 frames. ], tot_loss[loss=0.3361, simple_loss=0.3792, pruned_loss=0.1465, over 3287996.00 frames. ], batch size: 146, lr: 4.48e-02, grad_scale: 16.0 2023-04-27 13:33:28,189 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.73 vs. limit=2.0 2023-04-27 13:33:35,071 INFO [optim.py:368] (1/8) 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 13:33:54,064 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=3723.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 13:34:15,098 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=3737.0, num_to_drop=2, layers_to_drop={2, 3} 2023-04-27 13:34:21,413 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.6280, 2.6344, 2.4303, 1.8452, 2.5573, 2.5644, 2.4873, 2.5809], device='cuda:1'), covar=tensor([0.0189, 0.0168, 0.0299, 0.1462, 0.0234, 0.0185, 0.0163, 0.0200], device='cuda:1'), in_proj_covar=tensor([0.0046, 0.0047, 0.0051, 0.0094, 0.0045, 0.0047, 0.0043, 0.0053], device='cuda:1'), out_proj_covar=tensor([5.4678e-05, 5.7232e-05, 6.4194e-05, 1.1336e-04, 6.2356e-05, 5.9138e-05, 6.1524e-05, 6.0634e-05], device='cuda:1') 2023-04-27 13:34:33,710 INFO [train.py:904] (1/8) Epoch 1, batch 3750, loss[loss=0.3287, simple_loss=0.3656, pruned_loss=0.1459, over 16833.00 frames. ], tot_loss[loss=0.335, simple_loss=0.3773, pruned_loss=0.1463, over 3282655.94 frames. ], batch size: 96, lr: 4.47e-02, grad_scale: 16.0 2023-04-27 13:34:35,833 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([5.1083, 5.4523, 5.1578, 5.5121, 4.8860, 5.3454, 5.0811, 5.6111], device='cuda:1'), covar=tensor([0.0452, 0.0468, 0.0490, 0.0242, 0.0489, 0.0268, 0.0402, 0.0212], device='cuda:1'), in_proj_covar=tensor([0.0141, 0.0159, 0.0136, 0.0104, 0.0137, 0.0117, 0.0142, 0.0098], device='cuda:1'), out_proj_covar=tensor([1.2895e-04, 1.5265e-04, 1.2169e-04, 9.0912e-05, 1.2208e-04, 1.0362e-04, 1.3534e-04, 9.6236e-05], device='cuda:1') 2023-04-27 13:35:33,411 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=2.03 vs. limit=2.0 2023-04-27 13:35:46,283 INFO [train.py:904] (1/8) Epoch 1, batch 3800, loss[loss=0.2919, simple_loss=0.3451, pruned_loss=0.1193, over 16553.00 frames. ], tot_loss[loss=0.3362, simple_loss=0.3773, pruned_loss=0.1475, over 3267777.56 frames. ], batch size: 75, lr: 4.46e-02, grad_scale: 16.0 2023-04-27 13:36:00,501 INFO [optim.py:368] (1/8) 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 13:36:24,297 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=3.98 vs. limit=5.0 2023-04-27 13:36:53,916 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.6642, 3.5905, 3.5826, 2.4634, 3.1280, 2.1575, 3.6578, 4.0307], device='cuda:1'), covar=tensor([0.0428, 0.0421, 0.0368, 0.1894, 0.0937, 0.1666, 0.0373, 0.0283], device='cuda:1'), in_proj_covar=tensor([0.0042, 0.0038, 0.0060, 0.0103, 0.0089, 0.0093, 0.0050, 0.0034], device='cuda:1'), out_proj_covar=tensor([5.7758e-05, 5.1198e-05, 6.5836e-05, 1.0934e-04, 9.5834e-05, 9.6673e-05, 5.9406e-05, 4.6272e-05], device='cuda:1') 2023-04-27 13:36:57,572 INFO [train.py:904] (1/8) Epoch 1, batch 3850, loss[loss=0.2868, simple_loss=0.3515, pruned_loss=0.111, over 17117.00 frames. ], tot_loss[loss=0.3324, simple_loss=0.3744, pruned_loss=0.1452, over 3281761.32 frames. ], batch size: 48, lr: 4.45e-02, grad_scale: 16.0 2023-04-27 13:37:03,321 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.7371, 3.7398, 3.6447, 2.6505, 3.1089, 2.1362, 3.6609, 4.2739], device='cuda:1'), covar=tensor([0.0408, 0.0384, 0.0333, 0.1917, 0.1015, 0.1657, 0.0396, 0.0195], device='cuda:1'), in_proj_covar=tensor([0.0043, 0.0038, 0.0061, 0.0104, 0.0090, 0.0094, 0.0051, 0.0034], device='cuda:1'), out_proj_covar=tensor([5.8575e-05, 5.1540e-05, 6.6633e-05, 1.1053e-04, 9.6794e-05, 9.7685e-05, 6.0248e-05, 4.6755e-05], device='cuda:1') 2023-04-27 13:38:09,799 INFO [train.py:904] (1/8) Epoch 1, batch 3900, loss[loss=0.326, simple_loss=0.3634, pruned_loss=0.1443, over 16441.00 frames. ], tot_loss[loss=0.329, simple_loss=0.3711, pruned_loss=0.1434, over 3273323.36 frames. ], batch size: 146, lr: 4.44e-02, grad_scale: 16.0 2023-04-27 13:38:11,798 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=3902.0, num_to_drop=1, layers_to_drop={2} 2023-04-27 13:38:24,737 INFO [optim.py:368] (1/8) 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 13:38:46,377 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=3926.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 13:38:49,612 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.0328, 2.5857, 3.1055, 2.6932, 3.5388, 3.0644, 3.7109, 3.3659], device='cuda:1'), covar=tensor([0.0078, 0.0471, 0.0235, 0.0153, 0.0121, 0.0236, 0.0079, 0.0087], device='cuda:1'), in_proj_covar=tensor([0.0028, 0.0045, 0.0038, 0.0035, 0.0031, 0.0034, 0.0035, 0.0033], device='cuda:1'), out_proj_covar=tensor([4.0771e-05, 6.6466e-05, 5.5330e-05, 4.2985e-05, 4.1869e-05, 4.5505e-05, 4.3720e-05, 4.2833e-05], device='cuda:1') 2023-04-27 13:39:09,947 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.0029, 3.9422, 4.1336, 4.2740, 4.3717, 3.9480, 4.0248, 4.2284], device='cuda:1'), covar=tensor([0.0278, 0.0256, 0.0426, 0.0344, 0.0327, 0.0313, 0.0430, 0.0278], device='cuda:1'), in_proj_covar=tensor([0.0114, 0.0099, 0.0127, 0.0125, 0.0133, 0.0111, 0.0118, 0.0098], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0001, 0.0002, 0.0001, 0.0001, 0.0001], device='cuda:1') 2023-04-27 13:39:11,914 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=3944.0, num_to_drop=1, layers_to_drop={2} 2023-04-27 13:39:21,373 INFO [train.py:904] (1/8) Epoch 1, batch 3950, loss[loss=0.332, simple_loss=0.3589, pruned_loss=0.1526, over 16763.00 frames. ], tot_loss[loss=0.327, simple_loss=0.3686, pruned_loss=0.1427, over 3282569.05 frames. ], batch size: 124, lr: 4.43e-02, grad_scale: 16.0 2023-04-27 13:39:32,463 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.91 vs. limit=2.0 2023-04-27 13:40:13,472 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=3987.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 13:40:20,469 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=3992.0, num_to_drop=1, layers_to_drop={1} 2023-04-27 13:40:37,499 INFO [train.py:904] (1/8) Epoch 1, batch 4000, loss[loss=0.3329, simple_loss=0.37, pruned_loss=0.1479, over 16836.00 frames. ], tot_loss[loss=0.3253, simple_loss=0.3672, pruned_loss=0.1417, over 3290363.13 frames. ], batch size: 116, lr: 4.42e-02, grad_scale: 16.0 2023-04-27 13:40:52,170 INFO [optim.py:368] (1/8) 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 13:41:24,001 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=4032.0, num_to_drop=1, layers_to_drop={1} 2023-04-27 13:41:50,490 INFO [train.py:904] (1/8) Epoch 1, batch 4050, loss[loss=0.25, simple_loss=0.3267, pruned_loss=0.0866, over 16478.00 frames. ], tot_loss[loss=0.3159, simple_loss=0.3619, pruned_loss=0.135, over 3301014.15 frames. ], batch size: 75, lr: 4.41e-02, grad_scale: 16.0 2023-04-27 13:42:43,213 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.4443, 3.3920, 3.0528, 3.1305, 3.3996, 3.2899, 3.1349, 3.2313], device='cuda:1'), covar=tensor([0.0088, 0.0086, 0.0110, 0.0247, 0.0063, 0.0131, 0.0093, 0.0116], device='cuda:1'), in_proj_covar=tensor([0.0050, 0.0042, 0.0059, 0.0068, 0.0044, 0.0053, 0.0057, 0.0056], device='cuda:1'), out_proj_covar=tensor([8.2752e-05, 6.6814e-05, 1.0666e-04, 1.1052e-04, 6.3784e-05, 8.4799e-05, 9.9035e-05, 1.0100e-04], device='cuda:1') 2023-04-27 13:43:04,572 INFO [train.py:904] (1/8) Epoch 1, batch 4100, loss[loss=0.3404, simple_loss=0.4005, pruned_loss=0.1402, over 16665.00 frames. ], tot_loss[loss=0.3111, simple_loss=0.3601, pruned_loss=0.131, over 3285337.31 frames. ], batch size: 83, lr: 4.40e-02, grad_scale: 32.0 2023-04-27 13:43:10,278 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.70 vs. limit=2.0 2023-04-27 13:43:17,361 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.69 vs. limit=2.0 2023-04-27 13:43:18,961 INFO [optim.py:368] (1/8) 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 13:44:04,981 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=4141.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 13:44:20,297 INFO [train.py:904] (1/8) Epoch 1, batch 4150, loss[loss=0.3352, simple_loss=0.3859, pruned_loss=0.1423, over 16812.00 frames. ], tot_loss[loss=0.3219, simple_loss=0.3698, pruned_loss=0.1371, over 3241475.64 frames. ], batch size: 42, lr: 4.39e-02, grad_scale: 32.0 2023-04-27 13:44:55,145 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.8983, 2.7778, 2.6678, 3.2443, 2.5366, 2.6675, 2.3601, 2.7091], device='cuda:1'), covar=tensor([0.0308, 0.0396, 0.0316, 0.0241, 0.1174, 0.0417, 0.0770, 0.0766], device='cuda:1'), in_proj_covar=tensor([0.0049, 0.0054, 0.0046, 0.0044, 0.0092, 0.0050, 0.0068, 0.0053], device='cuda:1'), out_proj_covar=tensor([5.0746e-05, 5.4893e-05, 4.5729e-05, 4.9111e-05, 9.7282e-05, 5.1998e-05, 6.7452e-05, 6.0651e-05], device='cuda:1') 2023-04-27 13:45:37,087 INFO [train.py:904] (1/8) Epoch 1, batch 4200, loss[loss=0.3445, simple_loss=0.4148, pruned_loss=0.1372, over 16617.00 frames. ], tot_loss[loss=0.3295, simple_loss=0.3787, pruned_loss=0.1401, over 3211271.41 frames. ], batch size: 76, lr: 4.38e-02, grad_scale: 16.0 2023-04-27 13:45:39,580 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=4202.0, num_to_drop=1, layers_to_drop={2} 2023-04-27 13:45:39,688 INFO [zipformer.py:625] (1/8) 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] (1/8) 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 13:46:19,502 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.2028, 3.6943, 2.8374, 4.2007, 4.4973, 4.2070, 2.8437, 4.1826], device='cuda:1'), covar=tensor([0.1997, 0.0177, 0.1031, 0.0085, 0.0064, 0.0267, 0.0608, 0.0117], device='cuda:1'), in_proj_covar=tensor([0.0144, 0.0062, 0.0108, 0.0039, 0.0038, 0.0058, 0.0082, 0.0059], device='cuda:1'), out_proj_covar=tensor([1.5351e-04, 7.0693e-05, 1.2551e-04, 5.1052e-05, 5.0308e-05, 7.9578e-05, 9.5942e-05, 6.8342e-05], device='cuda:1') 2023-04-27 13:46:49,603 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=4250.0, num_to_drop=1, layers_to_drop={1} 2023-04-27 13:46:50,857 INFO [train.py:904] (1/8) Epoch 1, batch 4250, loss[loss=0.3405, simple_loss=0.3915, pruned_loss=0.1448, over 16969.00 frames. ], tot_loss[loss=0.3295, simple_loss=0.3809, pruned_loss=0.1391, over 3209308.71 frames. ], batch size: 41, lr: 4.36e-02, grad_scale: 16.0 2023-04-27 13:47:36,657 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=4282.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 13:47:45,332 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=4288.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 13:48:04,725 INFO [train.py:904] (1/8) Epoch 1, batch 4300, loss[loss=0.3236, simple_loss=0.3876, pruned_loss=0.1298, over 16276.00 frames. ], tot_loss[loss=0.3275, simple_loss=0.3811, pruned_loss=0.137, over 3203254.21 frames. ], batch size: 165, lr: 4.35e-02, grad_scale: 16.0 2023-04-27 13:48:21,414 INFO [optim.py:368] (1/8) 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 13:48:52,493 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=4332.0, num_to_drop=1, layers_to_drop={0} 2023-04-27 13:48:55,448 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.2535, 4.7140, 4.3947, 4.5728, 4.6997, 5.0148, 4.9164, 4.5474], device='cuda:1'), covar=tensor([0.0678, 0.0825, 0.0758, 0.1102, 0.1348, 0.0552, 0.0555, 0.1429], device='cuda:1'), in_proj_covar=tensor([0.0132, 0.0173, 0.0135, 0.0144, 0.0179, 0.0130, 0.0134, 0.0189], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0002, 0.0001, 0.0001, 0.0002, 0.0001, 0.0001, 0.0002], device='cuda:1') 2023-04-27 13:49:17,773 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=4349.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 13:49:20,275 INFO [train.py:904] (1/8) Epoch 1, batch 4350, loss[loss=0.3466, simple_loss=0.399, pruned_loss=0.1471, over 16866.00 frames. ], tot_loss[loss=0.3309, simple_loss=0.3851, pruned_loss=0.1384, over 3213724.78 frames. ], batch size: 116, lr: 4.34e-02, grad_scale: 16.0 2023-04-27 13:50:04,954 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 1, batch 4400, loss[loss=0.3256, simple_loss=0.3809, pruned_loss=0.1351, over 15532.00 frames. ], tot_loss[loss=0.3326, simple_loss=0.3868, pruned_loss=0.1392, over 3208468.47 frames. ], batch size: 190, lr: 4.33e-02, grad_scale: 16.0 2023-04-27 13:50:51,962 INFO [optim.py:368] (1/8) 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 13:51:36,193 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.4974, 3.6486, 3.8645, 3.8684, 3.9580, 3.6005, 3.7228, 3.8664], device='cuda:1'), covar=tensor([0.0282, 0.0221, 0.0385, 0.0375, 0.0296, 0.0346, 0.0359, 0.0192], device='cuda:1'), in_proj_covar=tensor([0.0110, 0.0096, 0.0120, 0.0117, 0.0127, 0.0110, 0.0119, 0.0096], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0001, 0.0002, 0.0002, 0.0001, 0.0001], device='cuda:1') 2023-04-27 13:51:48,408 INFO [train.py:904] (1/8) Epoch 1, batch 4450, loss[loss=0.3119, simple_loss=0.3829, pruned_loss=0.1204, over 16752.00 frames. ], tot_loss[loss=0.3328, simple_loss=0.3889, pruned_loss=0.1383, over 3209565.81 frames. ], batch size: 89, lr: 4.32e-02, grad_scale: 16.0 2023-04-27 13:51:54,409 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=4.96 vs. limit=5.0 2023-04-27 13:52:57,210 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 1, batch 4500, loss[loss=0.3254, simple_loss=0.3863, pruned_loss=0.1322, over 17126.00 frames. ], tot_loss[loss=0.328, simple_loss=0.3857, pruned_loss=0.1351, over 3227257.97 frames. ], batch size: 47, lr: 4.31e-02, grad_scale: 8.0 2023-04-27 13:53:20,026 INFO [optim.py:368] (1/8) 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 13:53:46,965 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.63 vs. limit=2.0 2023-04-27 13:54:14,091 INFO [train.py:904] (1/8) Epoch 1, batch 4550, loss[loss=0.3719, simple_loss=0.4033, pruned_loss=0.1703, over 11883.00 frames. ], tot_loss[loss=0.3278, simple_loss=0.3854, pruned_loss=0.1351, over 3220896.93 frames. ], batch size: 246, lr: 4.30e-02, grad_scale: 8.0 2023-04-27 13:54:57,996 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=4582.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 13:55:25,725 INFO [train.py:904] (1/8) Epoch 1, batch 4600, loss[loss=0.3243, simple_loss=0.3854, pruned_loss=0.1316, over 15489.00 frames. ], tot_loss[loss=0.3265, simple_loss=0.3849, pruned_loss=0.134, over 3220177.43 frames. ], batch size: 190, lr: 4.29e-02, grad_scale: 8.0 2023-04-27 13:55:43,312 INFO [optim.py:368] (1/8) 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] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=4630.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 13:56:27,601 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 1, batch 4650, loss[loss=0.2838, simple_loss=0.3465, pruned_loss=0.1106, over 17039.00 frames. ], tot_loss[loss=0.3225, simple_loss=0.3817, pruned_loss=0.1317, over 3235553.08 frames. ], batch size: 50, lr: 4.28e-02, grad_scale: 8.0 2023-04-27 13:56:40,164 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([1.8970, 1.3528, 1.6629, 2.0517, 1.3706, 2.2594, 1.8001, 2.2062], device='cuda:1'), covar=tensor([0.0124, 0.0513, 0.0186, 0.0162, 0.0169, 0.0107, 0.0178, 0.0188], device='cuda:1'), in_proj_covar=tensor([0.0023, 0.0039, 0.0026, 0.0026, 0.0024, 0.0027, 0.0025, 0.0028], device='cuda:1'), out_proj_covar=tensor([2.5423e-05, 4.7116e-05, 2.8105e-05, 2.8950e-05, 2.3920e-05, 2.6156e-05, 2.6877e-05, 2.8793e-05], device='cuda:1') 2023-04-27 13:57:01,827 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.54 vs. limit=2.0 2023-04-27 13:57:15,578 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.6932, 3.0026, 2.9596, 4.0349, 3.1433, 3.7246, 3.1421, 3.1515], device='cuda:1'), covar=tensor([0.0346, 0.0625, 0.0503, 0.0266, 0.1196, 0.0340, 0.0725, 0.1050], device='cuda:1'), in_proj_covar=tensor([0.0062, 0.0068, 0.0059, 0.0061, 0.0122, 0.0066, 0.0084, 0.0072], device='cuda:1'), out_proj_covar=tensor([6.9643e-05, 7.3831e-05, 6.2651e-05, 7.3099e-05, 1.3604e-04, 7.3505e-05, 8.5599e-05, 8.7539e-05], device='cuda:1') 2023-04-27 13:57:50,180 INFO [train.py:904] (1/8) Epoch 1, batch 4700, loss[loss=0.3159, simple_loss=0.3562, pruned_loss=0.1378, over 11428.00 frames. ], tot_loss[loss=0.3186, simple_loss=0.3779, pruned_loss=0.1296, over 3229233.76 frames. ], batch size: 248, lr: 4.27e-02, grad_scale: 8.0 2023-04-27 13:58:07,983 INFO [optim.py:368] (1/8) 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 13:59:02,499 INFO [train.py:904] (1/8) Epoch 1, batch 4750, loss[loss=0.2814, simple_loss=0.3521, pruned_loss=0.1053, over 16851.00 frames. ], tot_loss[loss=0.3147, simple_loss=0.3741, pruned_loss=0.1276, over 3219300.92 frames. ], batch size: 102, lr: 4.26e-02, grad_scale: 8.0 2023-04-27 14:00:05,625 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.99 vs. limit=2.0 2023-04-27 14:00:11,480 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=4797.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 14:00:11,571 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.7090, 3.8932, 1.5534, 4.0919, 2.9774, 4.1722, 2.3378, 2.9314], device='cuda:1'), covar=tensor([0.0101, 0.0208, 0.1266, 0.0069, 0.0345, 0.0061, 0.0808, 0.0379], device='cuda:1'), in_proj_covar=tensor([0.0042, 0.0039, 0.0080, 0.0042, 0.0065, 0.0037, 0.0085, 0.0054], device='cuda:1'), out_proj_covar=tensor([5.2202e-05, 5.3523e-05, 1.1290e-04, 5.2411e-05, 8.1157e-05, 5.5926e-05, 1.1281e-04, 7.5510e-05], device='cuda:1') 2023-04-27 14:00:16,558 INFO [train.py:904] (1/8) Epoch 1, batch 4800, loss[loss=0.3936, simple_loss=0.4177, pruned_loss=0.1848, over 12075.00 frames. ], tot_loss[loss=0.3111, simple_loss=0.3707, pruned_loss=0.1258, over 3217270.88 frames. ], batch size: 246, lr: 4.25e-02, grad_scale: 8.0 2023-04-27 14:00:18,110 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.52 vs. limit=2.0 2023-04-27 14:00:34,035 INFO [optim.py:368] (1/8) Clipping_scale=2.0, grad-norm quartiles 2.251e+02 4.322e+02 5.200e+02 6.651e+02 1.076e+03, threshold=1.040e+03, percent-clipped=0.0 2023-04-27 14:01:22,953 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=4845.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 14:01:31,749 INFO [train.py:904] (1/8) Epoch 1, batch 4850, loss[loss=0.2732, simple_loss=0.3389, pruned_loss=0.1038, over 17021.00 frames. ], tot_loss[loss=0.3133, simple_loss=0.3733, pruned_loss=0.1266, over 3199023.10 frames. ], batch size: 53, lr: 4.24e-02, grad_scale: 8.0 2023-04-27 14:02:49,096 INFO [train.py:904] (1/8) Epoch 1, batch 4900, loss[loss=0.2902, simple_loss=0.3692, pruned_loss=0.1056, over 16847.00 frames. ], tot_loss[loss=0.313, simple_loss=0.3733, pruned_loss=0.1263, over 3159067.74 frames. ], batch size: 102, lr: 4.23e-02, grad_scale: 8.0 2023-04-27 14:03:07,685 INFO [optim.py:368] (1/8) 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 14:03:12,648 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=3.29 vs. limit=5.0 2023-04-27 14:03:35,190 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.7785, 4.3474, 4.6511, 4.9149, 4.0780, 4.5367, 4.2558, 4.3242], device='cuda:1'), covar=tensor([0.0232, 0.0151, 0.0157, 0.0077, 0.0679, 0.0257, 0.0165, 0.0163], device='cuda:1'), in_proj_covar=tensor([0.0081, 0.0059, 0.0107, 0.0080, 0.0131, 0.0086, 0.0074, 0.0085], device='cuda:1'), out_proj_covar=tensor([1.4124e-04, 9.8111e-05, 1.8434e-04, 1.2799e-04, 1.8723e-04, 1.5571e-04, 1.2768e-04, 1.4942e-04], device='cuda:1') 2023-04-27 14:03:54,985 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=4944.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 14:04:04,718 INFO [train.py:904] (1/8) Epoch 1, batch 4950, loss[loss=0.3063, simple_loss=0.3689, pruned_loss=0.1218, over 16625.00 frames. ], tot_loss[loss=0.3142, simple_loss=0.3744, pruned_loss=0.127, over 3162295.60 frames. ], batch size: 62, lr: 4.21e-02, grad_scale: 8.0 2023-04-27 14:05:04,811 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 1, batch 5000, loss[loss=0.3677, simple_loss=0.4054, pruned_loss=0.165, over 12170.00 frames. ], tot_loss[loss=0.3139, simple_loss=0.3752, pruned_loss=0.1263, over 3183050.46 frames. ], batch size: 247, lr: 4.20e-02, grad_scale: 8.0 2023-04-27 14:05:35,392 INFO [optim.py:368] (1/8) Clipping_scale=2.0, grad-norm quartiles 2.744e+02 4.820e+02 5.822e+02 7.344e+02 1.526e+03, threshold=1.164e+03, percent-clipped=12.0 2023-04-27 14:06:31,100 INFO [train.py:904] (1/8) Epoch 1, batch 5050, loss[loss=0.2745, simple_loss=0.347, pruned_loss=0.1009, over 17026.00 frames. ], tot_loss[loss=0.3123, simple_loss=0.3745, pruned_loss=0.1251, over 3192588.65 frames. ], batch size: 55, lr: 4.19e-02, grad_scale: 8.0 2023-04-27 14:06:35,516 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=4.90 vs. limit=5.0 2023-04-27 14:07:13,059 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=3.37 vs. limit=5.0 2023-04-27 14:07:42,590 INFO [train.py:904] (1/8) Epoch 1, batch 5100, loss[loss=0.2574, simple_loss=0.3209, pruned_loss=0.09695, over 17252.00 frames. ], tot_loss[loss=0.3091, simple_loss=0.3715, pruned_loss=0.1233, over 3196388.56 frames. ], batch size: 52, lr: 4.18e-02, grad_scale: 8.0 2023-04-27 14:07:59,829 INFO [optim.py:368] (1/8) Clipping_scale=2.0, grad-norm quartiles 2.617e+02 4.246e+02 5.535e+02 6.678e+02 1.197e+03, threshold=1.107e+03, percent-clipped=1.0 2023-04-27 14:08:58,107 INFO [train.py:904] (1/8) Epoch 1, batch 5150, loss[loss=0.3416, simple_loss=0.4052, pruned_loss=0.1389, over 16949.00 frames. ], tot_loss[loss=0.3092, simple_loss=0.3721, pruned_loss=0.1231, over 3179695.07 frames. ], batch size: 109, lr: 4.17e-02, grad_scale: 8.0 2023-04-27 14:10:12,903 INFO [train.py:904] (1/8) Epoch 1, batch 5200, loss[loss=0.297, simple_loss=0.3566, pruned_loss=0.1187, over 16910.00 frames. ], tot_loss[loss=0.3098, simple_loss=0.3718, pruned_loss=0.1239, over 3170133.00 frames. ], batch size: 90, lr: 4.16e-02, grad_scale: 8.0 2023-04-27 14:10:14,703 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.5431, 3.3247, 2.6774, 3.0264, 2.4121, 2.0617, 3.3862, 3.6711], device='cuda:1'), covar=tensor([0.1368, 0.0435, 0.0956, 0.0373, 0.1321, 0.1247, 0.0276, 0.0071], device='cuda:1'), in_proj_covar=tensor([0.0164, 0.0120, 0.0155, 0.0090, 0.0140, 0.0128, 0.0087, 0.0051], device='cuda:1'), out_proj_covar=tensor([1.9296e-04, 1.4570e-04, 1.6572e-04, 9.9941e-05, 1.7063e-04, 1.4178e-04, 1.0243e-04, 6.2366e-05], device='cuda:1') 2023-04-27 14:10:22,045 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.0850, 3.9593, 4.4229, 4.3757, 4.5786, 4.1055, 4.1279, 4.2808], device='cuda:1'), covar=tensor([0.0232, 0.0274, 0.0372, 0.0340, 0.0231, 0.0246, 0.0442, 0.0205], device='cuda:1'), in_proj_covar=tensor([0.0112, 0.0103, 0.0132, 0.0130, 0.0140, 0.0113, 0.0132, 0.0101], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001], device='cuda:1') 2023-04-27 14:10:30,245 INFO [optim.py:368] (1/8) 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] (1/8) Epoch 1, batch 5250, loss[loss=0.293, simple_loss=0.366, pruned_loss=0.11, over 16305.00 frames. ], tot_loss[loss=0.3076, simple_loss=0.3687, pruned_loss=0.1232, over 3187952.94 frames. ], batch size: 165, lr: 4.15e-02, grad_scale: 8.0 2023-04-27 14:12:34,563 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=4.59 vs. limit=5.0 2023-04-27 14:12:37,170 INFO [train.py:904] (1/8) Epoch 1, batch 5300, loss[loss=0.2799, simple_loss=0.3368, pruned_loss=0.1114, over 16666.00 frames. ], tot_loss[loss=0.3034, simple_loss=0.3642, pruned_loss=0.1213, over 3181797.85 frames. ], batch size: 57, lr: 4.14e-02, grad_scale: 8.0 2023-04-27 14:12:43,544 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([5.9206, 6.1587, 5.7207, 6.2924, 5.5589, 5.3088, 5.7732, 6.3944], device='cuda:1'), covar=tensor([0.0274, 0.0323, 0.0433, 0.0161, 0.0430, 0.0233, 0.0325, 0.0179], device='cuda:1'), in_proj_covar=tensor([0.0146, 0.0173, 0.0165, 0.0116, 0.0143, 0.0117, 0.0157, 0.0108], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0002, 0.0001, 0.0001, 0.0001, 0.0001, 0.0002, 0.0001], device='cuda:1') 2023-04-27 14:12:54,710 INFO [optim.py:368] (1/8) 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 14:13:10,325 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.98 vs. limit=2.0 2023-04-27 14:13:49,534 INFO [train.py:904] (1/8) Epoch 1, batch 5350, loss[loss=0.3059, simple_loss=0.371, pruned_loss=0.1204, over 16414.00 frames. ], tot_loss[loss=0.3022, simple_loss=0.3631, pruned_loss=0.1206, over 3172135.78 frames. ], batch size: 68, lr: 4.13e-02, grad_scale: 8.0 2023-04-27 14:14:41,653 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.95 vs. limit=2.0 2023-04-27 14:15:00,987 INFO [train.py:904] (1/8) Epoch 1, batch 5400, loss[loss=0.3136, simple_loss=0.383, pruned_loss=0.1221, over 16531.00 frames. ], tot_loss[loss=0.3061, simple_loss=0.3668, pruned_loss=0.1227, over 3158590.91 frames. ], batch size: 75, lr: 4.12e-02, grad_scale: 8.0 2023-04-27 14:15:18,320 INFO [optim.py:368] (1/8) 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 14:15:31,945 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.1969, 3.8355, 3.6738, 3.3235, 4.1505, 3.7805, 3.5530, 3.9445], device='cuda:1'), covar=tensor([0.0105, 0.0108, 0.0118, 0.0447, 0.0056, 0.0182, 0.0125, 0.0127], device='cuda:1'), in_proj_covar=tensor([0.0050, 0.0037, 0.0056, 0.0075, 0.0043, 0.0059, 0.0054, 0.0052], device='cuda:1'), out_proj_covar=tensor([1.1081e-04, 7.7769e-05, 1.2747e-04, 1.5050e-04, 8.2714e-05, 1.2187e-04, 1.1949e-04, 1.2388e-04], device='cuda:1') 2023-04-27 14:16:06,374 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.8651, 4.1029, 3.8088, 4.0565, 3.4682, 3.9428, 3.8342, 3.9316], device='cuda:1'), covar=tensor([0.0390, 0.0541, 0.0591, 0.0300, 0.0605, 0.0446, 0.0411, 0.0560], device='cuda:1'), in_proj_covar=tensor([0.0143, 0.0174, 0.0162, 0.0114, 0.0142, 0.0117, 0.0154, 0.0106], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0002, 0.0001, 0.0001, 0.0001, 0.0001, 0.0002, 0.0001], device='cuda:1') 2023-04-27 14:16:19,544 INFO [train.py:904] (1/8) Epoch 1, batch 5450, loss[loss=0.3306, simple_loss=0.3908, pruned_loss=0.1352, over 17187.00 frames. ], tot_loss[loss=0.3112, simple_loss=0.3713, pruned_loss=0.1256, over 3168142.15 frames. ], batch size: 44, lr: 4.11e-02, grad_scale: 8.0 2023-04-27 14:17:07,321 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=5.70 vs. limit=5.0 2023-04-27 14:17:37,159 INFO [train.py:904] (1/8) Epoch 1, batch 5500, loss[loss=0.3702, simple_loss=0.4191, pruned_loss=0.1607, over 16679.00 frames. ], tot_loss[loss=0.3269, simple_loss=0.3828, pruned_loss=0.1355, over 3136978.05 frames. ], batch size: 134, lr: 4.10e-02, grad_scale: 8.0 2023-04-27 14:17:56,200 INFO [optim.py:368] (1/8) 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 14:18:17,089 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.93 vs. limit=2.0 2023-04-27 14:18:57,385 INFO [train.py:904] (1/8) Epoch 1, batch 5550, loss[loss=0.3622, simple_loss=0.4118, pruned_loss=0.1563, over 16872.00 frames. ], tot_loss[loss=0.3411, simple_loss=0.393, pruned_loss=0.1446, over 3136567.32 frames. ], batch size: 116, lr: 4.09e-02, grad_scale: 8.0 2023-04-27 14:20:03,553 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.65 vs. limit=2.0 2023-04-27 14:20:13,362 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.92 vs. limit=2.0 2023-04-27 14:20:17,844 INFO [train.py:904] (1/8) Epoch 1, batch 5600, loss[loss=0.4444, simple_loss=0.4524, pruned_loss=0.2182, over 11441.00 frames. ], tot_loss[loss=0.3525, simple_loss=0.4004, pruned_loss=0.1523, over 3082946.69 frames. ], batch size: 250, lr: 4.08e-02, grad_scale: 8.0 2023-04-27 14:20:37,625 INFO [optim.py:368] (1/8) 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 14:21:34,697 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=2.99 vs. limit=5.0 2023-04-27 14:21:39,923 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=5650.0, num_to_drop=1, layers_to_drop={1} 2023-04-27 14:21:40,595 INFO [train.py:904] (1/8) Epoch 1, batch 5650, loss[loss=0.4728, simple_loss=0.4644, pruned_loss=0.2406, over 10982.00 frames. ], tot_loss[loss=0.3628, simple_loss=0.4074, pruned_loss=0.1591, over 3072948.16 frames. ], batch size: 247, lr: 4.07e-02, grad_scale: 8.0 2023-04-27 14:21:53,472 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.4926, 2.4427, 1.6187, 2.6819, 2.0033, 2.4320, 1.8895, 2.2602], device='cuda:1'), covar=tensor([0.0112, 0.0171, 0.0952, 0.0100, 0.0430, 0.0252, 0.0875, 0.0387], device='cuda:1'), in_proj_covar=tensor([0.0049, 0.0048, 0.0099, 0.0051, 0.0085, 0.0047, 0.0110, 0.0077], device='cuda:1'), out_proj_covar=tensor([6.8641e-05, 7.3477e-05, 1.4353e-04, 6.8762e-05, 1.1741e-04, 7.7741e-05, 1.5719e-04, 1.1848e-04], device='cuda:1') 2023-04-27 14:22:59,471 INFO [train.py:904] (1/8) Epoch 1, batch 5700, loss[loss=0.4155, simple_loss=0.454, pruned_loss=0.1885, over 16252.00 frames. ], tot_loss[loss=0.3679, simple_loss=0.4105, pruned_loss=0.1626, over 3045343.23 frames. ], batch size: 165, lr: 4.06e-02, grad_scale: 8.0 2023-04-27 14:23:15,442 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=5711.0, num_to_drop=1, layers_to_drop={0} 2023-04-27 14:23:17,958 INFO [optim.py:368] (1/8) 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 14:23:23,973 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.63 vs. limit=2.0 2023-04-27 14:23:29,230 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.1049, 4.0917, 2.4337, 4.4381, 4.4365, 4.3456, 3.0086, 4.2221], device='cuda:1'), covar=tensor([0.2485, 0.0176, 0.1810, 0.0076, 0.0089, 0.0225, 0.0674, 0.0157], device='cuda:1'), in_proj_covar=tensor([0.0165, 0.0078, 0.0146, 0.0050, 0.0050, 0.0063, 0.0109, 0.0085], device='cuda:1'), out_proj_covar=tensor([1.9679e-04, 1.0510e-04, 1.8068e-04, 7.5722e-05, 7.8101e-05, 1.0910e-04, 1.4683e-04, 1.1438e-04], device='cuda:1') 2023-04-27 14:24:21,185 INFO [train.py:904] (1/8) Epoch 1, batch 5750, loss[loss=0.3404, simple_loss=0.3979, pruned_loss=0.1414, over 16816.00 frames. ], tot_loss[loss=0.3693, simple_loss=0.4125, pruned_loss=0.1631, over 3040850.26 frames. ], batch size: 96, lr: 4.05e-02, grad_scale: 8.0 2023-04-27 14:24:28,302 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=5755.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 14:25:42,938 INFO [train.py:904] (1/8) Epoch 1, batch 5800, loss[loss=0.3196, simple_loss=0.385, pruned_loss=0.1271, over 16405.00 frames. ], tot_loss[loss=0.3666, simple_loss=0.4116, pruned_loss=0.1608, over 3041643.35 frames. ], batch size: 68, lr: 4.04e-02, grad_scale: 8.0 2023-04-27 14:26:01,838 INFO [optim.py:368] (1/8) 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 14:26:06,939 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=5816.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 14:27:02,341 INFO [train.py:904] (1/8) Epoch 1, batch 5850, loss[loss=0.3299, simple_loss=0.3926, pruned_loss=0.1336, over 16690.00 frames. ], tot_loss[loss=0.3621, simple_loss=0.4086, pruned_loss=0.1578, over 3059952.34 frames. ], batch size: 57, lr: 4.03e-02, grad_scale: 8.0 2023-04-27 14:27:05,712 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=4.63 vs. limit=5.0 2023-04-27 14:27:14,301 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 1, batch 5900, loss[loss=0.2852, simple_loss=0.3519, pruned_loss=0.1093, over 16911.00 frames. ], tot_loss[loss=0.36, simple_loss=0.4073, pruned_loss=0.1563, over 3074926.79 frames. ], batch size: 116, lr: 4.02e-02, grad_scale: 8.0 2023-04-27 14:28:38,044 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.5160, 2.6697, 2.4949, 1.7058, 2.5808, 2.6296, 2.3834, 2.5120], device='cuda:1'), covar=tensor([0.0190, 0.0103, 0.0211, 0.1617, 0.0180, 0.0109, 0.0224, 0.0224], device='cuda:1'), in_proj_covar=tensor([0.0067, 0.0063, 0.0060, 0.0141, 0.0061, 0.0055, 0.0061, 0.0077], device='cuda:1'), out_proj_covar=tensor([9.6559e-05, 9.0026e-05, 9.2424e-05, 2.0262e-04, 9.4751e-05, 8.4603e-05, 1.0356e-04, 1.1070e-04], device='cuda:1') 2023-04-27 14:28:48,065 INFO [optim.py:368] (1/8) 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 14:28:58,977 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=5919.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 14:29:26,277 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([1.5707, 2.4018, 2.3754, 2.1573, 2.4901, 2.4560, 2.5707, 2.0736], device='cuda:1'), covar=tensor([0.0901, 0.0152, 0.0137, 0.0288, 0.0106, 0.0110, 0.0104, 0.0324], device='cuda:1'), in_proj_covar=tensor([0.0089, 0.0036, 0.0038, 0.0057, 0.0035, 0.0035, 0.0041, 0.0053], device='cuda:1'), out_proj_covar=tensor([1.6262e-04, 7.1835e-05, 7.5645e-05, 1.0495e-04, 6.7870e-05, 7.6641e-05, 7.9045e-05, 1.0085e-04], device='cuda:1') 2023-04-27 14:29:49,226 INFO [train.py:904] (1/8) Epoch 1, batch 5950, loss[loss=0.3338, simple_loss=0.3901, pruned_loss=0.1388, over 16783.00 frames. ], tot_loss[loss=0.3565, simple_loss=0.4066, pruned_loss=0.1532, over 3089645.22 frames. ], batch size: 83, lr: 4.01e-02, grad_scale: 8.0 2023-04-27 14:30:36,378 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([1.8139, 1.7227, 1.7170, 1.6705, 2.0434, 2.1412, 2.1526, 2.4730], device='cuda:1'), covar=tensor([0.0097, 0.0367, 0.0148, 0.0207, 0.0125, 0.0141, 0.0213, 0.0132], device='cuda:1'), in_proj_covar=tensor([0.0030, 0.0055, 0.0036, 0.0033, 0.0034, 0.0038, 0.0029, 0.0032], device='cuda:1'), out_proj_covar=tensor([3.4196e-05, 7.6924e-05, 4.3520e-05, 4.1648e-05, 4.0070e-05, 4.5048e-05, 3.8268e-05, 4.0291e-05], device='cuda:1') 2023-04-27 14:30:52,037 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=5990.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 14:30:54,514 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.0602, 3.3842, 3.0710, 4.4819, 3.5220, 4.4530, 3.4564, 3.3558], device='cuda:1'), covar=tensor([0.0253, 0.0364, 0.0344, 0.0171, 0.0923, 0.0118, 0.0442, 0.0906], device='cuda:1'), in_proj_covar=tensor([0.0087, 0.0092, 0.0078, 0.0092, 0.0165, 0.0086, 0.0112, 0.0108], device='cuda:1'), out_proj_covar=tensor([1.1206e-04, 1.1263e-04, 9.6658e-05, 1.2020e-04, 2.0840e-04, 1.0628e-04, 1.2506e-04, 1.4291e-04], device='cuda:1') 2023-04-27 14:31:14,081 INFO [train.py:904] (1/8) Epoch 1, batch 6000, loss[loss=0.3181, simple_loss=0.3785, pruned_loss=0.1289, over 16855.00 frames. ], tot_loss[loss=0.3571, simple_loss=0.4067, pruned_loss=0.1537, over 3089337.19 frames. ], batch size: 96, lr: 4.00e-02, grad_scale: 8.0 2023-04-27 14:31:14,081 INFO [train.py:929] (1/8) Computing validation loss 2023-04-27 14:31:23,948 INFO [train.py:938] (1/8) Epoch 1, validation: loss=0.2762, simple_loss=0.3752, pruned_loss=0.08863, over 944034.00 frames. 2023-04-27 14:31:23,949 INFO [train.py:939] (1/8) Maximum memory allocated so far is 17407MB 2023-04-27 14:31:31,560 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=6006.0, num_to_drop=1, layers_to_drop={3} 2023-04-27 14:31:36,004 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([1.9698, 1.6527, 1.6973, 2.2460, 2.2186, 2.1843, 2.1905, 2.1624], device='cuda:1'), covar=tensor([0.0142, 0.1027, 0.0440, 0.0218, 0.0200, 0.0280, 0.0198, 0.0220], device='cuda:1'), in_proj_covar=tensor([0.0037, 0.0083, 0.0060, 0.0044, 0.0038, 0.0038, 0.0046, 0.0038], device='cuda:1'), out_proj_covar=tensor([6.2020e-05, 1.4851e-04, 1.0776e-04, 7.2897e-05, 6.3544e-05, 6.1688e-05, 7.1284e-05, 6.3784e-05], device='cuda:1') 2023-04-27 14:31:41,452 INFO [optim.py:368] (1/8) Clipping_scale=2.0, grad-norm quartiles 3.294e+02 5.611e+02 7.114e+02 9.006e+02 1.900e+03, threshold=1.423e+03, percent-clipped=2.0 2023-04-27 14:32:30,012 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=6043.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 14:32:42,911 INFO [train.py:904] (1/8) Epoch 1, batch 6050, loss[loss=0.3203, simple_loss=0.3973, pruned_loss=0.1216, over 16852.00 frames. ], tot_loss[loss=0.3538, simple_loss=0.4041, pruned_loss=0.1517, over 3106773.30 frames. ], batch size: 96, lr: 3.99e-02, grad_scale: 8.0 2023-04-27 14:32:44,261 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=6051.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 14:32:54,375 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.7683, 4.5952, 4.1225, 4.5888, 3.4879, 2.9042, 4.7320, 5.1555], device='cuda:1'), covar=tensor([0.1391, 0.0459, 0.0798, 0.0205, 0.1702, 0.1178, 0.0229, 0.0030], device='cuda:1'), in_proj_covar=tensor([0.0183, 0.0129, 0.0171, 0.0099, 0.0172, 0.0141, 0.0103, 0.0060], device='cuda:1'), out_proj_covar=tensor([2.1987e-04, 1.5820e-04, 1.8538e-04, 1.1350e-04, 2.0785e-04, 1.6320e-04, 1.2337e-04, 7.3845e-05], device='cuda:1') 2023-04-27 14:33:34,428 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.9049, 3.9636, 3.3442, 1.7714, 2.9361, 2.0357, 3.3005, 3.9349], device='cuda:1'), covar=tensor([0.0328, 0.0269, 0.0365, 0.2145, 0.0946, 0.1446, 0.0911, 0.0219], device='cuda:1'), in_proj_covar=tensor([0.0087, 0.0059, 0.0098, 0.0140, 0.0134, 0.0130, 0.0122, 0.0056], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001], device='cuda:1') 2023-04-27 14:34:03,525 INFO [train.py:904] (1/8) Epoch 1, batch 6100, loss[loss=0.3512, simple_loss=0.4061, pruned_loss=0.1482, over 16295.00 frames. ], tot_loss[loss=0.3494, simple_loss=0.4019, pruned_loss=0.1485, over 3118383.57 frames. ], batch size: 165, lr: 3.98e-02, grad_scale: 8.0 2023-04-27 14:34:08,700 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=6104.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 14:34:21,377 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=6111.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 14:34:24,017 INFO [optim.py:368] (1/8) Clipping_scale=2.0, grad-norm quartiles 2.820e+02 4.604e+02 6.189e+02 8.357e+02 1.892e+03, threshold=1.238e+03, percent-clipped=2.0 2023-04-27 14:34:59,819 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=6135.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 14:35:26,096 INFO [train.py:904] (1/8) Epoch 1, batch 6150, loss[loss=0.3176, simple_loss=0.3772, pruned_loss=0.129, over 16817.00 frames. ], tot_loss[loss=0.3471, simple_loss=0.3995, pruned_loss=0.1474, over 3109028.64 frames. ], batch size: 83, lr: 3.97e-02, grad_scale: 8.0 2023-04-27 14:35:57,125 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=6170.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 14:36:39,027 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=6196.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 14:36:46,261 INFO [train.py:904] (1/8) Epoch 1, batch 6200, loss[loss=0.3113, simple_loss=0.3658, pruned_loss=0.1284, over 16993.00 frames. ], tot_loss[loss=0.3456, simple_loss=0.3972, pruned_loss=0.147, over 3107083.75 frames. ], batch size: 55, lr: 3.96e-02, grad_scale: 8.0 2023-04-27 14:36:49,313 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=6203.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 14:37:03,220 INFO [zipformer.py:625] (1/8) 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] (1/8) 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 14:37:08,133 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=6214.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 14:37:33,335 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=6231.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 14:37:36,826 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.89 vs. limit=2.0 2023-04-27 14:38:02,374 INFO [train.py:904] (1/8) Epoch 1, batch 6250, loss[loss=0.3492, simple_loss=0.4063, pruned_loss=0.146, over 16482.00 frames. ], tot_loss[loss=0.3435, simple_loss=0.3959, pruned_loss=0.1456, over 3114775.81 frames. ], batch size: 146, lr: 3.95e-02, grad_scale: 8.0 2023-04-27 14:38:05,360 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([1.3840, 2.9962, 2.8510, 2.3644, 2.9804, 2.9780, 3.1918, 1.9762], device='cuda:1'), covar=tensor([0.1255, 0.0123, 0.0148, 0.0345, 0.0120, 0.0164, 0.0146, 0.0482], device='cuda:1'), in_proj_covar=tensor([0.0094, 0.0037, 0.0037, 0.0061, 0.0036, 0.0036, 0.0043, 0.0058], device='cuda:1'), out_proj_covar=tensor([1.7252e-04, 7.3141e-05, 7.6849e-05, 1.1581e-04, 7.2231e-05, 7.9732e-05, 8.5557e-05, 1.1311e-04], device='cuda:1') 2023-04-27 14:38:20,967 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=6264.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 14:38:24,192 INFO [zipformer.py:625] (1/8) 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] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=6272.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 14:39:17,026 INFO [train.py:904] (1/8) Epoch 1, batch 6300, loss[loss=0.3343, simple_loss=0.3885, pruned_loss=0.14, over 16894.00 frames. ], tot_loss[loss=0.3435, simple_loss=0.3958, pruned_loss=0.1456, over 3105637.63 frames. ], batch size: 109, lr: 3.94e-02, grad_scale: 8.0 2023-04-27 14:39:25,752 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=6306.0, num_to_drop=1, layers_to_drop={0} 2023-04-27 14:39:36,719 INFO [optim.py:368] (1/8) 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 14:39:40,461 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([1.8025, 4.0393, 2.1577, 4.4924, 4.6828, 4.4729, 2.6329, 3.9912], device='cuda:1'), covar=tensor([0.2402, 0.0201, 0.1848, 0.0073, 0.0075, 0.0244, 0.0790, 0.0243], device='cuda:1'), in_proj_covar=tensor([0.0162, 0.0081, 0.0148, 0.0050, 0.0054, 0.0066, 0.0114, 0.0092], device='cuda:1'), out_proj_covar=tensor([2.0162e-04, 1.1423e-04, 1.8998e-04, 8.1357e-05, 8.9060e-05, 1.1754e-04, 1.5820e-04, 1.3112e-04], device='cuda:1') 2023-04-27 14:39:57,823 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.5544, 3.1322, 3.2105, 3.8297, 2.9507, 3.6184, 3.1157, 2.7872], device='cuda:1'), covar=tensor([0.0276, 0.0312, 0.0243, 0.0231, 0.0923, 0.0206, 0.0444, 0.0764], device='cuda:1'), in_proj_covar=tensor([0.0091, 0.0093, 0.0078, 0.0096, 0.0166, 0.0090, 0.0114, 0.0112], device='cuda:1'), out_proj_covar=tensor([1.1852e-04, 1.1583e-04, 9.7546e-05, 1.2697e-04, 2.1205e-04, 1.1264e-04, 1.2832e-04, 1.4930e-04], device='cuda:1') 2023-04-27 14:39:59,155 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=6327.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 14:40:19,186 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.94 vs. limit=2.0 2023-04-27 14:40:28,622 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=6346.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 14:40:36,334 INFO [train.py:904] (1/8) Epoch 1, batch 6350, loss[loss=0.3823, simple_loss=0.4184, pruned_loss=0.1731, over 15245.00 frames. ], tot_loss[loss=0.3481, simple_loss=0.3985, pruned_loss=0.1489, over 3101096.16 frames. ], batch size: 190, lr: 3.93e-02, grad_scale: 8.0 2023-04-27 14:40:40,270 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=6354.0, num_to_drop=1, layers_to_drop={1} 2023-04-27 14:41:50,538 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=6399.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 14:41:53,485 INFO [train.py:904] (1/8) Epoch 1, batch 6400, loss[loss=0.3747, simple_loss=0.4128, pruned_loss=0.1683, over 15323.00 frames. ], tot_loss[loss=0.3496, simple_loss=0.3988, pruned_loss=0.1501, over 3099769.23 frames. ], batch size: 191, lr: 3.92e-02, grad_scale: 8.0 2023-04-27 14:42:08,527 INFO [zipformer.py:625] (1/8) 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] (1/8) 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 14:42:32,097 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=4.87 vs. limit=5.0 2023-04-27 14:43:06,336 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([1.5760, 2.6444, 2.4628, 2.2552, 2.4584, 2.5651, 2.6900, 1.8892], device='cuda:1'), covar=tensor([0.1020, 0.0091, 0.0110, 0.0324, 0.0139, 0.0140, 0.0076, 0.0480], device='cuda:1'), in_proj_covar=tensor([0.0101, 0.0039, 0.0039, 0.0065, 0.0038, 0.0039, 0.0043, 0.0063], device='cuda:1'), out_proj_covar=tensor([1.8804e-04, 7.8393e-05, 8.2948e-05, 1.2520e-04, 7.8214e-05, 8.8046e-05, 8.6335e-05, 1.2413e-04], device='cuda:1') 2023-04-27 14:43:09,619 INFO [train.py:904] (1/8) Epoch 1, batch 6450, loss[loss=0.2953, simple_loss=0.3642, pruned_loss=0.1132, over 16721.00 frames. ], tot_loss[loss=0.3446, simple_loss=0.3961, pruned_loss=0.1466, over 3118223.15 frames. ], batch size: 124, lr: 3.91e-02, grad_scale: 8.0 2023-04-27 14:43:22,170 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=6459.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 14:43:26,573 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=6462.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 14:43:33,811 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=6467.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 14:44:12,254 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=6491.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 14:44:17,212 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.87 vs. limit=2.0 2023-04-27 14:44:26,199 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.2177, 3.1311, 2.9808, 2.7199, 3.2618, 2.5847, 2.8410, 2.9613], device='cuda:1'), covar=tensor([0.0122, 0.0076, 0.0125, 0.0413, 0.0071, 0.0474, 0.0124, 0.0157], device='cuda:1'), in_proj_covar=tensor([0.0048, 0.0037, 0.0053, 0.0075, 0.0039, 0.0069, 0.0053, 0.0052], device='cuda:1'), out_proj_covar=tensor([1.1805e-04, 9.0254e-05, 1.3417e-04, 1.6678e-04, 9.0215e-05, 1.5792e-04, 1.3417e-04, 1.4122e-04], device='cuda:1') 2023-04-27 14:44:26,976 INFO [train.py:904] (1/8) Epoch 1, batch 6500, loss[loss=0.346, simple_loss=0.3723, pruned_loss=0.1599, over 11323.00 frames. ], tot_loss[loss=0.3423, simple_loss=0.3934, pruned_loss=0.1456, over 3106695.43 frames. ], batch size: 248, lr: 3.90e-02, grad_scale: 16.0 2023-04-27 14:44:45,181 INFO [optim.py:368] (1/8) 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 14:44:46,842 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=6514.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 14:44:59,776 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=6523.0, num_to_drop=1, layers_to_drop={3} 2023-04-27 14:45:04,515 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=6526.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 14:45:07,622 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 1, batch 6550, loss[loss=0.4073, simple_loss=0.43, pruned_loss=0.1923, over 11835.00 frames. ], tot_loss[loss=0.3454, simple_loss=0.3971, pruned_loss=0.1468, over 3114868.76 frames. ], batch size: 246, lr: 3.89e-02, grad_scale: 16.0 2023-04-27 14:45:55,587 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=6559.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 14:46:00,517 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=6562.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 14:46:08,765 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=6567.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 14:46:32,601 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.4707, 4.4458, 4.4240, 4.8969, 4.8521, 4.6357, 4.9415, 4.6348], device='cuda:1'), covar=tensor([0.0401, 0.0323, 0.0946, 0.0250, 0.0352, 0.0277, 0.0185, 0.0344], device='cuda:1'), in_proj_covar=tensor([0.0151, 0.0151, 0.0238, 0.0168, 0.0140, 0.0152, 0.0128, 0.0145], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-27 14:46:56,176 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.55 vs. limit=2.0 2023-04-27 14:46:59,693 INFO [train.py:904] (1/8) Epoch 1, batch 6600, loss[loss=0.3233, simple_loss=0.3867, pruned_loss=0.13, over 16716.00 frames. ], tot_loss[loss=0.3474, simple_loss=0.3995, pruned_loss=0.1477, over 3115382.98 frames. ], batch size: 124, lr: 3.89e-02, grad_scale: 16.0 2023-04-27 14:47:18,178 INFO [optim.py:368] (1/8) 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 14:47:32,782 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=6622.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 14:47:43,304 INFO [zipformer.py:625] (1/8) 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] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=6646.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 14:48:18,077 INFO [train.py:904] (1/8) Epoch 1, batch 6650, loss[loss=0.3243, simple_loss=0.3786, pruned_loss=0.135, over 16707.00 frames. ], tot_loss[loss=0.3472, simple_loss=0.3991, pruned_loss=0.1476, over 3117635.40 frames. ], batch size: 89, lr: 3.88e-02, grad_scale: 16.0 2023-04-27 14:48:50,112 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.0833, 1.5308, 1.4667, 1.3447, 2.0661, 1.9382, 2.2569, 2.3685], device='cuda:1'), covar=tensor([0.0086, 0.0416, 0.0218, 0.0249, 0.0140, 0.0167, 0.0171, 0.0133], device='cuda:1'), in_proj_covar=tensor([0.0030, 0.0061, 0.0041, 0.0040, 0.0037, 0.0041, 0.0031, 0.0033], device='cuda:1'), out_proj_covar=tensor([3.6091e-05, 8.7991e-05, 5.5443e-05, 5.5067e-05, 4.8089e-05, 5.3714e-05, 4.3577e-05, 4.4272e-05], device='cuda:1') 2023-04-27 14:49:18,723 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=6690.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 14:49:23,554 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=6694.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 14:49:32,154 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 1, batch 6700, loss[loss=0.3192, simple_loss=0.3714, pruned_loss=0.1335, over 16751.00 frames. ], tot_loss[loss=0.3468, simple_loss=0.398, pruned_loss=0.1478, over 3109669.24 frames. ], batch size: 39, lr: 3.87e-02, grad_scale: 16.0 2023-04-27 14:49:52,616 INFO [optim.py:368] (1/8) 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 14:50:24,888 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.0178, 3.6478, 3.6173, 2.9689, 3.6221, 3.6812, 3.9891, 2.1264], device='cuda:1'), covar=tensor([0.1179, 0.0121, 0.0129, 0.0309, 0.0113, 0.0131, 0.0110, 0.0579], device='cuda:1'), in_proj_covar=tensor([0.0107, 0.0042, 0.0042, 0.0068, 0.0039, 0.0040, 0.0045, 0.0070], device='cuda:1'), out_proj_covar=tensor([2.0296e-04, 8.5804e-05, 8.9664e-05, 1.3566e-04, 8.0252e-05, 9.0737e-05, 8.9932e-05, 1.3963e-04], device='cuda:1') 2023-04-27 14:50:31,064 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.3342, 3.1503, 3.0606, 2.8348, 3.2536, 2.6718, 2.8997, 3.0717], device='cuda:1'), covar=tensor([0.0076, 0.0069, 0.0096, 0.0306, 0.0066, 0.0409, 0.0085, 0.0107], device='cuda:1'), in_proj_covar=tensor([0.0047, 0.0037, 0.0052, 0.0075, 0.0040, 0.0071, 0.0052, 0.0052], device='cuda:1'), out_proj_covar=tensor([1.1932e-04, 9.5076e-05, 1.3351e-04, 1.6882e-04, 9.5041e-05, 1.6700e-04, 1.3288e-04, 1.4271e-04], device='cuda:1') 2023-04-27 14:50:45,161 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=6747.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 14:50:51,003 INFO [train.py:904] (1/8) Epoch 1, batch 6750, loss[loss=0.3158, simple_loss=0.3741, pruned_loss=0.1288, over 16730.00 frames. ], tot_loss[loss=0.3466, simple_loss=0.3971, pruned_loss=0.148, over 3102664.22 frames. ], batch size: 134, lr: 3.86e-02, grad_scale: 16.0 2023-04-27 14:51:12,082 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.4652, 4.9374, 4.8558, 4.8281, 4.9473, 5.3792, 5.1929, 4.8258], device='cuda:1'), covar=tensor([0.0724, 0.0996, 0.0831, 0.1246, 0.1534, 0.0542, 0.0636, 0.1706], device='cuda:1'), in_proj_covar=tensor([0.0149, 0.0205, 0.0172, 0.0181, 0.0224, 0.0170, 0.0165, 0.0240], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-27 14:51:50,791 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=6791.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 14:52:05,968 INFO [train.py:904] (1/8) Epoch 1, batch 6800, loss[loss=0.401, simple_loss=0.423, pruned_loss=0.1895, over 11246.00 frames. ], tot_loss[loss=0.3454, simple_loss=0.3963, pruned_loss=0.1473, over 3094611.90 frames. ], batch size: 248, lr: 3.85e-02, grad_scale: 16.0 2023-04-27 14:52:07,662 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.62 vs. limit=2.0 2023-04-27 14:52:24,954 INFO [optim.py:368] (1/8) 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 14:52:33,198 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=6818.0, num_to_drop=1, layers_to_drop={3} 2023-04-27 14:52:33,329 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([1.7106, 3.2502, 3.1276, 2.7650, 3.3258, 3.1179, 3.4087, 2.1733], device='cuda:1'), covar=tensor([0.1094, 0.0077, 0.0115, 0.0283, 0.0084, 0.0150, 0.0084, 0.0487], device='cuda:1'), in_proj_covar=tensor([0.0104, 0.0042, 0.0041, 0.0067, 0.0039, 0.0040, 0.0044, 0.0069], device='cuda:1'), out_proj_covar=tensor([1.9803e-04, 8.6143e-05, 8.8760e-05, 1.3389e-04, 7.9085e-05, 9.0042e-05, 8.8723e-05, 1.3864e-04], device='cuda:1') 2023-04-27 14:52:41,649 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=6823.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 14:52:46,861 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=6826.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 14:53:06,063 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=6839.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 14:53:10,653 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=2.65 vs. limit=5.0 2023-04-27 14:53:23,170 INFO [train.py:904] (1/8) Epoch 1, batch 6850, loss[loss=0.2978, simple_loss=0.381, pruned_loss=0.1073, over 16796.00 frames. ], tot_loss[loss=0.3461, simple_loss=0.3975, pruned_loss=0.1473, over 3107466.26 frames. ], batch size: 124, lr: 3.84e-02, grad_scale: 16.0 2023-04-27 14:53:35,303 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=6859.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 14:53:46,379 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=6867.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 14:53:57,190 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=6874.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 14:54:03,461 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=6878.0, num_to_drop=1, layers_to_drop={1} 2023-04-27 14:54:37,400 INFO [train.py:904] (1/8) Epoch 1, batch 6900, loss[loss=0.3233, simple_loss=0.3901, pruned_loss=0.1283, over 16751.00 frames. ], tot_loss[loss=0.3444, simple_loss=0.3989, pruned_loss=0.145, over 3134792.99 frames. ], batch size: 76, lr: 3.83e-02, grad_scale: 16.0 2023-04-27 14:54:47,054 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=6907.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 14:54:49,770 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=6909.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 14:54:49,874 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([1.8197, 3.1583, 1.8580, 3.7641, 3.9144, 3.7866, 2.2103, 3.2946], device='cuda:1'), covar=tensor([0.2349, 0.0312, 0.2080, 0.0104, 0.0125, 0.0315, 0.0952, 0.0365], device='cuda:1'), in_proj_covar=tensor([0.0163, 0.0084, 0.0154, 0.0052, 0.0057, 0.0070, 0.0121, 0.0098], device='cuda:1'), out_proj_covar=tensor([2.1343e-04, 1.2483e-04, 2.0500e-04, 8.7950e-05, 9.7909e-05, 1.2620e-04, 1.7319e-04, 1.4565e-04], device='cuda:1') 2023-04-27 14:54:55,475 INFO [optim.py:368] (1/8) 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 14:54:59,428 INFO [zipformer.py:625] (1/8) 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] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=6922.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 14:55:09,548 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.2504, 3.7205, 3.6390, 3.2062, 3.5953, 3.6225, 3.9411, 2.5247], device='cuda:1'), covar=tensor([0.0993, 0.0107, 0.0126, 0.0255, 0.0127, 0.0121, 0.0094, 0.0493], device='cuda:1'), in_proj_covar=tensor([0.0102, 0.0041, 0.0040, 0.0066, 0.0038, 0.0038, 0.0043, 0.0066], device='cuda:1'), out_proj_covar=tensor([1.9474e-04, 8.4671e-05, 8.8335e-05, 1.3269e-04, 8.0496e-05, 8.5575e-05, 8.6658e-05, 1.3436e-04], device='cuda:1') 2023-04-27 14:55:36,645 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=6939.0, num_to_drop=1, layers_to_drop={1} 2023-04-27 14:55:54,245 INFO [train.py:904] (1/8) Epoch 1, batch 6950, loss[loss=0.3365, simple_loss=0.3914, pruned_loss=0.1408, over 16427.00 frames. ], tot_loss[loss=0.3486, simple_loss=0.4013, pruned_loss=0.148, over 3117226.97 frames. ], batch size: 146, lr: 3.82e-02, grad_scale: 16.0 2023-04-27 14:56:23,226 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.0482, 3.4611, 3.3048, 1.6252, 3.5010, 3.4376, 2.8414, 3.0621], device='cuda:1'), covar=tensor([0.0207, 0.0087, 0.0150, 0.1736, 0.0091, 0.0102, 0.0241, 0.0206], device='cuda:1'), in_proj_covar=tensor([0.0077, 0.0063, 0.0061, 0.0140, 0.0062, 0.0055, 0.0066, 0.0079], device='cuda:1'), out_proj_covar=tensor([1.1953e-04, 9.9216e-05, 1.0234e-04, 2.1019e-04, 1.0219e-04, 8.8731e-05, 1.1916e-04, 1.2247e-04], device='cuda:1') 2023-04-27 14:56:25,069 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=6970.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 14:56:25,217 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=6970.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 14:56:26,488 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.0815, 4.3633, 2.1065, 5.0460, 4.7834, 4.5847, 3.5632, 3.9739], device='cuda:1'), covar=tensor([0.2370, 0.0232, 0.2103, 0.0096, 0.0098, 0.0221, 0.0650, 0.0345], device='cuda:1'), in_proj_covar=tensor([0.0165, 0.0086, 0.0157, 0.0054, 0.0058, 0.0071, 0.0123, 0.0102], device='cuda:1'), out_proj_covar=tensor([2.1738e-04, 1.2888e-04, 2.0859e-04, 9.0826e-05, 1.0001e-04, 1.2989e-04, 1.7622e-04, 1.5072e-04], device='cuda:1') 2023-04-27 14:56:47,685 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=6985.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 14:57:12,233 INFO [train.py:904] (1/8) Epoch 1, batch 7000, loss[loss=0.3637, simple_loss=0.4342, pruned_loss=0.1466, over 16791.00 frames. ], tot_loss[loss=0.3463, simple_loss=0.4003, pruned_loss=0.1461, over 3125349.92 frames. ], batch size: 83, lr: 3.81e-02, grad_scale: 16.0 2023-04-27 14:57:30,859 INFO [optim.py:368] (1/8) 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 14:58:27,027 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=7048.0, num_to_drop=1, layers_to_drop={1} 2023-04-27 14:58:31,408 INFO [train.py:904] (1/8) Epoch 1, batch 7050, loss[loss=0.32, simple_loss=0.3839, pruned_loss=0.1281, over 16628.00 frames. ], tot_loss[loss=0.3481, simple_loss=0.4018, pruned_loss=0.1472, over 3102568.35 frames. ], batch size: 57, lr: 3.80e-02, grad_scale: 16.0 2023-04-27 14:59:03,687 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=5.45 vs. limit=5.0 2023-04-27 14:59:20,008 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.30 vs. limit=2.0 2023-04-27 14:59:20,232 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.94 vs. limit=2.0 2023-04-27 14:59:35,207 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=4.74 vs. limit=5.0 2023-04-27 14:59:51,865 INFO [train.py:904] (1/8) Epoch 1, batch 7100, loss[loss=0.3395, simple_loss=0.3918, pruned_loss=0.1436, over 16256.00 frames. ], tot_loss[loss=0.3472, simple_loss=0.4001, pruned_loss=0.1471, over 3099487.35 frames. ], batch size: 165, lr: 3.79e-02, grad_scale: 16.0 2023-04-27 15:00:05,774 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=7109.0, num_to_drop=1, layers_to_drop={1} 2023-04-27 15:00:11,238 INFO [optim.py:368] (1/8) 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 15:00:18,857 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.9301, 3.8707, 3.7104, 4.3129, 4.1909, 4.1775, 4.1674, 4.1422], device='cuda:1'), covar=tensor([0.0451, 0.0382, 0.1211, 0.0332, 0.0629, 0.0355, 0.0420, 0.0358], device='cuda:1'), in_proj_covar=tensor([0.0160, 0.0161, 0.0248, 0.0173, 0.0150, 0.0160, 0.0130, 0.0152], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-27 15:00:19,910 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=7118.0, num_to_drop=1, layers_to_drop={0} 2023-04-27 15:00:27,699 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=7123.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 15:01:11,088 INFO [train.py:904] (1/8) Epoch 1, batch 7150, loss[loss=0.3172, simple_loss=0.3729, pruned_loss=0.1308, over 16539.00 frames. ], tot_loss[loss=0.3448, simple_loss=0.3975, pruned_loss=0.1461, over 3090095.00 frames. ], batch size: 62, lr: 3.78e-02, grad_scale: 8.0 2023-04-27 15:01:30,090 INFO [zipformer.py:625] (1/8) 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] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=7166.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 15:01:41,726 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=7171.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 15:02:00,253 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([1.8590, 3.0169, 1.9729, 3.5396, 3.6670, 3.6425, 2.2995, 3.1044], device='cuda:1'), covar=tensor([0.2223, 0.0329, 0.1724, 0.0108, 0.0117, 0.0256, 0.0889, 0.0416], device='cuda:1'), in_proj_covar=tensor([0.0165, 0.0087, 0.0155, 0.0056, 0.0060, 0.0073, 0.0125, 0.0104], device='cuda:1'), out_proj_covar=tensor([2.1923e-04, 1.3255e-04, 2.1054e-04, 9.4829e-05, 1.0357e-04, 1.3485e-04, 1.8116e-04, 1.5641e-04], device='cuda:1') 2023-04-27 15:02:27,366 INFO [train.py:904] (1/8) Epoch 1, batch 7200, loss[loss=0.2579, simple_loss=0.3278, pruned_loss=0.09397, over 16702.00 frames. ], tot_loss[loss=0.3393, simple_loss=0.3937, pruned_loss=0.1424, over 3096151.50 frames. ], batch size: 57, lr: 3.78e-02, grad_scale: 8.0 2023-04-27 15:02:32,003 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.3193, 4.1670, 4.2789, 4.6836, 4.5790, 4.4037, 4.6950, 4.4305], device='cuda:1'), covar=tensor([0.0383, 0.0313, 0.0864, 0.0221, 0.0362, 0.0366, 0.0211, 0.0266], device='cuda:1'), in_proj_covar=tensor([0.0158, 0.0158, 0.0246, 0.0173, 0.0150, 0.0161, 0.0132, 0.0150], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-27 15:02:46,727 INFO [optim.py:368] (1/8) Clipping_scale=2.0, grad-norm quartiles 3.108e+02 4.516e+02 5.501e+02 7.177e+02 1.508e+03, threshold=1.100e+03, percent-clipped=3.0 2023-04-27 15:02:58,061 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=2.08 vs. limit=2.0 2023-04-27 15:03:02,451 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=7224.0, num_to_drop=1, layers_to_drop={1} 2023-04-27 15:03:19,571 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=7234.0, num_to_drop=1, layers_to_drop={0} 2023-04-27 15:03:46,520 INFO [train.py:904] (1/8) Epoch 1, batch 7250, loss[loss=0.2946, simple_loss=0.348, pruned_loss=0.1206, over 16233.00 frames. ], tot_loss[loss=0.335, simple_loss=0.3899, pruned_loss=0.1401, over 3115305.88 frames. ], batch size: 35, lr: 3.77e-02, grad_scale: 8.0 2023-04-27 15:04:06,710 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=7265.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 15:04:20,675 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=7275.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 15:04:35,883 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=7285.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 15:04:55,376 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.95 vs. limit=2.0 2023-04-27 15:04:58,721 INFO [train.py:904] (1/8) Epoch 1, batch 7300, loss[loss=0.357, simple_loss=0.3869, pruned_loss=0.1635, over 11113.00 frames. ], tot_loss[loss=0.334, simple_loss=0.3881, pruned_loss=0.14, over 3085302.80 frames. ], batch size: 246, lr: 3.76e-02, grad_scale: 8.0 2023-04-27 15:05:03,038 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.4383, 4.9517, 4.8209, 4.9390, 4.8293, 5.3492, 5.1072, 4.8534], device='cuda:1'), covar=tensor([0.0662, 0.0802, 0.0724, 0.1043, 0.1739, 0.0527, 0.0626, 0.1462], device='cuda:1'), in_proj_covar=tensor([0.0138, 0.0197, 0.0171, 0.0179, 0.0217, 0.0165, 0.0159, 0.0230], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-27 15:05:19,382 INFO [optim.py:368] (1/8) 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 15:05:49,055 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=7333.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 15:05:53,628 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=7336.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 15:06:15,327 INFO [train.py:904] (1/8) Epoch 1, batch 7350, loss[loss=0.324, simple_loss=0.3865, pruned_loss=0.1308, over 16762.00 frames. ], tot_loss[loss=0.3327, simple_loss=0.387, pruned_loss=0.1392, over 3071304.03 frames. ], batch size: 76, lr: 3.75e-02, grad_scale: 8.0 2023-04-27 15:06:23,606 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.0966, 4.1340, 3.3346, 3.8862, 3.1789, 2.4710, 4.5092, 4.8554], device='cuda:1'), covar=tensor([0.1641, 0.0471, 0.0965, 0.0318, 0.1941, 0.1210, 0.0153, 0.0041], device='cuda:1'), in_proj_covar=tensor([0.0206, 0.0160, 0.0191, 0.0116, 0.0208, 0.0151, 0.0124, 0.0068], device='cuda:1'), out_proj_covar=tensor([2.4798e-04, 1.9305e-04, 2.1090e-04, 1.3727e-04, 2.5135e-04, 1.7939e-04, 1.5164e-04, 8.5343e-05], device='cuda:1') 2023-04-27 15:07:30,952 INFO [train.py:904] (1/8) Epoch 1, batch 7400, loss[loss=0.2958, simple_loss=0.3707, pruned_loss=0.1105, over 16897.00 frames. ], tot_loss[loss=0.3349, simple_loss=0.3891, pruned_loss=0.1404, over 3071998.96 frames. ], batch size: 102, lr: 3.74e-02, grad_scale: 8.0 2023-04-27 15:07:36,102 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=7404.0, num_to_drop=1, layers_to_drop={1} 2023-04-27 15:07:37,304 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.7672, 5.1064, 4.7888, 5.0013, 4.4765, 4.5715, 4.5584, 5.1954], device='cuda:1'), covar=tensor([0.0378, 0.0515, 0.0812, 0.0312, 0.0549, 0.0340, 0.0395, 0.0341], device='cuda:1'), in_proj_covar=tensor([0.0162, 0.0203, 0.0195, 0.0126, 0.0160, 0.0129, 0.0180, 0.0134], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001, 0.0002, 0.0001], device='cuda:1') 2023-04-27 15:07:40,494 INFO [zipformer.py:625] (1/8) 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] (1/8) Clipping_scale=2.0, grad-norm quartiles 3.431e+02 4.999e+02 6.259e+02 7.546e+02 1.554e+03, threshold=1.252e+03, percent-clipped=1.0 2023-04-27 15:08:34,223 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.6150, 3.3150, 3.1925, 1.4254, 3.3757, 3.3841, 2.9782, 3.0864], device='cuda:1'), covar=tensor([0.0371, 0.0091, 0.0222, 0.1885, 0.0100, 0.0075, 0.0243, 0.0188], device='cuda:1'), in_proj_covar=tensor([0.0082, 0.0063, 0.0065, 0.0141, 0.0063, 0.0055, 0.0068, 0.0079], device='cuda:1'), out_proj_covar=tensor([1.3016e-04, 1.0292e-04, 1.0961e-04, 2.1631e-04, 1.0738e-04, 9.3138e-05, 1.2301e-04, 1.2708e-04], device='cuda:1') 2023-04-27 15:08:47,922 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([1.4790, 2.5756, 2.3872, 2.1693, 2.4938, 2.5995, 2.5770, 1.6957], device='cuda:1'), covar=tensor([0.1180, 0.0135, 0.0156, 0.0386, 0.0143, 0.0103, 0.0137, 0.0684], device='cuda:1'), in_proj_covar=tensor([0.0114, 0.0044, 0.0045, 0.0077, 0.0043, 0.0042, 0.0048, 0.0081], device='cuda:1'), out_proj_covar=tensor([2.2299e-04, 9.5184e-05, 1.0244e-04, 1.5935e-04, 9.1962e-05, 9.6471e-05, 1.0112e-04, 1.6795e-04], device='cuda:1') 2023-04-27 15:08:52,556 INFO [train.py:904] (1/8) Epoch 1, batch 7450, loss[loss=0.323, simple_loss=0.3841, pruned_loss=0.1309, over 16653.00 frames. ], tot_loss[loss=0.3386, simple_loss=0.3912, pruned_loss=0.143, over 3043001.21 frames. ], batch size: 62, lr: 3.73e-02, grad_scale: 8.0 2023-04-27 15:09:22,498 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=7468.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 15:10:10,760 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.1154, 4.5963, 2.4628, 5.3692, 5.2485, 4.9116, 3.4334, 4.6556], device='cuda:1'), covar=tensor([0.2251, 0.0220, 0.1763, 0.0062, 0.0067, 0.0176, 0.0738, 0.0259], device='cuda:1'), in_proj_covar=tensor([0.0166, 0.0088, 0.0159, 0.0057, 0.0062, 0.0076, 0.0128, 0.0108], device='cuda:1'), out_proj_covar=tensor([2.2487e-04, 1.3627e-04, 2.1929e-04, 9.7983e-05, 1.1111e-04, 1.4499e-04, 1.8745e-04, 1.6342e-04], device='cuda:1') 2023-04-27 15:10:15,193 INFO [train.py:904] (1/8) Epoch 1, batch 7500, loss[loss=0.3273, simple_loss=0.3714, pruned_loss=0.1415, over 16999.00 frames. ], tot_loss[loss=0.3386, simple_loss=0.3922, pruned_loss=0.1425, over 3068224.79 frames. ], batch size: 55, lr: 3.72e-02, grad_scale: 8.0 2023-04-27 15:10:35,037 INFO [optim.py:368] (1/8) 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] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=7519.0, num_to_drop=1, layers_to_drop={0} 2023-04-27 15:11:06,815 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=7534.0, num_to_drop=1, layers_to_drop={2} 2023-04-27 15:11:19,557 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.5285, 4.3806, 4.0371, 3.3546, 4.4977, 2.5249, 4.0684, 4.4418], device='cuda:1'), covar=tensor([0.0108, 0.0102, 0.0111, 0.0463, 0.0055, 0.0972, 0.0093, 0.0129], device='cuda:1'), in_proj_covar=tensor([0.0047, 0.0038, 0.0055, 0.0079, 0.0040, 0.0079, 0.0052, 0.0053], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001, 0.0002], device='cuda:1') 2023-04-27 15:11:31,890 INFO [train.py:904] (1/8) Epoch 1, batch 7550, loss[loss=0.3995, simple_loss=0.4176, pruned_loss=0.1907, over 11160.00 frames. ], tot_loss[loss=0.3379, simple_loss=0.3912, pruned_loss=0.1423, over 3071125.94 frames. ], batch size: 247, lr: 3.72e-02, grad_scale: 4.0 2023-04-27 15:11:32,799 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.94 vs. limit=2.0 2023-04-27 15:11:54,176 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=7565.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 15:12:01,168 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.5780, 2.5703, 1.6781, 2.6413, 1.8913, 2.5657, 1.7437, 2.2337], device='cuda:1'), covar=tensor([0.0102, 0.0164, 0.1177, 0.0094, 0.0653, 0.0244, 0.1105, 0.0499], device='cuda:1'), in_proj_covar=tensor([0.0060, 0.0064, 0.0135, 0.0061, 0.0116, 0.0070, 0.0145, 0.0105], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0001, 0.0002, 0.0001, 0.0002, 0.0002], device='cuda:1') 2023-04-27 15:12:21,502 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=7582.0, num_to_drop=1, layers_to_drop={0} 2023-04-27 15:12:50,412 INFO [train.py:904] (1/8) Epoch 1, batch 7600, loss[loss=0.4163, simple_loss=0.427, pruned_loss=0.2028, over 11374.00 frames. ], tot_loss[loss=0.3381, simple_loss=0.3908, pruned_loss=0.1427, over 3067121.32 frames. ], batch size: 248, lr: 3.71e-02, grad_scale: 8.0 2023-04-27 15:13:07,566 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.1913, 3.9539, 3.7577, 3.4344, 4.0693, 2.5028, 3.8121, 3.9312], device='cuda:1'), covar=tensor([0.0083, 0.0078, 0.0095, 0.0351, 0.0052, 0.0700, 0.0075, 0.0093], device='cuda:1'), in_proj_covar=tensor([0.0048, 0.0039, 0.0056, 0.0080, 0.0041, 0.0081, 0.0053, 0.0054], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001, 0.0002], device='cuda:1') 2023-04-27 15:13:10,045 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=7613.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 15:13:12,792 INFO [optim.py:368] (1/8) 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 15:13:20,368 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.0477, 4.0655, 1.7784, 4.2119, 2.4178, 3.9873, 1.7341, 2.8719], device='cuda:1'), covar=tensor([0.0058, 0.0160, 0.1788, 0.0041, 0.0823, 0.0147, 0.1680, 0.0615], device='cuda:1'), in_proj_covar=tensor([0.0060, 0.0064, 0.0139, 0.0061, 0.0119, 0.0070, 0.0147, 0.0107], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0001, 0.0002, 0.0001, 0.0002, 0.0002], device='cuda:1') 2023-04-27 15:13:29,655 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=7624.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 15:13:39,790 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=7631.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 15:13:47,463 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.6291, 3.3481, 2.7839, 4.0645, 3.0024, 4.0900, 3.1111, 2.7192], device='cuda:1'), covar=tensor([0.0293, 0.0285, 0.0338, 0.0228, 0.0983, 0.0157, 0.0471, 0.1069], device='cuda:1'), in_proj_covar=tensor([0.0112, 0.0109, 0.0091, 0.0124, 0.0191, 0.0108, 0.0131, 0.0141], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0002, 0.0003, 0.0001, 0.0002, 0.0002], device='cuda:1') 2023-04-27 15:14:13,721 INFO [train.py:904] (1/8) Epoch 1, batch 7650, loss[loss=0.3516, simple_loss=0.4074, pruned_loss=0.1479, over 16238.00 frames. ], tot_loss[loss=0.3404, simple_loss=0.3924, pruned_loss=0.1442, over 3080376.48 frames. ], batch size: 165, lr: 3.70e-02, grad_scale: 8.0 2023-04-27 15:14:15,614 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.5712, 2.4810, 1.6394, 2.6774, 1.9289, 2.5517, 1.7228, 2.2584], device='cuda:1'), covar=tensor([0.0110, 0.0185, 0.1130, 0.0095, 0.0677, 0.0306, 0.1110, 0.0499], device='cuda:1'), in_proj_covar=tensor([0.0060, 0.0063, 0.0136, 0.0060, 0.0117, 0.0069, 0.0144, 0.0106], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0001, 0.0002, 0.0001, 0.0002, 0.0002], device='cuda:1') 2023-04-27 15:14:22,450 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=4.61 vs. limit=5.0 2023-04-27 15:14:28,808 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.97 vs. limit=2.0 2023-04-27 15:14:32,870 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.0773, 3.9927, 4.0935, 4.4558, 4.4445, 4.2734, 4.4673, 4.2519], device='cuda:1'), covar=tensor([0.0476, 0.0367, 0.1112, 0.0322, 0.0337, 0.0309, 0.0247, 0.0299], device='cuda:1'), in_proj_covar=tensor([0.0168, 0.0166, 0.0258, 0.0179, 0.0154, 0.0160, 0.0139, 0.0152], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-27 15:15:14,465 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=7685.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 15:15:39,928 INFO [train.py:904] (1/8) Epoch 1, batch 7700, loss[loss=0.2982, simple_loss=0.3647, pruned_loss=0.1159, over 16913.00 frames. ], tot_loss[loss=0.3409, simple_loss=0.3925, pruned_loss=0.1447, over 3096918.60 frames. ], batch size: 96, lr: 3.69e-02, grad_scale: 8.0 2023-04-27 15:15:45,475 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=7704.0, num_to_drop=1, layers_to_drop={2} 2023-04-27 15:16:00,934 INFO [optim.py:368] (1/8) 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 15:16:05,647 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([1.8943, 3.2292, 2.0179, 3.6975, 3.6719, 3.5693, 2.1884, 3.2893], device='cuda:1'), covar=tensor([0.2057, 0.0306, 0.1819, 0.0099, 0.0132, 0.0220, 0.0940, 0.0369], device='cuda:1'), in_proj_covar=tensor([0.0165, 0.0091, 0.0160, 0.0057, 0.0063, 0.0079, 0.0130, 0.0111], device='cuda:1'), out_proj_covar=tensor([2.2704e-04, 1.4376e-04, 2.2421e-04, 9.9503e-05, 1.1347e-04, 1.4868e-04, 1.9171e-04, 1.6970e-04], device='cuda:1') 2023-04-27 15:16:45,682 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=7743.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 15:16:57,055 INFO [train.py:904] (1/8) Epoch 1, batch 7750, loss[loss=0.3349, simple_loss=0.3949, pruned_loss=0.1374, over 16743.00 frames. ], tot_loss[loss=0.34, simple_loss=0.3924, pruned_loss=0.1438, over 3118817.04 frames. ], batch size: 124, lr: 3.68e-02, grad_scale: 8.0 2023-04-27 15:16:59,253 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=7752.0, num_to_drop=1, layers_to_drop={0} 2023-04-27 15:17:16,058 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=7763.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 15:18:08,944 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.9729, 2.9867, 3.3725, 3.4135, 3.4367, 3.0380, 3.2099, 3.3564], device='cuda:1'), covar=tensor([0.0370, 0.0501, 0.0483, 0.0467, 0.0358, 0.0497, 0.0713, 0.0334], device='cuda:1'), in_proj_covar=tensor([0.0124, 0.0117, 0.0136, 0.0134, 0.0147, 0.0123, 0.0173, 0.0120], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-27 15:18:14,290 INFO [train.py:904] (1/8) Epoch 1, batch 7800, loss[loss=0.2938, simple_loss=0.3664, pruned_loss=0.1106, over 16256.00 frames. ], tot_loss[loss=0.3414, simple_loss=0.3937, pruned_loss=0.1446, over 3136436.98 frames. ], batch size: 35, lr: 3.67e-02, grad_scale: 8.0 2023-04-27 15:18:16,736 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=7802.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 15:18:19,353 INFO [zipformer.py:625] (1/8) 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] (1/8) 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 15:18:43,836 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=7819.0, num_to_drop=1, layers_to_drop={2} 2023-04-27 15:19:31,859 INFO [train.py:904] (1/8) Epoch 1, batch 7850, loss[loss=0.4103, simple_loss=0.4325, pruned_loss=0.1941, over 11573.00 frames. ], tot_loss[loss=0.3438, simple_loss=0.3956, pruned_loss=0.146, over 3103072.16 frames. ], batch size: 247, lr: 3.66e-02, grad_scale: 8.0 2023-04-27 15:19:51,364 INFO [zipformer.py:625] (1/8) 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] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=7867.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 15:20:49,085 INFO [train.py:904] (1/8) Epoch 1, batch 7900, loss[loss=0.3361, simple_loss=0.3892, pruned_loss=0.1415, over 16792.00 frames. ], tot_loss[loss=0.3406, simple_loss=0.3935, pruned_loss=0.1439, over 3114325.22 frames. ], batch size: 124, lr: 3.66e-02, grad_scale: 8.0 2023-04-27 15:21:11,992 INFO [optim.py:368] (1/8) 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 15:21:39,309 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=7931.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 15:22:08,690 INFO [train.py:904] (1/8) Epoch 1, batch 7950, loss[loss=0.3493, simple_loss=0.3948, pruned_loss=0.1519, over 15319.00 frames. ], tot_loss[loss=0.3413, simple_loss=0.3938, pruned_loss=0.1444, over 3095199.12 frames. ], batch size: 190, lr: 3.65e-02, grad_scale: 8.0 2023-04-27 15:22:21,764 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.9757, 4.8883, 4.5234, 2.0766, 3.3168, 2.6608, 4.0055, 4.9739], device='cuda:1'), covar=tensor([0.0254, 0.0243, 0.0263, 0.2100, 0.0976, 0.1374, 0.0673, 0.0355], device='cuda:1'), in_proj_covar=tensor([0.0105, 0.0077, 0.0116, 0.0155, 0.0151, 0.0141, 0.0139, 0.0070], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001], device='cuda:1') 2023-04-27 15:22:54,218 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=7979.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 15:22:56,122 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=7980.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 15:23:30,895 INFO [train.py:904] (1/8) Epoch 1, batch 8000, loss[loss=0.313, simple_loss=0.3772, pruned_loss=0.1244, over 16691.00 frames. ], tot_loss[loss=0.3418, simple_loss=0.3936, pruned_loss=0.145, over 3087454.49 frames. ], batch size: 57, lr: 3.64e-02, grad_scale: 8.0 2023-04-27 15:23:51,331 INFO [optim.py:368] (1/8) 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 15:24:05,639 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.69 vs. limit=2.0 2023-04-27 15:24:22,515 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.93 vs. limit=2.0 2023-04-27 15:24:45,916 INFO [train.py:904] (1/8) Epoch 1, batch 8050, loss[loss=0.3708, simple_loss=0.3952, pruned_loss=0.1732, over 11776.00 frames. ], tot_loss[loss=0.3399, simple_loss=0.3924, pruned_loss=0.1437, over 3085010.20 frames. ], batch size: 246, lr: 3.63e-02, grad_scale: 8.0 2023-04-27 15:25:04,618 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=8063.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 15:25:08,458 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.4348, 3.2814, 2.7134, 3.1683, 2.4834, 1.9039, 3.3564, 3.7882], device='cuda:1'), covar=tensor([0.1824, 0.0716, 0.1032, 0.0375, 0.1854, 0.1469, 0.0358, 0.0076], device='cuda:1'), in_proj_covar=tensor([0.0215, 0.0170, 0.0200, 0.0123, 0.0221, 0.0154, 0.0134, 0.0074], device='cuda:1'), out_proj_covar=tensor([2.5671e-04, 2.0386e-04, 2.1968e-04, 1.4386e-04, 2.6590e-04, 1.8505e-04, 1.6236e-04, 9.3560e-05], device='cuda:1') 2023-04-27 15:25:57,835 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=8099.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 15:26:00,726 INFO [train.py:904] (1/8) Epoch 1, batch 8100, loss[loss=0.332, simple_loss=0.389, pruned_loss=0.1375, over 16396.00 frames. ], tot_loss[loss=0.339, simple_loss=0.3923, pruned_loss=0.1428, over 3097557.21 frames. ], batch size: 146, lr: 3.62e-02, grad_scale: 4.0 2023-04-27 15:26:15,157 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=8111.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 15:26:19,369 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.91 vs. limit=2.0 2023-04-27 15:26:22,781 INFO [optim.py:368] (1/8) 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 15:26:31,509 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.8217, 3.3442, 2.7847, 4.1570, 2.8532, 4.0137, 3.0905, 2.7798], device='cuda:1'), covar=tensor([0.0308, 0.0334, 0.0327, 0.0251, 0.1179, 0.0216, 0.0604, 0.1134], device='cuda:1'), in_proj_covar=tensor([0.0123, 0.0115, 0.0095, 0.0131, 0.0200, 0.0113, 0.0137, 0.0152], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0001, 0.0001, 0.0002, 0.0003, 0.0001, 0.0002, 0.0002], device='cuda:1') 2023-04-27 15:26:52,941 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.45 vs. limit=2.0 2023-04-27 15:26:53,921 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.2594, 4.5144, 4.2329, 4.4361, 3.9005, 4.2026, 4.1788, 4.5108], device='cuda:1'), covar=tensor([0.0465, 0.0624, 0.0853, 0.0348, 0.0627, 0.0483, 0.0506, 0.0545], device='cuda:1'), in_proj_covar=tensor([0.0168, 0.0220, 0.0210, 0.0133, 0.0164, 0.0137, 0.0186, 0.0146], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001, 0.0002, 0.0002], device='cuda:1') 2023-04-27 15:27:16,439 INFO [train.py:904] (1/8) Epoch 1, batch 8150, loss[loss=0.3051, simple_loss=0.3636, pruned_loss=0.1233, over 16882.00 frames. ], tot_loss[loss=0.3362, simple_loss=0.3894, pruned_loss=0.1415, over 3099998.19 frames. ], batch size: 83, lr: 3.62e-02, grad_scale: 4.0 2023-04-27 15:27:27,358 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 1, batch 8200, loss[loss=0.3659, simple_loss=0.3932, pruned_loss=0.1693, over 11273.00 frames. ], tot_loss[loss=0.3342, simple_loss=0.3872, pruned_loss=0.1406, over 3104644.78 frames. ], batch size: 246, lr: 3.61e-02, grad_scale: 4.0 2023-04-27 15:28:33,925 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.36 vs. limit=2.0 2023-04-27 15:28:43,847 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.6463, 3.5321, 1.5284, 3.4884, 2.1643, 3.4124, 1.9352, 2.7749], device='cuda:1'), covar=tensor([0.0065, 0.0156, 0.1713, 0.0078, 0.0833, 0.0314, 0.1303, 0.0547], device='cuda:1'), in_proj_covar=tensor([0.0063, 0.0071, 0.0150, 0.0068, 0.0127, 0.0079, 0.0153, 0.0117], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-27 15:28:56,620 INFO [optim.py:368] (1/8) 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 15:29:53,199 INFO [train.py:904] (1/8) Epoch 1, batch 8250, loss[loss=0.31, simple_loss=0.3775, pruned_loss=0.1212, over 16616.00 frames. ], tot_loss[loss=0.3317, simple_loss=0.3864, pruned_loss=0.1385, over 3080757.39 frames. ], batch size: 62, lr: 3.60e-02, grad_scale: 4.0 2023-04-27 15:30:15,878 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=8265.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 15:30:41,225 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=8280.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 15:30:44,871 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=3.85 vs. limit=5.0 2023-04-27 15:31:14,606 INFO [train.py:904] (1/8) Epoch 1, batch 8300, loss[loss=0.2602, simple_loss=0.3485, pruned_loss=0.08595, over 16910.00 frames. ], tot_loss[loss=0.3227, simple_loss=0.3804, pruned_loss=0.1325, over 3065715.53 frames. ], batch size: 96, lr: 3.59e-02, grad_scale: 4.0 2023-04-27 15:31:40,147 INFO [optim.py:368] (1/8) 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 15:31:56,725 INFO [zipformer.py:625] (1/8) 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] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=8328.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 15:32:36,033 INFO [train.py:904] (1/8) Epoch 1, batch 8350, loss[loss=0.2954, simple_loss=0.3703, pruned_loss=0.1102, over 16852.00 frames. ], tot_loss[loss=0.3146, simple_loss=0.3756, pruned_loss=0.1267, over 3057654.82 frames. ], batch size: 116, lr: 3.58e-02, grad_scale: 4.0 2023-04-27 15:32:49,257 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.48 vs. limit=2.0 2023-04-27 15:33:08,669 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.1819, 4.2449, 4.5839, 4.6130, 4.6775, 4.2403, 4.2106, 4.4312], device='cuda:1'), covar=tensor([0.0268, 0.0266, 0.0394, 0.0348, 0.0381, 0.0297, 0.0720, 0.0274], device='cuda:1'), in_proj_covar=tensor([0.0124, 0.0114, 0.0135, 0.0131, 0.0142, 0.0121, 0.0168, 0.0118], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-27 15:33:43,206 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.5968, 3.5325, 3.2257, 1.7433, 2.8168, 2.0441, 2.9832, 3.5157], device='cuda:1'), covar=tensor([0.0254, 0.0369, 0.0348, 0.2085, 0.0943, 0.1355, 0.1070, 0.0333], device='cuda:1'), in_proj_covar=tensor([0.0098, 0.0078, 0.0116, 0.0152, 0.0145, 0.0139, 0.0134, 0.0069], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001], device='cuda:1') 2023-04-27 15:33:55,088 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 1, batch 8400, loss[loss=0.3228, simple_loss=0.3731, pruned_loss=0.1363, over 12206.00 frames. ], tot_loss[loss=0.3082, simple_loss=0.371, pruned_loss=0.1227, over 3047757.14 frames. ], batch size: 249, lr: 3.58e-02, grad_scale: 8.0 2023-04-27 15:34:13,102 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.4158, 3.2480, 3.1699, 2.9592, 3.3083, 2.1595, 3.1095, 3.1158], device='cuda:1'), covar=tensor([0.0100, 0.0081, 0.0101, 0.0272, 0.0069, 0.0935, 0.0087, 0.0106], device='cuda:1'), in_proj_covar=tensor([0.0047, 0.0038, 0.0055, 0.0071, 0.0040, 0.0087, 0.0053, 0.0052], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001, 0.0002], device='cuda:1') 2023-04-27 15:34:21,545 INFO [optim.py:368] (1/8) 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 15:34:52,374 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.7806, 3.7900, 3.7063, 1.7391, 3.7228, 3.7645, 3.2271, 3.4434], device='cuda:1'), covar=tensor([0.0410, 0.0100, 0.0133, 0.1811, 0.0128, 0.0098, 0.0279, 0.0167], device='cuda:1'), in_proj_covar=tensor([0.0092, 0.0064, 0.0066, 0.0151, 0.0065, 0.0059, 0.0077, 0.0087], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0001, 0.0001, 0.0002, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:1') 2023-04-27 15:35:12,363 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=8447.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 15:35:18,231 INFO [train.py:904] (1/8) Epoch 1, batch 8450, loss[loss=0.2585, simple_loss=0.3408, pruned_loss=0.08809, over 17264.00 frames. ], tot_loss[loss=0.3043, simple_loss=0.3681, pruned_loss=0.1202, over 3034432.41 frames. ], batch size: 52, lr: 3.57e-02, grad_scale: 8.0 2023-04-27 15:35:30,870 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=8458.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 15:36:06,793 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=3.81 vs. limit=5.0 2023-04-27 15:36:31,059 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.82 vs. limit=2.0 2023-04-27 15:36:39,471 INFO [train.py:904] (1/8) Epoch 1, batch 8500, loss[loss=0.2634, simple_loss=0.316, pruned_loss=0.1054, over 11954.00 frames. ], tot_loss[loss=0.2961, simple_loss=0.3615, pruned_loss=0.1153, over 3014119.86 frames. ], batch size: 247, lr: 3.56e-02, grad_scale: 8.0 2023-04-27 15:36:49,163 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=8506.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 15:37:04,603 INFO [optim.py:368] (1/8) 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 15:37:54,795 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.4338, 3.3972, 3.0691, 3.0584, 2.4873, 2.0389, 3.3049, 3.7885], device='cuda:1'), covar=tensor([0.1491, 0.0447, 0.0699, 0.0315, 0.1574, 0.1367, 0.0279, 0.0075], device='cuda:1'), in_proj_covar=tensor([0.0208, 0.0170, 0.0193, 0.0117, 0.0181, 0.0156, 0.0129, 0.0068], device='cuda:1'), out_proj_covar=tensor([2.4782e-04, 1.9987e-04, 2.1235e-04, 1.3818e-04, 2.1906e-04, 1.8866e-04, 1.5580e-04, 8.7741e-05], device='cuda:1') 2023-04-27 15:38:03,992 INFO [train.py:904] (1/8) Epoch 1, batch 8550, loss[loss=0.2903, simple_loss=0.3699, pruned_loss=0.1053, over 16873.00 frames. ], tot_loss[loss=0.2919, simple_loss=0.3581, pruned_loss=0.1129, over 3014783.79 frames. ], batch size: 90, lr: 3.55e-02, grad_scale: 8.0 2023-04-27 15:38:11,194 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([1.8406, 3.4880, 1.9815, 4.3460, 4.2023, 3.9520, 2.4127, 3.4708], device='cuda:1'), covar=tensor([0.2237, 0.0320, 0.2060, 0.0083, 0.0102, 0.0395, 0.1120, 0.0432], device='cuda:1'), in_proj_covar=tensor([0.0160, 0.0091, 0.0161, 0.0059, 0.0067, 0.0080, 0.0136, 0.0115], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0001, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-27 15:39:44,567 INFO [train.py:904] (1/8) Epoch 1, batch 8600, loss[loss=0.2787, simple_loss=0.3513, pruned_loss=0.1031, over 16321.00 frames. ], tot_loss[loss=0.2915, simple_loss=0.3589, pruned_loss=0.1121, over 3014204.62 frames. ], batch size: 35, lr: 3.54e-02, grad_scale: 8.0 2023-04-27 15:40:17,874 INFO [optim.py:368] (1/8) 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] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=8621.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 15:40:47,654 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.66 vs. limit=2.0 2023-04-27 15:41:26,180 INFO [train.py:904] (1/8) Epoch 1, batch 8650, loss[loss=0.2401, simple_loss=0.3238, pruned_loss=0.07826, over 15171.00 frames. ], tot_loss[loss=0.2863, simple_loss=0.3556, pruned_loss=0.1086, over 3034002.87 frames. ], batch size: 190, lr: 3.54e-02, grad_scale: 8.0 2023-04-27 15:41:42,715 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.6639, 2.5616, 1.5349, 2.5655, 1.9049, 2.6222, 1.8579, 2.4050], device='cuda:1'), covar=tensor([0.0085, 0.0172, 0.1477, 0.0102, 0.0733, 0.0302, 0.1233, 0.0528], device='cuda:1'), in_proj_covar=tensor([0.0062, 0.0067, 0.0148, 0.0066, 0.0124, 0.0076, 0.0152, 0.0119], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0001, 0.0002, 0.0002, 0.0003, 0.0002], device='cuda:1') 2023-04-27 15:42:17,634 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.99 vs. limit=2.0 2023-04-27 15:42:47,137 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.7834, 3.2304, 2.6926, 4.0585, 2.5754, 3.9903, 2.8903, 2.4056], device='cuda:1'), covar=tensor([0.0224, 0.0287, 0.0279, 0.0179, 0.1178, 0.0143, 0.0463, 0.1136], device='cuda:1'), in_proj_covar=tensor([0.0121, 0.0117, 0.0094, 0.0126, 0.0193, 0.0113, 0.0133, 0.0149], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0001, 0.0001, 0.0002, 0.0003, 0.0001, 0.0002, 0.0002], device='cuda:1') 2023-04-27 15:43:13,573 INFO [train.py:904] (1/8) Epoch 1, batch 8700, loss[loss=0.2796, simple_loss=0.3356, pruned_loss=0.1119, over 12573.00 frames. ], tot_loss[loss=0.2808, simple_loss=0.3507, pruned_loss=0.1054, over 3055365.98 frames. ], batch size: 248, lr: 3.53e-02, grad_scale: 8.0 2023-04-27 15:43:41,221 INFO [optim.py:368] (1/8) 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 15:44:49,920 INFO [train.py:904] (1/8) Epoch 1, batch 8750, loss[loss=0.2989, simple_loss=0.3734, pruned_loss=0.1122, over 16767.00 frames. ], tot_loss[loss=0.2782, simple_loss=0.3491, pruned_loss=0.1036, over 3062665.37 frames. ], batch size: 134, lr: 3.52e-02, grad_scale: 8.0 2023-04-27 15:45:14,094 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=8760.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 15:46:42,406 INFO [train.py:904] (1/8) Epoch 1, batch 8800, loss[loss=0.3043, simple_loss=0.371, pruned_loss=0.1188, over 16799.00 frames. ], tot_loss[loss=0.2755, simple_loss=0.3472, pruned_loss=0.1019, over 3077742.88 frames. ], batch size: 124, lr: 3.51e-02, grad_scale: 8.0 2023-04-27 15:46:45,982 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.81 vs. limit=2.0 2023-04-27 15:47:13,493 INFO [optim.py:368] (1/8) 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 15:47:25,135 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=8821.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 15:48:26,697 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.2972, 2.8690, 2.3358, 3.4973, 2.3016, 3.3271, 2.6374, 2.2686], device='cuda:1'), covar=tensor([0.0277, 0.0294, 0.0273, 0.0242, 0.1200, 0.0196, 0.0463, 0.1064], device='cuda:1'), in_proj_covar=tensor([0.0129, 0.0124, 0.0098, 0.0136, 0.0199, 0.0118, 0.0138, 0.0156], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-27 15:48:27,335 INFO [train.py:904] (1/8) Epoch 1, batch 8850, loss[loss=0.2648, simple_loss=0.3472, pruned_loss=0.09125, over 15331.00 frames. ], tot_loss[loss=0.2757, simple_loss=0.3492, pruned_loss=0.1011, over 3063248.21 frames. ], batch size: 191, lr: 3.51e-02, grad_scale: 8.0 2023-04-27 15:48:33,198 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=5.25 vs. limit=5.0 2023-04-27 15:50:09,024 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.1785, 2.7135, 2.6163, 1.6434, 2.8260, 2.7745, 2.4337, 2.6610], device='cuda:1'), covar=tensor([0.0486, 0.0129, 0.0213, 0.1569, 0.0115, 0.0099, 0.0322, 0.0199], device='cuda:1'), in_proj_covar=tensor([0.0092, 0.0065, 0.0061, 0.0143, 0.0061, 0.0058, 0.0074, 0.0085], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0001, 0.0001, 0.0002, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:1') 2023-04-27 15:50:13,540 INFO [train.py:904] (1/8) Epoch 1, batch 8900, loss[loss=0.3143, simple_loss=0.3776, pruned_loss=0.1255, over 15369.00 frames. ], tot_loss[loss=0.275, simple_loss=0.3495, pruned_loss=0.1002, over 3055183.13 frames. ], batch size: 190, lr: 3.50e-02, grad_scale: 8.0 2023-04-27 15:50:19,600 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.4381, 3.5852, 2.8709, 3.2516, 2.7743, 2.1197, 3.6258, 4.0682], device='cuda:1'), covar=tensor([0.1791, 0.0525, 0.1038, 0.0333, 0.1139, 0.1252, 0.0255, 0.0069], device='cuda:1'), in_proj_covar=tensor([0.0211, 0.0178, 0.0198, 0.0123, 0.0166, 0.0159, 0.0133, 0.0074], device='cuda:1'), out_proj_covar=tensor([2.5011e-04, 2.0947e-04, 2.1756e-04, 1.4478e-04, 2.0204e-04, 1.9323e-04, 1.6149e-04, 9.4952e-05], device='cuda:1') 2023-04-27 15:50:42,917 INFO [optim.py:368] (1/8) 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 15:50:56,768 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=8921.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 15:51:13,339 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=8927.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 15:52:19,949 INFO [train.py:904] (1/8) Epoch 1, batch 8950, loss[loss=0.2546, simple_loss=0.3365, pruned_loss=0.08632, over 16730.00 frames. ], tot_loss[loss=0.2761, simple_loss=0.3497, pruned_loss=0.1012, over 3041614.02 frames. ], batch size: 134, lr: 3.49e-02, grad_scale: 8.0 2023-04-27 15:53:00,567 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=8969.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 15:53:38,095 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=8986.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 15:53:40,100 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.7944, 4.0549, 3.4817, 3.7374, 2.9769, 2.2171, 4.2354, 4.5025], device='cuda:1'), covar=tensor([0.1774, 0.0492, 0.0816, 0.0299, 0.1449, 0.1386, 0.0204, 0.0054], device='cuda:1'), in_proj_covar=tensor([0.0215, 0.0180, 0.0199, 0.0123, 0.0166, 0.0159, 0.0136, 0.0075], device='cuda:1'), out_proj_covar=tensor([2.5442e-04, 2.1013e-04, 2.1819e-04, 1.4502e-04, 2.0180e-04, 1.9286e-04, 1.6469e-04, 9.6521e-05], device='cuda:1') 2023-04-27 15:53:42,107 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=8988.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 15:54:08,281 INFO [train.py:904] (1/8) Epoch 1, batch 9000, loss[loss=0.2171, simple_loss=0.3034, pruned_loss=0.06544, over 16852.00 frames. ], tot_loss[loss=0.2708, simple_loss=0.3451, pruned_loss=0.0983, over 3055898.16 frames. ], batch size: 90, lr: 3.48e-02, grad_scale: 8.0 2023-04-27 15:54:08,282 INFO [train.py:929] (1/8) Computing validation loss 2023-04-27 15:54:19,192 INFO [train.py:938] (1/8) Epoch 1, validation: loss=0.2299, simple_loss=0.3267, pruned_loss=0.06658, over 944034.00 frames. 2023-04-27 15:54:19,194 INFO [train.py:939] (1/8) Maximum memory allocated so far is 17407MB 2023-04-27 15:54:46,208 INFO [zipformer.py:625] (1/8) 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] (1/8) 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 15:54:56,994 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.34 vs. limit=2.0 2023-04-27 15:55:56,689 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=9047.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 15:56:02,938 INFO [train.py:904] (1/8) Epoch 1, batch 9050, loss[loss=0.2857, simple_loss=0.3517, pruned_loss=0.1098, over 16868.00 frames. ], tot_loss[loss=0.2728, simple_loss=0.3467, pruned_loss=0.09944, over 3072549.64 frames. ], batch size: 96, lr: 3.48e-02, grad_scale: 8.0 2023-04-27 15:56:48,387 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=9074.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 15:57:42,681 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.2928, 3.0960, 2.7222, 2.5782, 2.2496, 1.9870, 3.0861, 3.2549], device='cuda:1'), covar=tensor([0.1355, 0.0571, 0.0816, 0.0394, 0.1462, 0.1325, 0.0267, 0.0110], device='cuda:1'), in_proj_covar=tensor([0.0216, 0.0181, 0.0203, 0.0125, 0.0169, 0.0162, 0.0136, 0.0074], device='cuda:1'), out_proj_covar=tensor([2.5425e-04, 2.1198e-04, 2.2228e-04, 1.4753e-04, 2.0474e-04, 1.9727e-04, 1.6338e-04, 9.5079e-05], device='cuda:1') 2023-04-27 15:57:45,071 INFO [train.py:904] (1/8) Epoch 1, batch 9100, loss[loss=0.2585, simple_loss=0.3412, pruned_loss=0.08789, over 16922.00 frames. ], tot_loss[loss=0.2729, simple_loss=0.3461, pruned_loss=0.09983, over 3065275.67 frames. ], batch size: 90, lr: 3.47e-02, grad_scale: 8.0 2023-04-27 15:58:14,295 INFO [optim.py:368] (1/8) 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 15:58:14,899 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=9116.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 15:58:26,356 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([5.2719, 5.5451, 5.2588, 5.4760, 4.8550, 4.9989, 5.2007, 5.6296], device='cuda:1'), covar=tensor([0.0346, 0.0599, 0.0622, 0.0267, 0.0522, 0.0254, 0.0365, 0.0399], device='cuda:1'), in_proj_covar=tensor([0.0150, 0.0204, 0.0179, 0.0122, 0.0151, 0.0120, 0.0168, 0.0130], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001, 0.0002, 0.0001], device='cuda:1') 2023-04-27 15:59:42,636 INFO [train.py:904] (1/8) Epoch 1, batch 9150, loss[loss=0.2733, simple_loss=0.3427, pruned_loss=0.102, over 16317.00 frames. ], tot_loss[loss=0.2724, simple_loss=0.3462, pruned_loss=0.09935, over 3044987.43 frames. ], batch size: 146, lr: 3.46e-02, grad_scale: 8.0 2023-04-27 15:59:56,113 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.7983, 3.4344, 3.6963, 3.8104, 3.2892, 3.7325, 3.5717, 3.4637], device='cuda:1'), covar=tensor([0.0282, 0.0277, 0.0218, 0.0136, 0.0601, 0.0157, 0.0483, 0.0225], device='cuda:1'), in_proj_covar=tensor([0.0088, 0.0066, 0.0116, 0.0090, 0.0138, 0.0089, 0.0081, 0.0097], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-27 16:01:13,286 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([1.5192, 3.0948, 2.9296, 2.2554, 3.0777, 2.9287, 3.0709, 1.7430], device='cuda:1'), covar=tensor([0.1205, 0.0090, 0.0075, 0.0481, 0.0068, 0.0114, 0.0069, 0.0794], device='cuda:1'), in_proj_covar=tensor([0.0116, 0.0046, 0.0050, 0.0086, 0.0047, 0.0049, 0.0050, 0.0095], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0001, 0.0001, 0.0002, 0.0001, 0.0001, 0.0001, 0.0002], device='cuda:1') 2023-04-27 16:01:27,880 INFO [train.py:904] (1/8) Epoch 1, batch 9200, loss[loss=0.2345, simple_loss=0.3054, pruned_loss=0.08176, over 12406.00 frames. ], tot_loss[loss=0.2679, simple_loss=0.3406, pruned_loss=0.09755, over 3056574.45 frames. ], batch size: 246, lr: 3.45e-02, grad_scale: 8.0 2023-04-27 16:01:54,927 INFO [optim.py:368] (1/8) 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] (1/8) Epoch 1, batch 9250, loss[loss=0.269, simple_loss=0.3476, pruned_loss=0.09518, over 16183.00 frames. ], tot_loss[loss=0.2676, simple_loss=0.3404, pruned_loss=0.09739, over 3065250.51 frames. ], batch size: 165, lr: 3.45e-02, grad_scale: 8.0 2023-04-27 16:04:16,909 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=9283.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 16:04:58,123 INFO [train.py:904] (1/8) Epoch 1, batch 9300, loss[loss=0.2486, simple_loss=0.3273, pruned_loss=0.08489, over 16739.00 frames. ], tot_loss[loss=0.2636, simple_loss=0.3368, pruned_loss=0.0952, over 3051158.96 frames. ], batch size: 134, lr: 3.44e-02, grad_scale: 8.0 2023-04-27 16:05:33,416 INFO [optim.py:368] (1/8) Clipping_scale=2.0, grad-norm quartiles 2.238e+02 3.939e+02 4.601e+02 5.465e+02 1.094e+03, threshold=9.201e+02, percent-clipped=0.0 2023-04-27 16:06:27,288 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=9342.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 16:06:44,333 INFO [train.py:904] (1/8) Epoch 1, batch 9350, loss[loss=0.2471, simple_loss=0.3273, pruned_loss=0.0834, over 16819.00 frames. ], tot_loss[loss=0.2615, simple_loss=0.3354, pruned_loss=0.09381, over 3060972.80 frames. ], batch size: 90, lr: 3.43e-02, grad_scale: 8.0 2023-04-27 16:07:24,444 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=9369.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 16:08:26,592 INFO [train.py:904] (1/8) Epoch 1, batch 9400, loss[loss=0.2818, simple_loss=0.3553, pruned_loss=0.1042, over 15310.00 frames. ], tot_loss[loss=0.2611, simple_loss=0.3346, pruned_loss=0.09373, over 3035373.31 frames. ], batch size: 191, lr: 3.43e-02, grad_scale: 8.0 2023-04-27 16:08:54,918 INFO [zipformer.py:625] (1/8) 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] (1/8) 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 16:08:57,386 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=9416.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 16:08:59,801 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([1.8270, 3.0519, 2.2228, 3.6340, 3.5896, 3.6511, 2.1891, 2.9982], device='cuda:1'), covar=tensor([0.1999, 0.0364, 0.1618, 0.0114, 0.0172, 0.0264, 0.1081, 0.0528], device='cuda:1'), in_proj_covar=tensor([0.0156, 0.0095, 0.0159, 0.0062, 0.0071, 0.0082, 0.0136, 0.0121], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-27 16:10:08,181 INFO [train.py:904] (1/8) Epoch 1, batch 9450, loss[loss=0.2551, simple_loss=0.3229, pruned_loss=0.09363, over 12187.00 frames. ], tot_loss[loss=0.2632, simple_loss=0.3376, pruned_loss=0.09447, over 3034803.82 frames. ], batch size: 248, lr: 3.42e-02, grad_scale: 8.0 2023-04-27 16:10:33,964 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=9464.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 16:10:58,519 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=9475.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 16:11:50,036 INFO [train.py:904] (1/8) Epoch 1, batch 9500, loss[loss=0.2743, simple_loss=0.3495, pruned_loss=0.09956, over 15417.00 frames. ], tot_loss[loss=0.2614, simple_loss=0.3362, pruned_loss=0.09329, over 3049019.91 frames. ], batch size: 191, lr: 3.41e-02, grad_scale: 8.0 2023-04-27 16:12:21,411 INFO [optim.py:368] (1/8) 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 16:13:02,255 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([1.8789, 3.5542, 2.5779, 4.4681, 4.3676, 4.2249, 2.4947, 3.7009], device='cuda:1'), covar=tensor([0.2217, 0.0405, 0.1513, 0.0080, 0.0084, 0.0281, 0.1039, 0.0428], device='cuda:1'), in_proj_covar=tensor([0.0159, 0.0097, 0.0161, 0.0061, 0.0070, 0.0085, 0.0140, 0.0122], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-27 16:13:37,007 INFO [train.py:904] (1/8) Epoch 1, batch 9550, loss[loss=0.2554, simple_loss=0.3262, pruned_loss=0.09232, over 12546.00 frames. ], tot_loss[loss=0.2619, simple_loss=0.3364, pruned_loss=0.09366, over 3048297.55 frames. ], batch size: 250, lr: 3.41e-02, grad_scale: 8.0 2023-04-27 16:14:46,646 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=9583.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 16:14:48,702 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=9584.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 16:15:14,323 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.7450, 2.6573, 1.5236, 2.7364, 1.8699, 2.7262, 1.6729, 2.2680], device='cuda:1'), covar=tensor([0.0090, 0.0161, 0.1370, 0.0097, 0.0829, 0.0235, 0.1267, 0.0567], device='cuda:1'), in_proj_covar=tensor([0.0067, 0.0077, 0.0160, 0.0071, 0.0138, 0.0087, 0.0167, 0.0130], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0002, 0.0003, 0.0001, 0.0002, 0.0002, 0.0003, 0.0002], device='cuda:1') 2023-04-27 16:15:18,643 INFO [train.py:904] (1/8) Epoch 1, batch 9600, loss[loss=0.309, simple_loss=0.3817, pruned_loss=0.1182, over 15288.00 frames. ], tot_loss[loss=0.2648, simple_loss=0.3385, pruned_loss=0.09549, over 3036907.59 frames. ], batch size: 191, lr: 3.40e-02, grad_scale: 8.0 2023-04-27 16:15:48,621 INFO [optim.py:368] (1/8) 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 16:15:54,143 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([1.9311, 3.4866, 2.1827, 4.2964, 4.3295, 4.2690, 1.9481, 3.5285], device='cuda:1'), covar=tensor([0.1903, 0.0326, 0.1784, 0.0080, 0.0099, 0.0308, 0.1254, 0.0413], device='cuda:1'), in_proj_covar=tensor([0.0157, 0.0097, 0.0160, 0.0061, 0.0071, 0.0086, 0.0139, 0.0122], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-27 16:16:20,294 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=9631.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 16:16:25,353 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.66 vs. limit=2.0 2023-04-27 16:16:45,905 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=9642.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 16:16:53,245 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=9645.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 16:17:08,555 INFO [train.py:904] (1/8) Epoch 1, batch 9650, loss[loss=0.2308, simple_loss=0.3161, pruned_loss=0.07274, over 16536.00 frames. ], tot_loss[loss=0.2675, simple_loss=0.3419, pruned_loss=0.09657, over 3048067.13 frames. ], batch size: 68, lr: 3.39e-02, grad_scale: 8.0 2023-04-27 16:17:26,826 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=9658.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 16:17:52,713 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=9669.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 16:18:35,173 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=9690.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 16:18:57,043 INFO [train.py:904] (1/8) Epoch 1, batch 9700, loss[loss=0.2882, simple_loss=0.3605, pruned_loss=0.108, over 16218.00 frames. ], tot_loss[loss=0.2659, simple_loss=0.3405, pruned_loss=0.09565, over 3067474.70 frames. ], batch size: 165, lr: 3.38e-02, grad_scale: 8.0 2023-04-27 16:19:24,736 INFO [optim.py:368] (1/8) 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] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=9717.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 16:19:33,541 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=9719.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 16:20:27,182 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 1, batch 9750, loss[loss=0.2475, simple_loss=0.3228, pruned_loss=0.0861, over 17151.00 frames. ], tot_loss[loss=0.2641, simple_loss=0.338, pruned_loss=0.09505, over 3056223.07 frames. ], batch size: 48, lr: 3.38e-02, grad_scale: 8.0 2023-04-27 16:20:55,205 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.3976, 3.0633, 2.7700, 2.6387, 2.0939, 1.8347, 3.0286, 3.3349], device='cuda:1'), covar=tensor([0.1410, 0.0559, 0.0861, 0.0336, 0.1571, 0.1436, 0.0279, 0.0136], device='cuda:1'), in_proj_covar=tensor([0.0218, 0.0192, 0.0213, 0.0131, 0.0174, 0.0166, 0.0146, 0.0079], device='cuda:1'), out_proj_covar=tensor([2.5535e-04, 2.2242e-04, 2.3281e-04, 1.5390e-04, 2.0783e-04, 2.0082e-04, 1.7466e-04, 9.9851e-05], device='cuda:1') 2023-04-27 16:21:13,609 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.4641, 1.5917, 1.7685, 1.3242, 2.2283, 2.0088, 2.2878, 2.4239], device='cuda:1'), covar=tensor([0.0027, 0.0288, 0.0151, 0.0281, 0.0082, 0.0178, 0.0042, 0.0063], device='cuda:1'), in_proj_covar=tensor([0.0032, 0.0075, 0.0059, 0.0066, 0.0053, 0.0064, 0.0034, 0.0041], device='cuda:1'), out_proj_covar=tensor([4.3527e-05, 1.1634e-04, 8.9616e-05, 1.0080e-04, 8.1441e-05, 9.8022e-05, 5.3167e-05, 6.6452e-05], device='cuda:1') 2023-04-27 16:21:17,335 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=9770.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 16:22:10,630 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.92 vs. limit=2.0 2023-04-27 16:22:19,290 INFO [train.py:904] (1/8) Epoch 1, batch 9800, loss[loss=0.255, simple_loss=0.3369, pruned_loss=0.08655, over 16584.00 frames. ], tot_loss[loss=0.2619, simple_loss=0.3373, pruned_loss=0.09326, over 3070083.01 frames. ], batch size: 57, lr: 3.37e-02, grad_scale: 8.0 2023-04-27 16:22:26,146 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=9804.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 16:22:35,435 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.4159, 1.5700, 1.6952, 1.4780, 2.0084, 1.9754, 2.1526, 2.3062], device='cuda:1'), covar=tensor([0.0024, 0.0238, 0.0130, 0.0177, 0.0081, 0.0141, 0.0049, 0.0053], device='cuda:1'), in_proj_covar=tensor([0.0032, 0.0075, 0.0059, 0.0065, 0.0052, 0.0065, 0.0034, 0.0042], device='cuda:1'), out_proj_covar=tensor([4.3688e-05, 1.1638e-04, 8.9716e-05, 1.0023e-04, 8.0504e-05, 9.9834e-05, 5.2097e-05, 6.7512e-05], device='cuda:1') 2023-04-27 16:22:47,902 INFO [optim.py:368] (1/8) 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 16:24:05,875 INFO [train.py:904] (1/8) Epoch 1, batch 9850, loss[loss=0.2766, simple_loss=0.3506, pruned_loss=0.1013, over 15287.00 frames. ], tot_loss[loss=0.2629, simple_loss=0.3392, pruned_loss=0.09327, over 3066585.59 frames. ], batch size: 191, lr: 3.36e-02, grad_scale: 8.0 2023-04-27 16:25:58,731 INFO [train.py:904] (1/8) Epoch 1, batch 9900, loss[loss=0.2527, simple_loss=0.3422, pruned_loss=0.08157, over 16304.00 frames. ], tot_loss[loss=0.2631, simple_loss=0.3401, pruned_loss=0.09309, over 3078443.83 frames. ], batch size: 165, lr: 3.36e-02, grad_scale: 8.0 2023-04-27 16:26:03,920 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.2015, 4.2374, 4.5020, 4.5294, 4.6520, 4.2010, 4.3132, 4.3643], device='cuda:1'), covar=tensor([0.0183, 0.0203, 0.0439, 0.0428, 0.0280, 0.0251, 0.0463, 0.0234], device='cuda:1'), in_proj_covar=tensor([0.0113, 0.0109, 0.0122, 0.0126, 0.0132, 0.0114, 0.0159, 0.0113], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-27 16:26:18,752 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.95 vs. limit=2.0 2023-04-27 16:26:31,897 INFO [optim.py:368] (1/8) 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 16:26:58,784 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.92 vs. limit=2.0 2023-04-27 16:27:21,744 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.6090, 3.6009, 3.0439, 3.2961, 2.5928, 2.1609, 3.7809, 4.1217], device='cuda:1'), covar=tensor([0.1678, 0.0591, 0.0952, 0.0355, 0.1417, 0.1227, 0.0266, 0.0063], device='cuda:1'), in_proj_covar=tensor([0.0214, 0.0194, 0.0210, 0.0131, 0.0171, 0.0164, 0.0148, 0.0078], device='cuda:1'), out_proj_covar=tensor([2.4935e-04, 2.2492e-04, 2.2841e-04, 1.5458e-04, 2.0449e-04, 1.9834e-04, 1.7582e-04, 9.7223e-05], device='cuda:1') 2023-04-27 16:27:32,218 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=9940.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 16:27:57,270 INFO [train.py:904] (1/8) Epoch 1, batch 9950, loss[loss=0.2343, simple_loss=0.325, pruned_loss=0.07185, over 16934.00 frames. ], tot_loss[loss=0.2639, simple_loss=0.3415, pruned_loss=0.0932, over 3071685.34 frames. ], batch size: 102, lr: 3.35e-02, grad_scale: 8.0 2023-04-27 16:30:02,395 INFO [train.py:904] (1/8) Epoch 1, batch 10000, loss[loss=0.2278, simple_loss=0.3128, pruned_loss=0.07145, over 16631.00 frames. ], tot_loss[loss=0.262, simple_loss=0.3397, pruned_loss=0.09213, over 3091911.03 frames. ], batch size: 62, lr: 3.34e-02, grad_scale: 8.0 2023-04-27 16:30:30,092 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=10014.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 16:30:32,719 INFO [optim.py:368] (1/8) 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 16:30:37,699 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.56 vs. limit=2.0 2023-04-27 16:31:21,937 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=10039.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 16:31:43,672 INFO [train.py:904] (1/8) Epoch 1, batch 10050, loss[loss=0.2725, simple_loss=0.3463, pruned_loss=0.09935, over 12165.00 frames. ], tot_loss[loss=0.2613, simple_loss=0.3392, pruned_loss=0.09164, over 3068187.77 frames. ], batch size: 248, lr: 3.34e-02, grad_scale: 8.0 2023-04-27 16:31:55,291 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.82 vs. limit=2.0 2023-04-27 16:32:06,036 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=10062.0, num_to_drop=1, layers_to_drop={0} 2023-04-27 16:32:23,069 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=10070.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 16:33:16,684 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=10099.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 16:33:18,996 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=10100.0, num_to_drop=1, layers_to_drop={3} 2023-04-27 16:33:19,582 INFO [train.py:904] (1/8) Epoch 1, batch 10100, loss[loss=0.2424, simple_loss=0.3246, pruned_loss=0.08006, over 16602.00 frames. ], tot_loss[loss=0.2625, simple_loss=0.3401, pruned_loss=0.09238, over 3080655.17 frames. ], batch size: 57, lr: 3.33e-02, grad_scale: 16.0 2023-04-27 16:33:49,433 INFO [optim.py:368] (1/8) 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 16:33:53,060 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=10118.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 16:34:06,036 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=10123.0, num_to_drop=1, layers_to_drop={1} 2023-04-27 16:35:05,148 INFO [train.py:904] (1/8) Epoch 2, batch 0, loss[loss=0.312, simple_loss=0.3694, pruned_loss=0.1273, over 17273.00 frames. ], tot_loss[loss=0.312, simple_loss=0.3694, pruned_loss=0.1273, over 17273.00 frames. ], batch size: 52, lr: 3.26e-02, grad_scale: 8.0 2023-04-27 16:35:05,148 INFO [train.py:929] (1/8) Computing validation loss 2023-04-27 16:35:12,743 INFO [train.py:938] (1/8) Epoch 2, validation: loss=0.2235, simple_loss=0.3221, pruned_loss=0.06242, over 944034.00 frames. 2023-04-27 16:35:12,743 INFO [train.py:939] (1/8) Maximum memory allocated so far is 17407MB 2023-04-27 16:35:57,155 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.80 vs. limit=2.0 2023-04-27 16:36:07,568 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.5190, 4.7193, 4.9831, 5.0647, 5.1645, 4.5688, 4.7037, 4.7936], device='cuda:1'), covar=tensor([0.0229, 0.0182, 0.0347, 0.0338, 0.0313, 0.0220, 0.0566, 0.0239], device='cuda:1'), in_proj_covar=tensor([0.0125, 0.0117, 0.0132, 0.0137, 0.0143, 0.0123, 0.0179, 0.0121], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-27 16:36:22,853 INFO [train.py:904] (1/8) Epoch 2, batch 50, loss[loss=0.3733, simple_loss=0.3987, pruned_loss=0.174, over 16540.00 frames. ], tot_loss[loss=0.3181, simple_loss=0.3695, pruned_loss=0.1333, over 755540.28 frames. ], batch size: 75, lr: 3.25e-02, grad_scale: 4.0 2023-04-27 16:36:45,677 INFO [optim.py:368] (1/8) Clipping_scale=2.0, grad-norm quartiles 3.307e+02 5.068e+02 6.475e+02 7.712e+02 1.660e+03, threshold=1.295e+03, percent-clipped=4.0 2023-04-27 16:37:16,247 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=10240.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 16:37:30,898 INFO [train.py:904] (1/8) Epoch 2, batch 100, loss[loss=0.252, simple_loss=0.3137, pruned_loss=0.09519, over 16822.00 frames. ], tot_loss[loss=0.3089, simple_loss=0.3622, pruned_loss=0.1278, over 1317836.56 frames. ], batch size: 39, lr: 3.25e-02, grad_scale: 4.0 2023-04-27 16:38:22,578 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=10288.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 16:38:24,088 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.83 vs. limit=2.0 2023-04-27 16:38:38,636 INFO [train.py:904] (1/8) Epoch 2, batch 150, loss[loss=0.2933, simple_loss=0.3615, pruned_loss=0.1126, over 17244.00 frames. ], tot_loss[loss=0.3009, simple_loss=0.3575, pruned_loss=0.1221, over 1774318.05 frames. ], batch size: 52, lr: 3.24e-02, grad_scale: 4.0 2023-04-27 16:38:56,059 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=10314.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 16:38:59,694 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.9899, 3.9158, 1.8846, 4.0349, 2.4613, 4.0268, 1.9986, 3.0418], device='cuda:1'), covar=tensor([0.0046, 0.0116, 0.1497, 0.0039, 0.0722, 0.0179, 0.1262, 0.0531], device='cuda:1'), in_proj_covar=tensor([0.0069, 0.0082, 0.0162, 0.0072, 0.0138, 0.0096, 0.0163, 0.0138], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0002, 0.0003, 0.0001, 0.0003, 0.0002, 0.0003, 0.0003], device='cuda:1') 2023-04-27 16:39:01,491 INFO [optim.py:368] (1/8) 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 16:39:47,448 INFO [train.py:904] (1/8) Epoch 2, batch 200, loss[loss=0.3056, simple_loss=0.3445, pruned_loss=0.1333, over 16874.00 frames. ], tot_loss[loss=0.3011, simple_loss=0.3572, pruned_loss=0.1225, over 2108134.58 frames. ], batch size: 96, lr: 3.23e-02, grad_scale: 4.0 2023-04-27 16:40:03,410 INFO [zipformer.py:625] (1/8) 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] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=10395.0, num_to_drop=1, layers_to_drop={3} 2023-04-27 16:40:55,026 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 2, batch 250, loss[loss=0.3145, simple_loss=0.3589, pruned_loss=0.135, over 16330.00 frames. ], tot_loss[loss=0.2987, simple_loss=0.3548, pruned_loss=0.1213, over 2384300.42 frames. ], batch size: 165, lr: 3.23e-02, grad_scale: 4.0 2023-04-27 16:41:10,381 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=10409.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 16:41:22,651 INFO [optim.py:368] (1/8) 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 16:41:22,935 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=10418.0, num_to_drop=1, layers_to_drop={3} 2023-04-27 16:42:03,850 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 2, batch 300, loss[loss=0.2719, simple_loss=0.3284, pruned_loss=0.1077, over 16503.00 frames. ], tot_loss[loss=0.2926, simple_loss=0.35, pruned_loss=0.1176, over 2591590.10 frames. ], batch size: 75, lr: 3.22e-02, grad_scale: 4.0 2023-04-27 16:42:14,789 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=10454.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 16:42:36,332 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 2, batch 350, loss[loss=0.2651, simple_loss=0.3373, pruned_loss=0.09644, over 17178.00 frames. ], tot_loss[loss=0.2881, simple_loss=0.3464, pruned_loss=0.1149, over 2744871.72 frames. ], batch size: 46, lr: 3.21e-02, grad_scale: 4.0 2023-04-27 16:43:39,868 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=10515.0, num_to_drop=1, layers_to_drop={0} 2023-04-27 16:43:42,893 INFO [optim.py:368] (1/8) 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 16:44:27,074 INFO [train.py:904] (1/8) Epoch 2, batch 400, loss[loss=0.3313, simple_loss=0.3604, pruned_loss=0.1511, over 16859.00 frames. ], tot_loss[loss=0.2866, simple_loss=0.3444, pruned_loss=0.1144, over 2875868.52 frames. ], batch size: 90, lr: 3.21e-02, grad_scale: 8.0 2023-04-27 16:45:36,495 INFO [train.py:904] (1/8) Epoch 2, batch 450, loss[loss=0.2915, simple_loss=0.3352, pruned_loss=0.124, over 16172.00 frames. ], tot_loss[loss=0.2836, simple_loss=0.3421, pruned_loss=0.1125, over 2984116.88 frames. ], batch size: 165, lr: 3.20e-02, grad_scale: 8.0 2023-04-27 16:45:59,585 INFO [optim.py:368] (1/8) 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 16:46:03,009 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.6129, 4.2032, 4.4914, 4.6247, 3.9450, 4.4336, 4.3924, 4.1833], device='cuda:1'), covar=tensor([0.0289, 0.0227, 0.0159, 0.0105, 0.0862, 0.0180, 0.0217, 0.0204], device='cuda:1'), in_proj_covar=tensor([0.0117, 0.0086, 0.0154, 0.0122, 0.0190, 0.0120, 0.0107, 0.0136], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-04-27 16:46:28,501 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.6041, 3.4560, 1.7356, 3.5206, 2.4468, 3.5639, 1.8812, 2.7204], device='cuda:1'), covar=tensor([0.0049, 0.0153, 0.1425, 0.0056, 0.0641, 0.0199, 0.1267, 0.0580], device='cuda:1'), in_proj_covar=tensor([0.0072, 0.0084, 0.0160, 0.0074, 0.0141, 0.0104, 0.0165, 0.0141], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0002, 0.0003, 0.0001, 0.0003, 0.0002, 0.0003, 0.0003], device='cuda:1') 2023-04-27 16:46:44,100 INFO [train.py:904] (1/8) Epoch 2, batch 500, loss[loss=0.2312, simple_loss=0.3075, pruned_loss=0.07744, over 17136.00 frames. ], tot_loss[loss=0.282, simple_loss=0.3403, pruned_loss=0.1119, over 3044871.49 frames. ], batch size: 47, lr: 3.20e-02, grad_scale: 8.0 2023-04-27 16:47:44,945 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=10695.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 16:47:52,720 INFO [train.py:904] (1/8) Epoch 2, batch 550, loss[loss=0.2443, simple_loss=0.3186, pruned_loss=0.08497, over 17230.00 frames. ], tot_loss[loss=0.2804, simple_loss=0.3391, pruned_loss=0.1109, over 3097919.05 frames. ], batch size: 45, lr: 3.19e-02, grad_scale: 8.0 2023-04-27 16:48:17,047 INFO [optim.py:368] (1/8) 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 16:48:17,341 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=10718.0, num_to_drop=1, layers_to_drop={2} 2023-04-27 16:48:51,485 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=10743.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 16:48:52,754 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=10744.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 16:49:02,282 INFO [train.py:904] (1/8) Epoch 2, batch 600, loss[loss=0.2627, simple_loss=0.3356, pruned_loss=0.09495, over 16708.00 frames. ], tot_loss[loss=0.2793, simple_loss=0.3379, pruned_loss=0.1104, over 3136244.50 frames. ], batch size: 57, lr: 3.18e-02, grad_scale: 8.0 2023-04-27 16:49:21,916 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=10765.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 16:49:23,782 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=10766.0, num_to_drop=1, layers_to_drop={1} 2023-04-27 16:49:33,715 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 2, batch 650, loss[loss=0.2576, simple_loss=0.3131, pruned_loss=0.101, over 16912.00 frames. ], tot_loss[loss=0.2765, simple_loss=0.3352, pruned_loss=0.1089, over 3173926.17 frames. ], batch size: 96, lr: 3.18e-02, grad_scale: 8.0 2023-04-27 16:50:16,440 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=10805.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 16:50:23,981 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=10810.0, num_to_drop=1, layers_to_drop={3} 2023-04-27 16:50:33,023 INFO [optim.py:368] (1/8) Clipping_scale=2.0, grad-norm quartiles 2.483e+02 4.152e+02 4.979e+02 6.118e+02 1.196e+03, threshold=9.959e+02, percent-clipped=5.0 2023-04-27 16:50:48,241 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.76 vs. limit=2.0 2023-04-27 16:50:57,032 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=10835.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 16:51:19,285 INFO [train.py:904] (1/8) Epoch 2, batch 700, loss[loss=0.2674, simple_loss=0.3322, pruned_loss=0.1013, over 16727.00 frames. ], tot_loss[loss=0.275, simple_loss=0.3342, pruned_loss=0.1079, over 3210847.17 frames. ], batch size: 62, lr: 3.17e-02, grad_scale: 8.0 2023-04-27 16:52:06,369 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=10885.0, num_to_drop=1, layers_to_drop={1} 2023-04-27 16:52:19,019 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([1.9744, 3.9052, 2.2870, 4.7577, 4.5373, 4.4786, 2.3023, 3.5017], device='cuda:1'), covar=tensor([0.1516, 0.0283, 0.1461, 0.0067, 0.0137, 0.0226, 0.1027, 0.0422], device='cuda:1'), in_proj_covar=tensor([0.0154, 0.0106, 0.0164, 0.0063, 0.0082, 0.0091, 0.0145, 0.0129], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0003, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-27 16:52:26,661 INFO [train.py:904] (1/8) Epoch 2, batch 750, loss[loss=0.2257, simple_loss=0.303, pruned_loss=0.07415, over 17215.00 frames. ], tot_loss[loss=0.2747, simple_loss=0.3344, pruned_loss=0.1075, over 3235817.41 frames. ], batch size: 44, lr: 3.17e-02, grad_scale: 8.0 2023-04-27 16:52:50,206 INFO [optim.py:368] (1/8) 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 16:53:11,007 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=3.74 vs. limit=5.0 2023-04-27 16:53:28,035 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=10946.0, num_to_drop=1, layers_to_drop={2} 2023-04-27 16:53:33,379 INFO [train.py:904] (1/8) Epoch 2, batch 800, loss[loss=0.2696, simple_loss=0.3221, pruned_loss=0.1086, over 16736.00 frames. ], tot_loss[loss=0.273, simple_loss=0.3333, pruned_loss=0.1064, over 3255390.15 frames. ], batch size: 134, lr: 3.16e-02, grad_scale: 8.0 2023-04-27 16:54:43,055 INFO [train.py:904] (1/8) Epoch 2, batch 850, loss[loss=0.2029, simple_loss=0.2742, pruned_loss=0.06585, over 16788.00 frames. ], tot_loss[loss=0.2711, simple_loss=0.3319, pruned_loss=0.1052, over 3273796.49 frames. ], batch size: 39, lr: 3.15e-02, grad_scale: 8.0 2023-04-27 16:55:06,011 INFO [optim.py:368] (1/8) 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] (1/8) Epoch 2, batch 900, loss[loss=0.2514, simple_loss=0.3353, pruned_loss=0.08377, over 17121.00 frames. ], tot_loss[loss=0.2684, simple_loss=0.3305, pruned_loss=0.1032, over 3287239.44 frames. ], batch size: 49, lr: 3.15e-02, grad_scale: 8.0 2023-04-27 16:56:09,987 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=11065.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 16:56:56,879 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=11100.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 16:56:58,312 INFO [train.py:904] (1/8) Epoch 2, batch 950, loss[loss=0.2657, simple_loss=0.3239, pruned_loss=0.1037, over 16868.00 frames. ], tot_loss[loss=0.2682, simple_loss=0.3306, pruned_loss=0.1029, over 3300268.69 frames. ], batch size: 42, lr: 3.14e-02, grad_scale: 8.0 2023-04-27 16:57:00,572 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.3983, 3.6539, 3.0701, 4.7214, 2.7718, 4.2468, 3.0225, 2.6644], device='cuda:1'), covar=tensor([0.0198, 0.0267, 0.0274, 0.0151, 0.1181, 0.0167, 0.0513, 0.1302], device='cuda:1'), in_proj_covar=tensor([0.0167, 0.0148, 0.0121, 0.0169, 0.0225, 0.0139, 0.0159, 0.0203], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003], device='cuda:1') 2023-04-27 16:57:10,482 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=11110.0, num_to_drop=1, layers_to_drop={1} 2023-04-27 16:57:13,954 INFO [zipformer.py:625] (1/8) 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] (1/8) Clipping_scale=2.0, grad-norm quartiles 2.464e+02 3.927e+02 4.944e+02 5.832e+02 1.325e+03, threshold=9.889e+02, percent-clipped=4.0 2023-04-27 16:57:38,004 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=11130.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 16:57:46,327 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.2867, 3.1513, 2.9980, 1.9359, 2.7218, 2.0469, 2.7899, 3.1220], device='cuda:1'), covar=tensor([0.0338, 0.0412, 0.0331, 0.1574, 0.0697, 0.1040, 0.0715, 0.0306], device='cuda:1'), in_proj_covar=tensor([0.0128, 0.0097, 0.0137, 0.0157, 0.0147, 0.0139, 0.0145, 0.0093], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0002], device='cuda:1') 2023-04-27 16:58:06,961 INFO [train.py:904] (1/8) Epoch 2, batch 1000, loss[loss=0.2382, simple_loss=0.3151, pruned_loss=0.08064, over 17208.00 frames. ], tot_loss[loss=0.2667, simple_loss=0.3289, pruned_loss=0.1023, over 3309546.89 frames. ], batch size: 46, lr: 3.14e-02, grad_scale: 8.0 2023-04-27 16:58:16,902 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=11158.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 16:58:26,156 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.6730, 4.7164, 5.1358, 5.2010, 5.2804, 4.7206, 4.9379, 4.9502], device='cuda:1'), covar=tensor([0.0262, 0.0285, 0.0473, 0.0429, 0.0376, 0.0293, 0.0605, 0.0293], device='cuda:1'), in_proj_covar=tensor([0.0149, 0.0142, 0.0160, 0.0158, 0.0184, 0.0149, 0.0231, 0.0144], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0002], device='cuda:1') 2023-04-27 16:58:37,481 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 2, batch 1050, loss[loss=0.2864, simple_loss=0.3543, pruned_loss=0.1092, over 17068.00 frames. ], tot_loss[loss=0.267, simple_loss=0.3293, pruned_loss=0.1024, over 3306525.67 frames. ], batch size: 55, lr: 3.13e-02, grad_scale: 8.0 2023-04-27 16:59:26,446 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=2.05 vs. limit=2.0 2023-04-27 16:59:36,187 INFO [optim.py:368] (1/8) 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 17:00:00,085 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=11235.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 17:00:08,424 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=11241.0, num_to_drop=1, layers_to_drop={2} 2023-04-27 17:00:21,895 INFO [train.py:904] (1/8) Epoch 2, batch 1100, loss[loss=0.275, simple_loss=0.3277, pruned_loss=0.1112, over 16694.00 frames. ], tot_loss[loss=0.2659, simple_loss=0.3283, pruned_loss=0.1017, over 3309589.48 frames. ], batch size: 83, lr: 3.12e-02, grad_scale: 8.0 2023-04-27 17:00:22,290 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.8962, 3.8843, 3.7761, 3.0929, 3.9491, 1.9419, 3.6755, 3.8072], device='cuda:1'), covar=tensor([0.0144, 0.0088, 0.0104, 0.0475, 0.0080, 0.1483, 0.0114, 0.0137], device='cuda:1'), in_proj_covar=tensor([0.0061, 0.0052, 0.0073, 0.0097, 0.0055, 0.0104, 0.0071, 0.0074], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0002, 0.0002], device='cuda:1') 2023-04-27 17:00:49,047 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([5.0105, 5.7840, 5.7256, 5.5390, 5.6701, 6.0719, 6.0385, 5.6683], device='cuda:1'), covar=tensor([0.0505, 0.0895, 0.0771, 0.1424, 0.1948, 0.0715, 0.0561, 0.1618], device='cuda:1'), in_proj_covar=tensor([0.0172, 0.0268, 0.0232, 0.0233, 0.0290, 0.0235, 0.0201, 0.0293], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003], device='cuda:1') 2023-04-27 17:01:28,364 INFO [train.py:904] (1/8) Epoch 2, batch 1150, loss[loss=0.2907, simple_loss=0.3399, pruned_loss=0.1207, over 16733.00 frames. ], tot_loss[loss=0.2652, simple_loss=0.328, pruned_loss=0.1012, over 3317281.02 frames. ], batch size: 124, lr: 3.12e-02, grad_scale: 8.0 2023-04-27 17:01:52,662 INFO [optim.py:368] (1/8) 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 17:01:55,429 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.2128, 3.0381, 3.1845, 3.5026, 3.4152, 3.2428, 3.3933, 3.3881], device='cuda:1'), covar=tensor([0.0356, 0.0437, 0.0931, 0.0333, 0.0428, 0.0961, 0.0423, 0.0384], device='cuda:1'), in_proj_covar=tensor([0.0215, 0.0241, 0.0358, 0.0250, 0.0207, 0.0204, 0.0186, 0.0208], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-27 17:02:39,328 INFO [train.py:904] (1/8) Epoch 2, batch 1200, loss[loss=0.286, simple_loss=0.3347, pruned_loss=0.1186, over 12099.00 frames. ], tot_loss[loss=0.2647, simple_loss=0.3277, pruned_loss=0.1008, over 3301246.43 frames. ], batch size: 247, lr: 3.11e-02, grad_scale: 8.0 2023-04-27 17:03:01,021 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.9825, 2.5781, 2.1225, 3.2336, 2.3595, 3.0303, 2.5156, 2.2600], device='cuda:1'), covar=tensor([0.0270, 0.0306, 0.0313, 0.0236, 0.0934, 0.0192, 0.0473, 0.0941], device='cuda:1'), in_proj_covar=tensor([0.0169, 0.0149, 0.0124, 0.0173, 0.0229, 0.0142, 0.0158, 0.0205], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003], device='cuda:1') 2023-04-27 17:03:46,800 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 2, batch 1250, loss[loss=0.3076, simple_loss=0.3736, pruned_loss=0.1209, over 17278.00 frames. ], tot_loss[loss=0.2658, simple_loss=0.3283, pruned_loss=0.1016, over 3300541.89 frames. ], batch size: 52, lr: 3.11e-02, grad_scale: 8.0 2023-04-27 17:04:10,333 INFO [optim.py:368] (1/8) 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 17:04:26,797 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=11430.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 17:04:39,461 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.97 vs. limit=2.0 2023-04-27 17:04:49,846 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=11448.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 17:04:53,724 INFO [train.py:904] (1/8) Epoch 2, batch 1300, loss[loss=0.272, simple_loss=0.3265, pruned_loss=0.1087, over 16701.00 frames. ], tot_loss[loss=0.2673, simple_loss=0.329, pruned_loss=0.1028, over 3304831.87 frames. ], batch size: 89, lr: 3.10e-02, grad_scale: 8.0 2023-04-27 17:05:26,302 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=2.00 vs. limit=2.0 2023-04-27 17:05:30,716 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 2, batch 1350, loss[loss=0.2553, simple_loss=0.3144, pruned_loss=0.09805, over 15541.00 frames. ], tot_loss[loss=0.2661, simple_loss=0.3285, pruned_loss=0.1018, over 3303028.16 frames. ], batch size: 190, lr: 3.10e-02, grad_scale: 8.0 2023-04-27 17:06:24,753 INFO [optim.py:368] (1/8) 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 17:06:41,095 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=11530.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 17:06:45,033 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=11532.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 17:06:57,662 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 2, batch 1400, loss[loss=0.2526, simple_loss=0.3117, pruned_loss=0.0967, over 16858.00 frames. ], tot_loss[loss=0.265, simple_loss=0.3273, pruned_loss=0.1013, over 3308237.73 frames. ], batch size: 96, lr: 3.09e-02, grad_scale: 8.0 2023-04-27 17:08:03,054 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=11589.0, num_to_drop=1, layers_to_drop={1} 2023-04-27 17:08:08,132 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=11593.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 17:08:17,067 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.8252, 4.8560, 5.3902, 5.4156, 5.4524, 5.0527, 5.0796, 5.1437], device='cuda:1'), covar=tensor([0.0239, 0.0257, 0.0313, 0.0344, 0.0274, 0.0221, 0.0674, 0.0223], device='cuda:1'), in_proj_covar=tensor([0.0157, 0.0145, 0.0170, 0.0167, 0.0197, 0.0159, 0.0243, 0.0154], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0002], device='cuda:1') 2023-04-27 17:08:19,413 INFO [train.py:904] (1/8) Epoch 2, batch 1450, loss[loss=0.2871, simple_loss=0.3328, pruned_loss=0.1207, over 16887.00 frames. ], tot_loss[loss=0.2634, simple_loss=0.3262, pruned_loss=0.1003, over 3317166.14 frames. ], batch size: 116, lr: 3.08e-02, grad_scale: 8.0 2023-04-27 17:08:43,788 INFO [optim.py:368] (1/8) 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 17:09:06,433 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=11635.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 17:09:14,302 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.76 vs. limit=2.0 2023-04-27 17:09:26,111 INFO [train.py:904] (1/8) Epoch 2, batch 1500, loss[loss=0.2562, simple_loss=0.3298, pruned_loss=0.09128, over 16727.00 frames. ], tot_loss[loss=0.2643, simple_loss=0.3265, pruned_loss=0.101, over 3323848.32 frames. ], batch size: 62, lr: 3.08e-02, grad_scale: 8.0 2023-04-27 17:09:58,990 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.0717, 4.5220, 2.5322, 5.0823, 5.2690, 4.6009, 2.8755, 4.0366], device='cuda:1'), covar=tensor([0.1775, 0.0282, 0.1554, 0.0140, 0.0118, 0.0380, 0.0992, 0.0434], device='cuda:1'), in_proj_covar=tensor([0.0153, 0.0108, 0.0162, 0.0068, 0.0089, 0.0099, 0.0148, 0.0133], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0003, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-27 17:10:09,295 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.3019, 3.0842, 2.8792, 1.9434, 2.6214, 2.0460, 2.7653, 3.1320], device='cuda:1'), covar=tensor([0.0237, 0.0310, 0.0333, 0.1393, 0.0655, 0.0900, 0.0550, 0.0231], device='cuda:1'), in_proj_covar=tensor([0.0125, 0.0097, 0.0138, 0.0156, 0.0145, 0.0138, 0.0149, 0.0094], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0002], device='cuda:1') 2023-04-27 17:10:29,989 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=11696.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 17:10:33,605 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=11699.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 17:10:36,150 INFO [train.py:904] (1/8) Epoch 2, batch 1550, loss[loss=0.3354, simple_loss=0.3827, pruned_loss=0.1441, over 16478.00 frames. ], tot_loss[loss=0.2684, simple_loss=0.3292, pruned_loss=0.1038, over 3317841.10 frames. ], batch size: 75, lr: 3.07e-02, grad_scale: 8.0 2023-04-27 17:10:58,964 INFO [optim.py:368] (1/8) 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 17:11:07,007 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=3.34 vs. limit=5.0 2023-04-27 17:11:44,738 INFO [train.py:904] (1/8) Epoch 2, batch 1600, loss[loss=0.2856, simple_loss=0.3376, pruned_loss=0.1168, over 16898.00 frames. ], tot_loss[loss=0.2691, simple_loss=0.33, pruned_loss=0.1041, over 3327873.88 frames. ], batch size: 109, lr: 3.07e-02, grad_scale: 8.0 2023-04-27 17:11:57,691 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=11760.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 17:12:04,737 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=2.02 vs. limit=2.0 2023-04-27 17:12:37,691 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.9069, 3.9056, 3.1671, 3.4182, 2.8312, 2.1479, 3.9612, 4.5020], device='cuda:1'), covar=tensor([0.1378, 0.0451, 0.0809, 0.0359, 0.1817, 0.1279, 0.0283, 0.0118], device='cuda:1'), in_proj_covar=tensor([0.0237, 0.0217, 0.0229, 0.0153, 0.0238, 0.0173, 0.0169, 0.0112], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0001], device='cuda:1') 2023-04-27 17:12:53,455 INFO [train.py:904] (1/8) Epoch 2, batch 1650, loss[loss=0.2215, simple_loss=0.2976, pruned_loss=0.07271, over 16836.00 frames. ], tot_loss[loss=0.2684, simple_loss=0.3299, pruned_loss=0.1035, over 3332118.13 frames. ], batch size: 42, lr: 3.06e-02, grad_scale: 8.0 2023-04-27 17:13:04,799 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.99 vs. limit=2.0 2023-04-27 17:13:15,784 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=11816.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 17:13:17,239 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.0936, 3.5255, 2.5902, 4.4656, 2.4421, 4.4103, 3.1292, 2.6406], device='cuda:1'), covar=tensor([0.0250, 0.0289, 0.0309, 0.0173, 0.1239, 0.0125, 0.0471, 0.1407], device='cuda:1'), in_proj_covar=tensor([0.0172, 0.0150, 0.0124, 0.0177, 0.0228, 0.0141, 0.0161, 0.0207], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003], device='cuda:1') 2023-04-27 17:13:17,808 INFO [optim.py:368] (1/8) 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 17:13:35,544 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=11830.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 17:13:56,914 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=11846.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 17:14:03,718 INFO [train.py:904] (1/8) Epoch 2, batch 1700, loss[loss=0.284, simple_loss=0.3399, pruned_loss=0.1141, over 16799.00 frames. ], tot_loss[loss=0.2709, simple_loss=0.3326, pruned_loss=0.1046, over 3335345.96 frames. ], batch size: 102, lr: 3.06e-02, grad_scale: 8.0 2023-04-27 17:14:40,637 INFO [zipformer.py:625] (1/8) 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] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=11878.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 17:14:55,343 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=11888.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 17:14:58,600 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.2286, 4.5375, 4.1611, 4.3494, 3.8251, 4.1034, 4.1086, 4.4917], device='cuda:1'), covar=tensor([0.0481, 0.0573, 0.0864, 0.0376, 0.0671, 0.0603, 0.0501, 0.0586], device='cuda:1'), in_proj_covar=tensor([0.0214, 0.0294, 0.0254, 0.0179, 0.0202, 0.0170, 0.0234, 0.0193], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-27 17:15:13,200 INFO [train.py:904] (1/8) Epoch 2, batch 1750, loss[loss=0.2617, simple_loss=0.3266, pruned_loss=0.0984, over 17223.00 frames. ], tot_loss[loss=0.2733, simple_loss=0.3349, pruned_loss=0.1059, over 3327747.71 frames. ], batch size: 45, lr: 3.05e-02, grad_scale: 8.0 2023-04-27 17:15:21,941 INFO [zipformer.py:625] (1/8) 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] (1/8) 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 17:15:39,649 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=11920.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 17:16:17,772 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 2, batch 1800, loss[loss=0.3389, simple_loss=0.3792, pruned_loss=0.1494, over 15439.00 frames. ], tot_loss[loss=0.2737, simple_loss=0.336, pruned_loss=0.1057, over 3329521.43 frames. ], batch size: 190, lr: 3.05e-02, grad_scale: 8.0 2023-04-27 17:16:22,040 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([1.9997, 3.9397, 2.5945, 4.9467, 4.8035, 4.4153, 2.2529, 3.5845], device='cuda:1'), covar=tensor([0.1564, 0.0307, 0.1423, 0.0071, 0.0151, 0.0270, 0.1072, 0.0448], device='cuda:1'), in_proj_covar=tensor([0.0149, 0.0104, 0.0162, 0.0066, 0.0091, 0.0099, 0.0144, 0.0129], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0003, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-27 17:17:03,304 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=11981.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 17:17:07,149 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=11984.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 17:17:16,340 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 2, batch 1850, loss[loss=0.2469, simple_loss=0.3206, pruned_loss=0.08662, over 17172.00 frames. ], tot_loss[loss=0.2749, simple_loss=0.3375, pruned_loss=0.1062, over 3328956.72 frames. ], batch size: 46, lr: 3.04e-02, grad_scale: 8.0 2023-04-27 17:17:45,644 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=12009.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 17:17:49,667 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.54 vs. limit=2.0 2023-04-27 17:17:52,191 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.9066, 5.5150, 5.5086, 5.3726, 5.3463, 5.8916, 5.8204, 5.4818], device='cuda:1'), covar=tensor([0.0532, 0.1066, 0.0768, 0.1536, 0.2212, 0.0682, 0.0598, 0.1762], device='cuda:1'), in_proj_covar=tensor([0.0177, 0.0265, 0.0237, 0.0232, 0.0294, 0.0237, 0.0206, 0.0300], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003], device='cuda:1') 2023-04-27 17:17:57,887 INFO [optim.py:368] (1/8) 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 17:18:24,850 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.7853, 3.4585, 2.5217, 3.7705, 3.7580, 3.8627, 3.6267, 3.7541], device='cuda:1'), covar=tensor([0.0743, 0.0910, 0.3453, 0.1086, 0.1053, 0.0926, 0.1120, 0.0979], device='cuda:1'), in_proj_covar=tensor([0.0229, 0.0255, 0.0377, 0.0273, 0.0219, 0.0215, 0.0195, 0.0219], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:1') 2023-04-27 17:18:35,106 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 2, batch 1900, loss[loss=0.2728, simple_loss=0.3269, pruned_loss=0.1094, over 16862.00 frames. ], tot_loss[loss=0.2739, simple_loss=0.3369, pruned_loss=0.1054, over 3316091.30 frames. ], batch size: 102, lr: 3.04e-02, grad_scale: 8.0 2023-04-27 17:18:48,961 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=12055.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 17:19:09,609 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.6596, 4.3427, 4.4404, 4.9571, 4.9575, 4.5232, 5.0104, 4.9387], device='cuda:1'), covar=tensor([0.0453, 0.0524, 0.1152, 0.0360, 0.0370, 0.0575, 0.0237, 0.0282], device='cuda:1'), in_proj_covar=tensor([0.0235, 0.0260, 0.0387, 0.0278, 0.0224, 0.0220, 0.0199, 0.0224], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0004, 0.0003, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-04-27 17:19:17,209 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=12076.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 17:19:18,810 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=5.42 vs. limit=5.0 2023-04-27 17:19:51,447 INFO [train.py:904] (1/8) Epoch 2, batch 1950, loss[loss=0.2535, simple_loss=0.3149, pruned_loss=0.09607, over 16816.00 frames. ], tot_loss[loss=0.2722, simple_loss=0.336, pruned_loss=0.1042, over 3315198.25 frames. ], batch size: 39, lr: 3.03e-02, grad_scale: 8.0 2023-04-27 17:20:14,620 INFO [optim.py:368] (1/8) 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 17:20:40,727 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=12137.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 17:20:59,923 INFO [train.py:904] (1/8) Epoch 2, batch 2000, loss[loss=0.246, simple_loss=0.323, pruned_loss=0.08448, over 17211.00 frames. ], tot_loss[loss=0.2701, simple_loss=0.3347, pruned_loss=0.1028, over 3321335.89 frames. ], batch size: 44, lr: 3.02e-02, grad_scale: 8.0 2023-04-27 17:21:28,839 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=12172.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 17:21:51,884 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=12188.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 17:22:09,011 INFO [train.py:904] (1/8) Epoch 2, batch 2050, loss[loss=0.3041, simple_loss=0.3534, pruned_loss=0.1274, over 16861.00 frames. ], tot_loss[loss=0.2701, simple_loss=0.3347, pruned_loss=0.1028, over 3318371.43 frames. ], batch size: 116, lr: 3.02e-02, grad_scale: 16.0 2023-04-27 17:22:10,417 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=12202.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 17:22:12,748 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=4.69 vs. limit=5.0 2023-04-27 17:22:32,914 INFO [optim.py:368] (1/8) 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] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=12236.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 17:23:06,420 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.0350, 4.0092, 1.7658, 3.9536, 2.5908, 4.0434, 1.9458, 3.0270], device='cuda:1'), covar=tensor([0.0067, 0.0126, 0.1582, 0.0067, 0.0723, 0.0195, 0.1394, 0.0528], device='cuda:1'), in_proj_covar=tensor([0.0083, 0.0098, 0.0167, 0.0075, 0.0152, 0.0118, 0.0170, 0.0148], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:1') 2023-04-27 17:23:20,403 INFO [train.py:904] (1/8) Epoch 2, batch 2100, loss[loss=0.3011, simple_loss=0.3518, pruned_loss=0.1252, over 16894.00 frames. ], tot_loss[loss=0.2727, simple_loss=0.3363, pruned_loss=0.1046, over 3319597.49 frames. ], batch size: 116, lr: 3.01e-02, grad_scale: 16.0 2023-04-27 17:23:28,492 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([1.9776, 3.8474, 2.5664, 4.8306, 4.5153, 4.5306, 1.6174, 3.7221], device='cuda:1'), covar=tensor([0.1420, 0.0280, 0.1224, 0.0051, 0.0167, 0.0185, 0.1202, 0.0397], device='cuda:1'), in_proj_covar=tensor([0.0150, 0.0107, 0.0162, 0.0065, 0.0094, 0.0101, 0.0146, 0.0131], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0003, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-27 17:23:53,737 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=12276.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 17:24:15,646 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=12291.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 17:24:28,744 INFO [train.py:904] (1/8) Epoch 2, batch 2150, loss[loss=0.2717, simple_loss=0.3299, pruned_loss=0.1067, over 16760.00 frames. ], tot_loss[loss=0.2729, simple_loss=0.3365, pruned_loss=0.1047, over 3325797.77 frames. ], batch size: 102, lr: 3.01e-02, grad_scale: 16.0 2023-04-27 17:24:33,512 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=12304.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 17:24:51,835 INFO [optim.py:368] (1/8) Clipping_scale=2.0, grad-norm quartiles 2.440e+02 4.527e+02 5.287e+02 6.148e+02 1.100e+03, threshold=1.057e+03, percent-clipped=4.0 2023-04-27 17:25:22,227 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=12339.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 17:25:23,341 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=12340.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 17:25:33,088 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.4660, 3.9199, 4.2512, 3.4711, 4.3639, 4.3176, 4.5650, 2.4455], device='cuda:1'), covar=tensor([0.0765, 0.0063, 0.0063, 0.0310, 0.0032, 0.0060, 0.0039, 0.0655], device='cuda:1'), in_proj_covar=tensor([0.0116, 0.0050, 0.0058, 0.0102, 0.0049, 0.0058, 0.0060, 0.0103], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0003], device='cuda:1') 2023-04-27 17:25:38,416 INFO [train.py:904] (1/8) Epoch 2, batch 2200, loss[loss=0.2821, simple_loss=0.3497, pruned_loss=0.1072, over 17222.00 frames. ], tot_loss[loss=0.2738, simple_loss=0.3373, pruned_loss=0.1052, over 3319055.98 frames. ], batch size: 45, lr: 3.00e-02, grad_scale: 16.0 2023-04-27 17:25:44,399 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=12355.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 17:26:25,604 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.7319, 4.5305, 2.1262, 4.5840, 2.7812, 4.6072, 2.4486, 3.2324], device='cuda:1'), covar=tensor([0.0068, 0.0131, 0.1409, 0.0034, 0.0756, 0.0174, 0.1109, 0.0525], device='cuda:1'), in_proj_covar=tensor([0.0079, 0.0097, 0.0165, 0.0075, 0.0149, 0.0119, 0.0165, 0.0147], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:1') 2023-04-27 17:26:48,779 INFO [train.py:904] (1/8) Epoch 2, batch 2250, loss[loss=0.2369, simple_loss=0.3097, pruned_loss=0.082, over 17208.00 frames. ], tot_loss[loss=0.2739, simple_loss=0.3373, pruned_loss=0.1052, over 3324868.17 frames. ], batch size: 44, lr: 3.00e-02, grad_scale: 16.0 2023-04-27 17:26:51,418 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=12403.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 17:27:12,082 INFO [optim.py:368] (1/8) Clipping_scale=2.0, grad-norm quartiles 2.514e+02 4.351e+02 5.087e+02 6.619e+02 9.905e+02, threshold=1.017e+03, percent-clipped=0.0 2023-04-27 17:27:32,273 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=12432.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 17:27:51,310 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=12446.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 17:27:57,826 INFO [train.py:904] (1/8) Epoch 2, batch 2300, loss[loss=0.2865, simple_loss=0.363, pruned_loss=0.105, over 17065.00 frames. ], tot_loss[loss=0.2741, simple_loss=0.3375, pruned_loss=0.1053, over 3321445.79 frames. ], batch size: 55, lr: 2.99e-02, grad_scale: 8.0 2023-04-27 17:28:04,106 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.5103, 3.3911, 1.6751, 3.3110, 2.3773, 3.4639, 1.7210, 2.6662], device='cuda:1'), covar=tensor([0.0066, 0.0133, 0.1377, 0.0073, 0.0638, 0.0223, 0.1211, 0.0485], device='cuda:1'), in_proj_covar=tensor([0.0081, 0.0097, 0.0164, 0.0074, 0.0150, 0.0119, 0.0166, 0.0146], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:1') 2023-04-27 17:28:26,482 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=12472.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 17:29:06,505 INFO [train.py:904] (1/8) Epoch 2, batch 2350, loss[loss=0.2424, simple_loss=0.3163, pruned_loss=0.08422, over 16216.00 frames. ], tot_loss[loss=0.2747, simple_loss=0.3381, pruned_loss=0.1056, over 3329565.47 frames. ], batch size: 36, lr: 2.99e-02, grad_scale: 8.0 2023-04-27 17:29:07,831 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=12502.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 17:29:15,000 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=12507.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 17:29:31,566 INFO [optim.py:368] (1/8) 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 17:29:33,871 INFO [zipformer.py:625] (1/8) 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] (1/8) 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] (1/8) Epoch 2, batch 2400, loss[loss=0.2218, simple_loss=0.2974, pruned_loss=0.07304, over 16845.00 frames. ], tot_loss[loss=0.2749, simple_loss=0.3389, pruned_loss=0.1054, over 3325943.77 frames. ], batch size: 42, lr: 2.98e-02, grad_scale: 8.0 2023-04-27 17:30:52,443 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=12576.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 17:31:26,492 INFO [train.py:904] (1/8) Epoch 2, batch 2450, loss[loss=0.2678, simple_loss=0.3459, pruned_loss=0.09485, over 17095.00 frames. ], tot_loss[loss=0.2748, simple_loss=0.3397, pruned_loss=0.105, over 3323288.20 frames. ], batch size: 49, lr: 2.98e-02, grad_scale: 8.0 2023-04-27 17:31:31,813 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=12604.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 17:31:51,020 INFO [optim.py:368] (1/8) 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 17:31:59,230 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=12624.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 17:32:20,346 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=12640.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 17:32:26,721 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([5.1482, 4.9702, 5.0166, 5.4386, 5.5586, 5.0742, 5.4754, 5.3511], device='cuda:1'), covar=tensor([0.0433, 0.0430, 0.1080, 0.0352, 0.0340, 0.0513, 0.0245, 0.0293], device='cuda:1'), in_proj_covar=tensor([0.0229, 0.0265, 0.0381, 0.0281, 0.0220, 0.0216, 0.0197, 0.0219], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-27 17:32:26,870 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.8497, 4.7107, 4.8098, 2.0872, 3.4494, 2.8961, 3.8602, 4.8551], device='cuda:1'), covar=tensor([0.0238, 0.0510, 0.0222, 0.1699, 0.0700, 0.0986, 0.0685, 0.0472], device='cuda:1'), in_proj_covar=tensor([0.0133, 0.0102, 0.0139, 0.0156, 0.0144, 0.0138, 0.0144, 0.0095], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0002], device='cuda:1') 2023-04-27 17:32:35,554 INFO [train.py:904] (1/8) Epoch 2, batch 2500, loss[loss=0.2614, simple_loss=0.3468, pruned_loss=0.08806, over 17030.00 frames. ], tot_loss[loss=0.274, simple_loss=0.3389, pruned_loss=0.1045, over 3322582.62 frames. ], batch size: 50, lr: 2.97e-02, grad_scale: 8.0 2023-04-27 17:32:36,842 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=12652.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 17:32:41,946 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=4.72 vs. limit=5.0 2023-04-27 17:33:07,394 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.0968, 3.5120, 3.8642, 3.2939, 4.0379, 4.0514, 4.1115, 2.0720], device='cuda:1'), covar=tensor([0.0797, 0.0143, 0.0086, 0.0300, 0.0042, 0.0058, 0.0060, 0.0651], device='cuda:1'), in_proj_covar=tensor([0.0116, 0.0049, 0.0058, 0.0102, 0.0051, 0.0058, 0.0062, 0.0103], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0003], device='cuda:1') 2023-04-27 17:33:26,473 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=12688.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 17:33:32,704 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=5.42 vs. limit=5.0 2023-04-27 17:33:43,145 INFO [train.py:904] (1/8) Epoch 2, batch 2550, loss[loss=0.2389, simple_loss=0.304, pruned_loss=0.08692, over 16753.00 frames. ], tot_loss[loss=0.2737, simple_loss=0.3382, pruned_loss=0.1047, over 3323298.69 frames. ], batch size: 39, lr: 2.97e-02, grad_scale: 8.0 2023-04-27 17:34:08,154 INFO [optim.py:368] (1/8) 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 17:34:27,593 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 2, batch 2600, loss[loss=0.245, simple_loss=0.3081, pruned_loss=0.091, over 16691.00 frames. ], tot_loss[loss=0.2729, simple_loss=0.3378, pruned_loss=0.104, over 3314735.06 frames. ], batch size: 89, lr: 2.96e-02, grad_scale: 8.0 2023-04-27 17:34:56,607 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.3417, 4.2216, 4.3357, 3.7335, 4.3391, 4.3005, 4.5697, 2.3044], device='cuda:1'), covar=tensor([0.0753, 0.0056, 0.0062, 0.0276, 0.0050, 0.0099, 0.0053, 0.0697], device='cuda:1'), in_proj_covar=tensor([0.0114, 0.0048, 0.0058, 0.0101, 0.0052, 0.0059, 0.0061, 0.0103], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0003], device='cuda:1') 2023-04-27 17:34:58,209 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=4.89 vs. limit=5.0 2023-04-27 17:35:01,657 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=3.20 vs. limit=5.0 2023-04-27 17:35:08,908 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=12763.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 17:35:32,888 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=12780.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 17:36:01,535 INFO [train.py:904] (1/8) Epoch 2, batch 2650, loss[loss=0.276, simple_loss=0.3298, pruned_loss=0.1111, over 15587.00 frames. ], tot_loss[loss=0.2747, simple_loss=0.3398, pruned_loss=0.1048, over 3317689.85 frames. ], batch size: 191, lr: 2.96e-02, grad_scale: 8.0 2023-04-27 17:36:03,297 INFO [zipformer.py:625] (1/8) 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] (1/8) 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] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=12824.0, num_to_drop=1, layers_to_drop={1} 2023-04-27 17:37:09,358 INFO [train.py:904] (1/8) Epoch 2, batch 2700, loss[loss=0.3076, simple_loss=0.3553, pruned_loss=0.1299, over 16243.00 frames. ], tot_loss[loss=0.2725, simple_loss=0.3387, pruned_loss=0.1031, over 3329659.66 frames. ], batch size: 165, lr: 2.95e-02, grad_scale: 8.0 2023-04-27 17:37:25,838 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.9207, 4.5920, 4.5716, 3.6845, 4.6124, 1.9605, 4.4158, 4.7023], device='cuda:1'), covar=tensor([0.0070, 0.0088, 0.0093, 0.0425, 0.0069, 0.1356, 0.0081, 0.0127], device='cuda:1'), in_proj_covar=tensor([0.0064, 0.0056, 0.0080, 0.0102, 0.0060, 0.0102, 0.0073, 0.0080], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-04-27 17:38:19,507 INFO [train.py:904] (1/8) Epoch 2, batch 2750, loss[loss=0.2645, simple_loss=0.3326, pruned_loss=0.09816, over 16835.00 frames. ], tot_loss[loss=0.2699, simple_loss=0.3371, pruned_loss=0.1013, over 3333205.06 frames. ], batch size: 90, lr: 2.95e-02, grad_scale: 8.0 2023-04-27 17:38:22,784 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.8983, 4.5817, 4.6362, 4.7996, 4.1776, 4.7267, 4.5989, 4.3588], device='cuda:1'), covar=tensor([0.0233, 0.0151, 0.0175, 0.0104, 0.0757, 0.0157, 0.0241, 0.0223], device='cuda:1'), in_proj_covar=tensor([0.0128, 0.0090, 0.0166, 0.0132, 0.0202, 0.0136, 0.0118, 0.0143], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0004, 0.0003, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-04-27 17:38:41,652 INFO [optim.py:368] (1/8) 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 17:38:57,798 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=2.83 vs. limit=5.0 2023-04-27 17:39:26,263 INFO [train.py:904] (1/8) Epoch 2, batch 2800, loss[loss=0.2341, simple_loss=0.3095, pruned_loss=0.07929, over 17212.00 frames. ], tot_loss[loss=0.2706, simple_loss=0.3374, pruned_loss=0.102, over 3332059.30 frames. ], batch size: 44, lr: 2.94e-02, grad_scale: 8.0 2023-04-27 17:40:33,622 INFO [train.py:904] (1/8) Epoch 2, batch 2850, loss[loss=0.2839, simple_loss=0.3549, pruned_loss=0.1064, over 17058.00 frames. ], tot_loss[loss=0.2697, simple_loss=0.3361, pruned_loss=0.1016, over 3333861.84 frames. ], batch size: 50, lr: 2.94e-02, grad_scale: 8.0 2023-04-27 17:40:57,333 INFO [optim.py:368] (1/8) 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 17:41:13,267 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([1.9383, 2.7068, 2.7095, 1.7296, 2.8089, 2.8288, 2.4341, 2.5564], device='cuda:1'), covar=tensor([0.0667, 0.0135, 0.0183, 0.1125, 0.0111, 0.0077, 0.0273, 0.0264], device='cuda:1'), in_proj_covar=tensor([0.0126, 0.0079, 0.0078, 0.0148, 0.0071, 0.0068, 0.0095, 0.0107], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0001, 0.0002, 0.0002], device='cuda:1') 2023-04-27 17:41:18,130 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.9836, 3.6914, 3.5268, 1.6583, 2.6432, 2.0362, 3.5100, 3.7748], device='cuda:1'), covar=tensor([0.0282, 0.0390, 0.0330, 0.1693, 0.0755, 0.1109, 0.0576, 0.0336], device='cuda:1'), in_proj_covar=tensor([0.0135, 0.0108, 0.0141, 0.0154, 0.0146, 0.0139, 0.0150, 0.0102], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:1') 2023-04-27 17:41:21,504 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.8199, 4.0313, 3.6652, 3.9100, 3.4604, 3.6568, 3.7605, 3.9555], device='cuda:1'), covar=tensor([0.0665, 0.0843, 0.1111, 0.0470, 0.0712, 0.0798, 0.0586, 0.0845], device='cuda:1'), in_proj_covar=tensor([0.0222, 0.0307, 0.0271, 0.0189, 0.0211, 0.0177, 0.0243, 0.0210], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-27 17:41:41,132 INFO [train.py:904] (1/8) Epoch 2, batch 2900, loss[loss=0.2358, simple_loss=0.3046, pruned_loss=0.08349, over 16996.00 frames. ], tot_loss[loss=0.2699, simple_loss=0.3352, pruned_loss=0.1023, over 3331037.88 frames. ], batch size: 41, lr: 2.93e-02, grad_scale: 8.0 2023-04-27 17:41:59,085 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.92 vs. limit=2.0 2023-04-27 17:42:49,006 INFO [train.py:904] (1/8) Epoch 2, batch 2950, loss[loss=0.2346, simple_loss=0.3132, pruned_loss=0.07803, over 17077.00 frames. ], tot_loss[loss=0.269, simple_loss=0.3342, pruned_loss=0.1019, over 3336010.99 frames. ], batch size: 53, lr: 2.93e-02, grad_scale: 8.0 2023-04-27 17:42:50,385 INFO [zipformer.py:625] (1/8) 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] (1/8) 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 17:43:14,206 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=13119.0, num_to_drop=1, layers_to_drop={3} 2023-04-27 17:43:53,664 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=13150.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 17:43:54,570 INFO [train.py:904] (1/8) Epoch 2, batch 3000, loss[loss=0.2228, simple_loss=0.2932, pruned_loss=0.07619, over 16891.00 frames. ], tot_loss[loss=0.2702, simple_loss=0.3348, pruned_loss=0.1028, over 3328336.79 frames. ], batch size: 42, lr: 2.92e-02, grad_scale: 8.0 2023-04-27 17:43:54,570 INFO [train.py:929] (1/8) Computing validation loss 2023-04-27 17:44:03,914 INFO [train.py:938] (1/8) Epoch 2, validation: loss=0.1858, simple_loss=0.2917, pruned_loss=0.04, over 944034.00 frames. 2023-04-27 17:44:03,915 INFO [train.py:939] (1/8) Maximum memory allocated so far is 17407MB 2023-04-27 17:45:05,141 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([5.4078, 5.2149, 5.1627, 5.2360, 4.6980, 5.1587, 5.1637, 4.8015], device='cuda:1'), covar=tensor([0.0235, 0.0109, 0.0160, 0.0099, 0.0670, 0.0153, 0.0111, 0.0190], device='cuda:1'), in_proj_covar=tensor([0.0131, 0.0094, 0.0168, 0.0135, 0.0206, 0.0139, 0.0118, 0.0148], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0004, 0.0003, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-04-27 17:45:09,406 INFO [train.py:904] (1/8) Epoch 2, batch 3050, loss[loss=0.2975, simple_loss=0.3501, pruned_loss=0.1224, over 15613.00 frames. ], tot_loss[loss=0.2698, simple_loss=0.3347, pruned_loss=0.1025, over 3333067.75 frames. ], batch size: 191, lr: 2.92e-02, grad_scale: 8.0 2023-04-27 17:45:33,139 INFO [optim.py:368] (1/8) 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 17:46:15,658 INFO [train.py:904] (1/8) Epoch 2, batch 3100, loss[loss=0.2783, simple_loss=0.3276, pruned_loss=0.1145, over 16534.00 frames. ], tot_loss[loss=0.2703, simple_loss=0.3347, pruned_loss=0.1029, over 3331619.82 frames. ], batch size: 68, lr: 2.91e-02, grad_scale: 8.0 2023-04-27 17:46:51,914 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.0414, 3.8492, 4.0809, 4.4640, 4.4576, 4.1403, 4.1402, 4.3211], device='cuda:1'), covar=tensor([0.0564, 0.0601, 0.1093, 0.0296, 0.0423, 0.0630, 0.0789, 0.0376], device='cuda:1'), in_proj_covar=tensor([0.0243, 0.0286, 0.0408, 0.0293, 0.0229, 0.0225, 0.0216, 0.0232], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0004, 0.0003, 0.0003, 0.0002, 0.0002, 0.0003], device='cuda:1') 2023-04-27 17:47:22,151 INFO [train.py:904] (1/8) Epoch 2, batch 3150, loss[loss=0.2651, simple_loss=0.3409, pruned_loss=0.09465, over 17087.00 frames. ], tot_loss[loss=0.2692, simple_loss=0.3332, pruned_loss=0.1026, over 3327200.64 frames. ], batch size: 49, lr: 2.91e-02, grad_scale: 8.0 2023-04-27 17:47:27,112 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.7366, 4.4829, 4.7366, 5.0634, 5.1130, 4.5609, 5.1393, 5.0432], device='cuda:1'), covar=tensor([0.0470, 0.0499, 0.1014, 0.0333, 0.0315, 0.0486, 0.0258, 0.0296], device='cuda:1'), in_proj_covar=tensor([0.0238, 0.0282, 0.0403, 0.0292, 0.0228, 0.0223, 0.0215, 0.0230], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0004, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:1') 2023-04-27 17:47:44,587 INFO [optim.py:368] (1/8) 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 17:48:02,024 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=4.52 vs. limit=5.0 2023-04-27 17:48:06,906 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=4.66 vs. limit=5.0 2023-04-27 17:48:28,283 INFO [train.py:904] (1/8) Epoch 2, batch 3200, loss[loss=0.282, simple_loss=0.3331, pruned_loss=0.1154, over 16860.00 frames. ], tot_loss[loss=0.2673, simple_loss=0.3313, pruned_loss=0.1016, over 3329480.26 frames. ], batch size: 116, lr: 2.90e-02, grad_scale: 8.0 2023-04-27 17:49:34,225 INFO [train.py:904] (1/8) Epoch 2, batch 3250, loss[loss=0.2677, simple_loss=0.321, pruned_loss=0.1072, over 16854.00 frames. ], tot_loss[loss=0.2676, simple_loss=0.3317, pruned_loss=0.1017, over 3337001.30 frames. ], batch size: 96, lr: 2.90e-02, grad_scale: 8.0 2023-04-27 17:49:50,624 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([1.7482, 3.0499, 3.4024, 2.7066, 3.3937, 3.4426, 3.7724, 1.9229], device='cuda:1'), covar=tensor([0.0777, 0.0196, 0.0083, 0.0341, 0.0081, 0.0083, 0.0047, 0.0585], device='cuda:1'), in_proj_covar=tensor([0.0114, 0.0048, 0.0057, 0.0100, 0.0052, 0.0057, 0.0059, 0.0101], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0001, 0.0002, 0.0003, 0.0001, 0.0002, 0.0002, 0.0003], device='cuda:1') 2023-04-27 17:49:58,157 INFO [optim.py:368] (1/8) 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 17:49:58,550 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=13419.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 17:50:42,665 INFO [train.py:904] (1/8) Epoch 2, batch 3300, loss[loss=0.2668, simple_loss=0.3421, pruned_loss=0.09572, over 17131.00 frames. ], tot_loss[loss=0.2681, simple_loss=0.3327, pruned_loss=0.1018, over 3331314.29 frames. ], batch size: 47, lr: 2.89e-02, grad_scale: 8.0 2023-04-27 17:51:02,650 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=13467.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 17:51:04,472 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.6359, 5.0998, 5.1159, 5.0796, 5.0316, 5.5861, 5.3227, 5.0409], device='cuda:1'), covar=tensor([0.0866, 0.1086, 0.1048, 0.1374, 0.2335, 0.0768, 0.0748, 0.1562], device='cuda:1'), in_proj_covar=tensor([0.0196, 0.0282, 0.0256, 0.0242, 0.0327, 0.0262, 0.0215, 0.0321], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003], device='cuda:1') 2023-04-27 17:51:48,084 INFO [train.py:904] (1/8) Epoch 2, batch 3350, loss[loss=0.2494, simple_loss=0.3366, pruned_loss=0.08111, over 17055.00 frames. ], tot_loss[loss=0.268, simple_loss=0.3331, pruned_loss=0.1015, over 3331583.44 frames. ], batch size: 55, lr: 2.89e-02, grad_scale: 8.0 2023-04-27 17:52:13,283 INFO [optim.py:368] (1/8) 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 17:52:56,206 INFO [train.py:904] (1/8) Epoch 2, batch 3400, loss[loss=0.2383, simple_loss=0.3027, pruned_loss=0.08698, over 16989.00 frames. ], tot_loss[loss=0.2665, simple_loss=0.3323, pruned_loss=0.1003, over 3331474.46 frames. ], batch size: 41, lr: 2.88e-02, grad_scale: 8.0 2023-04-27 17:54:05,266 INFO [train.py:904] (1/8) Epoch 2, batch 3450, loss[loss=0.2462, simple_loss=0.3152, pruned_loss=0.08865, over 17013.00 frames. ], tot_loss[loss=0.2661, simple_loss=0.3318, pruned_loss=0.1002, over 3334601.83 frames. ], batch size: 41, lr: 2.88e-02, grad_scale: 8.0 2023-04-27 17:54:29,795 INFO [optim.py:368] (1/8) 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 17:55:00,448 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.62 vs. limit=2.0 2023-04-27 17:55:11,512 INFO [train.py:904] (1/8) Epoch 2, batch 3500, loss[loss=0.3165, simple_loss=0.3622, pruned_loss=0.1354, over 11549.00 frames. ], tot_loss[loss=0.2654, simple_loss=0.3303, pruned_loss=0.1003, over 3321858.83 frames. ], batch size: 246, lr: 2.88e-02, grad_scale: 8.0 2023-04-27 17:55:18,347 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=5.75 vs. limit=5.0 2023-04-27 17:56:04,072 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.58 vs. limit=2.0 2023-04-27 17:56:21,453 INFO [train.py:904] (1/8) Epoch 2, batch 3550, loss[loss=0.2516, simple_loss=0.3399, pruned_loss=0.08171, over 17038.00 frames. ], tot_loss[loss=0.2636, simple_loss=0.3292, pruned_loss=0.099, over 3321723.02 frames. ], batch size: 53, lr: 2.87e-02, grad_scale: 8.0 2023-04-27 17:56:27,060 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=4.97 vs. limit=5.0 2023-04-27 17:56:45,025 INFO [optim.py:368] (1/8) 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 17:56:45,448 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.4445, 3.9378, 3.8622, 1.6491, 4.0003, 4.0261, 3.4430, 3.4174], device='cuda:1'), covar=tensor([0.0626, 0.0066, 0.0143, 0.1505, 0.0064, 0.0041, 0.0209, 0.0257], device='cuda:1'), in_proj_covar=tensor([0.0126, 0.0081, 0.0077, 0.0145, 0.0073, 0.0067, 0.0095, 0.0108], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0001, 0.0002, 0.0002], device='cuda:1') 2023-04-27 17:57:10,827 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=13737.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 17:57:29,797 INFO [train.py:904] (1/8) Epoch 2, batch 3600, loss[loss=0.3002, simple_loss=0.3428, pruned_loss=0.1288, over 11865.00 frames. ], tot_loss[loss=0.2614, simple_loss=0.3272, pruned_loss=0.09776, over 3317083.72 frames. ], batch size: 247, lr: 2.87e-02, grad_scale: 8.0 2023-04-27 17:57:39,512 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=4.72 vs. limit=5.0 2023-04-27 17:58:37,312 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 2, batch 3650, loss[loss=0.2514, simple_loss=0.319, pruned_loss=0.09189, over 15441.00 frames. ], tot_loss[loss=0.2596, simple_loss=0.3248, pruned_loss=0.09721, over 3325508.72 frames. ], batch size: 191, lr: 2.86e-02, grad_scale: 8.0 2023-04-27 17:58:57,795 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.3077, 4.3637, 4.3751, 4.3598, 4.2232, 4.7845, 4.6384, 4.2234], device='cuda:1'), covar=tensor([0.1005, 0.1062, 0.1008, 0.1364, 0.2206, 0.0881, 0.0708, 0.1760], device='cuda:1'), in_proj_covar=tensor([0.0185, 0.0271, 0.0246, 0.0230, 0.0293, 0.0253, 0.0200, 0.0300], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003], device='cuda:1') 2023-04-27 17:59:08,122 INFO [optim.py:368] (1/8) 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 17:59:46,310 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.46 vs. limit=2.0 2023-04-27 17:59:54,876 INFO [train.py:904] (1/8) Epoch 2, batch 3700, loss[loss=0.2709, simple_loss=0.3222, pruned_loss=0.1098, over 16662.00 frames. ], tot_loss[loss=0.2611, simple_loss=0.3239, pruned_loss=0.09916, over 3284393.46 frames. ], batch size: 89, lr: 2.86e-02, grad_scale: 8.0 2023-04-27 18:00:21,431 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.5321, 4.1197, 3.9672, 1.8112, 4.1290, 4.1799, 3.3601, 3.2865], device='cuda:1'), covar=tensor([0.0726, 0.0076, 0.0180, 0.1517, 0.0081, 0.0051, 0.0314, 0.0308], device='cuda:1'), in_proj_covar=tensor([0.0130, 0.0081, 0.0077, 0.0147, 0.0075, 0.0069, 0.0100, 0.0111], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0001, 0.0002, 0.0002], device='cuda:1') 2023-04-27 18:00:23,963 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.7868, 2.9542, 2.6243, 4.1116, 2.1509, 3.8824, 2.5148, 2.4037], device='cuda:1'), covar=tensor([0.0275, 0.0366, 0.0322, 0.0174, 0.1413, 0.0155, 0.0617, 0.1055], device='cuda:1'), in_proj_covar=tensor([0.0192, 0.0166, 0.0138, 0.0198, 0.0246, 0.0154, 0.0173, 0.0226], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003], device='cuda:1') 2023-04-27 18:01:04,349 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.7390, 1.7608, 2.3585, 2.5043, 2.8713, 2.4790, 1.8275, 2.4431], device='cuda:1'), covar=tensor([0.0043, 0.0297, 0.0152, 0.0154, 0.0038, 0.0146, 0.0214, 0.0098], device='cuda:1'), in_proj_covar=tensor([0.0071, 0.0112, 0.0093, 0.0084, 0.0057, 0.0060, 0.0093, 0.0056], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0001], device='cuda:1') 2023-04-27 18:01:07,710 INFO [train.py:904] (1/8) Epoch 2, batch 3750, loss[loss=0.311, simple_loss=0.3623, pruned_loss=0.1298, over 11235.00 frames. ], tot_loss[loss=0.2632, simple_loss=0.3244, pruned_loss=0.101, over 3268994.57 frames. ], batch size: 246, lr: 2.85e-02, grad_scale: 8.0 2023-04-27 18:01:33,975 INFO [optim.py:368] (1/8) Clipping_scale=2.0, grad-norm quartiles 2.812e+02 4.240e+02 4.945e+02 6.119e+02 1.020e+03, threshold=9.889e+02, percent-clipped=4.0 2023-04-27 18:02:23,400 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.2285, 3.9902, 4.1276, 4.4986, 4.4391, 4.2081, 4.3221, 4.3997], device='cuda:1'), covar=tensor([0.0375, 0.0435, 0.0890, 0.0289, 0.0378, 0.0532, 0.0479, 0.0298], device='cuda:1'), in_proj_covar=tensor([0.0230, 0.0269, 0.0376, 0.0283, 0.0223, 0.0212, 0.0215, 0.0218], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-27 18:02:24,606 INFO [train.py:904] (1/8) Epoch 2, batch 3800, loss[loss=0.3089, simple_loss=0.3491, pruned_loss=0.1344, over 16905.00 frames. ], tot_loss[loss=0.2656, simple_loss=0.3256, pruned_loss=0.1028, over 3268480.59 frames. ], batch size: 116, lr: 2.85e-02, grad_scale: 8.0 2023-04-27 18:02:55,515 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.9717, 4.0725, 1.7466, 4.0734, 2.9250, 4.0458, 2.2213, 2.8454], device='cuda:1'), covar=tensor([0.0075, 0.0136, 0.1703, 0.0056, 0.0573, 0.0280, 0.1380, 0.0591], device='cuda:1'), in_proj_covar=tensor([0.0078, 0.0099, 0.0170, 0.0078, 0.0152, 0.0127, 0.0171, 0.0147], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:1') 2023-04-27 18:03:40,205 INFO [train.py:904] (1/8) Epoch 2, batch 3850, loss[loss=0.2284, simple_loss=0.2878, pruned_loss=0.08443, over 16858.00 frames. ], tot_loss[loss=0.2648, simple_loss=0.3244, pruned_loss=0.1026, over 3262228.45 frames. ], batch size: 102, lr: 2.84e-02, grad_scale: 8.0 2023-04-27 18:03:53,447 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.98 vs. limit=2.0 2023-04-27 18:04:07,044 INFO [optim.py:368] (1/8) 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 18:04:11,519 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.6095, 4.9114, 4.8476, 4.9465, 4.7641, 5.2988, 5.1142, 4.8055], device='cuda:1'), covar=tensor([0.0715, 0.0915, 0.0792, 0.1230, 0.2004, 0.0676, 0.0694, 0.1500], device='cuda:1'), in_proj_covar=tensor([0.0192, 0.0268, 0.0251, 0.0236, 0.0296, 0.0250, 0.0206, 0.0306], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003], device='cuda:1') 2023-04-27 18:04:53,516 INFO [train.py:904] (1/8) Epoch 2, batch 3900, loss[loss=0.2648, simple_loss=0.3344, pruned_loss=0.0976, over 16446.00 frames. ], tot_loss[loss=0.264, simple_loss=0.3234, pruned_loss=0.1022, over 3272240.09 frames. ], batch size: 35, lr: 2.84e-02, grad_scale: 8.0 2023-04-27 18:05:05,816 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([5.2636, 5.0016, 4.8880, 3.7320, 5.1222, 2.2836, 4.7875, 5.1164], device='cuda:1'), covar=tensor([0.0088, 0.0074, 0.0086, 0.0525, 0.0054, 0.1568, 0.0084, 0.0091], device='cuda:1'), in_proj_covar=tensor([0.0063, 0.0055, 0.0080, 0.0101, 0.0063, 0.0105, 0.0076, 0.0079], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-04-27 18:05:25,136 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=14072.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 18:05:48,073 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([5.5526, 5.8566, 5.5305, 5.8578, 5.0746, 4.8679, 5.4692, 5.9840], device='cuda:1'), covar=tensor([0.0440, 0.0589, 0.0692, 0.0290, 0.0536, 0.0347, 0.0392, 0.0447], device='cuda:1'), in_proj_covar=tensor([0.0228, 0.0304, 0.0268, 0.0194, 0.0213, 0.0182, 0.0246, 0.0215], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-27 18:05:51,623 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.0275, 3.2337, 3.4553, 3.4635, 3.4682, 3.1609, 3.2801, 3.3889], device='cuda:1'), covar=tensor([0.0315, 0.0357, 0.0404, 0.0453, 0.0379, 0.0337, 0.0619, 0.0305], device='cuda:1'), in_proj_covar=tensor([0.0171, 0.0162, 0.0183, 0.0176, 0.0211, 0.0174, 0.0267, 0.0162], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0002], device='cuda:1') 2023-04-27 18:05:55,289 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 2, batch 3950, loss[loss=0.2995, simple_loss=0.3447, pruned_loss=0.1271, over 16277.00 frames. ], tot_loss[loss=0.263, simple_loss=0.3224, pruned_loss=0.1018, over 3273947.18 frames. ], batch size: 165, lr: 2.83e-02, grad_scale: 8.0 2023-04-27 18:06:08,867 INFO [zipformer.py:625] (1/8) 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] (1/8) 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 18:06:34,053 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=14120.0, num_to_drop=1, layers_to_drop={0} 2023-04-27 18:06:52,485 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 2, batch 4000, loss[loss=0.2416, simple_loss=0.3042, pruned_loss=0.08949, over 16794.00 frames. ], tot_loss[loss=0.2632, simple_loss=0.3222, pruned_loss=0.1021, over 3271306.63 frames. ], batch size: 102, lr: 2.83e-02, grad_scale: 8.0 2023-04-27 18:07:37,742 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=14164.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 18:08:02,172 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=14181.0, num_to_drop=1, layers_to_drop={2} 2023-04-27 18:08:10,370 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.7069, 3.7494, 3.2129, 2.9244, 2.7636, 2.1817, 3.9837, 4.5382], device='cuda:1'), covar=tensor([0.1733, 0.0578, 0.0846, 0.0577, 0.1746, 0.1189, 0.0320, 0.0121], device='cuda:1'), in_proj_covar=tensor([0.0249, 0.0220, 0.0234, 0.0165, 0.0261, 0.0182, 0.0186, 0.0134], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-27 18:08:31,014 INFO [train.py:904] (1/8) Epoch 2, batch 4050, loss[loss=0.2911, simple_loss=0.3476, pruned_loss=0.1173, over 12064.00 frames. ], tot_loss[loss=0.2594, simple_loss=0.3204, pruned_loss=0.09919, over 3271602.16 frames. ], batch size: 246, lr: 2.83e-02, grad_scale: 8.0 2023-04-27 18:08:56,318 INFO [optim.py:368] (1/8) Clipping_scale=2.0, grad-norm quartiles 2.111e+02 3.521e+02 4.461e+02 5.736e+02 1.012e+03, threshold=8.922e+02, percent-clipped=2.0 2023-04-27 18:09:43,227 INFO [train.py:904] (1/8) Epoch 2, batch 4100, loss[loss=0.2933, simple_loss=0.3595, pruned_loss=0.1136, over 16648.00 frames. ], tot_loss[loss=0.257, simple_loss=0.32, pruned_loss=0.097, over 3268904.96 frames. ], batch size: 62, lr: 2.82e-02, grad_scale: 8.0 2023-04-27 18:10:25,866 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([1.9145, 3.0246, 3.0241, 1.5558, 3.3043, 3.2903, 2.8295, 2.6456], device='cuda:1'), covar=tensor([0.0902, 0.0128, 0.0271, 0.1507, 0.0079, 0.0061, 0.0270, 0.0368], device='cuda:1'), in_proj_covar=tensor([0.0132, 0.0080, 0.0077, 0.0152, 0.0073, 0.0072, 0.0102, 0.0115], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0001, 0.0002, 0.0002], device='cuda:1') 2023-04-27 18:10:59,418 INFO [train.py:904] (1/8) Epoch 2, batch 4150, loss[loss=0.2757, simple_loss=0.3529, pruned_loss=0.09923, over 16984.00 frames. ], tot_loss[loss=0.2655, simple_loss=0.3288, pruned_loss=0.1011, over 3249050.02 frames. ], batch size: 41, lr: 2.82e-02, grad_scale: 8.0 2023-04-27 18:11:25,024 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=2.04 vs. limit=2.0 2023-04-27 18:11:25,252 INFO [optim.py:368] (1/8) Clipping_scale=2.0, grad-norm quartiles 2.427e+02 3.836e+02 4.498e+02 5.328e+02 1.203e+03, threshold=8.995e+02, percent-clipped=3.0 2023-04-27 18:11:31,639 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.5145, 4.1912, 4.2659, 4.3888, 3.8694, 4.3010, 4.2614, 3.9550], device='cuda:1'), covar=tensor([0.0197, 0.0137, 0.0166, 0.0100, 0.0603, 0.0157, 0.0214, 0.0206], device='cuda:1'), in_proj_covar=tensor([0.0117, 0.0082, 0.0151, 0.0116, 0.0175, 0.0122, 0.0104, 0.0129], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-04-27 18:12:14,210 INFO [train.py:904] (1/8) Epoch 2, batch 4200, loss[loss=0.2832, simple_loss=0.3506, pruned_loss=0.1079, over 16790.00 frames. ], tot_loss[loss=0.2728, simple_loss=0.3372, pruned_loss=0.1042, over 3219351.68 frames. ], batch size: 39, lr: 2.81e-02, grad_scale: 8.0 2023-04-27 18:13:15,554 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=14393.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 18:13:26,254 INFO [train.py:904] (1/8) Epoch 2, batch 4250, loss[loss=0.2603, simple_loss=0.3456, pruned_loss=0.08749, over 16857.00 frames. ], tot_loss[loss=0.2767, simple_loss=0.3417, pruned_loss=0.1058, over 3181777.18 frames. ], batch size: 96, lr: 2.81e-02, grad_scale: 8.0 2023-04-27 18:13:44,118 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=14413.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 18:13:50,318 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.8455, 4.6893, 4.5351, 4.0893, 4.7537, 1.9119, 4.3357, 4.6318], device='cuda:1'), covar=tensor([0.0038, 0.0038, 0.0055, 0.0226, 0.0034, 0.1159, 0.0060, 0.0075], device='cuda:1'), in_proj_covar=tensor([0.0058, 0.0050, 0.0073, 0.0093, 0.0056, 0.0101, 0.0069, 0.0073], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002], device='cuda:1') 2023-04-27 18:13:52,832 INFO [optim.py:368] (1/8) 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 18:14:05,660 INFO [zipformer.py:625] (1/8) 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] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=14441.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 18:14:33,646 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.77 vs. limit=2.0 2023-04-27 18:14:38,545 INFO [train.py:904] (1/8) Epoch 2, batch 4300, loss[loss=0.2859, simple_loss=0.3599, pruned_loss=0.106, over 16714.00 frames. ], tot_loss[loss=0.275, simple_loss=0.3418, pruned_loss=0.1041, over 3168686.66 frames. ], batch size: 62, lr: 2.80e-02, grad_scale: 16.0 2023-04-27 18:14:49,276 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=14459.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 18:15:10,475 INFO [zipformer.py:625] (1/8) 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] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=14476.0, num_to_drop=1, layers_to_drop={2} 2023-04-27 18:15:29,945 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=14487.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 18:15:49,381 INFO [train.py:904] (1/8) Epoch 2, batch 4350, loss[loss=0.2778, simple_loss=0.3482, pruned_loss=0.1037, over 15544.00 frames. ], tot_loss[loss=0.2788, simple_loss=0.3462, pruned_loss=0.1057, over 3165351.98 frames. ], batch size: 191, lr: 2.80e-02, grad_scale: 16.0 2023-04-27 18:16:13,768 INFO [optim.py:368] (1/8) 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 18:16:56,253 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 2, batch 4400, loss[loss=0.2557, simple_loss=0.3327, pruned_loss=0.0894, over 16986.00 frames. ], tot_loss[loss=0.2796, simple_loss=0.3476, pruned_loss=0.1058, over 3187553.93 frames. ], batch size: 55, lr: 2.79e-02, grad_scale: 16.0 2023-04-27 18:17:42,934 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.9397, 2.3671, 2.1838, 3.1092, 2.1935, 2.9379, 2.3103, 2.0247], device='cuda:1'), covar=tensor([0.0236, 0.0367, 0.0283, 0.0196, 0.1071, 0.0196, 0.0498, 0.1099], device='cuda:1'), in_proj_covar=tensor([0.0196, 0.0172, 0.0145, 0.0202, 0.0256, 0.0158, 0.0179, 0.0239], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003], device='cuda:1') 2023-04-27 18:17:47,918 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.71 vs. limit=2.0 2023-04-27 18:18:09,487 INFO [train.py:904] (1/8) Epoch 2, batch 4450, loss[loss=0.2754, simple_loss=0.347, pruned_loss=0.1019, over 17069.00 frames. ], tot_loss[loss=0.2803, simple_loss=0.3498, pruned_loss=0.1054, over 3205859.56 frames. ], batch size: 41, lr: 2.79e-02, grad_scale: 16.0 2023-04-27 18:18:34,420 INFO [optim.py:368] (1/8) 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 18:18:47,984 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.6773, 2.4289, 1.8968, 2.2421, 3.0549, 2.7701, 3.8110, 3.5257], device='cuda:1'), covar=tensor([0.0011, 0.0132, 0.0164, 0.0143, 0.0064, 0.0114, 0.0015, 0.0038], device='cuda:1'), in_proj_covar=tensor([0.0042, 0.0092, 0.0091, 0.0093, 0.0082, 0.0091, 0.0053, 0.0064], device='cuda:1'), out_proj_covar=tensor([6.9965e-05, 1.4552e-04, 1.4111e-04, 1.5180e-04, 1.3630e-04, 1.4854e-04, 8.7550e-05, 1.0822e-04], device='cuda:1') 2023-04-27 18:19:05,061 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.63 vs. limit=2.0 2023-04-27 18:19:17,772 INFO [train.py:904] (1/8) Epoch 2, batch 4500, loss[loss=0.2989, simple_loss=0.3638, pruned_loss=0.117, over 15359.00 frames. ], tot_loss[loss=0.2777, simple_loss=0.3481, pruned_loss=0.1037, over 3214073.00 frames. ], batch size: 190, lr: 2.79e-02, grad_scale: 16.0 2023-04-27 18:19:35,077 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=4.99 vs. limit=5.0 2023-04-27 18:20:28,459 INFO [train.py:904] (1/8) Epoch 2, batch 4550, loss[loss=0.3065, simple_loss=0.3746, pruned_loss=0.1192, over 16729.00 frames. ], tot_loss[loss=0.2773, simple_loss=0.3479, pruned_loss=0.1033, over 3235815.02 frames. ], batch size: 124, lr: 2.78e-02, grad_scale: 16.0 2023-04-27 18:20:49,309 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([1.7298, 2.7181, 2.5649, 1.6150, 2.7222, 2.6626, 2.3391, 2.3852], device='cuda:1'), covar=tensor([0.0947, 0.0128, 0.0219, 0.1230, 0.0128, 0.0125, 0.0422, 0.0398], device='cuda:1'), in_proj_covar=tensor([0.0134, 0.0078, 0.0078, 0.0148, 0.0072, 0.0071, 0.0099, 0.0114], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0001, 0.0002, 0.0002], device='cuda:1') 2023-04-27 18:20:54,773 INFO [optim.py:368] (1/8) 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 18:21:08,320 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=14728.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 18:21:25,090 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=14740.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 18:21:43,175 INFO [train.py:904] (1/8) Epoch 2, batch 4600, loss[loss=0.2666, simple_loss=0.3498, pruned_loss=0.09169, over 16407.00 frames. ], tot_loss[loss=0.2765, simple_loss=0.3479, pruned_loss=0.1026, over 3234536.06 frames. ], batch size: 146, lr: 2.78e-02, grad_scale: 16.0 2023-04-27 18:21:53,345 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=14759.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 18:22:10,139 INFO [zipformer.py:625] (1/8) 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] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=14776.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 18:22:18,664 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=14776.0, num_to_drop=1, layers_to_drop={1} 2023-04-27 18:22:52,493 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.33 vs. limit=2.0 2023-04-27 18:22:54,478 INFO [train.py:904] (1/8) Epoch 2, batch 4650, loss[loss=0.2979, simple_loss=0.3646, pruned_loss=0.1157, over 15495.00 frames. ], tot_loss[loss=0.2763, simple_loss=0.3472, pruned_loss=0.1027, over 3214076.46 frames. ], batch size: 191, lr: 2.77e-02, grad_scale: 16.0 2023-04-27 18:22:55,049 INFO [zipformer.py:625] (1/8) 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] (1/8) 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] (1/8) 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 18:23:27,860 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=14824.0, num_to_drop=1, layers_to_drop={0} 2023-04-27 18:23:32,499 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.4383, 4.3132, 4.1964, 4.7276, 4.6859, 4.2370, 4.6742, 4.6479], device='cuda:1'), covar=tensor([0.0382, 0.0385, 0.1037, 0.0298, 0.0380, 0.0452, 0.0341, 0.0291], device='cuda:1'), in_proj_covar=tensor([0.0221, 0.0253, 0.0360, 0.0260, 0.0206, 0.0202, 0.0199, 0.0202], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-27 18:23:54,901 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=14843.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 18:24:00,661 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.1297, 4.0401, 4.2636, 3.0319, 4.0176, 3.9988, 4.3270, 1.8499], device='cuda:1'), covar=tensor([0.0578, 0.0028, 0.0035, 0.0282, 0.0036, 0.0073, 0.0026, 0.0578], device='cuda:1'), in_proj_covar=tensor([0.0116, 0.0050, 0.0057, 0.0106, 0.0053, 0.0057, 0.0057, 0.0104], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0001, 0.0002, 0.0003, 0.0001, 0.0002, 0.0002, 0.0003], device='cuda:1') 2023-04-27 18:24:06,750 INFO [train.py:904] (1/8) Epoch 2, batch 4700, loss[loss=0.2799, simple_loss=0.3414, pruned_loss=0.1092, over 16866.00 frames. ], tot_loss[loss=0.273, simple_loss=0.3438, pruned_loss=0.1011, over 3225246.68 frames. ], batch size: 109, lr: 2.77e-02, grad_scale: 16.0 2023-04-27 18:25:02,946 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.9571, 4.0044, 3.4253, 3.0478, 3.3660, 2.5361, 4.5505, 5.0718], device='cuda:1'), covar=tensor([0.1630, 0.0550, 0.0858, 0.0521, 0.1569, 0.1028, 0.0204, 0.0065], device='cuda:1'), in_proj_covar=tensor([0.0248, 0.0214, 0.0231, 0.0164, 0.0275, 0.0178, 0.0183, 0.0123], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-27 18:25:20,206 INFO [train.py:904] (1/8) Epoch 2, batch 4750, loss[loss=0.2564, simple_loss=0.3213, pruned_loss=0.0957, over 16205.00 frames. ], tot_loss[loss=0.2707, simple_loss=0.3411, pruned_loss=0.1002, over 3199927.33 frames. ], batch size: 35, lr: 2.77e-02, grad_scale: 16.0 2023-04-27 18:25:27,228 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.61 vs. limit=2.0 2023-04-27 18:25:45,828 INFO [optim.py:368] (1/8) 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 18:26:30,658 INFO [train.py:904] (1/8) Epoch 2, batch 4800, loss[loss=0.2764, simple_loss=0.3488, pruned_loss=0.102, over 16742.00 frames. ], tot_loss[loss=0.2664, simple_loss=0.3372, pruned_loss=0.09781, over 3204737.46 frames. ], batch size: 124, lr: 2.76e-02, grad_scale: 16.0 2023-04-27 18:27:10,738 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.2672, 4.0710, 3.8542, 4.5899, 4.6094, 4.2061, 4.5699, 4.5358], device='cuda:1'), covar=tensor([0.0436, 0.0515, 0.1505, 0.0484, 0.0486, 0.0490, 0.0462, 0.0420], device='cuda:1'), in_proj_covar=tensor([0.0221, 0.0255, 0.0358, 0.0264, 0.0210, 0.0198, 0.0197, 0.0202], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-27 18:27:12,368 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.92 vs. limit=2.0 2023-04-27 18:27:43,243 INFO [train.py:904] (1/8) Epoch 2, batch 4850, loss[loss=0.2273, simple_loss=0.3149, pruned_loss=0.06986, over 16720.00 frames. ], tot_loss[loss=0.2681, simple_loss=0.3392, pruned_loss=0.09849, over 3183260.55 frames. ], batch size: 89, lr: 2.76e-02, grad_scale: 16.0 2023-04-27 18:28:11,112 INFO [optim.py:368] (1/8) 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] (1/8) Epoch 2, batch 4900, loss[loss=0.2459, simple_loss=0.3292, pruned_loss=0.08131, over 16847.00 frames. ], tot_loss[loss=0.2668, simple_loss=0.3386, pruned_loss=0.09754, over 3171131.94 frames. ], batch size: 116, lr: 2.75e-02, grad_scale: 16.0 2023-04-27 18:29:22,443 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.2898, 4.1953, 1.6065, 4.1832, 2.6631, 4.0411, 2.0069, 2.8600], device='cuda:1'), covar=tensor([0.0018, 0.0084, 0.1658, 0.0032, 0.0704, 0.0149, 0.1270, 0.0533], device='cuda:1'), in_proj_covar=tensor([0.0071, 0.0096, 0.0164, 0.0076, 0.0154, 0.0119, 0.0170, 0.0146], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:1') 2023-04-27 18:29:24,596 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=15069.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 18:30:04,991 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=15096.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 18:30:10,701 INFO [train.py:904] (1/8) Epoch 2, batch 4950, loss[loss=0.2668, simple_loss=0.3425, pruned_loss=0.09557, over 16516.00 frames. ], tot_loss[loss=0.2667, simple_loss=0.3383, pruned_loss=0.09753, over 3181941.18 frames. ], batch size: 75, lr: 2.75e-02, grad_scale: 16.0 2023-04-27 18:30:34,534 INFO [zipformer.py:625] (1/8) 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] (1/8) 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 18:31:12,817 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 2, batch 5000, loss[loss=0.2815, simple_loss=0.354, pruned_loss=0.1045, over 16351.00 frames. ], tot_loss[loss=0.2681, simple_loss=0.3405, pruned_loss=0.09787, over 3180846.61 frames. ], batch size: 165, lr: 2.74e-02, grad_scale: 16.0 2023-04-27 18:32:21,458 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=15191.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 18:32:26,335 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=15195.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 18:32:34,519 INFO [train.py:904] (1/8) Epoch 2, batch 5050, loss[loss=0.2402, simple_loss=0.3166, pruned_loss=0.08191, over 17251.00 frames. ], tot_loss[loss=0.2667, simple_loss=0.3399, pruned_loss=0.09681, over 3190071.65 frames. ], batch size: 52, lr: 2.74e-02, grad_scale: 16.0 2023-04-27 18:33:00,585 INFO [optim.py:368] (1/8) 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 18:33:31,523 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.2614, 4.0289, 1.6969, 4.1579, 2.6583, 4.1007, 1.9106, 3.0075], device='cuda:1'), covar=tensor([0.0026, 0.0094, 0.1686, 0.0027, 0.0657, 0.0168, 0.1507, 0.0525], device='cuda:1'), in_proj_covar=tensor([0.0072, 0.0096, 0.0160, 0.0076, 0.0149, 0.0118, 0.0165, 0.0143], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:1') 2023-04-27 18:33:42,885 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.47 vs. limit=2.0 2023-04-27 18:33:45,520 INFO [train.py:904] (1/8) Epoch 2, batch 5100, loss[loss=0.2523, simple_loss=0.3266, pruned_loss=0.08902, over 16347.00 frames. ], tot_loss[loss=0.2646, simple_loss=0.338, pruned_loss=0.09558, over 3193059.85 frames. ], batch size: 146, lr: 2.74e-02, grad_scale: 16.0 2023-04-27 18:33:53,407 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=15256.0, num_to_drop=1, layers_to_drop={3} 2023-04-27 18:34:56,469 INFO [train.py:904] (1/8) Epoch 2, batch 5150, loss[loss=0.2468, simple_loss=0.3147, pruned_loss=0.08945, over 16776.00 frames. ], tot_loss[loss=0.2632, simple_loss=0.3376, pruned_loss=0.09441, over 3202864.63 frames. ], batch size: 39, lr: 2.73e-02, grad_scale: 16.0 2023-04-27 18:35:22,256 INFO [optim.py:368] (1/8) Clipping_scale=2.0, grad-norm quartiles 2.454e+02 3.560e+02 4.122e+02 5.032e+02 1.152e+03, threshold=8.244e+02, percent-clipped=1.0 2023-04-27 18:35:23,748 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.3847, 3.2552, 2.6519, 2.6786, 2.4918, 2.1170, 3.3471, 3.7108], device='cuda:1'), covar=tensor([0.1630, 0.0568, 0.0865, 0.0542, 0.1353, 0.1061, 0.0310, 0.0114], device='cuda:1'), in_proj_covar=tensor([0.0242, 0.0215, 0.0226, 0.0165, 0.0253, 0.0173, 0.0187, 0.0124], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-27 18:35:38,021 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([5.0246, 5.3722, 5.0356, 5.2064, 4.6825, 4.4677, 4.8172, 5.4117], device='cuda:1'), covar=tensor([0.0353, 0.0467, 0.0697, 0.0282, 0.0476, 0.0367, 0.0388, 0.0471], device='cuda:1'), in_proj_covar=tensor([0.0205, 0.0281, 0.0261, 0.0182, 0.0194, 0.0174, 0.0232, 0.0195], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-27 18:35:41,647 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.5587, 5.3649, 5.3712, 5.2790, 5.2341, 5.7936, 5.6304, 5.3452], device='cuda:1'), covar=tensor([0.0553, 0.1080, 0.0794, 0.1389, 0.2033, 0.0763, 0.0606, 0.1623], device='cuda:1'), in_proj_covar=tensor([0.0185, 0.0259, 0.0237, 0.0226, 0.0295, 0.0247, 0.0189, 0.0301], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:1') 2023-04-27 18:36:10,018 INFO [train.py:904] (1/8) Epoch 2, batch 5200, loss[loss=0.2965, simple_loss=0.3627, pruned_loss=0.1151, over 15452.00 frames. ], tot_loss[loss=0.2635, simple_loss=0.3375, pruned_loss=0.09477, over 3217488.44 frames. ], batch size: 191, lr: 2.73e-02, grad_scale: 16.0 2023-04-27 18:36:53,411 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.0424, 3.7988, 3.7560, 3.9151, 3.3456, 3.8171, 3.7654, 3.6425], device='cuda:1'), covar=tensor([0.0301, 0.0223, 0.0219, 0.0149, 0.0896, 0.0278, 0.0422, 0.0283], device='cuda:1'), in_proj_covar=tensor([0.0124, 0.0095, 0.0161, 0.0123, 0.0190, 0.0133, 0.0110, 0.0141], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-04-27 18:37:02,361 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.9867, 4.7440, 4.7824, 2.3447, 5.0245, 4.9533, 3.5159, 3.7498], device='cuda:1'), covar=tensor([0.0695, 0.0065, 0.0089, 0.1390, 0.0031, 0.0026, 0.0227, 0.0319], device='cuda:1'), in_proj_covar=tensor([0.0139, 0.0082, 0.0079, 0.0154, 0.0076, 0.0073, 0.0103, 0.0118], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-27 18:37:16,302 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 2, batch 5250, loss[loss=0.2251, simple_loss=0.312, pruned_loss=0.0691, over 16880.00 frames. ], tot_loss[loss=0.2613, simple_loss=0.3345, pruned_loss=0.09401, over 3205795.45 frames. ], batch size: 90, lr: 2.72e-02, grad_scale: 16.0 2023-04-27 18:37:48,867 INFO [optim.py:368] (1/8) 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 18:38:23,202 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 2, batch 5300, loss[loss=0.2021, simple_loss=0.2821, pruned_loss=0.06106, over 16950.00 frames. ], tot_loss[loss=0.2563, simple_loss=0.3294, pruned_loss=0.0916, over 3216844.94 frames. ], batch size: 96, lr: 2.72e-02, grad_scale: 16.0 2023-04-27 18:39:25,834 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.1925, 4.1552, 4.1517, 3.3094, 4.0723, 4.0639, 4.3042, 2.0878], device='cuda:1'), covar=tensor([0.0558, 0.0021, 0.0032, 0.0231, 0.0046, 0.0066, 0.0033, 0.0503], device='cuda:1'), in_proj_covar=tensor([0.0118, 0.0050, 0.0059, 0.0110, 0.0053, 0.0057, 0.0057, 0.0104], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0001, 0.0002, 0.0003, 0.0001, 0.0002, 0.0002, 0.0003], device='cuda:1') 2023-04-27 18:39:43,334 INFO [train.py:904] (1/8) Epoch 2, batch 5350, loss[loss=0.3025, simple_loss=0.3705, pruned_loss=0.1172, over 16165.00 frames. ], tot_loss[loss=0.2538, simple_loss=0.3269, pruned_loss=0.09036, over 3226115.49 frames. ], batch size: 165, lr: 2.72e-02, grad_scale: 16.0 2023-04-27 18:40:09,981 INFO [optim.py:368] (1/8) 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 18:40:38,835 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([5.2398, 5.1421, 5.0030, 5.1585, 3.9403, 5.0684, 5.1527, 4.6512], device='cuda:1'), covar=tensor([0.0329, 0.0203, 0.0264, 0.0158, 0.1440, 0.0216, 0.0125, 0.0225], device='cuda:1'), in_proj_covar=tensor([0.0126, 0.0095, 0.0159, 0.0124, 0.0190, 0.0131, 0.0109, 0.0140], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-04-27 18:40:42,326 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.40 vs. limit=2.0 2023-04-27 18:40:56,836 INFO [train.py:904] (1/8) Epoch 2, batch 5400, loss[loss=0.2592, simple_loss=0.3248, pruned_loss=0.0968, over 17120.00 frames. ], tot_loss[loss=0.2571, simple_loss=0.3304, pruned_loss=0.0919, over 3224157.25 frames. ], batch size: 47, lr: 2.71e-02, grad_scale: 16.0 2023-04-27 18:40:57,153 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=15551.0, num_to_drop=1, layers_to_drop={2} 2023-04-27 18:42:14,668 INFO [train.py:904] (1/8) Epoch 2, batch 5450, loss[loss=0.3424, simple_loss=0.3895, pruned_loss=0.1477, over 12128.00 frames. ], tot_loss[loss=0.2636, simple_loss=0.3356, pruned_loss=0.09578, over 3179588.33 frames. ], batch size: 247, lr: 2.71e-02, grad_scale: 16.0 2023-04-27 18:42:43,047 INFO [optim.py:368] (1/8) Clipping_scale=2.0, grad-norm quartiles 2.524e+02 4.170e+02 4.882e+02 6.060e+02 1.199e+03, threshold=9.765e+02, percent-clipped=3.0 2023-04-27 18:43:07,484 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.8434, 2.1451, 1.4744, 1.9596, 2.9208, 2.8321, 3.6789, 3.3660], device='cuda:1'), covar=tensor([0.0007, 0.0113, 0.0157, 0.0134, 0.0050, 0.0104, 0.0018, 0.0037], device='cuda:1'), in_proj_covar=tensor([0.0041, 0.0094, 0.0094, 0.0098, 0.0084, 0.0096, 0.0053, 0.0067], device='cuda:1'), out_proj_covar=tensor([6.0670e-05, 1.4808e-04, 1.4385e-04, 1.5720e-04, 1.3801e-04, 1.5775e-04, 8.4909e-05, 1.1365e-04], device='cuda:1') 2023-04-27 18:43:34,028 INFO [train.py:904] (1/8) Epoch 2, batch 5500, loss[loss=0.3285, simple_loss=0.3903, pruned_loss=0.1334, over 16745.00 frames. ], tot_loss[loss=0.2793, simple_loss=0.3476, pruned_loss=0.1055, over 3159244.51 frames. ], batch size: 124, lr: 2.70e-02, grad_scale: 16.0 2023-04-27 18:43:56,308 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.5212, 3.8434, 1.5262, 3.8850, 2.4547, 3.9494, 1.9402, 2.6609], device='cuda:1'), covar=tensor([0.0058, 0.0121, 0.1747, 0.0043, 0.0774, 0.0200, 0.1444, 0.0631], device='cuda:1'), in_proj_covar=tensor([0.0075, 0.0102, 0.0169, 0.0078, 0.0158, 0.0127, 0.0176, 0.0152], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0004, 0.0003], device='cuda:1') 2023-04-27 18:43:59,508 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.8102, 3.1302, 3.4121, 3.4626, 3.4350, 3.0845, 3.1807, 3.3132], device='cuda:1'), covar=tensor([0.0345, 0.0351, 0.0360, 0.0389, 0.0357, 0.0349, 0.0840, 0.0346], device='cuda:1'), in_proj_covar=tensor([0.0155, 0.0148, 0.0167, 0.0165, 0.0196, 0.0159, 0.0248, 0.0156], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0003, 0.0003, 0.0002, 0.0003, 0.0002], device='cuda:1') 2023-04-27 18:44:06,028 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.5521, 3.0497, 2.6744, 2.4145, 2.1922, 1.9993, 3.0187, 3.2436], device='cuda:1'), covar=tensor([0.1059, 0.0422, 0.0694, 0.0446, 0.1604, 0.1037, 0.0274, 0.0151], device='cuda:1'), in_proj_covar=tensor([0.0252, 0.0227, 0.0241, 0.0176, 0.0279, 0.0182, 0.0197, 0.0133], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0003, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-27 18:44:32,538 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.24 vs. limit=2.0 2023-04-27 18:44:50,087 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.6650, 2.3988, 1.8854, 2.0719, 2.8433, 2.7740, 3.4429, 3.2093], device='cuda:1'), covar=tensor([0.0008, 0.0137, 0.0160, 0.0154, 0.0068, 0.0121, 0.0031, 0.0057], device='cuda:1'), in_proj_covar=tensor([0.0041, 0.0094, 0.0094, 0.0098, 0.0084, 0.0094, 0.0053, 0.0067], device='cuda:1'), out_proj_covar=tensor([6.0134e-05, 1.4721e-04, 1.4418e-04, 1.5650e-04, 1.3850e-04, 1.5357e-04, 8.5761e-05, 1.1327e-04], device='cuda:1') 2023-04-27 18:44:50,704 INFO [train.py:904] (1/8) Epoch 2, batch 5550, loss[loss=0.318, simple_loss=0.3836, pruned_loss=0.1262, over 17009.00 frames. ], tot_loss[loss=0.2911, simple_loss=0.3565, pruned_loss=0.1129, over 3145141.02 frames. ], batch size: 55, lr: 2.70e-02, grad_scale: 16.0 2023-04-27 18:45:19,403 INFO [optim.py:368] (1/8) 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 18:45:58,658 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([1.7378, 2.6747, 2.5639, 1.9982, 2.4610, 2.4310, 2.5043, 1.6308], device='cuda:1'), covar=tensor([0.0427, 0.0052, 0.0084, 0.0249, 0.0066, 0.0105, 0.0066, 0.0438], device='cuda:1'), in_proj_covar=tensor([0.0119, 0.0053, 0.0058, 0.0109, 0.0053, 0.0058, 0.0059, 0.0106], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0001, 0.0002, 0.0003, 0.0001, 0.0002, 0.0002, 0.0003], device='cuda:1') 2023-04-27 18:46:11,049 INFO [train.py:904] (1/8) Epoch 2, batch 5600, loss[loss=0.3159, simple_loss=0.3805, pruned_loss=0.1257, over 16703.00 frames. ], tot_loss[loss=0.3033, simple_loss=0.3647, pruned_loss=0.1209, over 3090477.36 frames. ], batch size: 89, lr: 2.70e-02, grad_scale: 16.0 2023-04-27 18:46:47,112 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.2173, 3.0290, 2.8103, 2.0440, 2.6185, 2.1132, 2.8747, 3.1257], device='cuda:1'), covar=tensor([0.0281, 0.0331, 0.0369, 0.1237, 0.0578, 0.0888, 0.0564, 0.0261], device='cuda:1'), in_proj_covar=tensor([0.0133, 0.0108, 0.0149, 0.0153, 0.0147, 0.0137, 0.0151, 0.0096], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0002], device='cuda:1') 2023-04-27 18:47:34,289 INFO [train.py:904] (1/8) Epoch 2, batch 5650, loss[loss=0.3986, simple_loss=0.4369, pruned_loss=0.1801, over 15280.00 frames. ], tot_loss[loss=0.3128, simple_loss=0.3716, pruned_loss=0.127, over 3084128.70 frames. ], batch size: 190, lr: 2.69e-02, grad_scale: 16.0 2023-04-27 18:48:01,991 INFO [optim.py:368] (1/8) 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 18:48:21,414 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([1.8912, 3.6228, 2.0226, 4.9285, 4.7693, 4.2449, 1.7797, 3.0911], device='cuda:1'), covar=tensor([0.1462, 0.0386, 0.1518, 0.0047, 0.0127, 0.0208, 0.1221, 0.0629], device='cuda:1'), in_proj_covar=tensor([0.0149, 0.0119, 0.0163, 0.0065, 0.0101, 0.0114, 0.0150, 0.0147], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0003, 0.0001, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:1') 2023-04-27 18:48:53,334 INFO [train.py:904] (1/8) Epoch 2, batch 5700, loss[loss=0.3481, simple_loss=0.4114, pruned_loss=0.1424, over 16225.00 frames. ], tot_loss[loss=0.3139, simple_loss=0.3722, pruned_loss=0.1278, over 3098904.26 frames. ], batch size: 165, lr: 2.69e-02, grad_scale: 16.0 2023-04-27 18:48:53,758 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=15851.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 18:49:21,365 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.9221, 3.8464, 3.7085, 3.2928, 3.7916, 1.8139, 3.6635, 3.7118], device='cuda:1'), covar=tensor([0.0068, 0.0058, 0.0074, 0.0255, 0.0053, 0.1256, 0.0068, 0.0078], device='cuda:1'), in_proj_covar=tensor([0.0062, 0.0051, 0.0076, 0.0096, 0.0058, 0.0109, 0.0070, 0.0075], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0001, 0.0002, 0.0003, 0.0002, 0.0003, 0.0002, 0.0002], device='cuda:1') 2023-04-27 18:49:44,745 INFO [zipformer.py:625] (1/8) 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] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=15899.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 18:50:12,127 INFO [train.py:904] (1/8) Epoch 2, batch 5750, loss[loss=0.3217, simple_loss=0.3792, pruned_loss=0.1321, over 16807.00 frames. ], tot_loss[loss=0.3185, simple_loss=0.3762, pruned_loss=0.1304, over 3073679.41 frames. ], batch size: 116, lr: 2.69e-02, grad_scale: 8.0 2023-04-27 18:50:24,970 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.96 vs. limit=2.0 2023-04-27 18:50:42,011 INFO [optim.py:368] (1/8) 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 18:51:22,894 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 2, batch 5800, loss[loss=0.3197, simple_loss=0.3857, pruned_loss=0.1269, over 16421.00 frames. ], tot_loss[loss=0.3153, simple_loss=0.3749, pruned_loss=0.1279, over 3074843.53 frames. ], batch size: 68, lr: 2.68e-02, grad_scale: 8.0 2023-04-27 18:51:43,987 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.97 vs. limit=2.0 2023-04-27 18:52:30,733 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.71 vs. limit=2.0 2023-04-27 18:52:56,020 INFO [train.py:904] (1/8) Epoch 2, batch 5850, loss[loss=0.3204, simple_loss=0.3781, pruned_loss=0.1313, over 16650.00 frames. ], tot_loss[loss=0.3123, simple_loss=0.3728, pruned_loss=0.1259, over 3077178.17 frames. ], batch size: 134, lr: 2.68e-02, grad_scale: 8.0 2023-04-27 18:52:57,366 INFO [zipformer.py:625] (1/8) 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] (1/8) 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 18:54:18,253 INFO [train.py:904] (1/8) Epoch 2, batch 5900, loss[loss=0.28, simple_loss=0.3609, pruned_loss=0.09955, over 16455.00 frames. ], tot_loss[loss=0.3099, simple_loss=0.3714, pruned_loss=0.1242, over 3093558.21 frames. ], batch size: 68, lr: 2.67e-02, grad_scale: 8.0 2023-04-27 18:54:39,939 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=16062.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 18:55:18,022 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=16085.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 18:55:39,770 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=2.04 vs. limit=2.0 2023-04-27 18:55:42,172 INFO [train.py:904] (1/8) Epoch 2, batch 5950, loss[loss=0.3049, simple_loss=0.3756, pruned_loss=0.117, over 16769.00 frames. ], tot_loss[loss=0.3089, simple_loss=0.3725, pruned_loss=0.1226, over 3098008.95 frames. ], batch size: 83, lr: 2.67e-02, grad_scale: 8.0 2023-04-27 18:55:59,744 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.1494, 2.9239, 2.7928, 1.9315, 2.6282, 2.1300, 2.7527, 3.0678], device='cuda:1'), covar=tensor([0.0268, 0.0376, 0.0331, 0.1288, 0.0509, 0.0759, 0.0457, 0.0334], device='cuda:1'), in_proj_covar=tensor([0.0129, 0.0105, 0.0147, 0.0150, 0.0143, 0.0136, 0.0148, 0.0095], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0002], device='cuda:1') 2023-04-27 18:56:13,605 INFO [optim.py:368] (1/8) 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 18:56:56,124 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=16146.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 18:57:01,408 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=16149.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 18:57:04,557 INFO [train.py:904] (1/8) Epoch 2, batch 6000, loss[loss=0.3047, simple_loss=0.3636, pruned_loss=0.123, over 16688.00 frames. ], tot_loss[loss=0.3088, simple_loss=0.3719, pruned_loss=0.1228, over 3097005.34 frames. ], batch size: 124, lr: 2.67e-02, grad_scale: 8.0 2023-04-27 18:57:04,557 INFO [train.py:929] (1/8) Computing validation loss 2023-04-27 18:57:15,924 INFO [train.py:938] (1/8) Epoch 2, validation: loss=0.2302, simple_loss=0.3372, pruned_loss=0.06166, over 944034.00 frames. 2023-04-27 18:57:15,925 INFO [train.py:939] (1/8) Maximum memory allocated so far is 17676MB 2023-04-27 18:57:29,087 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=2.04 vs. limit=2.0 2023-04-27 18:58:34,798 INFO [train.py:904] (1/8) Epoch 2, batch 6050, loss[loss=0.3376, simple_loss=0.378, pruned_loss=0.1486, over 11713.00 frames. ], tot_loss[loss=0.3062, simple_loss=0.3696, pruned_loss=0.1214, over 3102750.69 frames. ], batch size: 246, lr: 2.66e-02, grad_scale: 8.0 2023-04-27 18:58:48,638 INFO [zipformer.py:625] (1/8) 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] (1/8) Clipping_scale=2.0, grad-norm quartiles 3.028e+02 4.577e+02 5.718e+02 7.421e+02 1.323e+03, threshold=1.144e+03, percent-clipped=1.0 2023-04-27 18:59:33,626 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=16239.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 18:59:49,979 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.1284, 3.1489, 2.3648, 4.5008, 2.2248, 4.4905, 2.5706, 2.8682], device='cuda:1'), covar=tensor([0.0262, 0.0378, 0.0367, 0.0158, 0.1409, 0.0131, 0.0602, 0.0996], device='cuda:1'), in_proj_covar=tensor([0.0206, 0.0177, 0.0150, 0.0203, 0.0257, 0.0163, 0.0180, 0.0238], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-27 18:59:52,049 INFO [train.py:904] (1/8) Epoch 2, batch 6100, loss[loss=0.3377, simple_loss=0.3872, pruned_loss=0.1441, over 15609.00 frames. ], tot_loss[loss=0.3048, simple_loss=0.3685, pruned_loss=0.1205, over 3085200.56 frames. ], batch size: 191, lr: 2.66e-02, grad_scale: 8.0 2023-04-27 19:01:01,304 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.43 vs. limit=2.0 2023-04-27 19:01:09,758 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.1768, 3.8018, 3.6196, 1.5691, 2.8313, 2.2870, 3.5842, 3.9886], device='cuda:1'), covar=tensor([0.0271, 0.0504, 0.0420, 0.2060, 0.0812, 0.1093, 0.0816, 0.0416], device='cuda:1'), in_proj_covar=tensor([0.0130, 0.0106, 0.0149, 0.0151, 0.0145, 0.0137, 0.0150, 0.0099], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:1') 2023-04-27 19:01:17,143 INFO [train.py:904] (1/8) Epoch 2, batch 6150, loss[loss=0.3079, simple_loss=0.3758, pruned_loss=0.12, over 16542.00 frames. ], tot_loss[loss=0.3019, simple_loss=0.3657, pruned_loss=0.119, over 3078190.59 frames. ], batch size: 146, lr: 2.66e-02, grad_scale: 8.0 2023-04-27 19:01:45,850 INFO [optim.py:368] (1/8) 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 19:02:11,163 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.2956, 1.3877, 1.8506, 2.0504, 2.1133, 2.1721, 1.3464, 2.0566], device='cuda:1'), covar=tensor([0.0039, 0.0226, 0.0104, 0.0091, 0.0038, 0.0054, 0.0142, 0.0036], device='cuda:1'), in_proj_covar=tensor([0.0068, 0.0106, 0.0092, 0.0079, 0.0058, 0.0056, 0.0091, 0.0053], device='cuda:1'), out_proj_covar=tensor([1.1960e-04, 1.8809e-04, 1.6882e-04, 1.4695e-04, 9.9749e-05, 1.0076e-04, 1.5615e-04, 9.1203e-05], device='cuda:1') 2023-04-27 19:02:34,532 INFO [train.py:904] (1/8) Epoch 2, batch 6200, loss[loss=0.3198, simple_loss=0.3665, pruned_loss=0.1365, over 11692.00 frames. ], tot_loss[loss=0.3006, simple_loss=0.3635, pruned_loss=0.1189, over 3069793.52 frames. ], batch size: 248, lr: 2.65e-02, grad_scale: 8.0 2023-04-27 19:02:45,547 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=16357.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 19:03:06,768 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=16371.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 19:03:32,960 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.54 vs. limit=2.0 2023-04-27 19:03:51,369 INFO [train.py:904] (1/8) Epoch 2, batch 6250, loss[loss=0.2608, simple_loss=0.3409, pruned_loss=0.09038, over 16686.00 frames. ], tot_loss[loss=0.3001, simple_loss=0.3632, pruned_loss=0.1185, over 3071459.23 frames. ], batch size: 76, lr: 2.65e-02, grad_scale: 8.0 2023-04-27 19:04:18,706 INFO [optim.py:368] (1/8) Clipping_scale=2.0, grad-norm quartiles 3.383e+02 4.905e+02 5.910e+02 7.685e+02 1.899e+03, threshold=1.182e+03, percent-clipped=6.0 2023-04-27 19:04:37,410 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=16432.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 19:04:51,245 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=16441.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 19:04:58,391 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.4642, 4.1535, 4.1685, 4.3896, 3.8331, 4.2068, 4.2370, 3.9286], device='cuda:1'), covar=tensor([0.0303, 0.0197, 0.0189, 0.0107, 0.0671, 0.0189, 0.0241, 0.0247], device='cuda:1'), in_proj_covar=tensor([0.0122, 0.0092, 0.0152, 0.0115, 0.0177, 0.0126, 0.0105, 0.0136], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-04-27 19:05:05,085 INFO [train.py:904] (1/8) Epoch 2, batch 6300, loss[loss=0.3387, simple_loss=0.3907, pruned_loss=0.1434, over 16984.00 frames. ], tot_loss[loss=0.2997, simple_loss=0.3634, pruned_loss=0.118, over 3095080.94 frames. ], batch size: 109, lr: 2.64e-02, grad_scale: 8.0 2023-04-27 19:05:10,849 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.2959, 3.9995, 1.4990, 4.0943, 2.6579, 4.1185, 2.0766, 2.7478], device='cuda:1'), covar=tensor([0.0047, 0.0154, 0.2035, 0.0040, 0.0818, 0.0281, 0.1517, 0.0814], device='cuda:1'), in_proj_covar=tensor([0.0074, 0.0104, 0.0172, 0.0076, 0.0158, 0.0135, 0.0177, 0.0153], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003, 0.0004, 0.0003], device='cuda:1') 2023-04-27 19:05:58,851 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.2053, 3.9114, 3.5326, 1.6237, 2.8940, 2.3725, 3.5734, 3.9464], device='cuda:1'), covar=tensor([0.0267, 0.0386, 0.0422, 0.1864, 0.0752, 0.1005, 0.0651, 0.0457], device='cuda:1'), in_proj_covar=tensor([0.0133, 0.0109, 0.0153, 0.0154, 0.0149, 0.0138, 0.0152, 0.0101], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:1') 2023-04-27 19:06:22,081 INFO [train.py:904] (1/8) Epoch 2, batch 6350, loss[loss=0.2744, simple_loss=0.3473, pruned_loss=0.1008, over 16875.00 frames. ], tot_loss[loss=0.3017, simple_loss=0.3647, pruned_loss=0.1193, over 3111732.08 frames. ], batch size: 102, lr: 2.64e-02, grad_scale: 8.0 2023-04-27 19:06:29,787 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=16505.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 19:06:52,065 INFO [optim.py:368] (1/8) 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 19:06:56,751 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.2555, 3.1403, 3.1467, 3.4144, 3.4014, 3.1890, 3.2980, 3.4126], device='cuda:1'), covar=tensor([0.0430, 0.0375, 0.0909, 0.0377, 0.0449, 0.0905, 0.0584, 0.0340], device='cuda:1'), in_proj_covar=tensor([0.0229, 0.0261, 0.0366, 0.0276, 0.0210, 0.0194, 0.0211, 0.0212], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-27 19:07:08,027 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.98 vs. limit=2.0 2023-04-27 19:07:21,376 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 2, batch 6400, loss[loss=0.2872, simple_loss=0.3557, pruned_loss=0.1094, over 16907.00 frames. ], tot_loss[loss=0.3027, simple_loss=0.3648, pruned_loss=0.1203, over 3112507.00 frames. ], batch size: 96, lr: 2.64e-02, grad_scale: 8.0 2023-04-27 19:08:34,558 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=16587.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 19:08:55,791 INFO [train.py:904] (1/8) Epoch 2, batch 6450, loss[loss=0.2709, simple_loss=0.3507, pruned_loss=0.09553, over 17230.00 frames. ], tot_loss[loss=0.3005, simple_loss=0.3635, pruned_loss=0.1187, over 3108300.67 frames. ], batch size: 43, lr: 2.63e-02, grad_scale: 8.0 2023-04-27 19:08:56,380 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.30 vs. limit=2.0 2023-04-27 19:09:25,707 INFO [optim.py:368] (1/8) Clipping_scale=2.0, grad-norm quartiles 3.052e+02 4.859e+02 5.740e+02 6.998e+02 1.216e+03, threshold=1.148e+03, percent-clipped=0.0 2023-04-27 19:09:33,343 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.2402, 3.3474, 3.1147, 3.2464, 2.7180, 3.2723, 3.1045, 2.9951], device='cuda:1'), covar=tensor([0.0442, 0.0206, 0.0265, 0.0205, 0.0918, 0.0218, 0.0789, 0.0285], device='cuda:1'), in_proj_covar=tensor([0.0120, 0.0092, 0.0148, 0.0115, 0.0177, 0.0123, 0.0105, 0.0138], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-04-27 19:10:13,794 INFO [train.py:904] (1/8) Epoch 2, batch 6500, loss[loss=0.2791, simple_loss=0.3424, pruned_loss=0.1079, over 17021.00 frames. ], tot_loss[loss=0.2979, simple_loss=0.361, pruned_loss=0.1173, over 3109364.66 frames. ], batch size: 50, lr: 2.63e-02, grad_scale: 8.0 2023-04-27 19:10:24,439 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=16657.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 19:10:31,655 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.89 vs. limit=2.0 2023-04-27 19:11:32,119 INFO [train.py:904] (1/8) Epoch 2, batch 6550, loss[loss=0.3152, simple_loss=0.389, pruned_loss=0.1207, over 16190.00 frames. ], tot_loss[loss=0.3015, simple_loss=0.365, pruned_loss=0.119, over 3099043.32 frames. ], batch size: 165, lr: 2.63e-02, grad_scale: 8.0 2023-04-27 19:11:37,688 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=16705.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 19:11:59,595 INFO [optim.py:368] (1/8) 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] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=16727.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 19:12:32,942 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=16741.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 19:12:47,617 INFO [train.py:904] (1/8) Epoch 2, batch 6600, loss[loss=0.2731, simple_loss=0.3431, pruned_loss=0.1016, over 16675.00 frames. ], tot_loss[loss=0.3058, simple_loss=0.3688, pruned_loss=0.1214, over 3080640.28 frames. ], batch size: 62, lr: 2.62e-02, grad_scale: 8.0 2023-04-27 19:13:46,766 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 2, batch 6650, loss[loss=0.361, simple_loss=0.3949, pruned_loss=0.1635, over 11465.00 frames. ], tot_loss[loss=0.3086, simple_loss=0.3701, pruned_loss=0.1235, over 3064917.33 frames. ], batch size: 248, lr: 2.62e-02, grad_scale: 8.0 2023-04-27 19:14:13,190 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.9055, 4.0411, 3.7916, 3.9374, 3.5588, 3.7625, 3.7662, 3.9840], device='cuda:1'), covar=tensor([0.0392, 0.0526, 0.0699, 0.0335, 0.0465, 0.0570, 0.0492, 0.0577], device='cuda:1'), in_proj_covar=tensor([0.0223, 0.0298, 0.0274, 0.0187, 0.0205, 0.0185, 0.0248, 0.0210], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-27 19:14:13,219 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=16805.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 19:14:35,762 INFO [optim.py:368] (1/8) 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 19:15:23,513 INFO [train.py:904] (1/8) Epoch 2, batch 6700, loss[loss=0.3772, simple_loss=0.4016, pruned_loss=0.1764, over 11547.00 frames. ], tot_loss[loss=0.3067, simple_loss=0.3677, pruned_loss=0.1228, over 3064979.66 frames. ], batch size: 246, lr: 2.61e-02, grad_scale: 8.0 2023-04-27 19:15:26,818 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=16853.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 19:15:50,015 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=16868.0, num_to_drop=1, layers_to_drop={0} 2023-04-27 19:16:38,379 INFO [train.py:904] (1/8) Epoch 2, batch 6750, loss[loss=0.2432, simple_loss=0.3243, pruned_loss=0.08104, over 16877.00 frames. ], tot_loss[loss=0.3061, simple_loss=0.3668, pruned_loss=0.1227, over 3068789.50 frames. ], batch size: 90, lr: 2.61e-02, grad_scale: 8.0 2023-04-27 19:16:46,580 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=16906.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 19:17:07,557 INFO [optim.py:368] (1/8) 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 19:17:22,316 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=16929.0, num_to_drop=1, layers_to_drop={0} 2023-04-27 19:17:53,311 INFO [train.py:904] (1/8) Epoch 2, batch 6800, loss[loss=0.3176, simple_loss=0.3815, pruned_loss=0.1268, over 16334.00 frames. ], tot_loss[loss=0.3039, simple_loss=0.3658, pruned_loss=0.1211, over 3085546.29 frames. ], batch size: 165, lr: 2.61e-02, grad_scale: 8.0 2023-04-27 19:17:58,960 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.1553, 1.6201, 2.2919, 2.8933, 3.0085, 2.9777, 1.6185, 3.0148], device='cuda:1'), covar=tensor([0.0030, 0.0232, 0.0114, 0.0068, 0.0027, 0.0052, 0.0178, 0.0035], device='cuda:1'), in_proj_covar=tensor([0.0068, 0.0107, 0.0092, 0.0081, 0.0059, 0.0057, 0.0092, 0.0052], device='cuda:1'), out_proj_covar=tensor([1.1884e-04, 1.8781e-04, 1.6734e-04, 1.4801e-04, 1.0106e-04, 1.0270e-04, 1.5771e-04, 9.0776e-05], device='cuda:1') 2023-04-27 19:18:18,720 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=16967.0, num_to_drop=1, layers_to_drop={2} 2023-04-27 19:19:10,217 INFO [train.py:904] (1/8) Epoch 2, batch 6850, loss[loss=0.292, simple_loss=0.3821, pruned_loss=0.1009, over 16759.00 frames. ], tot_loss[loss=0.3026, simple_loss=0.3656, pruned_loss=0.1198, over 3116598.61 frames. ], batch size: 124, lr: 2.60e-02, grad_scale: 8.0 2023-04-27 19:19:38,485 INFO [optim.py:368] (1/8) 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 19:19:49,700 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=17027.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 19:19:55,388 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([1.8611, 3.2843, 3.3781, 1.5156, 3.5031, 3.4627, 3.0846, 2.7076], device='cuda:1'), covar=tensor([0.0964, 0.0112, 0.0131, 0.1493, 0.0086, 0.0057, 0.0229, 0.0367], device='cuda:1'), in_proj_covar=tensor([0.0139, 0.0085, 0.0081, 0.0154, 0.0077, 0.0072, 0.0107, 0.0121], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-27 19:20:24,241 INFO [train.py:904] (1/8) Epoch 2, batch 6900, loss[loss=0.2984, simple_loss=0.3694, pruned_loss=0.1137, over 16707.00 frames. ], tot_loss[loss=0.3034, simple_loss=0.3679, pruned_loss=0.1194, over 3118745.59 frames. ], batch size: 134, lr: 2.60e-02, grad_scale: 8.0 2023-04-27 19:20:31,445 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.2862, 4.5103, 4.5343, 1.6736, 4.7815, 4.8134, 3.9706, 3.6203], device='cuda:1'), covar=tensor([0.1043, 0.0091, 0.0145, 0.1762, 0.0076, 0.0043, 0.0200, 0.0361], device='cuda:1'), in_proj_covar=tensor([0.0140, 0.0086, 0.0081, 0.0154, 0.0078, 0.0072, 0.0107, 0.0122], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:1') 2023-04-27 19:21:01,218 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=17075.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 19:21:40,737 INFO [train.py:904] (1/8) Epoch 2, batch 6950, loss[loss=0.293, simple_loss=0.3544, pruned_loss=0.1158, over 17224.00 frames. ], tot_loss[loss=0.3107, simple_loss=0.3724, pruned_loss=0.1245, over 3081936.91 frames. ], batch size: 44, lr: 2.60e-02, grad_scale: 8.0 2023-04-27 19:22:09,949 INFO [optim.py:368] (1/8) 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 19:22:27,110 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.9064, 3.7247, 3.9645, 4.2448, 4.2465, 3.8558, 4.2687, 4.1592], device='cuda:1'), covar=tensor([0.0532, 0.0503, 0.1005, 0.0360, 0.0383, 0.0617, 0.0344, 0.0336], device='cuda:1'), in_proj_covar=tensor([0.0242, 0.0278, 0.0382, 0.0280, 0.0224, 0.0201, 0.0221, 0.0225], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-27 19:22:54,958 INFO [train.py:904] (1/8) Epoch 2, batch 7000, loss[loss=0.3311, simple_loss=0.3751, pruned_loss=0.1435, over 11652.00 frames. ], tot_loss[loss=0.3095, simple_loss=0.3723, pruned_loss=0.1234, over 3083771.86 frames. ], batch size: 247, lr: 2.59e-02, grad_scale: 8.0 2023-04-27 19:23:53,020 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 2, batch 7050, loss[loss=0.3353, simple_loss=0.3881, pruned_loss=0.1412, over 16695.00 frames. ], tot_loss[loss=0.3103, simple_loss=0.3733, pruned_loss=0.1236, over 3081432.67 frames. ], batch size: 62, lr: 2.59e-02, grad_scale: 8.0 2023-04-27 19:24:37,723 INFO [optim.py:368] (1/8) 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 19:24:43,950 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=17224.0, num_to_drop=1, layers_to_drop={0} 2023-04-27 19:25:03,234 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.7039, 3.5798, 3.6880, 3.7426, 3.8044, 4.1627, 4.0048, 3.8182], device='cuda:1'), covar=tensor([0.1373, 0.1283, 0.1081, 0.1455, 0.1691, 0.0802, 0.0817, 0.1716], device='cuda:1'), in_proj_covar=tensor([0.0197, 0.0271, 0.0250, 0.0236, 0.0302, 0.0265, 0.0211, 0.0319], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:1') 2023-04-27 19:25:22,936 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=17250.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 19:25:23,662 INFO [train.py:904] (1/8) Epoch 2, batch 7100, loss[loss=0.2985, simple_loss=0.3618, pruned_loss=0.1176, over 16451.00 frames. ], tot_loss[loss=0.31, simple_loss=0.3722, pruned_loss=0.1239, over 3059458.44 frames. ], batch size: 146, lr: 2.59e-02, grad_scale: 8.0 2023-04-27 19:25:40,648 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=17262.0, num_to_drop=1, layers_to_drop={3} 2023-04-27 19:26:04,969 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.50 vs. limit=2.0 2023-04-27 19:26:27,084 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.86 vs. limit=2.0 2023-04-27 19:26:38,659 INFO [train.py:904] (1/8) Epoch 2, batch 7150, loss[loss=0.3076, simple_loss=0.3731, pruned_loss=0.121, over 16434.00 frames. ], tot_loss[loss=0.3078, simple_loss=0.3698, pruned_loss=0.1229, over 3063435.86 frames. ], batch size: 146, lr: 2.58e-02, grad_scale: 8.0 2023-04-27 19:27:07,391 INFO [optim.py:368] (1/8) 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 19:27:52,978 INFO [train.py:904] (1/8) Epoch 2, batch 7200, loss[loss=0.3485, simple_loss=0.3914, pruned_loss=0.1528, over 12240.00 frames. ], tot_loss[loss=0.3046, simple_loss=0.3671, pruned_loss=0.1211, over 3044613.94 frames. ], batch size: 247, lr: 2.58e-02, grad_scale: 8.0 2023-04-27 19:28:10,361 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=17362.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 19:28:44,464 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=17383.0, num_to_drop=1, layers_to_drop={0} 2023-04-27 19:29:13,591 INFO [train.py:904] (1/8) Epoch 2, batch 7250, loss[loss=0.244, simple_loss=0.3097, pruned_loss=0.0892, over 16622.00 frames. ], tot_loss[loss=0.2999, simple_loss=0.3634, pruned_loss=0.1182, over 3061775.91 frames. ], batch size: 62, lr: 2.58e-02, grad_scale: 8.0 2023-04-27 19:29:42,543 INFO [optim.py:368] (1/8) 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 19:29:47,349 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=17423.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 19:30:18,024 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.73 vs. limit=2.0 2023-04-27 19:30:19,164 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=17444.0, num_to_drop=1, layers_to_drop={0} 2023-04-27 19:30:29,421 INFO [train.py:904] (1/8) Epoch 2, batch 7300, loss[loss=0.3512, simple_loss=0.3867, pruned_loss=0.1578, over 11718.00 frames. ], tot_loss[loss=0.2978, simple_loss=0.3614, pruned_loss=0.1171, over 3053960.59 frames. ], batch size: 248, lr: 2.57e-02, grad_scale: 8.0 2023-04-27 19:30:43,321 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.8847, 2.3780, 2.1221, 3.0147, 2.0897, 2.8868, 2.2500, 1.9512], device='cuda:1'), covar=tensor([0.0328, 0.0451, 0.0322, 0.0229, 0.1294, 0.0243, 0.0671, 0.1181], device='cuda:1'), in_proj_covar=tensor([0.0215, 0.0186, 0.0156, 0.0214, 0.0267, 0.0169, 0.0187, 0.0243], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-27 19:31:46,484 INFO [train.py:904] (1/8) Epoch 2, batch 7350, loss[loss=0.2805, simple_loss=0.3504, pruned_loss=0.1054, over 16620.00 frames. ], tot_loss[loss=0.2966, simple_loss=0.3604, pruned_loss=0.1164, over 3054076.92 frames. ], batch size: 62, lr: 2.57e-02, grad_scale: 8.0 2023-04-27 19:32:16,421 INFO [optim.py:368] (1/8) 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 19:32:21,753 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=4.96 vs. limit=5.0 2023-04-27 19:32:22,406 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=17524.0, num_to_drop=1, layers_to_drop={0} 2023-04-27 19:32:56,947 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 2, batch 7400, loss[loss=0.2672, simple_loss=0.3441, pruned_loss=0.09509, over 16824.00 frames. ], tot_loss[loss=0.2997, simple_loss=0.3629, pruned_loss=0.1183, over 3050710.30 frames. ], batch size: 39, lr: 2.57e-02, grad_scale: 8.0 2023-04-27 19:33:25,374 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=17562.0, num_to_drop=1, layers_to_drop={1} 2023-04-27 19:33:39,625 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.2567, 3.1582, 1.6648, 3.2382, 2.3035, 3.2292, 1.7953, 2.5014], device='cuda:1'), covar=tensor([0.0052, 0.0246, 0.1359, 0.0050, 0.0736, 0.0326, 0.1339, 0.0530], device='cuda:1'), in_proj_covar=tensor([0.0073, 0.0110, 0.0173, 0.0073, 0.0160, 0.0137, 0.0178, 0.0152], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003, 0.0004, 0.0003], device='cuda:1') 2023-04-27 19:33:40,918 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=17572.0, num_to_drop=1, layers_to_drop={0} 2023-04-27 19:34:17,271 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.5263, 4.7431, 4.4239, 4.5261, 4.0852, 4.1598, 4.2062, 4.7883], device='cuda:1'), covar=tensor([0.0431, 0.0513, 0.0681, 0.0349, 0.0480, 0.0497, 0.0468, 0.0450], device='cuda:1'), in_proj_covar=tensor([0.0219, 0.0303, 0.0279, 0.0192, 0.0203, 0.0190, 0.0243, 0.0207], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-27 19:34:27,189 INFO [train.py:904] (1/8) Epoch 2, batch 7450, loss[loss=0.3299, simple_loss=0.3842, pruned_loss=0.1378, over 16635.00 frames. ], tot_loss[loss=0.3034, simple_loss=0.3653, pruned_loss=0.1208, over 3034636.48 frames. ], batch size: 62, lr: 2.56e-02, grad_scale: 8.0 2023-04-27 19:34:43,753 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=17610.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 19:34:55,319 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.8002, 3.7531, 1.4965, 3.8221, 2.4523, 3.8119, 1.8473, 2.6432], device='cuda:1'), covar=tensor([0.0041, 0.0153, 0.1678, 0.0049, 0.0782, 0.0287, 0.1404, 0.0665], device='cuda:1'), in_proj_covar=tensor([0.0074, 0.0110, 0.0173, 0.0073, 0.0160, 0.0138, 0.0175, 0.0151], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003, 0.0004, 0.0003], device='cuda:1') 2023-04-27 19:34:59,851 INFO [optim.py:368] (1/8) 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 19:35:49,111 INFO [train.py:904] (1/8) Epoch 2, batch 7500, loss[loss=0.312, simple_loss=0.3755, pruned_loss=0.1242, over 15386.00 frames. ], tot_loss[loss=0.3028, simple_loss=0.3657, pruned_loss=0.1199, over 3039226.54 frames. ], batch size: 191, lr: 2.56e-02, grad_scale: 8.0 2023-04-27 19:36:18,061 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.33 vs. limit=2.0 2023-04-27 19:36:25,676 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.85 vs. limit=2.0 2023-04-27 19:36:52,133 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.41 vs. limit=2.0 2023-04-27 19:37:05,013 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 2, batch 7550, loss[loss=0.3021, simple_loss=0.3684, pruned_loss=0.1179, over 15417.00 frames. ], tot_loss[loss=0.3016, simple_loss=0.3646, pruned_loss=0.1193, over 3048257.91 frames. ], batch size: 190, lr: 2.56e-02, grad_scale: 8.0 2023-04-27 19:37:06,279 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([1.6097, 3.0363, 3.1341, 2.1289, 3.0636, 3.0357, 3.1795, 1.7282], device='cuda:1'), covar=tensor([0.0487, 0.0032, 0.0055, 0.0269, 0.0040, 0.0069, 0.0024, 0.0432], device='cuda:1'), in_proj_covar=tensor([0.0117, 0.0055, 0.0060, 0.0111, 0.0054, 0.0061, 0.0058, 0.0107], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0001, 0.0002, 0.0003, 0.0001, 0.0002, 0.0002, 0.0003], device='cuda:1') 2023-04-27 19:37:32,528 INFO [zipformer.py:625] (1/8) 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] (1/8) 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 19:37:50,565 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=17730.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 19:38:04,540 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=17739.0, num_to_drop=1, layers_to_drop={3} 2023-04-27 19:38:22,209 INFO [train.py:904] (1/8) Epoch 2, batch 7600, loss[loss=0.2844, simple_loss=0.3532, pruned_loss=0.1078, over 17264.00 frames. ], tot_loss[loss=0.3002, simple_loss=0.3635, pruned_loss=0.1185, over 3078036.54 frames. ], batch size: 52, lr: 2.55e-02, grad_scale: 8.0 2023-04-27 19:38:35,931 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.3222, 4.5641, 4.2186, 4.3933, 3.9408, 4.0110, 4.1209, 4.5056], device='cuda:1'), covar=tensor([0.0432, 0.0565, 0.0761, 0.0321, 0.0514, 0.0621, 0.0457, 0.0608], device='cuda:1'), in_proj_covar=tensor([0.0219, 0.0305, 0.0277, 0.0193, 0.0203, 0.0191, 0.0244, 0.0205], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-27 19:38:37,825 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=17761.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 19:39:24,542 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=17791.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 19:39:35,787 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.75 vs. limit=2.0 2023-04-27 19:39:39,705 INFO [train.py:904] (1/8) Epoch 2, batch 7650, loss[loss=0.3813, simple_loss=0.4065, pruned_loss=0.178, over 11495.00 frames. ], tot_loss[loss=0.3029, simple_loss=0.3648, pruned_loss=0.1205, over 3057047.65 frames. ], batch size: 246, lr: 2.55e-02, grad_scale: 8.0 2023-04-27 19:40:10,727 INFO [optim.py:368] (1/8) 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 19:40:22,965 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.8066, 4.9642, 4.3605, 4.9535, 4.6286, 4.3543, 4.7083, 4.9933], device='cuda:1'), covar=tensor([0.0766, 0.0997, 0.1662, 0.0527, 0.0647, 0.0819, 0.0684, 0.0941], device='cuda:1'), in_proj_covar=tensor([0.0221, 0.0308, 0.0281, 0.0194, 0.0204, 0.0194, 0.0248, 0.0208], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-27 19:40:49,778 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=17845.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 19:40:58,860 INFO [train.py:904] (1/8) Epoch 2, batch 7700, loss[loss=0.2672, simple_loss=0.3497, pruned_loss=0.09235, over 16831.00 frames. ], tot_loss[loss=0.305, simple_loss=0.3661, pruned_loss=0.122, over 3054037.31 frames. ], batch size: 96, lr: 2.55e-02, grad_scale: 8.0 2023-04-27 19:41:20,331 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.89 vs. limit=2.0 2023-04-27 19:41:32,300 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.2679, 4.0873, 3.6944, 1.6332, 2.7604, 2.4282, 3.5988, 4.1273], device='cuda:1'), covar=tensor([0.0282, 0.0382, 0.0439, 0.1871, 0.0865, 0.1007, 0.0672, 0.0385], device='cuda:1'), in_proj_covar=tensor([0.0133, 0.0107, 0.0149, 0.0151, 0.0143, 0.0135, 0.0148, 0.0103], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:1') 2023-04-27 19:42:04,292 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=17893.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 19:42:16,832 INFO [train.py:904] (1/8) Epoch 2, batch 7750, loss[loss=0.3077, simple_loss=0.3715, pruned_loss=0.122, over 16781.00 frames. ], tot_loss[loss=0.3047, simple_loss=0.3661, pruned_loss=0.1217, over 3069122.60 frames. ], batch size: 39, lr: 2.54e-02, grad_scale: 16.0 2023-04-27 19:42:28,297 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([1.9908, 2.6848, 2.6519, 1.7328, 2.7914, 2.7784, 2.3550, 2.3936], device='cuda:1'), covar=tensor([0.0718, 0.0115, 0.0194, 0.1099, 0.0102, 0.0085, 0.0326, 0.0366], device='cuda:1'), in_proj_covar=tensor([0.0137, 0.0083, 0.0082, 0.0145, 0.0073, 0.0069, 0.0105, 0.0117], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-27 19:42:32,054 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([1.9969, 2.6217, 2.3609, 3.4149, 3.2571, 3.2821, 1.8895, 2.7046], device='cuda:1'), covar=tensor([0.1106, 0.0314, 0.0962, 0.0070, 0.0184, 0.0227, 0.0963, 0.0543], device='cuda:1'), in_proj_covar=tensor([0.0148, 0.0122, 0.0163, 0.0068, 0.0108, 0.0118, 0.0152, 0.0153], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:1') 2023-04-27 19:42:46,601 INFO [optim.py:368] (1/8) 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 19:43:31,900 INFO [train.py:904] (1/8) Epoch 2, batch 7800, loss[loss=0.3099, simple_loss=0.3828, pruned_loss=0.1186, over 16413.00 frames. ], tot_loss[loss=0.3045, simple_loss=0.3662, pruned_loss=0.1214, over 3073170.81 frames. ], batch size: 35, lr: 2.54e-02, grad_scale: 16.0 2023-04-27 19:43:47,078 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=17960.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 19:43:50,128 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([1.5762, 3.3836, 2.3340, 4.5687, 4.3993, 4.0677, 2.0530, 3.1779], device='cuda:1'), covar=tensor([0.2139, 0.0513, 0.1599, 0.0101, 0.0220, 0.0293, 0.1476, 0.0710], device='cuda:1'), in_proj_covar=tensor([0.0149, 0.0124, 0.0166, 0.0069, 0.0112, 0.0120, 0.0153, 0.0154], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:1') 2023-04-27 19:43:59,958 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.7492, 2.6582, 1.6634, 2.8012, 2.1368, 2.7454, 1.9166, 2.3743], device='cuda:1'), covar=tensor([0.0060, 0.0249, 0.1120, 0.0058, 0.0550, 0.0387, 0.1016, 0.0481], device='cuda:1'), in_proj_covar=tensor([0.0073, 0.0112, 0.0174, 0.0074, 0.0161, 0.0141, 0.0179, 0.0158], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003, 0.0004, 0.0003], device='cuda:1') 2023-04-27 19:44:40,918 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.93 vs. limit=2.0 2023-04-27 19:44:52,420 INFO [train.py:904] (1/8) Epoch 2, batch 7850, loss[loss=0.3053, simple_loss=0.3704, pruned_loss=0.1201, over 16315.00 frames. ], tot_loss[loss=0.3054, simple_loss=0.3677, pruned_loss=0.1216, over 3075759.08 frames. ], batch size: 146, lr: 2.54e-02, grad_scale: 16.0 2023-04-27 19:44:57,918 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.95 vs. limit=2.0 2023-04-27 19:45:09,551 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.3226, 2.6897, 2.2495, 3.7152, 1.8764, 3.6447, 2.4395, 2.1956], device='cuda:1'), covar=tensor([0.0336, 0.0465, 0.0394, 0.0222, 0.1499, 0.0184, 0.0669, 0.1241], device='cuda:1'), in_proj_covar=tensor([0.0219, 0.0188, 0.0159, 0.0220, 0.0264, 0.0169, 0.0190, 0.0247], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-27 19:45:18,017 INFO [zipformer.py:625] (1/8) 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] (1/8) 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 19:45:22,927 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=18021.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 19:45:48,797 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 2, batch 7900, loss[loss=0.3121, simple_loss=0.3784, pruned_loss=0.123, over 15404.00 frames. ], tot_loss[loss=0.3028, simple_loss=0.366, pruned_loss=0.1198, over 3098685.58 frames. ], batch size: 190, lr: 2.53e-02, grad_scale: 8.0 2023-04-27 19:46:13,958 INFO [zipformer.py:625] (1/8) 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] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=18066.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 19:47:03,301 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=18086.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 19:47:05,060 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=18087.0, num_to_drop=1, layers_to_drop={1} 2023-04-27 19:47:25,404 INFO [train.py:904] (1/8) Epoch 2, batch 7950, loss[loss=0.2796, simple_loss=0.3458, pruned_loss=0.1067, over 16559.00 frames. ], tot_loss[loss=0.3037, simple_loss=0.3663, pruned_loss=0.1206, over 3087281.69 frames. ], batch size: 68, lr: 2.53e-02, grad_scale: 8.0 2023-04-27 19:47:56,310 INFO [optim.py:368] (1/8) 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 19:48:08,268 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.5207, 4.8191, 4.5142, 4.6451, 4.1715, 4.2013, 4.3173, 4.9072], device='cuda:1'), covar=tensor([0.0460, 0.0611, 0.0769, 0.0317, 0.0532, 0.0540, 0.0484, 0.0529], device='cuda:1'), in_proj_covar=tensor([0.0225, 0.0315, 0.0283, 0.0199, 0.0210, 0.0200, 0.0259, 0.0215], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-27 19:48:32,573 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.3079, 4.5886, 4.3127, 4.3780, 3.9132, 3.9991, 4.1299, 4.6003], device='cuda:1'), covar=tensor([0.0417, 0.0548, 0.0734, 0.0320, 0.0541, 0.0629, 0.0469, 0.0504], device='cuda:1'), in_proj_covar=tensor([0.0222, 0.0310, 0.0280, 0.0198, 0.0208, 0.0198, 0.0255, 0.0213], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-27 19:48:39,045 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.46 vs. limit=2.0 2023-04-27 19:48:42,376 INFO [train.py:904] (1/8) Epoch 2, batch 8000, loss[loss=0.3118, simple_loss=0.366, pruned_loss=0.1288, over 16969.00 frames. ], tot_loss[loss=0.3049, simple_loss=0.3668, pruned_loss=0.1215, over 3064837.27 frames. ], batch size: 109, lr: 2.53e-02, grad_scale: 8.0 2023-04-27 19:49:16,785 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=18174.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 19:49:18,073 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 2, batch 8050, loss[loss=0.2745, simple_loss=0.3489, pruned_loss=0.1001, over 16813.00 frames. ], tot_loss[loss=0.304, simple_loss=0.3661, pruned_loss=0.1209, over 3056222.52 frames. ], batch size: 102, lr: 2.52e-02, grad_scale: 8.0 2023-04-27 19:50:24,927 INFO [optim.py:368] (1/8) 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 19:50:46,739 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=18235.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 19:50:48,697 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=18236.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 19:51:11,132 INFO [train.py:904] (1/8) Epoch 2, batch 8100, loss[loss=0.2896, simple_loss=0.3641, pruned_loss=0.1075, over 16840.00 frames. ], tot_loss[loss=0.3023, simple_loss=0.3649, pruned_loss=0.1198, over 3051104.99 frames. ], batch size: 116, lr: 2.52e-02, grad_scale: 8.0 2023-04-27 19:51:51,580 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=2.71 vs. limit=5.0 2023-04-27 19:52:29,017 INFO [train.py:904] (1/8) Epoch 2, batch 8150, loss[loss=0.2603, simple_loss=0.3335, pruned_loss=0.09352, over 16882.00 frames. ], tot_loss[loss=0.2985, simple_loss=0.3615, pruned_loss=0.1178, over 3066788.99 frames. ], batch size: 96, lr: 2.52e-02, grad_scale: 8.0 2023-04-27 19:52:52,885 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=18316.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 19:53:00,710 INFO [optim.py:368] (1/8) 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] (1/8) Epoch 2, batch 8200, loss[loss=0.2861, simple_loss=0.3557, pruned_loss=0.1082, over 16673.00 frames. ], tot_loss[loss=0.2957, simple_loss=0.3585, pruned_loss=0.1164, over 3077358.81 frames. ], batch size: 134, lr: 2.51e-02, grad_scale: 4.0 2023-04-27 19:53:57,253 INFO [zipformer.py:625] (1/8) 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] (1/8) 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] (1/8) Epoch 2, batch 8250, loss[loss=0.2512, simple_loss=0.3389, pruned_loss=0.08174, over 16821.00 frames. ], tot_loss[loss=0.2924, simple_loss=0.3573, pruned_loss=0.1137, over 3064280.40 frames. ], batch size: 102, lr: 2.51e-02, grad_scale: 4.0 2023-04-27 19:55:15,137 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=18404.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 19:55:30,783 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.6231, 4.6693, 4.7160, 3.7990, 4.5784, 4.5474, 4.9169, 2.7137], device='cuda:1'), covar=tensor([0.0390, 0.0026, 0.0024, 0.0164, 0.0029, 0.0050, 0.0017, 0.0352], device='cuda:1'), in_proj_covar=tensor([0.0114, 0.0053, 0.0057, 0.0106, 0.0052, 0.0057, 0.0057, 0.0104], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0001, 0.0002, 0.0003, 0.0001, 0.0002, 0.0002, 0.0003], device='cuda:1') 2023-04-27 19:55:44,640 INFO [optim.py:368] (1/8) 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 19:56:05,978 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=18434.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 19:56:23,331 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.80 vs. limit=2.0 2023-04-27 19:56:24,152 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.4905, 3.5801, 2.8020, 2.6853, 2.6306, 2.1122, 3.7133, 4.1990], device='cuda:1'), covar=tensor([0.1928, 0.0605, 0.1045, 0.0774, 0.1805, 0.1292, 0.0305, 0.0147], device='cuda:1'), in_proj_covar=tensor([0.0250, 0.0219, 0.0233, 0.0178, 0.0263, 0.0183, 0.0195, 0.0146], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-27 19:56:33,571 INFO [train.py:904] (1/8) Epoch 2, batch 8300, loss[loss=0.247, simple_loss=0.3397, pruned_loss=0.07719, over 16898.00 frames. ], tot_loss[loss=0.2843, simple_loss=0.3522, pruned_loss=0.1082, over 3070255.27 frames. ], batch size: 102, lr: 2.51e-02, grad_scale: 4.0 2023-04-27 19:56:48,888 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.4165, 4.1458, 4.2420, 3.7102, 4.2840, 1.6853, 3.9801, 4.1891], device='cuda:1'), covar=tensor([0.0049, 0.0056, 0.0057, 0.0223, 0.0045, 0.1357, 0.0063, 0.0093], device='cuda:1'), in_proj_covar=tensor([0.0063, 0.0051, 0.0078, 0.0094, 0.0058, 0.0110, 0.0070, 0.0076], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002], device='cuda:1') 2023-04-27 19:57:37,596 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=18490.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 19:57:54,223 INFO [train.py:904] (1/8) Epoch 2, batch 8350, loss[loss=0.2855, simple_loss=0.3414, pruned_loss=0.1148, over 11980.00 frames. ], tot_loss[loss=0.2791, simple_loss=0.3494, pruned_loss=0.1044, over 3062878.05 frames. ], batch size: 248, lr: 2.50e-02, grad_scale: 4.0 2023-04-27 19:58:06,643 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=18508.0, num_to_drop=1, layers_to_drop={1} 2023-04-27 19:58:22,518 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([1.6014, 1.2222, 1.5227, 1.6507, 1.8395, 1.8089, 1.3553, 1.7940], device='cuda:1'), covar=tensor([0.0063, 0.0203, 0.0109, 0.0127, 0.0038, 0.0063, 0.0188, 0.0065], device='cuda:1'), in_proj_covar=tensor([0.0070, 0.0110, 0.0095, 0.0080, 0.0060, 0.0057, 0.0098, 0.0055], device='cuda:1'), out_proj_covar=tensor([1.1969e-04, 1.8979e-04, 1.6894e-04, 1.4182e-04, 1.0168e-04, 9.6887e-05, 1.6648e-04, 9.2635e-05], device='cuda:1') 2023-04-27 19:58:28,817 INFO [optim.py:368] (1/8) 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 19:58:35,571 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([1.8669, 3.5319, 2.2676, 4.5997, 4.5260, 4.1479, 2.1949, 3.0691], device='cuda:1'), covar=tensor([0.1720, 0.0481, 0.1607, 0.0079, 0.0147, 0.0293, 0.1303, 0.0763], device='cuda:1'), in_proj_covar=tensor([0.0150, 0.0125, 0.0167, 0.0070, 0.0111, 0.0122, 0.0156, 0.0150], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:1') 2023-04-27 19:58:41,645 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=18530.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 19:58:44,088 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=18531.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 19:59:15,560 INFO [train.py:904] (1/8) Epoch 2, batch 8400, loss[loss=0.2847, simple_loss=0.3378, pruned_loss=0.1158, over 12500.00 frames. ], tot_loss[loss=0.2728, simple_loss=0.345, pruned_loss=0.1003, over 3081162.04 frames. ], batch size: 246, lr: 2.50e-02, grad_scale: 8.0 2023-04-27 19:59:16,068 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=18551.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 19:59:45,483 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=18569.0, num_to_drop=1, layers_to_drop={3} 2023-04-27 20:00:35,047 INFO [train.py:904] (1/8) Epoch 2, batch 8450, loss[loss=0.2392, simple_loss=0.328, pruned_loss=0.07524, over 16834.00 frames. ], tot_loss[loss=0.2694, simple_loss=0.3428, pruned_loss=0.09799, over 3092340.17 frames. ], batch size: 102, lr: 2.50e-02, grad_scale: 8.0 2023-04-27 20:01:00,314 INFO [zipformer.py:625] (1/8) 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] (1/8) Clipping_scale=2.0, grad-norm quartiles 2.749e+02 4.009e+02 5.012e+02 6.279e+02 1.629e+03, threshold=1.002e+03, percent-clipped=8.0 2023-04-27 20:01:55,557 INFO [train.py:904] (1/8) Epoch 2, batch 8500, loss[loss=0.2583, simple_loss=0.3298, pruned_loss=0.09346, over 15359.00 frames. ], tot_loss[loss=0.2629, simple_loss=0.3377, pruned_loss=0.09408, over 3101345.19 frames. ], batch size: 190, lr: 2.49e-02, grad_scale: 8.0 2023-04-27 20:02:17,755 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=18664.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 20:02:33,932 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=18674.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 20:03:19,433 INFO [train.py:904] (1/8) Epoch 2, batch 8550, loss[loss=0.2776, simple_loss=0.3572, pruned_loss=0.09903, over 16730.00 frames. ], tot_loss[loss=0.2601, simple_loss=0.3347, pruned_loss=0.09277, over 3067596.87 frames. ], batch size: 134, lr: 2.49e-02, grad_scale: 4.0 2023-04-27 20:03:23,274 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.94 vs. limit=2.0 2023-04-27 20:04:03,005 INFO [optim.py:368] (1/8) 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 20:04:23,318 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.2654, 1.4354, 1.6898, 2.3251, 2.3518, 2.1536, 1.5080, 2.1790], device='cuda:1'), covar=tensor([0.0042, 0.0218, 0.0137, 0.0094, 0.0044, 0.0086, 0.0175, 0.0079], device='cuda:1'), in_proj_covar=tensor([0.0072, 0.0112, 0.0095, 0.0085, 0.0063, 0.0058, 0.0099, 0.0056], device='cuda:1'), out_proj_covar=tensor([1.2174e-04, 1.9254e-04, 1.6969e-04, 1.5178e-04, 1.0570e-04, 1.0030e-04, 1.6798e-04, 9.4090e-05], device='cuda:1') 2023-04-27 20:04:26,398 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=18735.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 20:04:58,646 INFO [train.py:904] (1/8) Epoch 2, batch 8600, loss[loss=0.2603, simple_loss=0.3368, pruned_loss=0.09187, over 16506.00 frames. ], tot_loss[loss=0.2596, simple_loss=0.3351, pruned_loss=0.09204, over 3057801.23 frames. ], batch size: 68, lr: 2.49e-02, grad_scale: 2.0 2023-04-27 20:05:17,441 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.93 vs. limit=2.0 2023-04-27 20:06:37,153 INFO [train.py:904] (1/8) Epoch 2, batch 8650, loss[loss=0.2353, simple_loss=0.3244, pruned_loss=0.0731, over 16887.00 frames. ], tot_loss[loss=0.2547, simple_loss=0.3317, pruned_loss=0.08888, over 3063539.76 frames. ], batch size: 116, lr: 2.49e-02, grad_scale: 2.0 2023-04-27 20:07:29,096 INFO [optim.py:368] (1/8) 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 20:07:44,360 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=18830.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 20:07:46,110 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=18831.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 20:08:14,549 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=18846.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 20:08:23,064 INFO [train.py:904] (1/8) Epoch 2, batch 8700, loss[loss=0.2357, simple_loss=0.3066, pruned_loss=0.08245, over 12292.00 frames. ], tot_loss[loss=0.2499, simple_loss=0.327, pruned_loss=0.08638, over 3053416.52 frames. ], batch size: 250, lr: 2.48e-02, grad_scale: 2.0 2023-04-27 20:08:29,143 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.9142, 3.7341, 3.7326, 3.3504, 3.7863, 1.5570, 3.5858, 3.6536], device='cuda:1'), covar=tensor([0.0064, 0.0062, 0.0066, 0.0193, 0.0066, 0.1537, 0.0077, 0.0103], device='cuda:1'), in_proj_covar=tensor([0.0060, 0.0050, 0.0076, 0.0086, 0.0057, 0.0109, 0.0069, 0.0074], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0001, 0.0002, 0.0002, 0.0001, 0.0003, 0.0002, 0.0002], device='cuda:1') 2023-04-27 20:08:49,479 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=18864.0, num_to_drop=1, layers_to_drop={0} 2023-04-27 20:09:14,855 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=18878.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 20:09:17,054 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 2, batch 8750, loss[loss=0.2524, simple_loss=0.3354, pruned_loss=0.08468, over 16249.00 frames. ], tot_loss[loss=0.2475, simple_loss=0.3253, pruned_loss=0.08481, over 3055818.82 frames. ], batch size: 165, lr: 2.48e-02, grad_scale: 2.0 2023-04-27 20:10:57,517 INFO [optim.py:368] (1/8) 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 20:11:10,316 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 2, batch 8800, loss[loss=0.2492, simple_loss=0.3344, pruned_loss=0.08201, over 16845.00 frames. ], tot_loss[loss=0.2442, simple_loss=0.3223, pruned_loss=0.08307, over 3040163.33 frames. ], batch size: 96, lr: 2.48e-02, grad_scale: 4.0 2023-04-27 20:12:18,027 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.4071, 3.0108, 2.3892, 2.2963, 2.1952, 2.0073, 2.8850, 3.2540], device='cuda:1'), covar=tensor([0.1624, 0.0642, 0.1192, 0.0800, 0.1866, 0.1502, 0.0455, 0.0276], device='cuda:1'), in_proj_covar=tensor([0.0251, 0.0224, 0.0240, 0.0182, 0.0223, 0.0186, 0.0196, 0.0143], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-27 20:13:18,187 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=18991.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 20:13:38,731 INFO [train.py:904] (1/8) Epoch 2, batch 8850, loss[loss=0.2215, simple_loss=0.3201, pruned_loss=0.06145, over 16636.00 frames. ], tot_loss[loss=0.2434, simple_loss=0.324, pruned_loss=0.08139, over 3044171.32 frames. ], batch size: 62, lr: 2.47e-02, grad_scale: 4.0 2023-04-27 20:14:28,278 INFO [optim.py:368] (1/8) 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 20:14:43,425 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=19030.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 20:14:46,217 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.2836, 2.9901, 2.8520, 1.9270, 2.6536, 1.9117, 2.7230, 3.0457], device='cuda:1'), covar=tensor([0.0330, 0.0413, 0.0416, 0.1424, 0.0632, 0.1005, 0.0723, 0.0478], device='cuda:1'), in_proj_covar=tensor([0.0133, 0.0102, 0.0152, 0.0151, 0.0141, 0.0136, 0.0144, 0.0104], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0004, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:1') 2023-04-27 20:15:00,847 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.4895, 3.3121, 3.3998, 3.1257, 3.4173, 1.9258, 3.2486, 3.2426], device='cuda:1'), covar=tensor([0.0067, 0.0052, 0.0066, 0.0160, 0.0052, 0.1142, 0.0067, 0.0091], device='cuda:1'), in_proj_covar=tensor([0.0061, 0.0050, 0.0076, 0.0086, 0.0058, 0.0111, 0.0070, 0.0075], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0001, 0.0002, 0.0002, 0.0001, 0.0003, 0.0002, 0.0002], device='cuda:1') 2023-04-27 20:15:27,233 INFO [train.py:904] (1/8) Epoch 2, batch 8900, loss[loss=0.2313, simple_loss=0.3198, pruned_loss=0.07144, over 16233.00 frames. ], tot_loss[loss=0.2426, simple_loss=0.3242, pruned_loss=0.0805, over 3039050.13 frames. ], batch size: 165, lr: 2.47e-02, grad_scale: 4.0 2023-04-27 20:15:46,727 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.7769, 2.6591, 1.7082, 2.8311, 2.1527, 2.7499, 1.9604, 2.4293], device='cuda:1'), covar=tensor([0.0075, 0.0253, 0.1350, 0.0063, 0.0778, 0.0422, 0.1195, 0.0527], device='cuda:1'), in_proj_covar=tensor([0.0076, 0.0111, 0.0173, 0.0075, 0.0157, 0.0134, 0.0180, 0.0155], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003, 0.0004, 0.0003], device='cuda:1') 2023-04-27 20:16:30,515 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.2100, 1.7456, 2.0883, 3.0016, 2.9214, 3.0337, 1.7798, 2.8586], device='cuda:1'), covar=tensor([0.0035, 0.0199, 0.0135, 0.0061, 0.0036, 0.0070, 0.0176, 0.0073], device='cuda:1'), in_proj_covar=tensor([0.0070, 0.0109, 0.0094, 0.0082, 0.0063, 0.0057, 0.0097, 0.0056], device='cuda:1'), out_proj_covar=tensor([1.1812e-04, 1.8585e-04, 1.6748e-04, 1.4507e-04, 1.0476e-04, 9.6984e-05, 1.6349e-04, 9.2341e-05], device='cuda:1') 2023-04-27 20:17:32,222 INFO [train.py:904] (1/8) Epoch 2, batch 8950, loss[loss=0.2378, simple_loss=0.3215, pruned_loss=0.07701, over 16725.00 frames. ], tot_loss[loss=0.2426, simple_loss=0.3235, pruned_loss=0.08078, over 3042682.77 frames. ], batch size: 83, lr: 2.47e-02, grad_scale: 4.0 2023-04-27 20:17:48,330 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([1.7832, 3.1394, 2.2280, 4.0185, 4.0508, 4.0101, 1.6735, 3.0024], device='cuda:1'), covar=tensor([0.1453, 0.0407, 0.1426, 0.0079, 0.0129, 0.0200, 0.1346, 0.0631], device='cuda:1'), in_proj_covar=tensor([0.0148, 0.0124, 0.0169, 0.0069, 0.0110, 0.0118, 0.0154, 0.0152], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:1') 2023-04-27 20:18:08,423 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([1.7802, 2.7344, 2.7074, 2.0887, 2.5325, 2.5100, 2.7131, 1.7829], device='cuda:1'), covar=tensor([0.0366, 0.0036, 0.0051, 0.0209, 0.0038, 0.0058, 0.0038, 0.0339], device='cuda:1'), in_proj_covar=tensor([0.0115, 0.0054, 0.0057, 0.0105, 0.0052, 0.0058, 0.0057, 0.0106], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0001, 0.0002, 0.0003, 0.0001, 0.0002, 0.0002, 0.0003], device='cuda:1') 2023-04-27 20:18:20,856 INFO [optim.py:368] (1/8) 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 20:18:32,644 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=19129.0, num_to_drop=1, layers_to_drop={0} 2023-04-27 20:19:11,762 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=19146.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 20:19:16,267 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.3271, 2.3877, 1.7578, 1.8695, 3.0424, 2.7837, 3.5625, 3.3781], device='cuda:1'), covar=tensor([0.0013, 0.0132, 0.0171, 0.0179, 0.0068, 0.0119, 0.0020, 0.0052], device='cuda:1'), in_proj_covar=tensor([0.0048, 0.0103, 0.0104, 0.0104, 0.0093, 0.0102, 0.0056, 0.0076], device='cuda:1'), out_proj_covar=tensor([6.7093e-05, 1.5727e-04, 1.5469e-04, 1.5773e-04, 1.4792e-04, 1.6126e-04, 8.2191e-05, 1.2066e-04], device='cuda:1') 2023-04-27 20:19:21,318 INFO [train.py:904] (1/8) Epoch 2, batch 9000, loss[loss=0.2754, simple_loss=0.344, pruned_loss=0.1034, over 12317.00 frames. ], tot_loss[loss=0.2389, simple_loss=0.3201, pruned_loss=0.07887, over 3056303.09 frames. ], batch size: 248, lr: 2.46e-02, grad_scale: 2.0 2023-04-27 20:19:21,318 INFO [train.py:929] (1/8) Computing validation loss 2023-04-27 20:19:31,139 INFO [train.py:938] (1/8) Epoch 2, validation: loss=0.2044, simple_loss=0.3047, pruned_loss=0.05205, over 944034.00 frames. 2023-04-27 20:19:31,140 INFO [train.py:939] (1/8) Maximum memory allocated so far is 17676MB 2023-04-27 20:20:00,846 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=19164.0, num_to_drop=1, layers_to_drop={0} 2023-04-27 20:20:52,967 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=19190.0, num_to_drop=1, layers_to_drop={3} 2023-04-27 20:21:02,063 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=19194.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 20:21:13,856 INFO [train.py:904] (1/8) Epoch 2, batch 9050, loss[loss=0.2519, simple_loss=0.3281, pruned_loss=0.08781, over 12778.00 frames. ], tot_loss[loss=0.2416, simple_loss=0.3226, pruned_loss=0.0803, over 3067369.27 frames. ], batch size: 246, lr: 2.46e-02, grad_scale: 2.0 2023-04-27 20:21:37,536 INFO [zipformer.py:625] (1/8) 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] (1/8) 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 20:22:58,685 INFO [train.py:904] (1/8) Epoch 2, batch 9100, loss[loss=0.2289, simple_loss=0.32, pruned_loss=0.06888, over 16878.00 frames. ], tot_loss[loss=0.2416, simple_loss=0.3221, pruned_loss=0.08057, over 3080472.69 frames. ], batch size: 102, lr: 2.46e-02, grad_scale: 2.0 2023-04-27 20:23:06,511 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([1.6665, 3.2046, 3.0921, 2.1686, 3.0339, 3.0846, 3.0863, 1.6661], device='cuda:1'), covar=tensor([0.0505, 0.0031, 0.0053, 0.0269, 0.0029, 0.0048, 0.0033, 0.0422], device='cuda:1'), in_proj_covar=tensor([0.0119, 0.0055, 0.0060, 0.0110, 0.0053, 0.0062, 0.0060, 0.0108], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0001, 0.0002, 0.0003, 0.0001, 0.0002, 0.0002, 0.0003], device='cuda:1') 2023-04-27 20:24:22,917 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=19286.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 20:24:39,752 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.57 vs. limit=2.0 2023-04-27 20:24:55,898 INFO [train.py:904] (1/8) Epoch 2, batch 9150, loss[loss=0.2479, simple_loss=0.3231, pruned_loss=0.08633, over 12101.00 frames. ], tot_loss[loss=0.2422, simple_loss=0.3231, pruned_loss=0.08069, over 3071784.07 frames. ], batch size: 246, lr: 2.46e-02, grad_scale: 2.0 2023-04-27 20:25:49,332 INFO [optim.py:368] (1/8) 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 20:25:58,977 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 2, batch 9200, loss[loss=0.2224, simple_loss=0.3073, pruned_loss=0.06873, over 16694.00 frames. ], tot_loss[loss=0.2377, simple_loss=0.3177, pruned_loss=0.07884, over 3064965.58 frames. ], batch size: 76, lr: 2.45e-02, grad_scale: 4.0 2023-04-27 20:27:30,640 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=19378.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 20:28:03,168 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=19394.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 20:28:16,032 INFO [train.py:904] (1/8) Epoch 2, batch 9250, loss[loss=0.2095, simple_loss=0.2812, pruned_loss=0.0689, over 12451.00 frames. ], tot_loss[loss=0.2376, simple_loss=0.3174, pruned_loss=0.07891, over 3058145.19 frames. ], batch size: 250, lr: 2.45e-02, grad_scale: 4.0 2023-04-27 20:29:05,881 INFO [optim.py:368] (1/8) Clipping_scale=2.0, grad-norm quartiles 2.723e+02 4.011e+02 4.767e+02 6.782e+02 2.707e+03, threshold=9.534e+02, percent-clipped=7.0 2023-04-27 20:30:06,016 INFO [train.py:904] (1/8) Epoch 2, batch 9300, loss[loss=0.2048, simple_loss=0.2905, pruned_loss=0.05948, over 16761.00 frames. ], tot_loss[loss=0.2357, simple_loss=0.3153, pruned_loss=0.07807, over 3040910.00 frames. ], batch size: 83, lr: 2.45e-02, grad_scale: 4.0 2023-04-27 20:30:14,298 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=19455.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 20:31:20,828 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=19485.0, num_to_drop=1, layers_to_drop={0} 2023-04-27 20:31:49,020 INFO [train.py:904] (1/8) Epoch 2, batch 9350, loss[loss=0.2527, simple_loss=0.3276, pruned_loss=0.08892, over 16525.00 frames. ], tot_loss[loss=0.2357, simple_loss=0.3154, pruned_loss=0.07797, over 3055022.83 frames. ], batch size: 68, lr: 2.44e-02, grad_scale: 4.0 2023-04-27 20:32:37,319 INFO [optim.py:368] (1/8) 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 20:33:27,905 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.1348, 3.9302, 3.6974, 1.7775, 2.9065, 2.2100, 3.4321, 3.9861], device='cuda:1'), covar=tensor([0.0209, 0.0308, 0.0371, 0.1541, 0.0647, 0.1015, 0.0607, 0.0374], device='cuda:1'), in_proj_covar=tensor([0.0130, 0.0100, 0.0151, 0.0147, 0.0138, 0.0133, 0.0145, 0.0102], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:1') 2023-04-27 20:33:28,460 INFO [train.py:904] (1/8) Epoch 2, batch 9400, loss[loss=0.2475, simple_loss=0.3351, pruned_loss=0.07997, over 16918.00 frames. ], tot_loss[loss=0.2367, simple_loss=0.3166, pruned_loss=0.07839, over 3049604.12 frames. ], batch size: 116, lr: 2.44e-02, grad_scale: 4.0 2023-04-27 20:34:33,955 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.0895, 1.6287, 1.5209, 1.3374, 1.7999, 1.6411, 1.8594, 1.9361], device='cuda:1'), covar=tensor([0.0019, 0.0122, 0.0129, 0.0142, 0.0080, 0.0126, 0.0034, 0.0056], device='cuda:1'), in_proj_covar=tensor([0.0049, 0.0103, 0.0101, 0.0105, 0.0092, 0.0101, 0.0055, 0.0074], device='cuda:1'), out_proj_covar=tensor([6.8211e-05, 1.5663e-04, 1.4893e-04, 1.5957e-04, 1.4384e-04, 1.5780e-04, 7.9732e-05, 1.1645e-04], device='cuda:1') 2023-04-27 20:34:39,208 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=19586.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 20:35:08,266 INFO [train.py:904] (1/8) Epoch 2, batch 9450, loss[loss=0.2519, simple_loss=0.3185, pruned_loss=0.09261, over 12188.00 frames. ], tot_loss[loss=0.2377, simple_loss=0.3184, pruned_loss=0.07856, over 3044941.46 frames. ], batch size: 250, lr: 2.44e-02, grad_scale: 4.0 2023-04-27 20:35:15,451 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.51 vs. limit=2.0 2023-04-27 20:35:42,294 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=19618.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 20:35:56,416 INFO [optim.py:368] (1/8) 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] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=19634.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 20:36:48,218 INFO [train.py:904] (1/8) Epoch 2, batch 9500, loss[loss=0.2299, simple_loss=0.3118, pruned_loss=0.07397, over 16682.00 frames. ], tot_loss[loss=0.2354, simple_loss=0.3164, pruned_loss=0.07721, over 3043220.09 frames. ], batch size: 76, lr: 2.43e-02, grad_scale: 4.0 2023-04-27 20:37:12,145 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.5963, 4.4058, 4.1342, 1.9926, 3.3366, 2.5579, 3.7027, 4.4166], device='cuda:1'), covar=tensor([0.0251, 0.0323, 0.0344, 0.1555, 0.0596, 0.0903, 0.0703, 0.0450], device='cuda:1'), in_proj_covar=tensor([0.0132, 0.0100, 0.0151, 0.0148, 0.0137, 0.0133, 0.0144, 0.0103], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0004, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:1') 2023-04-27 20:37:45,616 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=19679.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 20:38:34,553 INFO [train.py:904] (1/8) Epoch 2, batch 9550, loss[loss=0.2738, simple_loss=0.3512, pruned_loss=0.09821, over 15250.00 frames. ], tot_loss[loss=0.2351, simple_loss=0.3157, pruned_loss=0.07728, over 3045873.68 frames. ], batch size: 191, lr: 2.43e-02, grad_scale: 4.0 2023-04-27 20:39:23,642 INFO [optim.py:368] (1/8) 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 20:39:30,223 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.0390, 3.9540, 3.8634, 3.9353, 3.5207, 3.9461, 3.8317, 3.6617], device='cuda:1'), covar=tensor([0.0291, 0.0190, 0.0168, 0.0128, 0.0553, 0.0190, 0.0333, 0.0274], device='cuda:1'), in_proj_covar=tensor([0.0114, 0.0093, 0.0137, 0.0115, 0.0166, 0.0124, 0.0099, 0.0134], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-27 20:40:12,586 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 2, batch 9600, loss[loss=0.2379, simple_loss=0.3172, pruned_loss=0.07927, over 16565.00 frames. ], tot_loss[loss=0.2381, simple_loss=0.3181, pruned_loss=0.07907, over 3047451.67 frames. ], batch size: 68, lr: 2.43e-02, grad_scale: 8.0 2023-04-27 20:41:22,040 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=19785.0, num_to_drop=1, layers_to_drop={1} 2023-04-27 20:41:29,633 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([1.7319, 3.0540, 2.4142, 4.0515, 3.9325, 3.9404, 1.7856, 2.8831], device='cuda:1'), covar=tensor([0.1564, 0.0486, 0.1107, 0.0085, 0.0132, 0.0277, 0.1181, 0.0662], device='cuda:1'), in_proj_covar=tensor([0.0148, 0.0126, 0.0165, 0.0068, 0.0111, 0.0122, 0.0154, 0.0154], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:1') 2023-04-27 20:42:00,142 INFO [train.py:904] (1/8) Epoch 2, batch 9650, loss[loss=0.2234, simple_loss=0.3121, pruned_loss=0.06731, over 17029.00 frames. ], tot_loss[loss=0.2405, simple_loss=0.3208, pruned_loss=0.08013, over 3033524.12 frames. ], batch size: 55, lr: 2.43e-02, grad_scale: 8.0 2023-04-27 20:42:46,451 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.67 vs. limit=2.0 2023-04-27 20:42:54,373 INFO [optim.py:368] (1/8) 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 20:42:59,864 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([1.7714, 1.2236, 1.5161, 1.7074, 1.7921, 1.7197, 1.4088, 1.8422], device='cuda:1'), covar=tensor([0.0060, 0.0169, 0.0094, 0.0111, 0.0043, 0.0080, 0.0125, 0.0043], device='cuda:1'), in_proj_covar=tensor([0.0075, 0.0112, 0.0098, 0.0086, 0.0063, 0.0060, 0.0100, 0.0054], device='cuda:1'), out_proj_covar=tensor([1.2804e-04, 1.8935e-04, 1.7158e-04, 1.4949e-04, 1.0447e-04, 9.8494e-05, 1.6520e-04, 8.7890e-05], device='cuda:1') 2023-04-27 20:43:09,972 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=19833.0, num_to_drop=1, layers_to_drop={0} 2023-04-27 20:43:10,612 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.62 vs. limit=2.0 2023-04-27 20:43:48,658 INFO [train.py:904] (1/8) Epoch 2, batch 9700, loss[loss=0.219, simple_loss=0.3091, pruned_loss=0.06444, over 16747.00 frames. ], tot_loss[loss=0.2396, simple_loss=0.3196, pruned_loss=0.07978, over 3028460.78 frames. ], batch size: 83, lr: 2.42e-02, grad_scale: 8.0 2023-04-27 20:44:53,522 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.61 vs. limit=2.0 2023-04-27 20:45:31,088 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=6.11 vs. limit=5.0 2023-04-27 20:45:33,548 INFO [train.py:904] (1/8) Epoch 2, batch 9750, loss[loss=0.1729, simple_loss=0.262, pruned_loss=0.04195, over 17131.00 frames. ], tot_loss[loss=0.2387, simple_loss=0.3178, pruned_loss=0.07979, over 3015944.53 frames. ], batch size: 47, lr: 2.42e-02, grad_scale: 8.0 2023-04-27 20:46:00,657 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=4.92 vs. limit=5.0 2023-04-27 20:46:21,255 INFO [optim.py:368] (1/8) Clipping_scale=2.0, grad-norm quartiles 2.379e+02 3.927e+02 4.879e+02 5.766e+02 1.284e+03, threshold=9.758e+02, percent-clipped=1.0 2023-04-27 20:46:53,942 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=2.67 vs. limit=5.0 2023-04-27 20:47:15,376 INFO [train.py:904] (1/8) Epoch 2, batch 9800, loss[loss=0.2273, simple_loss=0.3183, pruned_loss=0.06821, over 17032.00 frames. ], tot_loss[loss=0.236, simple_loss=0.3167, pruned_loss=0.0776, over 3043837.52 frames. ], batch size: 109, lr: 2.42e-02, grad_scale: 8.0 2023-04-27 20:47:34,550 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.5240, 3.4191, 3.4222, 3.1142, 3.3686, 2.0509, 3.3049, 3.2249], device='cuda:1'), covar=tensor([0.0052, 0.0045, 0.0059, 0.0167, 0.0049, 0.1143, 0.0061, 0.0098], device='cuda:1'), in_proj_covar=tensor([0.0059, 0.0050, 0.0074, 0.0081, 0.0056, 0.0110, 0.0067, 0.0073], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0001, 0.0002, 0.0002, 0.0001, 0.0003, 0.0002, 0.0002], device='cuda:1') 2023-04-27 20:47:47,864 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.3919, 3.5369, 3.9795, 3.9514, 3.8925, 3.5706, 3.6410, 3.7682], device='cuda:1'), covar=tensor([0.0275, 0.0311, 0.0307, 0.0345, 0.0399, 0.0302, 0.0694, 0.0277], device='cuda:1'), in_proj_covar=tensor([0.0160, 0.0143, 0.0160, 0.0164, 0.0190, 0.0170, 0.0239, 0.0158], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0003, 0.0003, 0.0002], device='cuda:1') 2023-04-27 20:48:00,422 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=19974.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 20:48:39,745 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.9969, 5.4039, 5.0855, 5.2260, 4.5631, 4.4493, 4.9109, 5.4849], device='cuda:1'), covar=tensor([0.0407, 0.0517, 0.0849, 0.0317, 0.0494, 0.0634, 0.0421, 0.0439], device='cuda:1'), in_proj_covar=tensor([0.0215, 0.0298, 0.0259, 0.0184, 0.0198, 0.0189, 0.0234, 0.0201], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-27 20:49:05,348 INFO [train.py:904] (1/8) Epoch 2, batch 9850, loss[loss=0.2399, simple_loss=0.3115, pruned_loss=0.08413, over 12292.00 frames. ], tot_loss[loss=0.2369, simple_loss=0.3184, pruned_loss=0.07767, over 3031677.72 frames. ], batch size: 247, lr: 2.41e-02, grad_scale: 8.0 2023-04-27 20:49:46,061 INFO [zipformer.py:625] (1/8) 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] (1/8) 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 20:49:56,779 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.7985, 3.6351, 3.8859, 4.1298, 4.1478, 3.7189, 4.1762, 4.1355], device='cuda:1'), covar=tensor([0.0546, 0.0579, 0.0868, 0.0335, 0.0333, 0.0544, 0.0297, 0.0307], device='cuda:1'), in_proj_covar=tensor([0.0235, 0.0267, 0.0351, 0.0266, 0.0207, 0.0184, 0.0208, 0.0215], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-27 20:50:58,133 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 2, batch 9900, loss[loss=0.2102, simple_loss=0.3083, pruned_loss=0.056, over 16760.00 frames. ], tot_loss[loss=0.237, simple_loss=0.319, pruned_loss=0.07747, over 3036174.90 frames. ], batch size: 76, lr: 2.41e-02, grad_scale: 8.0 2023-04-27 20:52:12,591 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=20081.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 20:52:31,312 INFO [zipformer.py:625] (1/8) 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] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=20098.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 20:52:56,914 INFO [train.py:904] (1/8) Epoch 2, batch 9950, loss[loss=0.235, simple_loss=0.3184, pruned_loss=0.07581, over 16787.00 frames. ], tot_loss[loss=0.2384, simple_loss=0.3213, pruned_loss=0.0778, over 3037777.10 frames. ], batch size: 124, lr: 2.41e-02, grad_scale: 8.0 2023-04-27 20:53:23,683 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=20112.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 20:53:55,568 INFO [optim.py:368] (1/8) 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 20:54:55,835 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 2, batch 10000, loss[loss=0.2539, simple_loss=0.3194, pruned_loss=0.09416, over 12525.00 frames. ], tot_loss[loss=0.2356, simple_loss=0.3184, pruned_loss=0.07639, over 3054456.69 frames. ], batch size: 250, lr: 2.41e-02, grad_scale: 8.0 2023-04-27 20:55:19,594 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.3576, 4.5580, 4.4375, 4.6258, 4.6198, 5.0528, 4.8162, 4.4660], device='cuda:1'), covar=tensor([0.0760, 0.1077, 0.0981, 0.1266, 0.1714, 0.0709, 0.0769, 0.2200], device='cuda:1'), in_proj_covar=tensor([0.0174, 0.0256, 0.0236, 0.0230, 0.0283, 0.0257, 0.0193, 0.0294], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-27 20:55:35,635 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.97 vs. limit=2.0 2023-04-27 20:55:41,182 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=20173.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 20:56:11,149 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.6648, 3.7428, 3.7542, 3.6996, 3.7402, 4.0967, 3.9591, 3.6239], device='cuda:1'), covar=tensor([0.1643, 0.1197, 0.0945, 0.1723, 0.2105, 0.1001, 0.0864, 0.2186], device='cuda:1'), in_proj_covar=tensor([0.0175, 0.0260, 0.0238, 0.0233, 0.0285, 0.0257, 0.0196, 0.0298], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-27 20:56:37,816 INFO [train.py:904] (1/8) Epoch 2, batch 10050, loss[loss=0.2584, simple_loss=0.3365, pruned_loss=0.09015, over 16810.00 frames. ], tot_loss[loss=0.2358, simple_loss=0.3188, pruned_loss=0.07641, over 3064348.25 frames. ], batch size: 124, lr: 2.40e-02, grad_scale: 8.0 2023-04-27 20:56:44,587 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=20204.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 20:57:24,816 INFO [optim.py:368] (1/8) 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 20:58:00,508 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=2.02 vs. limit=2.0 2023-04-27 20:58:10,190 INFO [train.py:904] (1/8) Epoch 2, batch 10100, loss[loss=0.2214, simple_loss=0.3005, pruned_loss=0.07115, over 16779.00 frames. ], tot_loss[loss=0.238, simple_loss=0.3204, pruned_loss=0.07782, over 3054209.06 frames. ], batch size: 124, lr: 2.40e-02, grad_scale: 8.0 2023-04-27 20:58:38,066 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=20265.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 20:58:58,349 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=20274.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 20:59:03,529 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.5219, 3.6145, 3.6943, 3.6145, 3.7843, 4.1234, 3.9305, 3.6028], device='cuda:1'), covar=tensor([0.1653, 0.1556, 0.1127, 0.1882, 0.2107, 0.1101, 0.0952, 0.2173], device='cuda:1'), in_proj_covar=tensor([0.0179, 0.0268, 0.0244, 0.0237, 0.0294, 0.0264, 0.0196, 0.0304], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-27 20:59:54,874 INFO [train.py:904] (1/8) Epoch 3, batch 0, loss[loss=0.3009, simple_loss=0.3498, pruned_loss=0.126, over 17015.00 frames. ], tot_loss[loss=0.3009, simple_loss=0.3498, pruned_loss=0.126, over 17015.00 frames. ], batch size: 41, lr: 2.28e-02, grad_scale: 8.0 2023-04-27 20:59:54,874 INFO [train.py:929] (1/8) Computing validation loss 2023-04-27 21:00:02,299 INFO [train.py:938] (1/8) Epoch 3, validation: loss=0.2012, simple_loss=0.3019, pruned_loss=0.05024, over 944034.00 frames. 2023-04-27 21:00:02,299 INFO [train.py:939] (1/8) Maximum memory allocated so far is 17676MB 2023-04-27 21:00:13,636 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.4887, 4.2459, 4.3949, 4.6824, 4.7469, 4.2791, 4.8752, 4.6824], device='cuda:1'), covar=tensor([0.0549, 0.0585, 0.1132, 0.0523, 0.0474, 0.0390, 0.0370, 0.0361], device='cuda:1'), in_proj_covar=tensor([0.0241, 0.0280, 0.0362, 0.0276, 0.0210, 0.0193, 0.0216, 0.0219], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-27 21:00:31,871 INFO [zipformer.py:625] (1/8) 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] (1/8) 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 21:00:53,993 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.6703, 4.9979, 4.9470, 5.0501, 4.9642, 5.5143, 5.1974, 4.9734], device='cuda:1'), covar=tensor([0.0688, 0.1094, 0.1063, 0.1222, 0.2336, 0.0665, 0.0913, 0.1587], device='cuda:1'), in_proj_covar=tensor([0.0191, 0.0286, 0.0261, 0.0254, 0.0317, 0.0279, 0.0214, 0.0329], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:1') 2023-04-27 21:01:12,891 INFO [train.py:904] (1/8) Epoch 3, batch 50, loss[loss=0.2322, simple_loss=0.3079, pruned_loss=0.07822, over 16771.00 frames. ], tot_loss[loss=0.2981, simple_loss=0.3534, pruned_loss=0.1213, over 748747.96 frames. ], batch size: 39, lr: 2.28e-02, grad_scale: 2.0 2023-04-27 21:01:45,734 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=20376.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:02:02,776 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=20388.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:02:19,623 INFO [train.py:904] (1/8) Epoch 3, batch 100, loss[loss=0.2116, simple_loss=0.2898, pruned_loss=0.06671, over 17226.00 frames. ], tot_loss[loss=0.2837, simple_loss=0.3443, pruned_loss=0.1116, over 1322234.62 frames. ], batch size: 45, lr: 2.27e-02, grad_scale: 2.0 2023-04-27 21:02:20,502 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=5.11 vs. limit=5.0 2023-04-27 21:02:43,345 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=20418.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:02:54,678 INFO [optim.py:368] (1/8) 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] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=20445.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:03:23,305 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.5876, 3.4664, 3.0201, 1.6872, 2.6233, 1.9440, 3.1492, 3.4321], device='cuda:1'), covar=tensor([0.0244, 0.0397, 0.0432, 0.1469, 0.0711, 0.0995, 0.0558, 0.0458], device='cuda:1'), in_proj_covar=tensor([0.0135, 0.0104, 0.0154, 0.0148, 0.0139, 0.0135, 0.0142, 0.0109], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0004, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:1') 2023-04-27 21:03:24,339 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 3, batch 150, loss[loss=0.2673, simple_loss=0.3421, pruned_loss=0.09625, over 17230.00 frames. ], tot_loss[loss=0.2764, simple_loss=0.3387, pruned_loss=0.107, over 1775073.98 frames. ], batch size: 52, lr: 2.27e-02, grad_scale: 2.0 2023-04-27 21:03:49,093 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=20468.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:03:54,822 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=4.91 vs. limit=5.0 2023-04-27 21:04:05,881 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=20479.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:04:35,430 INFO [train.py:904] (1/8) Epoch 3, batch 200, loss[loss=0.3359, simple_loss=0.3863, pruned_loss=0.1428, over 12051.00 frames. ], tot_loss[loss=0.274, simple_loss=0.3372, pruned_loss=0.1054, over 2109767.13 frames. ], batch size: 247, lr: 2.27e-02, grad_scale: 2.0 2023-04-27 21:05:09,760 INFO [optim.py:368] (1/8) 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] (1/8) Epoch 3, batch 250, loss[loss=0.2806, simple_loss=0.3323, pruned_loss=0.1144, over 16800.00 frames. ], tot_loss[loss=0.2702, simple_loss=0.3338, pruned_loss=0.1034, over 2384059.86 frames. ], batch size: 102, lr: 2.27e-02, grad_scale: 2.0 2023-04-27 21:05:58,359 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=20560.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:06:29,971 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.0204, 4.3226, 3.3002, 2.9411, 3.0337, 2.3768, 4.5808, 4.9420], device='cuda:1'), covar=tensor([0.1816, 0.0512, 0.1066, 0.0860, 0.1864, 0.1329, 0.0244, 0.0201], device='cuda:1'), in_proj_covar=tensor([0.0260, 0.0233, 0.0251, 0.0192, 0.0246, 0.0187, 0.0206, 0.0164], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0003, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-27 21:06:32,282 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.3155, 2.4752, 1.9441, 1.9905, 3.0043, 2.8962, 3.6091, 3.3584], device='cuda:1'), covar=tensor([0.0023, 0.0154, 0.0167, 0.0199, 0.0088, 0.0126, 0.0037, 0.0061], device='cuda:1'), in_proj_covar=tensor([0.0056, 0.0112, 0.0109, 0.0112, 0.0102, 0.0110, 0.0063, 0.0081], device='cuda:1'), out_proj_covar=tensor([7.7430e-05, 1.6868e-04, 1.5896e-04, 1.6724e-04, 1.5699e-04, 1.6977e-04, 9.3653e-05, 1.2725e-04], device='cuda:1') 2023-04-27 21:06:35,574 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=20588.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:06:54,935 INFO [train.py:904] (1/8) Epoch 3, batch 300, loss[loss=0.2265, simple_loss=0.3107, pruned_loss=0.07113, over 17123.00 frames. ], tot_loss[loss=0.2643, simple_loss=0.3286, pruned_loss=0.1001, over 2596153.31 frames. ], batch size: 49, lr: 2.26e-02, grad_scale: 2.0 2023-04-27 21:07:29,013 INFO [optim.py:368] (1/8) 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 21:07:48,982 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.7919, 2.7525, 2.4879, 4.0739, 2.0346, 3.8079, 2.3274, 2.5368], device='cuda:1'), covar=tensor([0.0274, 0.0545, 0.0352, 0.0181, 0.1466, 0.0183, 0.0765, 0.0985], device='cuda:1'), in_proj_covar=tensor([0.0230, 0.0207, 0.0171, 0.0234, 0.0279, 0.0178, 0.0203, 0.0267], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-27 21:07:59,107 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 3, batch 350, loss[loss=0.2327, simple_loss=0.2972, pruned_loss=0.08416, over 16250.00 frames. ], tot_loss[loss=0.2584, simple_loss=0.3234, pruned_loss=0.09669, over 2761990.64 frames. ], batch size: 165, lr: 2.26e-02, grad_scale: 2.0 2023-04-27 21:08:36,809 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=20676.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:09:04,570 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([1.8565, 3.2170, 3.5993, 2.6269, 3.6304, 3.6058, 3.6386, 1.6753], device='cuda:1'), covar=tensor([0.0382, 0.0084, 0.0047, 0.0210, 0.0045, 0.0050, 0.0031, 0.0390], device='cuda:1'), in_proj_covar=tensor([0.0110, 0.0057, 0.0060, 0.0106, 0.0055, 0.0061, 0.0060, 0.0105], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:1') 2023-04-27 21:09:09,381 INFO [train.py:904] (1/8) Epoch 3, batch 400, loss[loss=0.1857, simple_loss=0.2634, pruned_loss=0.05402, over 17008.00 frames. ], tot_loss[loss=0.255, simple_loss=0.321, pruned_loss=0.09448, over 2892563.59 frames. ], batch size: 41, lr: 2.26e-02, grad_scale: 4.0 2023-04-27 21:09:41,229 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=20724.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:09:45,244 INFO [optim.py:368] (1/8) 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 21:09:49,103 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([1.8632, 3.7873, 2.7205, 5.2931, 5.0846, 4.4114, 1.8099, 3.5565], device='cuda:1'), covar=tensor([0.1466, 0.0389, 0.1193, 0.0046, 0.0210, 0.0284, 0.1361, 0.0619], device='cuda:1'), in_proj_covar=tensor([0.0144, 0.0128, 0.0166, 0.0069, 0.0130, 0.0130, 0.0153, 0.0152], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:1') 2023-04-27 21:09:50,323 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.4120, 2.3258, 2.4189, 2.4489, 2.8695, 2.8587, 3.5139, 3.3122], device='cuda:1'), covar=tensor([0.0023, 0.0157, 0.0150, 0.0156, 0.0106, 0.0138, 0.0062, 0.0064], device='cuda:1'), in_proj_covar=tensor([0.0056, 0.0110, 0.0107, 0.0108, 0.0101, 0.0110, 0.0063, 0.0080], device='cuda:1'), out_proj_covar=tensor([7.7098e-05, 1.6546e-04, 1.5570e-04, 1.6120e-04, 1.5469e-04, 1.6791e-04, 9.2835e-05, 1.2441e-04], device='cuda:1') 2023-04-27 21:10:08,711 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=20744.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:10:10,752 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=20745.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:10:17,858 INFO [train.py:904] (1/8) Epoch 3, batch 450, loss[loss=0.2672, simple_loss=0.3514, pruned_loss=0.09147, over 17236.00 frames. ], tot_loss[loss=0.252, simple_loss=0.3187, pruned_loss=0.09268, over 2996550.00 frames. ], batch size: 52, lr: 2.25e-02, grad_scale: 4.0 2023-04-27 21:10:41,801 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=20768.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:10:49,445 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=20774.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:11:14,835 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=20793.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:11:24,055 INFO [train.py:904] (1/8) Epoch 3, batch 500, loss[loss=0.2178, simple_loss=0.3074, pruned_loss=0.06407, over 17123.00 frames. ], tot_loss[loss=0.2499, simple_loss=0.3177, pruned_loss=0.09102, over 3073506.39 frames. ], batch size: 48, lr: 2.25e-02, grad_scale: 4.0 2023-04-27 21:11:46,146 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=20816.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:12:00,585 INFO [optim.py:368] (1/8) 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 21:12:34,371 INFO [train.py:904] (1/8) Epoch 3, batch 550, loss[loss=0.2348, simple_loss=0.3177, pruned_loss=0.076, over 17255.00 frames. ], tot_loss[loss=0.2495, simple_loss=0.3166, pruned_loss=0.09122, over 3122669.21 frames. ], batch size: 52, lr: 2.25e-02, grad_scale: 4.0 2023-04-27 21:12:45,317 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=20860.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:13:40,833 INFO [train.py:904] (1/8) Epoch 3, batch 600, loss[loss=0.2678, simple_loss=0.3262, pruned_loss=0.1047, over 16448.00 frames. ], tot_loss[loss=0.2498, simple_loss=0.3153, pruned_loss=0.0921, over 3168484.41 frames. ], batch size: 68, lr: 2.25e-02, grad_scale: 4.0 2023-04-27 21:13:43,517 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([5.7795, 6.0377, 5.6978, 5.8081, 5.3024, 5.0335, 5.5111, 6.1526], device='cuda:1'), covar=tensor([0.0423, 0.0611, 0.0828, 0.0353, 0.0483, 0.0405, 0.0448, 0.0464], device='cuda:1'), in_proj_covar=tensor([0.0254, 0.0373, 0.0316, 0.0223, 0.0242, 0.0224, 0.0286, 0.0246], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-27 21:13:50,522 INFO [zipformer.py:625] (1/8) 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] (1/8) 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 21:14:38,328 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 3, batch 650, loss[loss=0.2125, simple_loss=0.2899, pruned_loss=0.06753, over 17195.00 frames. ], tot_loss[loss=0.2484, simple_loss=0.3138, pruned_loss=0.09148, over 3201171.91 frames. ], batch size: 46, lr: 2.24e-02, grad_scale: 4.0 2023-04-27 21:15:57,313 INFO [train.py:904] (1/8) Epoch 3, batch 700, loss[loss=0.2228, simple_loss=0.3005, pruned_loss=0.07257, over 17178.00 frames. ], tot_loss[loss=0.2459, simple_loss=0.3121, pruned_loss=0.08986, over 3223907.17 frames. ], batch size: 46, lr: 2.24e-02, grad_scale: 4.0 2023-04-27 21:16:30,866 INFO [optim.py:368] (1/8) 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 21:16:55,413 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=21044.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:17:03,823 INFO [train.py:904] (1/8) Epoch 3, batch 750, loss[loss=0.2384, simple_loss=0.3179, pruned_loss=0.07944, over 17074.00 frames. ], tot_loss[loss=0.2471, simple_loss=0.313, pruned_loss=0.09056, over 3241538.18 frames. ], batch size: 55, lr: 2.24e-02, grad_scale: 4.0 2023-04-27 21:17:06,015 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=21052.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:17:18,668 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([5.2396, 4.9794, 4.9987, 5.0837, 4.5571, 4.9496, 4.9957, 4.6993], device='cuda:1'), covar=tensor([0.0304, 0.0218, 0.0175, 0.0122, 0.0832, 0.0203, 0.0218, 0.0260], device='cuda:1'), in_proj_covar=tensor([0.0149, 0.0119, 0.0182, 0.0149, 0.0221, 0.0161, 0.0130, 0.0170], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-04-27 21:17:35,032 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=21074.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:17:57,835 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 3, batch 800, loss[loss=0.2031, simple_loss=0.2731, pruned_loss=0.06656, over 16545.00 frames. ], tot_loss[loss=0.2462, simple_loss=0.3124, pruned_loss=0.08999, over 3264732.83 frames. ], batch size: 68, lr: 2.24e-02, grad_scale: 8.0 2023-04-27 21:18:27,387 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=21113.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:18:34,854 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=21118.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:18:39,528 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=21122.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:18:47,522 INFO [optim.py:368] (1/8) Clipping_scale=2.0, grad-norm quartiles 2.745e+02 4.034e+02 4.669e+02 5.861e+02 1.175e+03, threshold=9.338e+02, percent-clipped=5.0 2023-04-27 21:19:20,166 INFO [train.py:904] (1/8) Epoch 3, batch 850, loss[loss=0.2435, simple_loss=0.3009, pruned_loss=0.0931, over 16783.00 frames. ], tot_loss[loss=0.247, simple_loss=0.3128, pruned_loss=0.09059, over 3271104.00 frames. ], batch size: 124, lr: 2.23e-02, grad_scale: 8.0 2023-04-27 21:19:58,159 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=21179.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:20:03,570 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.9266, 3.7777, 3.8032, 3.3580, 3.8020, 1.9255, 3.6099, 3.7290], device='cuda:1'), covar=tensor([0.0077, 0.0063, 0.0091, 0.0281, 0.0070, 0.1268, 0.0083, 0.0119], device='cuda:1'), in_proj_covar=tensor([0.0075, 0.0061, 0.0096, 0.0111, 0.0070, 0.0119, 0.0083, 0.0095], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-04-27 21:20:27,600 INFO [train.py:904] (1/8) Epoch 3, batch 900, loss[loss=0.2058, simple_loss=0.2865, pruned_loss=0.06253, over 17190.00 frames. ], tot_loss[loss=0.2433, simple_loss=0.3107, pruned_loss=0.08801, over 3287424.93 frames. ], batch size: 46, lr: 2.23e-02, grad_scale: 8.0 2023-04-27 21:20:57,866 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.85 vs. limit=2.0 2023-04-27 21:21:03,015 INFO [optim.py:368] (1/8) 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 21:21:23,804 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.2801, 2.4263, 1.7500, 2.0719, 2.9068, 2.8394, 3.2417, 3.0454], device='cuda:1'), covar=tensor([0.0044, 0.0117, 0.0148, 0.0141, 0.0072, 0.0087, 0.0043, 0.0062], device='cuda:1'), in_proj_covar=tensor([0.0057, 0.0110, 0.0110, 0.0109, 0.0102, 0.0110, 0.0065, 0.0083], device='cuda:1'), out_proj_covar=tensor([7.8401e-05, 1.6630e-04, 1.5833e-04, 1.6043e-04, 1.5626e-04, 1.6702e-04, 9.6291e-05, 1.2914e-04], device='cuda:1') 2023-04-27 21:21:26,687 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=21244.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:21:35,851 INFO [train.py:904] (1/8) Epoch 3, batch 950, loss[loss=0.2494, simple_loss=0.3064, pruned_loss=0.09619, over 16676.00 frames. ], tot_loss[loss=0.2439, simple_loss=0.3107, pruned_loss=0.08858, over 3297805.50 frames. ], batch size: 134, lr: 2.23e-02, grad_scale: 8.0 2023-04-27 21:22:22,724 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([5.3900, 5.7212, 5.3943, 5.5907, 4.9469, 4.8843, 5.2365, 5.7910], device='cuda:1'), covar=tensor([0.0430, 0.0629, 0.0894, 0.0318, 0.0496, 0.0438, 0.0438, 0.0565], device='cuda:1'), in_proj_covar=tensor([0.0256, 0.0382, 0.0323, 0.0224, 0.0244, 0.0225, 0.0288, 0.0252], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-27 21:22:34,059 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=21292.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:22:45,620 INFO [train.py:904] (1/8) Epoch 3, batch 1000, loss[loss=0.2319, simple_loss=0.2894, pruned_loss=0.08718, over 16836.00 frames. ], tot_loss[loss=0.2416, simple_loss=0.3091, pruned_loss=0.087, over 3304648.04 frames. ], batch size: 90, lr: 2.23e-02, grad_scale: 8.0 2023-04-27 21:23:01,338 INFO [zipformer.py:625] (1/8) 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] (1/8) 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 21:23:24,018 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=21328.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:23:55,381 INFO [train.py:904] (1/8) Epoch 3, batch 1050, loss[loss=0.2269, simple_loss=0.2866, pruned_loss=0.08356, over 16981.00 frames. ], tot_loss[loss=0.2403, simple_loss=0.3082, pruned_loss=0.08621, over 3308789.20 frames. ], batch size: 41, lr: 2.22e-02, grad_scale: 8.0 2023-04-27 21:24:25,348 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=21373.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:24:46,840 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=21389.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:25:02,814 INFO [train.py:904] (1/8) Epoch 3, batch 1100, loss[loss=0.2401, simple_loss=0.2994, pruned_loss=0.09038, over 16731.00 frames. ], tot_loss[loss=0.2402, simple_loss=0.3084, pruned_loss=0.08599, over 3311308.99 frames. ], batch size: 134, lr: 2.22e-02, grad_scale: 4.0 2023-04-27 21:25:12,306 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=21408.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:25:26,542 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([5.1969, 5.0941, 4.9425, 4.4159, 4.9238, 2.2283, 4.7247, 5.1417], device='cuda:1'), covar=tensor([0.0056, 0.0046, 0.0064, 0.0243, 0.0049, 0.1159, 0.0081, 0.0085], device='cuda:1'), in_proj_covar=tensor([0.0076, 0.0064, 0.0098, 0.0113, 0.0071, 0.0121, 0.0086, 0.0098], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-04-27 21:25:36,650 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.1118, 2.1551, 2.1152, 1.9153, 2.7039, 2.6082, 3.5290, 3.0823], device='cuda:1'), covar=tensor([0.0025, 0.0163, 0.0143, 0.0168, 0.0090, 0.0127, 0.0042, 0.0070], device='cuda:1'), in_proj_covar=tensor([0.0056, 0.0112, 0.0110, 0.0111, 0.0102, 0.0111, 0.0067, 0.0085], device='cuda:1'), out_proj_covar=tensor([7.9082e-05, 1.6857e-04, 1.5783e-04, 1.6472e-04, 1.5527e-04, 1.6911e-04, 1.0035e-04, 1.3119e-04], device='cuda:1') 2023-04-27 21:25:38,416 INFO [optim.py:368] (1/8) 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 21:26:10,401 INFO [train.py:904] (1/8) Epoch 3, batch 1150, loss[loss=0.2258, simple_loss=0.3105, pruned_loss=0.07054, over 17091.00 frames. ], tot_loss[loss=0.2386, simple_loss=0.3074, pruned_loss=0.08488, over 3314181.95 frames. ], batch size: 53, lr: 2.22e-02, grad_scale: 4.0 2023-04-27 21:26:42,688 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=21474.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:27:11,750 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.8009, 4.5360, 4.7209, 5.0860, 5.1190, 4.5125, 5.1317, 5.1105], device='cuda:1'), covar=tensor([0.0550, 0.0510, 0.1009, 0.0317, 0.0317, 0.0439, 0.0290, 0.0267], device='cuda:1'), in_proj_covar=tensor([0.0303, 0.0352, 0.0482, 0.0367, 0.0277, 0.0250, 0.0271, 0.0290], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-27 21:27:19,790 INFO [train.py:904] (1/8) Epoch 3, batch 1200, loss[loss=0.2505, simple_loss=0.3235, pruned_loss=0.08872, over 17110.00 frames. ], tot_loss[loss=0.2367, simple_loss=0.3055, pruned_loss=0.08396, over 3312225.48 frames. ], batch size: 48, lr: 2.22e-02, grad_scale: 8.0 2023-04-27 21:27:41,974 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.3794, 4.4817, 2.0971, 4.5913, 2.5876, 4.5377, 1.9662, 3.2134], device='cuda:1'), covar=tensor([0.0055, 0.0120, 0.1212, 0.0027, 0.0716, 0.0255, 0.1353, 0.0485], device='cuda:1'), in_proj_covar=tensor([0.0085, 0.0129, 0.0168, 0.0082, 0.0156, 0.0159, 0.0178, 0.0158], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0004, 0.0002, 0.0003, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-04-27 21:27:44,794 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.3849, 4.1448, 4.2461, 4.6081, 4.6481, 4.1858, 4.4972, 4.6241], device='cuda:1'), covar=tensor([0.0526, 0.0563, 0.1122, 0.0432, 0.0406, 0.0640, 0.0572, 0.0364], device='cuda:1'), in_proj_covar=tensor([0.0303, 0.0355, 0.0488, 0.0369, 0.0279, 0.0251, 0.0273, 0.0293], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:1') 2023-04-27 21:27:56,781 INFO [optim.py:368] (1/8) 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 21:28:22,337 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([1.8284, 2.9765, 3.5155, 2.3647, 3.4563, 3.5699, 3.5302, 1.6909], device='cuda:1'), covar=tensor([0.0410, 0.0128, 0.0061, 0.0253, 0.0052, 0.0075, 0.0058, 0.0432], device='cuda:1'), in_proj_covar=tensor([0.0115, 0.0059, 0.0063, 0.0109, 0.0055, 0.0063, 0.0062, 0.0108], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:1') 2023-04-27 21:28:27,750 INFO [train.py:904] (1/8) Epoch 3, batch 1250, loss[loss=0.292, simple_loss=0.3319, pruned_loss=0.1261, over 16895.00 frames. ], tot_loss[loss=0.2379, simple_loss=0.3057, pruned_loss=0.08498, over 3313568.84 frames. ], batch size: 116, lr: 2.21e-02, grad_scale: 8.0 2023-04-27 21:29:30,732 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=21595.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:29:39,772 INFO [train.py:904] (1/8) Epoch 3, batch 1300, loss[loss=0.239, simple_loss=0.3219, pruned_loss=0.07802, over 17115.00 frames. ], tot_loss[loss=0.2375, simple_loss=0.3056, pruned_loss=0.08473, over 3318056.94 frames. ], batch size: 49, lr: 2.21e-02, grad_scale: 8.0 2023-04-27 21:29:56,003 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=21613.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:30:17,377 INFO [optim.py:368] (1/8) 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 21:30:49,954 INFO [train.py:904] (1/8) Epoch 3, batch 1350, loss[loss=0.234, simple_loss=0.3191, pruned_loss=0.07445, over 17102.00 frames. ], tot_loss[loss=0.2368, simple_loss=0.3062, pruned_loss=0.08373, over 3326963.52 frames. ], batch size: 48, lr: 2.21e-02, grad_scale: 8.0 2023-04-27 21:30:56,176 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=21656.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:31:13,021 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=21668.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:31:20,967 INFO [zipformer.py:625] (1/8) 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] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=21684.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:31:58,964 INFO [train.py:904] (1/8) Epoch 3, batch 1400, loss[loss=0.2125, simple_loss=0.2919, pruned_loss=0.06657, over 17193.00 frames. ], tot_loss[loss=0.2357, simple_loss=0.3053, pruned_loss=0.08308, over 3319857.21 frames. ], batch size: 46, lr: 2.21e-02, grad_scale: 8.0 2023-04-27 21:32:09,231 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=21708.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:32:35,074 INFO [optim.py:368] (1/8) 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 21:33:07,164 INFO [train.py:904] (1/8) Epoch 3, batch 1450, loss[loss=0.2367, simple_loss=0.3158, pruned_loss=0.07881, over 17106.00 frames. ], tot_loss[loss=0.2362, simple_loss=0.3052, pruned_loss=0.08357, over 3319039.31 frames. ], batch size: 49, lr: 2.21e-02, grad_scale: 8.0 2023-04-27 21:33:13,877 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=21756.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:33:23,969 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=21763.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:33:39,397 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=21774.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:34:14,645 INFO [train.py:904] (1/8) Epoch 3, batch 1500, loss[loss=0.2229, simple_loss=0.2973, pruned_loss=0.07431, over 16673.00 frames. ], tot_loss[loss=0.2358, simple_loss=0.3044, pruned_loss=0.08358, over 3311069.38 frames. ], batch size: 62, lr: 2.20e-02, grad_scale: 8.0 2023-04-27 21:34:43,973 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=21822.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:34:47,275 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=21824.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:34:52,714 INFO [optim.py:368] (1/8) 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 21:35:23,679 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 3, batch 1550, loss[loss=0.2049, simple_loss=0.2807, pruned_loss=0.06458, over 16974.00 frames. ], tot_loss[loss=0.238, simple_loss=0.3068, pruned_loss=0.08457, over 3313325.09 frames. ], batch size: 41, lr: 2.20e-02, grad_scale: 8.0 2023-04-27 21:36:06,483 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.4625, 3.4777, 1.7239, 3.5643, 2.4335, 3.5398, 1.8227, 2.8179], device='cuda:1'), covar=tensor([0.0053, 0.0226, 0.1377, 0.0055, 0.0731, 0.0346, 0.1200, 0.0553], device='cuda:1'), in_proj_covar=tensor([0.0085, 0.0133, 0.0169, 0.0083, 0.0153, 0.0161, 0.0177, 0.0157], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0004, 0.0002, 0.0003, 0.0004, 0.0004, 0.0003], device='cuda:1') 2023-04-27 21:36:17,895 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.5899, 4.5493, 1.8285, 4.6565, 2.7449, 4.7115, 1.8374, 3.5025], device='cuda:1'), covar=tensor([0.0034, 0.0165, 0.1719, 0.0035, 0.0739, 0.0194, 0.1689, 0.0474], device='cuda:1'), in_proj_covar=tensor([0.0085, 0.0133, 0.0169, 0.0083, 0.0153, 0.0162, 0.0177, 0.0157], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0004, 0.0002, 0.0003, 0.0004, 0.0004, 0.0003], device='cuda:1') 2023-04-27 21:36:31,485 INFO [train.py:904] (1/8) Epoch 3, batch 1600, loss[loss=0.2244, simple_loss=0.303, pruned_loss=0.07293, over 16839.00 frames. ], tot_loss[loss=0.2403, simple_loss=0.3092, pruned_loss=0.08564, over 3313908.15 frames. ], batch size: 42, lr: 2.20e-02, grad_scale: 8.0 2023-04-27 21:36:46,201 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=21911.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:37:07,919 INFO [optim.py:368] (1/8) 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 21:37:12,806 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.3915, 4.2412, 3.8570, 1.9965, 2.9614, 2.5141, 3.7144, 4.2465], device='cuda:1'), covar=tensor([0.0274, 0.0546, 0.0430, 0.1495, 0.0703, 0.0936, 0.0610, 0.0473], device='cuda:1'), in_proj_covar=tensor([0.0140, 0.0123, 0.0154, 0.0146, 0.0139, 0.0131, 0.0147, 0.0121], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0003, 0.0004, 0.0004, 0.0003, 0.0003, 0.0004, 0.0003], device='cuda:1') 2023-04-27 21:37:32,651 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.59 vs. limit=2.0 2023-04-27 21:37:38,900 INFO [train.py:904] (1/8) Epoch 3, batch 1650, loss[loss=0.2583, simple_loss=0.3431, pruned_loss=0.0868, over 16729.00 frames. ], tot_loss[loss=0.2428, simple_loss=0.3118, pruned_loss=0.08687, over 3317450.19 frames. ], batch size: 57, lr: 2.20e-02, grad_scale: 8.0 2023-04-27 21:37:39,882 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=21951.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:38:01,342 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=21968.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:38:03,324 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=21969.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:38:04,755 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=21970.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:38:24,206 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=21984.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:38:38,127 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.9993, 4.7334, 4.9219, 5.3466, 5.4055, 4.4997, 5.4242, 5.3609], device='cuda:1'), covar=tensor([0.0487, 0.0629, 0.1186, 0.0369, 0.0330, 0.0486, 0.0327, 0.0306], device='cuda:1'), in_proj_covar=tensor([0.0300, 0.0355, 0.0490, 0.0368, 0.0274, 0.0254, 0.0280, 0.0292], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-27 21:38:50,891 INFO [train.py:904] (1/8) Epoch 3, batch 1700, loss[loss=0.2474, simple_loss=0.3173, pruned_loss=0.08873, over 17140.00 frames. ], tot_loss[loss=0.2446, simple_loss=0.3144, pruned_loss=0.08745, over 3323592.93 frames. ], batch size: 46, lr: 2.19e-02, grad_scale: 4.0 2023-04-27 21:39:11,657 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=22016.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:39:12,015 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.26 vs. limit=2.0 2023-04-27 21:39:23,658 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.7203, 3.5587, 3.5298, 4.0300, 4.0283, 3.7030, 3.7935, 4.0210], device='cuda:1'), covar=tensor([0.0590, 0.0676, 0.1439, 0.0571, 0.0511, 0.1081, 0.1036, 0.0525], device='cuda:1'), in_proj_covar=tensor([0.0299, 0.0354, 0.0485, 0.0365, 0.0273, 0.0252, 0.0280, 0.0290], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-27 21:39:28,008 INFO [optim.py:368] (1/8) 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 21:39:29,692 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([1.7944, 1.9703, 1.5299, 1.8407, 2.7861, 2.6487, 2.9353, 2.8016], device='cuda:1'), covar=tensor([0.0052, 0.0123, 0.0153, 0.0151, 0.0054, 0.0086, 0.0052, 0.0053], device='cuda:1'), in_proj_covar=tensor([0.0060, 0.0116, 0.0113, 0.0116, 0.0106, 0.0116, 0.0073, 0.0089], device='cuda:1'), out_proj_covar=tensor([8.6008e-05, 1.7196e-04, 1.6218e-04, 1.7078e-04, 1.6127e-04, 1.7586e-04, 1.0805e-04, 1.3721e-04], device='cuda:1') 2023-04-27 21:39:30,837 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=22031.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:39:33,271 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=22032.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:39:57,289 INFO [train.py:904] (1/8) Epoch 3, batch 1750, loss[loss=0.2804, simple_loss=0.3337, pruned_loss=0.1135, over 16258.00 frames. ], tot_loss[loss=0.2465, simple_loss=0.3158, pruned_loss=0.08858, over 3322309.93 frames. ], batch size: 165, lr: 2.19e-02, grad_scale: 4.0 2023-04-27 21:40:26,334 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.4531, 4.3717, 4.2226, 1.7645, 4.4173, 4.4194, 3.5684, 3.4818], device='cuda:1'), covar=tensor([0.0813, 0.0074, 0.0158, 0.1379, 0.0057, 0.0049, 0.0227, 0.0304], device='cuda:1'), in_proj_covar=tensor([0.0137, 0.0082, 0.0082, 0.0147, 0.0075, 0.0075, 0.0112, 0.0125], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003, 0.0003], device='cuda:1') 2023-04-27 21:41:06,920 INFO [train.py:904] (1/8) Epoch 3, batch 1800, loss[loss=0.2798, simple_loss=0.3496, pruned_loss=0.105, over 15667.00 frames. ], tot_loss[loss=0.2473, simple_loss=0.3173, pruned_loss=0.08869, over 3319449.54 frames. ], batch size: 191, lr: 2.19e-02, grad_scale: 4.0 2023-04-27 21:41:30,481 INFO [zipformer.py:625] (1/8) 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] (1/8) 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 21:41:45,818 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=22130.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:42:14,095 INFO [train.py:904] (1/8) Epoch 3, batch 1850, loss[loss=0.2409, simple_loss=0.3091, pruned_loss=0.08632, over 16321.00 frames. ], tot_loss[loss=0.2479, simple_loss=0.3184, pruned_loss=0.08871, over 3331016.98 frames. ], batch size: 165, lr: 2.19e-02, grad_scale: 4.0 2023-04-27 21:42:47,539 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=22176.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:42:51,830 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.4231, 3.7784, 3.8429, 1.5925, 3.9134, 3.9195, 3.3217, 3.0912], device='cuda:1'), covar=tensor([0.0728, 0.0096, 0.0122, 0.1251, 0.0070, 0.0053, 0.0275, 0.0320], device='cuda:1'), in_proj_covar=tensor([0.0136, 0.0081, 0.0081, 0.0147, 0.0075, 0.0075, 0.0113, 0.0126], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003, 0.0003], device='cuda:1') 2023-04-27 21:43:09,361 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=22191.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:43:11,891 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=3.51 vs. limit=5.0 2023-04-27 21:43:22,338 INFO [train.py:904] (1/8) Epoch 3, batch 1900, loss[loss=0.2437, simple_loss=0.3078, pruned_loss=0.08978, over 16410.00 frames. ], tot_loss[loss=0.2458, simple_loss=0.3169, pruned_loss=0.08734, over 3336165.45 frames. ], batch size: 146, lr: 2.18e-02, grad_scale: 4.0 2023-04-27 21:43:31,011 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=22206.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:43:42,679 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=22215.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:44:02,302 INFO [optim.py:368] (1/8) 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 21:44:12,210 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=22237.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:44:12,254 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=22237.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:44:32,423 INFO [train.py:904] (1/8) Epoch 3, batch 1950, loss[loss=0.2898, simple_loss=0.3597, pruned_loss=0.11, over 15631.00 frames. ], tot_loss[loss=0.2459, simple_loss=0.3172, pruned_loss=0.08731, over 3331987.74 frames. ], batch size: 191, lr: 2.18e-02, grad_scale: 4.0 2023-04-27 21:44:32,726 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=22251.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:44:43,384 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.6149, 3.4170, 2.9127, 1.7945, 2.5163, 2.0773, 3.1397, 3.4795], device='cuda:1'), covar=tensor([0.0279, 0.0514, 0.0492, 0.1469, 0.0720, 0.0957, 0.0608, 0.0439], device='cuda:1'), in_proj_covar=tensor([0.0140, 0.0121, 0.0154, 0.0146, 0.0138, 0.0131, 0.0146, 0.0119], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0003, 0.0004, 0.0004, 0.0003, 0.0003, 0.0004, 0.0003], device='cuda:1') 2023-04-27 21:44:47,567 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.9810, 4.4820, 3.5095, 2.6815, 3.1891, 2.3537, 4.7527, 5.0557], device='cuda:1'), covar=tensor([0.1727, 0.0456, 0.0942, 0.0885, 0.2091, 0.1176, 0.0229, 0.0209], device='cuda:1'), in_proj_covar=tensor([0.0257, 0.0234, 0.0249, 0.0200, 0.0275, 0.0188, 0.0209, 0.0198], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0003, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-27 21:44:58,120 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=22269.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:45:06,088 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=22276.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:45:21,882 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.0132, 4.0825, 3.8601, 4.0598, 3.2438, 4.0438, 3.8511, 3.7096], device='cuda:1'), covar=tensor([0.0555, 0.0329, 0.0365, 0.0233, 0.1375, 0.0296, 0.0743, 0.0391], device='cuda:1'), in_proj_covar=tensor([0.0157, 0.0132, 0.0191, 0.0159, 0.0227, 0.0172, 0.0140, 0.0183], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:1') 2023-04-27 21:45:35,321 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=22298.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:45:36,245 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 3, batch 2000, loss[loss=0.2132, simple_loss=0.2945, pruned_loss=0.06597, over 17166.00 frames. ], tot_loss[loss=0.2461, simple_loss=0.317, pruned_loss=0.08761, over 3325210.36 frames. ], batch size: 46, lr: 2.18e-02, grad_scale: 8.0 2023-04-27 21:46:01,054 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=22317.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:46:13,068 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=22326.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:46:17,542 INFO [optim.py:368] (1/8) 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 21:46:33,257 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.4179, 3.0283, 2.5344, 2.3701, 2.3134, 2.0819, 3.0488, 3.1115], device='cuda:1'), covar=tensor([0.1243, 0.0496, 0.0819, 0.0740, 0.1525, 0.1084, 0.0318, 0.0256], device='cuda:1'), in_proj_covar=tensor([0.0255, 0.0233, 0.0245, 0.0200, 0.0273, 0.0184, 0.0208, 0.0197], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-27 21:46:34,533 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.51 vs. limit=2.0 2023-04-27 21:46:46,961 INFO [train.py:904] (1/8) Epoch 3, batch 2050, loss[loss=0.2935, simple_loss=0.3446, pruned_loss=0.1212, over 16287.00 frames. ], tot_loss[loss=0.2461, simple_loss=0.3163, pruned_loss=0.08799, over 3318109.85 frames. ], batch size: 145, lr: 2.18e-02, grad_scale: 8.0 2023-04-27 21:47:14,198 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=4.61 vs. limit=5.0 2023-04-27 21:47:33,597 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 3, batch 2100, loss[loss=0.2502, simple_loss=0.3305, pruned_loss=0.08495, over 17029.00 frames. ], tot_loss[loss=0.2461, simple_loss=0.3158, pruned_loss=0.08822, over 3325435.72 frames. ], batch size: 53, lr: 2.17e-02, grad_scale: 8.0 2023-04-27 21:48:20,686 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=22419.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:48:31,785 INFO [optim.py:368] (1/8) 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 21:48:55,018 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 3, batch 2150, loss[loss=0.2963, simple_loss=0.3448, pruned_loss=0.1239, over 15556.00 frames. ], tot_loss[loss=0.247, simple_loss=0.3168, pruned_loss=0.08855, over 3318729.90 frames. ], batch size: 191, lr: 2.17e-02, grad_scale: 8.0 2023-04-27 21:49:22,488 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=22467.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:49:47,663 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 3, batch 2200, loss[loss=0.2199, simple_loss=0.2971, pruned_loss=0.07138, over 17193.00 frames. ], tot_loss[loss=0.2489, simple_loss=0.3182, pruned_loss=0.08975, over 3315975.80 frames. ], batch size: 44, lr: 2.17e-02, grad_scale: 4.0 2023-04-27 21:50:14,274 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=22506.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:50:20,114 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.5889, 3.8373, 4.0603, 4.0612, 4.0293, 3.7100, 3.3298, 3.7675], device='cuda:1'), covar=tensor([0.0455, 0.0421, 0.0426, 0.0537, 0.0638, 0.0505, 0.1330, 0.0517], device='cuda:1'), in_proj_covar=tensor([0.0188, 0.0180, 0.0197, 0.0195, 0.0237, 0.0198, 0.0301, 0.0182], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003], device='cuda:1') 2023-04-27 21:50:46,034 INFO [optim.py:368] (1/8) 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 21:50:48,906 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=22532.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:51:15,029 INFO [train.py:904] (1/8) Epoch 3, batch 2250, loss[loss=0.2516, simple_loss=0.3286, pruned_loss=0.08727, over 17064.00 frames. ], tot_loss[loss=0.2503, simple_loss=0.319, pruned_loss=0.09084, over 3315851.72 frames. ], batch size: 53, lr: 2.17e-02, grad_scale: 4.0 2023-04-27 21:51:18,565 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=22554.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:51:42,419 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=22571.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:52:06,521 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.2031, 4.0694, 4.2681, 4.5710, 4.5758, 4.1259, 4.4450, 4.5500], device='cuda:1'), covar=tensor([0.0547, 0.0530, 0.1098, 0.0392, 0.0395, 0.0634, 0.0689, 0.0306], device='cuda:1'), in_proj_covar=tensor([0.0298, 0.0355, 0.0478, 0.0365, 0.0274, 0.0256, 0.0277, 0.0290], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-27 21:52:09,944 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=22593.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:52:13,364 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([5.0008, 5.5895, 5.4849, 5.5579, 5.5202, 5.9916, 5.8050, 5.5120], device='cuda:1'), covar=tensor([0.0549, 0.1140, 0.1096, 0.1372, 0.2098, 0.0795, 0.0845, 0.2046], device='cuda:1'), in_proj_covar=tensor([0.0229, 0.0333, 0.0309, 0.0286, 0.0375, 0.0322, 0.0261, 0.0384], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-04-27 21:52:20,009 INFO [train.py:904] (1/8) Epoch 3, batch 2300, loss[loss=0.2678, simple_loss=0.3272, pruned_loss=0.1042, over 16413.00 frames. ], tot_loss[loss=0.2501, simple_loss=0.3192, pruned_loss=0.09052, over 3320062.05 frames. ], batch size: 75, lr: 2.17e-02, grad_scale: 4.0 2023-04-27 21:52:57,379 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=22626.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:53:01,630 INFO [optim.py:368] (1/8) Clipping_scale=2.0, grad-norm quartiles 2.100e+02 3.931e+02 4.776e+02 5.724e+02 1.077e+03, threshold=9.553e+02, percent-clipped=1.0 2023-04-27 21:53:12,011 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 3, batch 2350, loss[loss=0.2332, simple_loss=0.3134, pruned_loss=0.07653, over 17039.00 frames. ], tot_loss[loss=0.2502, simple_loss=0.3192, pruned_loss=0.09054, over 3323479.09 frames. ], batch size: 53, lr: 2.16e-02, grad_scale: 4.0 2023-04-27 21:54:01,191 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=22674.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:54:31,361 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.3063, 4.2443, 4.1504, 4.2682, 3.7472, 4.2407, 3.9592, 3.9393], device='cuda:1'), covar=tensor([0.0327, 0.0181, 0.0163, 0.0113, 0.0766, 0.0176, 0.0478, 0.0254], device='cuda:1'), in_proj_covar=tensor([0.0152, 0.0128, 0.0189, 0.0156, 0.0220, 0.0167, 0.0137, 0.0181], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-04-27 21:54:35,030 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=22700.0, num_to_drop=1, layers_to_drop={1} 2023-04-27 21:54:35,721 INFO [train.py:904] (1/8) Epoch 3, batch 2400, loss[loss=0.2638, simple_loss=0.3295, pruned_loss=0.099, over 16821.00 frames. ], tot_loss[loss=0.2499, simple_loss=0.3195, pruned_loss=0.09008, over 3326399.73 frames. ], batch size: 96, lr: 2.16e-02, grad_scale: 8.0 2023-04-27 21:55:17,391 INFO [optim.py:368] (1/8) 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 21:55:32,310 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 3, batch 2450, loss[loss=0.2356, simple_loss=0.2974, pruned_loss=0.08687, over 16758.00 frames. ], tot_loss[loss=0.2506, simple_loss=0.3204, pruned_loss=0.09037, over 3314009.20 frames. ], batch size: 89, lr: 2.16e-02, grad_scale: 8.0 2023-04-27 21:55:55,195 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([1.9839, 3.5320, 2.7363, 4.7506, 4.5899, 4.2926, 1.9314, 3.0920], device='cuda:1'), covar=tensor([0.1322, 0.0430, 0.1109, 0.0068, 0.0197, 0.0282, 0.1195, 0.0665], device='cuda:1'), in_proj_covar=tensor([0.0138, 0.0127, 0.0158, 0.0075, 0.0138, 0.0134, 0.0148, 0.0149], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:1') 2023-04-27 21:56:33,253 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=22786.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:56:46,230 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=22794.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:56:55,062 INFO [train.py:904] (1/8) Epoch 3, batch 2500, loss[loss=0.2563, simple_loss=0.3109, pruned_loss=0.1008, over 16711.00 frames. ], tot_loss[loss=0.2492, simple_loss=0.3195, pruned_loss=0.08945, over 3316691.53 frames. ], batch size: 124, lr: 2.16e-02, grad_scale: 8.0 2023-04-27 21:57:27,649 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=22825.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:57:33,674 INFO [optim.py:368] (1/8) Clipping_scale=2.0, grad-norm quartiles 2.850e+02 4.203e+02 4.842e+02 6.402e+02 1.699e+03, threshold=9.683e+02, percent-clipped=7.0 2023-04-27 21:57:37,019 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=22832.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:57:41,074 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=22834.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:58:03,439 INFO [train.py:904] (1/8) Epoch 3, batch 2550, loss[loss=0.25, simple_loss=0.3385, pruned_loss=0.08075, over 17105.00 frames. ], tot_loss[loss=0.2487, simple_loss=0.319, pruned_loss=0.08917, over 3326894.75 frames. ], batch size: 49, lr: 2.15e-02, grad_scale: 8.0 2023-04-27 21:58:10,261 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=22855.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:58:31,370 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=22871.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:58:44,554 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=22880.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:58:52,536 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=22886.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:59:01,449 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=22893.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:59:12,357 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.5423, 4.7132, 4.5082, 4.7845, 4.5627, 5.2321, 4.9481, 4.6231], device='cuda:1'), covar=tensor([0.1075, 0.1191, 0.1364, 0.1665, 0.2652, 0.0869, 0.0911, 0.2260], device='cuda:1'), in_proj_covar=tensor([0.0229, 0.0328, 0.0306, 0.0288, 0.0374, 0.0320, 0.0255, 0.0380], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-04-27 21:59:13,834 INFO [train.py:904] (1/8) Epoch 3, batch 2600, loss[loss=0.2623, simple_loss=0.3258, pruned_loss=0.09938, over 16853.00 frames. ], tot_loss[loss=0.2476, simple_loss=0.3186, pruned_loss=0.08832, over 3325916.55 frames. ], batch size: 116, lr: 2.15e-02, grad_scale: 8.0 2023-04-27 21:59:17,037 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=4.35 vs. limit=5.0 2023-04-27 21:59:38,826 INFO [zipformer.py:625] (1/8) 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] (1/8) 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 22:00:09,150 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=22941.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 22:00:22,556 INFO [train.py:904] (1/8) Epoch 3, batch 2650, loss[loss=0.3083, simple_loss=0.3597, pruned_loss=0.1285, over 16823.00 frames. ], tot_loss[loss=0.2482, simple_loss=0.3195, pruned_loss=0.08844, over 3332145.06 frames. ], batch size: 124, lr: 2.15e-02, grad_scale: 8.0 2023-04-27 22:00:54,754 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.6155, 4.4115, 1.5651, 4.5639, 2.5247, 4.5296, 2.1569, 3.1531], device='cuda:1'), covar=tensor([0.0029, 0.0139, 0.1561, 0.0028, 0.0770, 0.0218, 0.1215, 0.0507], device='cuda:1'), in_proj_covar=tensor([0.0090, 0.0137, 0.0172, 0.0086, 0.0160, 0.0170, 0.0178, 0.0161], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-04-27 22:01:22,673 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=22995.0, num_to_drop=1, layers_to_drop={0} 2023-04-27 22:01:30,282 INFO [train.py:904] (1/8) Epoch 3, batch 2700, loss[loss=0.2682, simple_loss=0.3473, pruned_loss=0.09452, over 17206.00 frames. ], tot_loss[loss=0.2478, simple_loss=0.3202, pruned_loss=0.08773, over 3334183.82 frames. ], batch size: 44, lr: 2.15e-02, grad_scale: 8.0 2023-04-27 22:01:58,252 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.89 vs. limit=2.0 2023-04-27 22:01:59,346 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.30 vs. limit=2.0 2023-04-27 22:02:09,771 INFO [optim.py:368] (1/8) 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 22:02:24,005 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 3, batch 2750, loss[loss=0.2495, simple_loss=0.3331, pruned_loss=0.08297, over 16745.00 frames. ], tot_loss[loss=0.246, simple_loss=0.3189, pruned_loss=0.0865, over 3337584.95 frames. ], batch size: 62, lr: 2.15e-02, grad_scale: 8.0 2023-04-27 22:02:52,033 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.44 vs. limit=2.0 2023-04-27 22:03:28,907 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=23089.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 22:03:45,517 INFO [train.py:904] (1/8) Epoch 3, batch 2800, loss[loss=0.224, simple_loss=0.3024, pruned_loss=0.07278, over 17236.00 frames. ], tot_loss[loss=0.2459, simple_loss=0.3187, pruned_loss=0.08654, over 3344341.62 frames. ], batch size: 45, lr: 2.14e-02, grad_scale: 8.0 2023-04-27 22:03:54,907 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=23108.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 22:04:25,542 INFO [optim.py:368] (1/8) 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 22:04:31,617 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.64 vs. limit=2.0 2023-04-27 22:04:54,741 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=23150.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 22:04:55,605 INFO [train.py:904] (1/8) Epoch 3, batch 2850, loss[loss=0.2485, simple_loss=0.3234, pruned_loss=0.08679, over 17236.00 frames. ], tot_loss[loss=0.2446, simple_loss=0.3181, pruned_loss=0.08551, over 3342940.29 frames. ], batch size: 45, lr: 2.14e-02, grad_scale: 8.0 2023-04-27 22:05:20,406 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=23169.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 22:05:36,684 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=23181.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 22:05:56,748 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.8786, 4.0860, 3.1717, 2.7067, 2.9956, 2.2187, 3.9328, 4.3610], device='cuda:1'), covar=tensor([0.1642, 0.0475, 0.0940, 0.0943, 0.1732, 0.1337, 0.0352, 0.0404], device='cuda:1'), in_proj_covar=tensor([0.0258, 0.0240, 0.0246, 0.0201, 0.0278, 0.0187, 0.0211, 0.0206], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-27 22:06:03,310 INFO [train.py:904] (1/8) Epoch 3, batch 2900, loss[loss=0.2266, simple_loss=0.3044, pruned_loss=0.07442, over 17198.00 frames. ], tot_loss[loss=0.2463, simple_loss=0.3174, pruned_loss=0.0876, over 3337497.81 frames. ], batch size: 46, lr: 2.14e-02, grad_scale: 8.0 2023-04-27 22:06:24,534 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.48 vs. limit=2.0 2023-04-27 22:06:25,813 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=23217.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 22:06:42,357 INFO [zipformer.py:625] (1/8) 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] (1/8) 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] (1/8) Epoch 3, batch 2950, loss[loss=0.2401, simple_loss=0.3275, pruned_loss=0.07638, over 17145.00 frames. ], tot_loss[loss=0.2455, simple_loss=0.3161, pruned_loss=0.08744, over 3334481.91 frames. ], batch size: 49, lr: 2.14e-02, grad_scale: 8.0 2023-04-27 22:07:49,227 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=23278.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 22:08:07,127 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=23290.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 22:08:12,715 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 3, batch 3000, loss[loss=0.222, simple_loss=0.3105, pruned_loss=0.0668, over 17196.00 frames. ], tot_loss[loss=0.2458, simple_loss=0.3162, pruned_loss=0.08769, over 3333228.07 frames. ], batch size: 45, lr: 2.13e-02, grad_scale: 8.0 2023-04-27 22:08:19,857 INFO [train.py:929] (1/8) Computing validation loss 2023-04-27 22:08:30,493 INFO [train.py:938] (1/8) Epoch 3, validation: loss=0.1721, simple_loss=0.279, pruned_loss=0.03262, over 944034.00 frames. 2023-04-27 22:08:30,494 INFO [train.py:939] (1/8) Maximum memory allocated so far is 17676MB 2023-04-27 22:08:42,027 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.3954, 4.3131, 1.6075, 4.2646, 2.5668, 4.3401, 1.9017, 3.0645], device='cuda:1'), covar=tensor([0.0034, 0.0136, 0.1511, 0.0044, 0.0752, 0.0237, 0.1333, 0.0545], device='cuda:1'), in_proj_covar=tensor([0.0092, 0.0138, 0.0173, 0.0088, 0.0162, 0.0174, 0.0185, 0.0163], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-04-27 22:09:10,133 INFO [optim.py:368] (1/8) 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] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=23343.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 22:09:37,531 INFO [train.py:904] (1/8) Epoch 3, batch 3050, loss[loss=0.263, simple_loss=0.3355, pruned_loss=0.09527, over 17034.00 frames. ], tot_loss[loss=0.2461, simple_loss=0.316, pruned_loss=0.0881, over 3335787.00 frames. ], batch size: 53, lr: 2.13e-02, grad_scale: 4.0 2023-04-27 22:09:40,306 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.0620, 3.8146, 4.0162, 4.2791, 4.2806, 3.8516, 4.0397, 4.3009], device='cuda:1'), covar=tensor([0.0517, 0.0531, 0.0851, 0.0361, 0.0403, 0.1073, 0.0785, 0.0293], device='cuda:1'), in_proj_covar=tensor([0.0303, 0.0372, 0.0487, 0.0376, 0.0279, 0.0266, 0.0293, 0.0302], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-27 22:09:54,881 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.42 vs. limit=2.0 2023-04-27 22:10:44,560 INFO [train.py:904] (1/8) Epoch 3, batch 3100, loss[loss=0.211, simple_loss=0.2945, pruned_loss=0.06374, over 17183.00 frames. ], tot_loss[loss=0.2464, simple_loss=0.3157, pruned_loss=0.08856, over 3318274.33 frames. ], batch size: 46, lr: 2.13e-02, grad_scale: 4.0 2023-04-27 22:11:28,255 INFO [optim.py:368] (1/8) Clipping_scale=2.0, grad-norm quartiles 2.500e+02 3.880e+02 4.394e+02 5.487e+02 1.419e+03, threshold=8.788e+02, percent-clipped=5.0 2023-04-27 22:11:53,212 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 3, batch 3150, loss[loss=0.2648, simple_loss=0.3291, pruned_loss=0.1003, over 16768.00 frames. ], tot_loss[loss=0.2454, simple_loss=0.3148, pruned_loss=0.08805, over 3308638.88 frames. ], batch size: 83, lr: 2.13e-02, grad_scale: 4.0 2023-04-27 22:12:12,269 INFO [zipformer.py:625] (1/8) 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] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=23481.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 22:12:44,065 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=23488.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 22:12:52,881 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.7142, 4.5856, 4.5795, 3.3617, 4.4133, 4.6169, 4.5190, 2.5649], device='cuda:1'), covar=tensor([0.0293, 0.0015, 0.0028, 0.0163, 0.0018, 0.0018, 0.0016, 0.0261], device='cuda:1'), in_proj_covar=tensor([0.0110, 0.0053, 0.0059, 0.0106, 0.0053, 0.0060, 0.0059, 0.0102], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0001, 0.0002, 0.0002, 0.0003], device='cuda:1') 2023-04-27 22:12:58,215 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 3, batch 3200, loss[loss=0.2115, simple_loss=0.2945, pruned_loss=0.06427, over 16967.00 frames. ], tot_loss[loss=0.2431, simple_loss=0.3133, pruned_loss=0.08643, over 3312192.65 frames. ], batch size: 41, lr: 2.13e-02, grad_scale: 8.0 2023-04-27 22:13:39,247 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=23529.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 22:13:42,225 INFO [optim.py:368] (1/8) 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 22:14:06,450 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=23549.0, num_to_drop=1, layers_to_drop={3} 2023-04-27 22:14:08,943 INFO [train.py:904] (1/8) Epoch 3, batch 3250, loss[loss=0.1989, simple_loss=0.2781, pruned_loss=0.05988, over 17180.00 frames. ], tot_loss[loss=0.2433, simple_loss=0.3133, pruned_loss=0.0867, over 3305060.04 frames. ], batch size: 40, lr: 2.12e-02, grad_scale: 8.0 2023-04-27 22:14:38,968 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=23573.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 22:14:56,591 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=23585.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 22:15:18,373 INFO [train.py:904] (1/8) Epoch 3, batch 3300, loss[loss=0.1851, simple_loss=0.2663, pruned_loss=0.05193, over 16989.00 frames. ], tot_loss[loss=0.2435, simple_loss=0.3138, pruned_loss=0.08665, over 3310517.34 frames. ], batch size: 41, lr: 2.12e-02, grad_scale: 8.0 2023-04-27 22:15:57,196 INFO [optim.py:368] (1/8) Clipping_scale=2.0, grad-norm quartiles 2.549e+02 3.983e+02 4.910e+02 5.801e+02 1.123e+03, threshold=9.819e+02, percent-clipped=5.0 2023-04-27 22:16:24,693 INFO [train.py:904] (1/8) Epoch 3, batch 3350, loss[loss=0.2671, simple_loss=0.3228, pruned_loss=0.1057, over 16938.00 frames. ], tot_loss[loss=0.2442, simple_loss=0.3145, pruned_loss=0.08689, over 3320930.99 frames. ], batch size: 109, lr: 2.12e-02, grad_scale: 8.0 2023-04-27 22:17:33,462 INFO [train.py:904] (1/8) Epoch 3, batch 3400, loss[loss=0.2679, simple_loss=0.3289, pruned_loss=0.1034, over 16378.00 frames. ], tot_loss[loss=0.2438, simple_loss=0.3148, pruned_loss=0.0864, over 3308880.54 frames. ], batch size: 146, lr: 2.12e-02, grad_scale: 8.0 2023-04-27 22:18:13,370 INFO [optim.py:368] (1/8) 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 22:18:23,619 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.63 vs. limit=2.0 2023-04-27 22:18:40,209 INFO [train.py:904] (1/8) Epoch 3, batch 3450, loss[loss=0.2638, simple_loss=0.3104, pruned_loss=0.1086, over 16862.00 frames. ], tot_loss[loss=0.2433, simple_loss=0.3142, pruned_loss=0.08619, over 3304761.24 frames. ], batch size: 96, lr: 2.11e-02, grad_scale: 8.0 2023-04-27 22:18:58,738 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=23764.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 22:19:47,201 INFO [train.py:904] (1/8) Epoch 3, batch 3500, loss[loss=0.2359, simple_loss=0.322, pruned_loss=0.07488, over 17268.00 frames. ], tot_loss[loss=0.241, simple_loss=0.3126, pruned_loss=0.08471, over 3310033.02 frames. ], batch size: 52, lr: 2.11e-02, grad_scale: 8.0 2023-04-27 22:20:04,593 INFO [zipformer.py:625] (1/8) 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] (1/8) 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 22:20:48,796 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=23844.0, num_to_drop=1, layers_to_drop={0} 2023-04-27 22:20:51,798 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.0802, 3.9315, 4.0475, 4.3771, 4.3864, 3.9635, 4.1350, 4.3612], device='cuda:1'), covar=tensor([0.0580, 0.0509, 0.1049, 0.0371, 0.0363, 0.0818, 0.0810, 0.0340], device='cuda:1'), in_proj_covar=tensor([0.0311, 0.0378, 0.0501, 0.0392, 0.0284, 0.0273, 0.0301, 0.0301], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-27 22:20:59,079 INFO [train.py:904] (1/8) Epoch 3, batch 3550, loss[loss=0.2395, simple_loss=0.3235, pruned_loss=0.07778, over 17065.00 frames. ], tot_loss[loss=0.2398, simple_loss=0.3113, pruned_loss=0.08412, over 3323721.20 frames. ], batch size: 53, lr: 2.11e-02, grad_scale: 8.0 2023-04-27 22:21:26,509 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.5717, 2.4107, 2.2339, 2.2204, 2.9028, 2.6949, 3.8673, 3.1853], device='cuda:1'), covar=tensor([0.0020, 0.0137, 0.0150, 0.0179, 0.0083, 0.0133, 0.0036, 0.0077], device='cuda:1'), in_proj_covar=tensor([0.0063, 0.0119, 0.0116, 0.0118, 0.0112, 0.0121, 0.0081, 0.0098], device='cuda:1'), out_proj_covar=tensor([9.2691e-05, 1.7263e-04, 1.6416e-04, 1.7142e-04, 1.6694e-04, 1.7925e-04, 1.2046e-04, 1.5006e-04], device='cuda:1') 2023-04-27 22:21:29,404 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=23873.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 22:21:31,750 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.9727, 4.6639, 4.3460, 5.1744, 5.2469, 4.6210, 5.3206, 5.1423], device='cuda:1'), covar=tensor([0.0677, 0.0759, 0.2409, 0.0677, 0.0624, 0.0592, 0.0530, 0.0609], device='cuda:1'), in_proj_covar=tensor([0.0316, 0.0383, 0.0512, 0.0401, 0.0290, 0.0275, 0.0304, 0.0305], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-27 22:21:45,158 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 3, batch 3600, loss[loss=0.2365, simple_loss=0.3205, pruned_loss=0.07626, over 17121.00 frames. ], tot_loss[loss=0.2385, simple_loss=0.3097, pruned_loss=0.08361, over 3317563.68 frames. ], batch size: 53, lr: 2.11e-02, grad_scale: 8.0 2023-04-27 22:22:11,049 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.6521, 3.3952, 2.5747, 2.3409, 2.5564, 1.8993, 3.3928, 3.7830], device='cuda:1'), covar=tensor([0.1499, 0.0533, 0.0984, 0.0966, 0.1717, 0.1445, 0.0361, 0.0417], device='cuda:1'), in_proj_covar=tensor([0.0259, 0.0243, 0.0250, 0.0205, 0.0281, 0.0189, 0.0214, 0.0212], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-27 22:22:33,415 INFO [zipformer.py:625] (1/8) 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] (1/8) 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] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=23933.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 22:23:14,443 INFO [train.py:904] (1/8) Epoch 3, batch 3650, loss[loss=0.2488, simple_loss=0.3064, pruned_loss=0.09563, over 16310.00 frames. ], tot_loss[loss=0.2386, simple_loss=0.3083, pruned_loss=0.08448, over 3304994.69 frames. ], batch size: 165, lr: 2.11e-02, grad_scale: 8.0 2023-04-27 22:23:29,153 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.2404, 4.2286, 1.9510, 4.1792, 2.8097, 4.3566, 2.0411, 3.1594], device='cuda:1'), covar=tensor([0.0054, 0.0177, 0.1770, 0.0056, 0.0819, 0.0244, 0.1451, 0.0701], device='cuda:1'), in_proj_covar=tensor([0.0090, 0.0141, 0.0172, 0.0084, 0.0163, 0.0173, 0.0182, 0.0162], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-04-27 22:23:51,105 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=2.03 vs. limit=2.0 2023-04-27 22:23:53,484 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=2.05 vs. limit=2.0 2023-04-27 22:24:29,887 INFO [train.py:904] (1/8) Epoch 3, batch 3700, loss[loss=0.2934, simple_loss=0.3405, pruned_loss=0.1231, over 11485.00 frames. ], tot_loss[loss=0.2406, simple_loss=0.3077, pruned_loss=0.08677, over 3254321.55 frames. ], batch size: 248, lr: 2.10e-02, grad_scale: 8.0 2023-04-27 22:24:41,162 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.9064, 3.7619, 3.9250, 4.1486, 4.1524, 3.7855, 3.9210, 4.1855], device='cuda:1'), covar=tensor([0.0580, 0.0522, 0.0934, 0.0413, 0.0467, 0.0946, 0.0902, 0.0351], device='cuda:1'), in_proj_covar=tensor([0.0289, 0.0354, 0.0464, 0.0366, 0.0271, 0.0254, 0.0285, 0.0289], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-27 22:25:13,795 INFO [optim.py:368] (1/8) 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 22:25:22,519 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=24037.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 22:25:24,057 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.8040, 4.6564, 4.2582, 1.8254, 3.4721, 2.4417, 4.0588, 4.7634], device='cuda:1'), covar=tensor([0.0189, 0.0320, 0.0322, 0.1725, 0.0570, 0.0879, 0.0612, 0.0341], device='cuda:1'), in_proj_covar=tensor([0.0139, 0.0122, 0.0154, 0.0144, 0.0136, 0.0128, 0.0146, 0.0122], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0003, 0.0004, 0.0004, 0.0003, 0.0003, 0.0004, 0.0003], device='cuda:1') 2023-04-27 22:25:42,967 INFO [train.py:904] (1/8) Epoch 3, batch 3750, loss[loss=0.2569, simple_loss=0.3174, pruned_loss=0.09823, over 16287.00 frames. ], tot_loss[loss=0.2432, simple_loss=0.309, pruned_loss=0.08869, over 3228739.46 frames. ], batch size: 165, lr: 2.10e-02, grad_scale: 8.0 2023-04-27 22:26:22,697 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([1.8463, 3.0558, 2.3805, 4.0191, 3.9219, 3.9612, 1.5495, 2.7503], device='cuda:1'), covar=tensor([0.1376, 0.0422, 0.1113, 0.0090, 0.0205, 0.0240, 0.1308, 0.0696], device='cuda:1'), in_proj_covar=tensor([0.0142, 0.0131, 0.0164, 0.0076, 0.0146, 0.0140, 0.0155, 0.0154], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:1') 2023-04-27 22:26:24,691 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.85 vs. limit=2.0 2023-04-27 22:26:33,985 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.5151, 1.4456, 2.1170, 2.4377, 2.5866, 2.5605, 1.7299, 2.5478], device='cuda:1'), covar=tensor([0.0035, 0.0183, 0.0102, 0.0067, 0.0042, 0.0051, 0.0136, 0.0034], device='cuda:1'), in_proj_covar=tensor([0.0086, 0.0123, 0.0109, 0.0104, 0.0086, 0.0069, 0.0112, 0.0066], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0001], device='cuda:1') 2023-04-27 22:26:46,876 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=24098.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 22:26:50,660 INFO [train.py:904] (1/8) Epoch 3, batch 3800, loss[loss=0.2423, simple_loss=0.3041, pruned_loss=0.09023, over 16356.00 frames. ], tot_loss[loss=0.2449, simple_loss=0.3099, pruned_loss=0.08994, over 3245611.58 frames. ], batch size: 165, lr: 2.10e-02, grad_scale: 8.0 2023-04-27 22:27:15,474 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.3602, 4.1595, 4.3392, 4.6729, 4.6734, 4.1541, 4.4401, 4.6532], device='cuda:1'), covar=tensor([0.0510, 0.0546, 0.0932, 0.0318, 0.0323, 0.0624, 0.0689, 0.0305], device='cuda:1'), in_proj_covar=tensor([0.0287, 0.0353, 0.0462, 0.0362, 0.0268, 0.0252, 0.0287, 0.0287], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-27 22:27:26,925 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.33 vs. limit=2.0 2023-04-27 22:27:34,020 INFO [optim.py:368] (1/8) 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 22:27:52,790 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=24144.0, num_to_drop=1, layers_to_drop={0} 2023-04-27 22:28:01,891 INFO [train.py:904] (1/8) Epoch 3, batch 3850, loss[loss=0.2351, simple_loss=0.3017, pruned_loss=0.08425, over 16730.00 frames. ], tot_loss[loss=0.2454, simple_loss=0.3098, pruned_loss=0.09051, over 3241821.35 frames. ], batch size: 83, lr: 2.10e-02, grad_scale: 8.0 2023-04-27 22:29:00,726 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 3, batch 3900, loss[loss=0.2223, simple_loss=0.2885, pruned_loss=0.07805, over 16716.00 frames. ], tot_loss[loss=0.2442, simple_loss=0.3086, pruned_loss=0.08993, over 3254015.85 frames. ], batch size: 89, lr: 2.10e-02, grad_scale: 8.0 2023-04-27 22:29:56,951 INFO [optim.py:368] (1/8) 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 22:30:11,339 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.0670, 3.9824, 1.9397, 3.9682, 2.8729, 4.0546, 2.1280, 3.0810], device='cuda:1'), covar=tensor([0.0044, 0.0194, 0.1403, 0.0047, 0.0487, 0.0283, 0.1270, 0.0518], device='cuda:1'), in_proj_covar=tensor([0.0090, 0.0137, 0.0170, 0.0082, 0.0160, 0.0167, 0.0179, 0.0160], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-04-27 22:30:25,201 INFO [train.py:904] (1/8) Epoch 3, batch 3950, loss[loss=0.2574, simple_loss=0.3117, pruned_loss=0.1015, over 16470.00 frames. ], tot_loss[loss=0.2431, simple_loss=0.307, pruned_loss=0.0896, over 3258769.58 frames. ], batch size: 146, lr: 2.09e-02, grad_scale: 8.0 2023-04-27 22:30:52,638 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.4866, 2.3117, 2.4097, 3.7928, 1.9036, 3.4455, 2.1188, 2.1529], device='cuda:1'), covar=tensor([0.0344, 0.0801, 0.0441, 0.0230, 0.1749, 0.0269, 0.1064, 0.1232], device='cuda:1'), in_proj_covar=tensor([0.0250, 0.0226, 0.0189, 0.0249, 0.0297, 0.0202, 0.0218, 0.0286], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-27 22:31:03,513 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([1.9976, 3.3553, 2.5216, 4.3573, 4.1649, 4.0182, 1.7936, 3.0192], device='cuda:1'), covar=tensor([0.1235, 0.0369, 0.1015, 0.0060, 0.0191, 0.0277, 0.1076, 0.0582], device='cuda:1'), in_proj_covar=tensor([0.0142, 0.0130, 0.0162, 0.0074, 0.0142, 0.0137, 0.0153, 0.0152], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:1') 2023-04-27 22:31:16,757 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.94 vs. limit=2.0 2023-04-27 22:31:34,849 INFO [train.py:904] (1/8) Epoch 3, batch 4000, loss[loss=0.2692, simple_loss=0.3182, pruned_loss=0.1101, over 16495.00 frames. ], tot_loss[loss=0.2433, simple_loss=0.3067, pruned_loss=0.08989, over 3265827.69 frames. ], batch size: 146, lr: 2.09e-02, grad_scale: 8.0 2023-04-27 22:31:36,335 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.5229, 3.6696, 3.9669, 3.9399, 3.8940, 3.5944, 3.5800, 3.6288], device='cuda:1'), covar=tensor([0.0293, 0.0353, 0.0310, 0.0406, 0.0376, 0.0304, 0.0785, 0.0372], device='cuda:1'), in_proj_covar=tensor([0.0198, 0.0188, 0.0202, 0.0198, 0.0239, 0.0204, 0.0307, 0.0179], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003], device='cuda:1') 2023-04-27 22:32:17,081 INFO [optim.py:368] (1/8) 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 22:32:45,200 INFO [train.py:904] (1/8) Epoch 3, batch 4050, loss[loss=0.2005, simple_loss=0.2792, pruned_loss=0.06087, over 16738.00 frames. ], tot_loss[loss=0.2396, simple_loss=0.3052, pruned_loss=0.087, over 3274485.33 frames. ], batch size: 83, lr: 2.09e-02, grad_scale: 8.0 2023-04-27 22:33:46,327 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 3, batch 4100, loss[loss=0.2816, simple_loss=0.3543, pruned_loss=0.1045, over 15403.00 frames. ], tot_loss[loss=0.2384, simple_loss=0.3058, pruned_loss=0.08555, over 3253567.86 frames. ], batch size: 191, lr: 2.09e-02, grad_scale: 8.0 2023-04-27 22:34:27,715 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.31 vs. limit=2.0 2023-04-27 22:34:42,994 INFO [optim.py:368] (1/8) 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 22:35:13,091 INFO [train.py:904] (1/8) Epoch 3, batch 4150, loss[loss=0.2684, simple_loss=0.3361, pruned_loss=0.1003, over 17062.00 frames. ], tot_loss[loss=0.2475, simple_loss=0.3149, pruned_loss=0.09003, over 3235657.43 frames. ], batch size: 53, lr: 2.09e-02, grad_scale: 8.0 2023-04-27 22:35:32,930 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=4.59 vs. limit=5.0 2023-04-27 22:36:27,612 INFO [train.py:904] (1/8) Epoch 3, batch 4200, loss[loss=0.3122, simple_loss=0.3777, pruned_loss=0.1234, over 16303.00 frames. ], tot_loss[loss=0.2543, simple_loss=0.3231, pruned_loss=0.09275, over 3217021.73 frames. ], batch size: 165, lr: 2.08e-02, grad_scale: 8.0 2023-04-27 22:37:10,966 INFO [optim.py:368] (1/8) Clipping_scale=2.0, grad-norm quartiles 2.707e+02 3.802e+02 4.415e+02 5.397e+02 1.092e+03, threshold=8.829e+02, percent-clipped=9.0 2023-04-27 22:37:35,718 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=24547.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 22:37:40,127 INFO [train.py:904] (1/8) Epoch 3, batch 4250, loss[loss=0.2574, simple_loss=0.3302, pruned_loss=0.09234, over 16789.00 frames. ], tot_loss[loss=0.2565, simple_loss=0.3265, pruned_loss=0.09321, over 3198712.89 frames. ], batch size: 39, lr: 2.08e-02, grad_scale: 8.0 2023-04-27 22:38:06,839 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=24569.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 22:38:53,670 INFO [train.py:904] (1/8) Epoch 3, batch 4300, loss[loss=0.2883, simple_loss=0.3573, pruned_loss=0.1096, over 16703.00 frames. ], tot_loss[loss=0.2546, simple_loss=0.3268, pruned_loss=0.09123, over 3185214.54 frames. ], batch size: 124, lr: 2.08e-02, grad_scale: 8.0 2023-04-27 22:38:53,942 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.6310, 5.3135, 5.2907, 5.1897, 5.1371, 5.7652, 5.4595, 5.2723], device='cuda:1'), covar=tensor([0.0676, 0.0896, 0.0716, 0.1282, 0.1907, 0.0616, 0.0834, 0.1808], device='cuda:1'), in_proj_covar=tensor([0.0226, 0.0306, 0.0281, 0.0269, 0.0348, 0.0300, 0.0248, 0.0360], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003], device='cuda:1') 2023-04-27 22:39:05,004 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=24608.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 22:39:37,552 INFO [zipformer.py:625] (1/8) 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] (1/8) 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] (1/8) Epoch 3, batch 4350, loss[loss=0.2625, simple_loss=0.3365, pruned_loss=0.09425, over 16649.00 frames. ], tot_loss[loss=0.2586, simple_loss=0.3309, pruned_loss=0.09315, over 3192599.74 frames. ], batch size: 62, lr: 2.08e-02, grad_scale: 8.0 2023-04-27 22:41:10,584 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 3, batch 4400, loss[loss=0.2575, simple_loss=0.3359, pruned_loss=0.08951, over 16381.00 frames. ], tot_loss[loss=0.2607, simple_loss=0.3331, pruned_loss=0.09415, over 3191515.25 frames. ], batch size: 68, lr: 2.08e-02, grad_scale: 8.0 2023-04-27 22:42:05,388 INFO [optim.py:368] (1/8) 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 22:42:20,526 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=24741.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 22:42:34,827 INFO [train.py:904] (1/8) Epoch 3, batch 4450, loss[loss=0.2459, simple_loss=0.3354, pruned_loss=0.07818, over 16963.00 frames. ], tot_loss[loss=0.2605, simple_loss=0.3349, pruned_loss=0.09308, over 3204438.11 frames. ], batch size: 41, lr: 2.07e-02, grad_scale: 8.0 2023-04-27 22:42:55,263 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=24765.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 22:43:27,599 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.6745, 5.1868, 5.2035, 5.0826, 5.2040, 5.7088, 5.2789, 5.0971], device='cuda:1'), covar=tensor([0.0704, 0.1109, 0.0817, 0.1456, 0.1902, 0.0694, 0.0791, 0.1819], device='cuda:1'), in_proj_covar=tensor([0.0230, 0.0310, 0.0285, 0.0274, 0.0354, 0.0307, 0.0248, 0.0367], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003], device='cuda:1') 2023-04-27 22:43:47,149 INFO [train.py:904] (1/8) Epoch 3, batch 4500, loss[loss=0.2346, simple_loss=0.3129, pruned_loss=0.07809, over 16534.00 frames. ], tot_loss[loss=0.2598, simple_loss=0.3344, pruned_loss=0.09259, over 3208778.15 frames. ], batch size: 75, lr: 2.07e-02, grad_scale: 8.0 2023-04-27 22:44:22,363 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=24826.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 22:44:29,086 INFO [optim.py:368] (1/8) 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 22:44:52,820 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 3, batch 4550, loss[loss=0.2938, simple_loss=0.3471, pruned_loss=0.1202, over 11777.00 frames. ], tot_loss[loss=0.2588, simple_loss=0.3334, pruned_loss=0.09214, over 3206406.80 frames. ], batch size: 247, lr: 2.07e-02, grad_scale: 8.0 2023-04-27 22:45:12,650 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.62 vs. limit=2.0 2023-04-27 22:46:07,316 INFO [train.py:904] (1/8) Epoch 3, batch 4600, loss[loss=0.2658, simple_loss=0.3411, pruned_loss=0.09521, over 16261.00 frames. ], tot_loss[loss=0.2591, simple_loss=0.3343, pruned_loss=0.09196, over 3214176.78 frames. ], batch size: 165, lr: 2.07e-02, grad_scale: 8.0 2023-04-27 22:46:10,655 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=24903.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 22:46:20,479 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=24909.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 22:46:44,267 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=24925.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 22:46:52,470 INFO [optim.py:368] (1/8) 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] (1/8) Epoch 3, batch 4650, loss[loss=0.252, simple_loss=0.3182, pruned_loss=0.09288, over 17259.00 frames. ], tot_loss[loss=0.2582, simple_loss=0.333, pruned_loss=0.09174, over 3226067.02 frames. ], batch size: 52, lr: 2.07e-02, grad_scale: 8.0 2023-04-27 22:47:35,488 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.55 vs. limit=2.0 2023-04-27 22:48:24,040 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=24995.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 22:48:32,748 INFO [train.py:904] (1/8) Epoch 3, batch 4700, loss[loss=0.2491, simple_loss=0.3212, pruned_loss=0.08853, over 16677.00 frames. ], tot_loss[loss=0.2555, simple_loss=0.3301, pruned_loss=0.09043, over 3209238.36 frames. ], batch size: 57, lr: 2.06e-02, grad_scale: 4.0 2023-04-27 22:49:17,306 INFO [optim.py:368] (1/8) 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 22:49:26,423 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=3.65 vs. limit=5.0 2023-04-27 22:49:31,911 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.7511, 2.7460, 1.6581, 2.8224, 2.1520, 2.7777, 1.8400, 2.4059], device='cuda:1'), covar=tensor([0.0089, 0.0220, 0.1148, 0.0059, 0.0581, 0.0383, 0.1030, 0.0476], device='cuda:1'), in_proj_covar=tensor([0.0086, 0.0127, 0.0169, 0.0077, 0.0159, 0.0157, 0.0178, 0.0159], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-04-27 22:49:45,101 INFO [train.py:904] (1/8) Epoch 3, batch 4750, loss[loss=0.3033, simple_loss=0.3569, pruned_loss=0.1248, over 11845.00 frames. ], tot_loss[loss=0.252, simple_loss=0.3265, pruned_loss=0.08879, over 3206826.93 frames. ], batch size: 248, lr: 2.06e-02, grad_scale: 4.0 2023-04-27 22:49:52,658 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=25056.0, num_to_drop=1, layers_to_drop={1} 2023-04-27 22:50:03,336 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.4468, 4.3030, 4.3554, 3.2115, 4.2572, 1.4856, 4.0924, 4.2523], device='cuda:1'), covar=tensor([0.0095, 0.0065, 0.0071, 0.0488, 0.0074, 0.1640, 0.0096, 0.0134], device='cuda:1'), in_proj_covar=tensor([0.0070, 0.0059, 0.0090, 0.0109, 0.0069, 0.0113, 0.0082, 0.0090], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002], device='cuda:1') 2023-04-27 22:50:21,608 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.1877, 4.0404, 4.0746, 3.3875, 3.9840, 1.5214, 3.7458, 4.0351], device='cuda:1'), covar=tensor([0.0072, 0.0057, 0.0062, 0.0348, 0.0062, 0.1420, 0.0088, 0.0096], device='cuda:1'), in_proj_covar=tensor([0.0070, 0.0059, 0.0091, 0.0110, 0.0069, 0.0114, 0.0082, 0.0091], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-04-27 22:50:58,609 INFO [train.py:904] (1/8) Epoch 3, batch 4800, loss[loss=0.2405, simple_loss=0.3196, pruned_loss=0.08069, over 17068.00 frames. ], tot_loss[loss=0.2472, simple_loss=0.3221, pruned_loss=0.08616, over 3210658.33 frames. ], batch size: 55, lr: 2.06e-02, grad_scale: 8.0 2023-04-27 22:51:15,596 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([1.9230, 2.6917, 2.7231, 3.8800, 3.6397, 3.7282, 1.4246, 3.0962], device='cuda:1'), covar=tensor([0.1220, 0.0438, 0.0770, 0.0058, 0.0117, 0.0220, 0.1300, 0.0469], device='cuda:1'), in_proj_covar=tensor([0.0143, 0.0130, 0.0164, 0.0071, 0.0130, 0.0135, 0.0154, 0.0154], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:1') 2023-04-27 22:51:28,791 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=25121.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 22:51:33,557 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.2667, 4.1727, 4.6947, 4.7725, 4.6898, 4.1252, 4.3232, 4.1461], device='cuda:1'), covar=tensor([0.0195, 0.0265, 0.0252, 0.0254, 0.0312, 0.0239, 0.0621, 0.0288], device='cuda:1'), in_proj_covar=tensor([0.0179, 0.0165, 0.0184, 0.0178, 0.0223, 0.0183, 0.0280, 0.0162], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0002], device='cuda:1') 2023-04-27 22:51:47,436 INFO [optim.py:368] (1/8) 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 22:52:13,127 INFO [train.py:904] (1/8) Epoch 3, batch 4850, loss[loss=0.2675, simple_loss=0.3403, pruned_loss=0.09738, over 16888.00 frames. ], tot_loss[loss=0.2485, simple_loss=0.3238, pruned_loss=0.08664, over 3187176.67 frames. ], batch size: 116, lr: 2.06e-02, grad_scale: 2.0 2023-04-27 22:53:25,029 INFO [train.py:904] (1/8) Epoch 3, batch 4900, loss[loss=0.2358, simple_loss=0.3208, pruned_loss=0.07545, over 17106.00 frames. ], tot_loss[loss=0.2469, simple_loss=0.3233, pruned_loss=0.08524, over 3194075.46 frames. ], batch size: 49, lr: 2.06e-02, grad_scale: 2.0 2023-04-27 22:53:28,449 INFO [zipformer.py:625] (1/8) 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] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=25204.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 22:53:32,441 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=4.89 vs. limit=5.0 2023-04-27 22:53:58,847 INFO [zipformer.py:625] (1/8) 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] (1/8) 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] (1/8) Epoch 3, batch 4950, loss[loss=0.3154, simple_loss=0.3717, pruned_loss=0.1296, over 11979.00 frames. ], tot_loss[loss=0.2471, simple_loss=0.3234, pruned_loss=0.08536, over 3190289.86 frames. ], batch size: 248, lr: 2.05e-02, grad_scale: 2.0 2023-04-27 22:54:35,558 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=25251.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 22:55:08,062 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=25273.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 22:55:48,336 INFO [train.py:904] (1/8) Epoch 3, batch 5000, loss[loss=0.2333, simple_loss=0.3089, pruned_loss=0.07888, over 16611.00 frames. ], tot_loss[loss=0.2469, simple_loss=0.324, pruned_loss=0.08484, over 3203914.12 frames. ], batch size: 62, lr: 2.05e-02, grad_scale: 2.0 2023-04-27 22:55:56,667 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.41 vs. limit=2.0 2023-04-27 22:56:35,247 INFO [optim.py:368] (1/8) 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] (1/8) Epoch 3, batch 5050, loss[loss=0.3102, simple_loss=0.358, pruned_loss=0.1312, over 12012.00 frames. ], tot_loss[loss=0.2465, simple_loss=0.3241, pruned_loss=0.08448, over 3210290.43 frames. ], batch size: 250, lr: 2.05e-02, grad_scale: 2.0 2023-04-27 22:56:59,955 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=25351.0, num_to_drop=1, layers_to_drop={0} 2023-04-27 22:58:08,622 INFO [train.py:904] (1/8) Epoch 3, batch 5100, loss[loss=0.2088, simple_loss=0.2952, pruned_loss=0.0612, over 16839.00 frames. ], tot_loss[loss=0.2441, simple_loss=0.3217, pruned_loss=0.0832, over 3207897.45 frames. ], batch size: 96, lr: 2.05e-02, grad_scale: 2.0 2023-04-27 22:58:26,519 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.6518, 1.8814, 1.6132, 1.6412, 2.3639, 2.1991, 2.5224, 2.5830], device='cuda:1'), covar=tensor([0.0016, 0.0145, 0.0161, 0.0173, 0.0089, 0.0114, 0.0031, 0.0063], device='cuda:1'), in_proj_covar=tensor([0.0053, 0.0119, 0.0121, 0.0121, 0.0114, 0.0124, 0.0073, 0.0097], device='cuda:1'), out_proj_covar=tensor([7.5485e-05, 1.7131e-04, 1.7003e-04, 1.7433e-04, 1.6659e-04, 1.8060e-04, 1.0399e-04, 1.4568e-04], device='cuda:1') 2023-04-27 22:58:32,567 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.67 vs. limit=2.0 2023-04-27 22:58:38,828 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=25421.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 22:58:47,848 INFO [zipformer.py:625] (1/8) 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] (1/8) 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 22:59:23,207 INFO [train.py:904] (1/8) Epoch 3, batch 5150, loss[loss=0.2332, simple_loss=0.3125, pruned_loss=0.07689, over 17029.00 frames. ], tot_loss[loss=0.2434, simple_loss=0.322, pruned_loss=0.08244, over 3213723.92 frames. ], batch size: 53, lr: 2.05e-02, grad_scale: 2.0 2023-04-27 22:59:40,724 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=3.32 vs. limit=5.0 2023-04-27 22:59:46,716 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.98 vs. limit=2.0 2023-04-27 22:59:50,322 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=25469.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 23:00:03,127 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.91 vs. limit=2.0 2023-04-27 23:00:19,368 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=25489.0, num_to_drop=1, layers_to_drop={1} 2023-04-27 23:00:35,983 INFO [train.py:904] (1/8) Epoch 3, batch 5200, loss[loss=0.3048, simple_loss=0.3569, pruned_loss=0.1263, over 12282.00 frames. ], tot_loss[loss=0.2436, simple_loss=0.3215, pruned_loss=0.08289, over 3210670.75 frames. ], batch size: 247, lr: 2.04e-02, grad_scale: 4.0 2023-04-27 23:00:40,461 INFO [zipformer.py:625] (1/8) 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] (1/8) 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] (1/8) Epoch 3, batch 5250, loss[loss=0.227, simple_loss=0.3102, pruned_loss=0.07196, over 16835.00 frames. ], tot_loss[loss=0.2428, simple_loss=0.3197, pruned_loss=0.08291, over 3208413.98 frames. ], batch size: 102, lr: 2.04e-02, grad_scale: 4.0 2023-04-27 23:01:47,884 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=25552.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 23:02:03,651 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.45 vs. limit=2.0 2023-04-27 23:02:09,279 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.54 vs. limit=2.0 2023-04-27 23:02:43,212 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.6107, 1.4101, 1.9405, 2.6553, 2.6115, 2.7831, 1.4434, 2.8353], device='cuda:1'), covar=tensor([0.0044, 0.0218, 0.0120, 0.0095, 0.0057, 0.0052, 0.0175, 0.0035], device='cuda:1'), in_proj_covar=tensor([0.0086, 0.0120, 0.0108, 0.0102, 0.0089, 0.0067, 0.0112, 0.0063], device='cuda:1'), out_proj_covar=tensor([1.3434e-04, 1.9086e-04, 1.7735e-04, 1.6767e-04, 1.4115e-04, 1.0368e-04, 1.7639e-04, 9.8242e-05], device='cuda:1') 2023-04-27 23:02:56,098 INFO [train.py:904] (1/8) Epoch 3, batch 5300, loss[loss=0.2219, simple_loss=0.2987, pruned_loss=0.07252, over 16754.00 frames. ], tot_loss[loss=0.2393, simple_loss=0.316, pruned_loss=0.08128, over 3212113.47 frames. ], batch size: 83, lr: 2.04e-02, grad_scale: 4.0 2023-04-27 23:03:16,423 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.63 vs. limit=2.0 2023-04-27 23:03:24,005 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.9039, 4.8009, 4.5871, 4.0003, 4.7954, 1.9996, 4.4569, 4.7130], device='cuda:1'), covar=tensor([0.0045, 0.0043, 0.0059, 0.0288, 0.0039, 0.1188, 0.0065, 0.0085], device='cuda:1'), in_proj_covar=tensor([0.0072, 0.0061, 0.0095, 0.0113, 0.0071, 0.0117, 0.0083, 0.0095], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0001, 0.0002, 0.0003, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-04-27 23:03:43,229 INFO [optim.py:368] (1/8) 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] (1/8) Epoch 3, batch 5350, loss[loss=0.257, simple_loss=0.3307, pruned_loss=0.09168, over 16638.00 frames. ], tot_loss[loss=0.2375, simple_loss=0.3142, pruned_loss=0.08037, over 3217021.53 frames. ], batch size: 62, lr: 2.04e-02, grad_scale: 4.0 2023-04-27 23:04:08,378 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=25651.0, num_to_drop=1, layers_to_drop={2} 2023-04-27 23:04:12,204 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([1.7413, 1.1650, 1.4799, 1.7967, 1.7635, 1.8632, 1.3743, 1.7774], device='cuda:1'), covar=tensor([0.0082, 0.0167, 0.0098, 0.0091, 0.0058, 0.0050, 0.0144, 0.0039], device='cuda:1'), in_proj_covar=tensor([0.0088, 0.0121, 0.0109, 0.0103, 0.0090, 0.0068, 0.0115, 0.0063], device='cuda:1'), out_proj_covar=tensor([1.3714e-04, 1.9228e-04, 1.7912e-04, 1.6950e-04, 1.4264e-04, 1.0513e-04, 1.8076e-04, 9.9525e-05], device='cuda:1') 2023-04-27 23:04:16,315 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=4.95 vs. limit=5.0 2023-04-27 23:05:16,510 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=25699.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 23:05:19,279 INFO [train.py:904] (1/8) Epoch 3, batch 5400, loss[loss=0.2339, simple_loss=0.3167, pruned_loss=0.07557, over 16506.00 frames. ], tot_loss[loss=0.2395, simple_loss=0.3167, pruned_loss=0.0811, over 3221820.51 frames. ], batch size: 68, lr: 2.04e-02, grad_scale: 4.0 2023-04-27 23:06:00,637 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.32 vs. limit=2.0 2023-04-27 23:06:07,786 INFO [optim.py:368] (1/8) 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 23:06:34,321 INFO [train.py:904] (1/8) Epoch 3, batch 5450, loss[loss=0.2908, simple_loss=0.3488, pruned_loss=0.1164, over 16305.00 frames. ], tot_loss[loss=0.2441, simple_loss=0.3208, pruned_loss=0.08368, over 3203492.37 frames. ], batch size: 35, lr: 2.03e-02, grad_scale: 4.0 2023-04-27 23:07:24,678 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=25784.0, num_to_drop=1, layers_to_drop={0} 2023-04-27 23:07:49,276 INFO [train.py:904] (1/8) Epoch 3, batch 5500, loss[loss=0.377, simple_loss=0.4119, pruned_loss=0.171, over 11839.00 frames. ], tot_loss[loss=0.2568, simple_loss=0.3306, pruned_loss=0.09152, over 3182898.84 frames. ], batch size: 247, lr: 2.03e-02, grad_scale: 4.0 2023-04-27 23:08:18,227 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.2829, 4.2578, 1.8324, 4.4860, 2.7383, 4.4378, 2.2254, 2.9061], device='cuda:1'), covar=tensor([0.0046, 0.0139, 0.1535, 0.0027, 0.0669, 0.0181, 0.1215, 0.0599], device='cuda:1'), in_proj_covar=tensor([0.0088, 0.0132, 0.0173, 0.0081, 0.0162, 0.0163, 0.0179, 0.0162], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-04-27 23:08:39,226 INFO [optim.py:368] (1/8) 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 23:09:04,976 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.2693, 3.7024, 2.9931, 3.5226, 3.1674, 3.3977, 3.4509, 3.6756], device='cuda:1'), covar=tensor([0.2042, 0.1365, 0.3077, 0.0990, 0.1365, 0.1619, 0.1351, 0.1352], device='cuda:1'), in_proj_covar=tensor([0.0260, 0.0364, 0.0322, 0.0230, 0.0236, 0.0231, 0.0296, 0.0252], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-27 23:09:06,197 INFO [train.py:904] (1/8) Epoch 3, batch 5550, loss[loss=0.4641, simple_loss=0.4526, pruned_loss=0.2378, over 10961.00 frames. ], tot_loss[loss=0.2706, simple_loss=0.3407, pruned_loss=0.1003, over 3145631.22 frames. ], batch size: 248, lr: 2.03e-02, grad_scale: 4.0 2023-04-27 23:09:17,555 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=25858.0, num_to_drop=1, layers_to_drop={1} 2023-04-27 23:10:24,442 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.9597, 3.9213, 1.8486, 3.9881, 2.6272, 4.0450, 2.0383, 2.8882], device='cuda:1'), covar=tensor([0.0047, 0.0164, 0.1540, 0.0034, 0.0688, 0.0289, 0.1338, 0.0568], device='cuda:1'), in_proj_covar=tensor([0.0088, 0.0131, 0.0173, 0.0080, 0.0161, 0.0162, 0.0179, 0.0161], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-04-27 23:10:25,184 INFO [train.py:904] (1/8) Epoch 3, batch 5600, loss[loss=0.2794, simple_loss=0.3549, pruned_loss=0.1019, over 16696.00 frames. ], tot_loss[loss=0.2797, simple_loss=0.3471, pruned_loss=0.1062, over 3105020.13 frames. ], batch size: 89, lr: 2.03e-02, grad_scale: 8.0 2023-04-27 23:10:56,267 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=25919.0, num_to_drop=1, layers_to_drop={3} 2023-04-27 23:11:21,465 INFO [optim.py:368] (1/8) 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 23:11:48,513 INFO [train.py:904] (1/8) Epoch 3, batch 5650, loss[loss=0.3464, simple_loss=0.3993, pruned_loss=0.1467, over 15417.00 frames. ], tot_loss[loss=0.2902, simple_loss=0.3545, pruned_loss=0.1129, over 3079467.96 frames. ], batch size: 190, lr: 2.03e-02, grad_scale: 8.0 2023-04-27 23:12:05,120 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.6445, 4.3833, 4.2731, 4.8161, 4.7790, 4.3826, 4.8408, 4.7787], device='cuda:1'), covar=tensor([0.0619, 0.0584, 0.1528, 0.0520, 0.0687, 0.0578, 0.0725, 0.0470], device='cuda:1'), in_proj_covar=tensor([0.0275, 0.0334, 0.0435, 0.0337, 0.0255, 0.0238, 0.0273, 0.0275], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-27 23:12:36,004 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.49 vs. limit=2.0 2023-04-27 23:13:09,633 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.99 vs. limit=2.0 2023-04-27 23:13:10,045 INFO [train.py:904] (1/8) Epoch 3, batch 5700, loss[loss=0.2598, simple_loss=0.3448, pruned_loss=0.08741, over 16750.00 frames. ], tot_loss[loss=0.2912, simple_loss=0.3553, pruned_loss=0.1135, over 3085691.47 frames. ], batch size: 39, lr: 2.02e-02, grad_scale: 8.0 2023-04-27 23:13:16,120 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.4407, 3.4062, 3.3024, 2.8692, 3.2624, 2.0987, 3.1315, 3.1425], device='cuda:1'), covar=tensor([0.0069, 0.0059, 0.0088, 0.0219, 0.0060, 0.1137, 0.0076, 0.0106], device='cuda:1'), in_proj_covar=tensor([0.0070, 0.0059, 0.0093, 0.0109, 0.0070, 0.0117, 0.0081, 0.0093], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002], device='cuda:1') 2023-04-27 23:13:34,072 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.0069, 3.9383, 3.9499, 2.7429, 3.7349, 3.8513, 4.0250, 1.9040], device='cuda:1'), covar=tensor([0.0364, 0.0017, 0.0023, 0.0205, 0.0029, 0.0057, 0.0014, 0.0313], device='cuda:1'), in_proj_covar=tensor([0.0111, 0.0052, 0.0058, 0.0109, 0.0051, 0.0063, 0.0056, 0.0104], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0001, 0.0002, 0.0003, 0.0001, 0.0002, 0.0002, 0.0003], device='cuda:1') 2023-04-27 23:14:00,537 INFO [optim.py:368] (1/8) 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 23:14:27,440 INFO [train.py:904] (1/8) Epoch 3, batch 5750, loss[loss=0.358, simple_loss=0.384, pruned_loss=0.166, over 11242.00 frames. ], tot_loss[loss=0.2946, simple_loss=0.3584, pruned_loss=0.1154, over 3051624.85 frames. ], batch size: 246, lr: 2.02e-02, grad_scale: 8.0 2023-04-27 23:15:21,997 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=26084.0, num_to_drop=1, layers_to_drop={0} 2023-04-27 23:15:47,177 INFO [train.py:904] (1/8) Epoch 3, batch 5800, loss[loss=0.2592, simple_loss=0.3318, pruned_loss=0.09335, over 16725.00 frames. ], tot_loss[loss=0.2932, simple_loss=0.3582, pruned_loss=0.1141, over 3043987.50 frames. ], batch size: 57, lr: 2.02e-02, grad_scale: 8.0 2023-04-27 23:15:49,012 INFO [zipformer.py:625] (1/8) 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] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=26132.0, num_to_drop=1, layers_to_drop={1} 2023-04-27 23:16:38,587 INFO [optim.py:368] (1/8) Clipping_scale=2.0, grad-norm quartiles 2.751e+02 5.049e+02 6.109e+02 8.114e+02 1.629e+03, threshold=1.222e+03, percent-clipped=2.0 2023-04-27 23:17:05,080 INFO [train.py:904] (1/8) Epoch 3, batch 5850, loss[loss=0.2805, simple_loss=0.3464, pruned_loss=0.1073, over 16925.00 frames. ], tot_loss[loss=0.2905, simple_loss=0.3562, pruned_loss=0.1125, over 3023970.02 frames. ], batch size: 109, lr: 2.02e-02, grad_scale: 8.0 2023-04-27 23:17:24,154 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=26163.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 23:17:31,353 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=26168.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 23:18:23,197 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.7528, 3.8830, 3.1662, 2.5236, 2.9576, 2.3944, 4.2319, 4.3115], device='cuda:1'), covar=tensor([0.1991, 0.0612, 0.1065, 0.1087, 0.1866, 0.1121, 0.0322, 0.0316], device='cuda:1'), in_proj_covar=tensor([0.0265, 0.0237, 0.0253, 0.0209, 0.0291, 0.0189, 0.0216, 0.0201], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0003, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-27 23:18:26,956 INFO [train.py:904] (1/8) Epoch 3, batch 5900, loss[loss=0.2672, simple_loss=0.3469, pruned_loss=0.09374, over 17002.00 frames. ], tot_loss[loss=0.2892, simple_loss=0.3552, pruned_loss=0.1116, over 3033171.85 frames. ], batch size: 41, lr: 2.02e-02, grad_scale: 8.0 2023-04-27 23:18:51,418 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=26214.0, num_to_drop=1, layers_to_drop={1} 2023-04-27 23:19:15,285 INFO [zipformer.py:625] (1/8) 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] (1/8) 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 23:19:47,845 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=26250.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 23:19:48,574 INFO [train.py:904] (1/8) Epoch 3, batch 5950, loss[loss=0.3076, simple_loss=0.373, pruned_loss=0.1211, over 16738.00 frames. ], tot_loss[loss=0.288, simple_loss=0.3563, pruned_loss=0.1099, over 3046689.33 frames. ], batch size: 89, lr: 2.02e-02, grad_scale: 8.0 2023-04-27 23:21:07,922 INFO [train.py:904] (1/8) Epoch 3, batch 6000, loss[loss=0.3248, simple_loss=0.3698, pruned_loss=0.1399, over 11679.00 frames. ], tot_loss[loss=0.2869, simple_loss=0.3551, pruned_loss=0.1093, over 3063531.14 frames. ], batch size: 246, lr: 2.01e-02, grad_scale: 8.0 2023-04-27 23:21:07,923 INFO [train.py:929] (1/8) Computing validation loss 2023-04-27 23:21:18,890 INFO [train.py:938] (1/8) Epoch 3, validation: loss=0.2097, simple_loss=0.3184, pruned_loss=0.05055, over 944034.00 frames. 2023-04-27 23:21:18,890 INFO [train.py:939] (1/8) Maximum memory allocated so far is 17676MB 2023-04-27 23:21:33,996 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=26311.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 23:22:07,342 INFO [optim.py:368] (1/8) Clipping_scale=2.0, grad-norm quartiles 2.939e+02 4.548e+02 5.519e+02 7.491e+02 1.813e+03, threshold=1.104e+03, percent-clipped=4.0 2023-04-27 23:22:36,223 INFO [train.py:904] (1/8) Epoch 3, batch 6050, loss[loss=0.2641, simple_loss=0.3587, pruned_loss=0.08472, over 16872.00 frames. ], tot_loss[loss=0.2841, simple_loss=0.3528, pruned_loss=0.1077, over 3071612.78 frames. ], batch size: 96, lr: 2.01e-02, grad_scale: 8.0 2023-04-27 23:23:51,471 INFO [train.py:904] (1/8) Epoch 3, batch 6100, loss[loss=0.2361, simple_loss=0.3166, pruned_loss=0.07779, over 16483.00 frames. ], tot_loss[loss=0.282, simple_loss=0.3518, pruned_loss=0.1061, over 3083857.89 frames. ], batch size: 75, lr: 2.01e-02, grad_scale: 8.0 2023-04-27 23:24:12,682 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=26414.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 23:24:42,456 INFO [optim.py:368] (1/8) 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 23:25:11,818 INFO [train.py:904] (1/8) Epoch 3, batch 6150, loss[loss=0.2488, simple_loss=0.3272, pruned_loss=0.08522, over 16880.00 frames. ], tot_loss[loss=0.2809, simple_loss=0.35, pruned_loss=0.1059, over 3072669.82 frames. ], batch size: 116, lr: 2.01e-02, grad_scale: 4.0 2023-04-27 23:25:23,377 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=26458.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 23:25:50,226 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 3, batch 6200, loss[loss=0.271, simple_loss=0.3535, pruned_loss=0.09425, over 16705.00 frames. ], tot_loss[loss=0.2782, simple_loss=0.3473, pruned_loss=0.1045, over 3070902.48 frames. ], batch size: 62, lr: 2.01e-02, grad_scale: 4.0 2023-04-27 23:26:48,628 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=26514.0, num_to_drop=1, layers_to_drop={2} 2023-04-27 23:27:03,445 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=26524.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 23:27:07,475 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.2197, 3.7958, 3.7579, 1.6602, 4.0413, 4.0496, 3.1020, 2.9899], device='cuda:1'), covar=tensor([0.0854, 0.0099, 0.0150, 0.1349, 0.0048, 0.0033, 0.0274, 0.0414], device='cuda:1'), in_proj_covar=tensor([0.0138, 0.0080, 0.0079, 0.0144, 0.0072, 0.0070, 0.0114, 0.0124], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003, 0.0003], device='cuda:1') 2023-04-27 23:27:13,268 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.72 vs. limit=2.0 2023-04-27 23:27:18,519 INFO [optim.py:368] (1/8) 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 23:27:41,889 INFO [train.py:904] (1/8) Epoch 3, batch 6250, loss[loss=0.2739, simple_loss=0.3504, pruned_loss=0.09866, over 16661.00 frames. ], tot_loss[loss=0.2785, simple_loss=0.3478, pruned_loss=0.1047, over 3080561.83 frames. ], batch size: 134, lr: 2.00e-02, grad_scale: 4.0 2023-04-27 23:27:58,186 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=26562.0, num_to_drop=1, layers_to_drop={0} 2023-04-27 23:28:54,857 INFO [train.py:904] (1/8) Epoch 3, batch 6300, loss[loss=0.2839, simple_loss=0.3446, pruned_loss=0.1115, over 17021.00 frames. ], tot_loss[loss=0.2778, simple_loss=0.3476, pruned_loss=0.104, over 3103197.97 frames. ], batch size: 55, lr: 2.00e-02, grad_scale: 4.0 2023-04-27 23:29:02,861 INFO [zipformer.py:625] (1/8) 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] (1/8) 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 23:30:11,907 INFO [train.py:904] (1/8) Epoch 3, batch 6350, loss[loss=0.3255, simple_loss=0.3812, pruned_loss=0.1349, over 15340.00 frames. ], tot_loss[loss=0.2817, simple_loss=0.35, pruned_loss=0.1067, over 3093256.82 frames. ], batch size: 190, lr: 2.00e-02, grad_scale: 4.0 2023-04-27 23:31:13,689 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.4636, 2.9623, 2.5488, 2.5131, 2.3112, 2.0838, 2.9505, 3.0894], device='cuda:1'), covar=tensor([0.1308, 0.0571, 0.0851, 0.0833, 0.1631, 0.1269, 0.0385, 0.0567], device='cuda:1'), in_proj_covar=tensor([0.0266, 0.0237, 0.0253, 0.0210, 0.0292, 0.0189, 0.0216, 0.0204], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-27 23:31:25,029 INFO [train.py:904] (1/8) Epoch 3, batch 6400, loss[loss=0.2446, simple_loss=0.3275, pruned_loss=0.08089, over 16747.00 frames. ], tot_loss[loss=0.2833, simple_loss=0.3506, pruned_loss=0.108, over 3078237.59 frames. ], batch size: 89, lr: 2.00e-02, grad_scale: 8.0 2023-04-27 23:31:37,280 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.1192, 1.6928, 2.1361, 2.8582, 2.7232, 3.2421, 1.7565, 3.2240], device='cuda:1'), covar=tensor([0.0035, 0.0174, 0.0116, 0.0087, 0.0062, 0.0035, 0.0168, 0.0024], device='cuda:1'), in_proj_covar=tensor([0.0085, 0.0121, 0.0105, 0.0102, 0.0093, 0.0068, 0.0115, 0.0061], device='cuda:1'), out_proj_covar=tensor([1.3067e-04, 1.9147e-04, 1.7036e-04, 1.6576e-04, 1.4542e-04, 1.0465e-04, 1.7828e-04, 9.4344e-05], device='cuda:1') 2023-04-27 23:32:12,498 INFO [optim.py:368] (1/8) Clipping_scale=2.0, grad-norm quartiles 2.974e+02 5.048e+02 6.146e+02 7.949e+02 2.287e+03, threshold=1.229e+03, percent-clipped=11.0 2023-04-27 23:32:35,287 INFO [train.py:904] (1/8) Epoch 3, batch 6450, loss[loss=0.2159, simple_loss=0.3031, pruned_loss=0.06439, over 16683.00 frames. ], tot_loss[loss=0.2817, simple_loss=0.3497, pruned_loss=0.1069, over 3073841.72 frames. ], batch size: 83, lr: 2.00e-02, grad_scale: 8.0 2023-04-27 23:32:47,293 INFO [zipformer.py:625] (1/8) 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] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=26770.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 23:33:07,183 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.8604, 4.0833, 3.7860, 3.9569, 3.5581, 3.7492, 3.8692, 4.0109], device='cuda:1'), covar=tensor([0.0609, 0.0832, 0.1004, 0.0414, 0.0616, 0.0975, 0.0568, 0.0917], device='cuda:1'), in_proj_covar=tensor([0.0277, 0.0382, 0.0337, 0.0244, 0.0245, 0.0242, 0.0301, 0.0263], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-27 23:33:53,754 INFO [train.py:904] (1/8) Epoch 3, batch 6500, loss[loss=0.2422, simple_loss=0.322, pruned_loss=0.08118, over 16645.00 frames. ], tot_loss[loss=0.2784, simple_loss=0.346, pruned_loss=0.1055, over 3076133.17 frames. ], batch size: 134, lr: 2.00e-02, grad_scale: 8.0 2023-04-27 23:34:01,036 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=26806.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 23:34:16,489 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.5290, 1.8783, 1.6108, 1.6647, 2.3626, 2.0714, 2.4058, 2.4590], device='cuda:1'), covar=tensor([0.0016, 0.0133, 0.0167, 0.0188, 0.0083, 0.0134, 0.0049, 0.0079], device='cuda:1'), in_proj_covar=tensor([0.0054, 0.0121, 0.0121, 0.0121, 0.0114, 0.0124, 0.0078, 0.0100], device='cuda:1'), out_proj_covar=tensor([7.3411e-05, 1.7222e-04, 1.6691e-04, 1.7212e-04, 1.6595e-04, 1.7731e-04, 1.1095e-04, 1.4705e-04], device='cuda:1') 2023-04-27 23:34:29,494 INFO [zipformer.py:625] (1/8) 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] (1/8) Clipping_scale=2.0, grad-norm quartiles 2.692e+02 4.621e+02 5.604e+02 6.997e+02 1.424e+03, threshold=1.121e+03, percent-clipped=2.0 2023-04-27 23:34:49,326 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.9908, 1.9811, 1.6507, 1.8520, 2.7662, 2.5624, 3.1406, 3.1057], device='cuda:1'), covar=tensor([0.0021, 0.0205, 0.0211, 0.0199, 0.0102, 0.0126, 0.0072, 0.0073], device='cuda:1'), in_proj_covar=tensor([0.0055, 0.0123, 0.0124, 0.0123, 0.0117, 0.0127, 0.0080, 0.0102], device='cuda:1'), out_proj_covar=tensor([7.4657e-05, 1.7576e-04, 1.7056e-04, 1.7562e-04, 1.6982e-04, 1.8182e-04, 1.1393e-04, 1.4916e-04], device='cuda:1') 2023-04-27 23:35:12,111 INFO [train.py:904] (1/8) Epoch 3, batch 6550, loss[loss=0.2873, simple_loss=0.3746, pruned_loss=0.1, over 16520.00 frames. ], tot_loss[loss=0.2818, simple_loss=0.3498, pruned_loss=0.1069, over 3067215.17 frames. ], batch size: 62, lr: 1.99e-02, grad_scale: 8.0 2023-04-27 23:35:45,477 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=26872.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 23:36:06,915 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.0663, 4.0650, 4.5718, 4.5183, 4.5192, 4.1169, 4.1562, 4.0672], device='cuda:1'), covar=tensor([0.0222, 0.0261, 0.0282, 0.0356, 0.0443, 0.0270, 0.0750, 0.0361], device='cuda:1'), in_proj_covar=tensor([0.0180, 0.0167, 0.0184, 0.0181, 0.0221, 0.0189, 0.0284, 0.0172], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0002], device='cuda:1') 2023-04-27 23:36:28,903 INFO [train.py:904] (1/8) Epoch 3, batch 6600, loss[loss=0.3348, simple_loss=0.3836, pruned_loss=0.143, over 11595.00 frames. ], tot_loss[loss=0.2837, simple_loss=0.3524, pruned_loss=0.1076, over 3069073.77 frames. ], batch size: 246, lr: 1.99e-02, grad_scale: 8.0 2023-04-27 23:36:33,650 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=26904.0, num_to_drop=1, layers_to_drop={1} 2023-04-27 23:36:37,607 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=26906.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 23:37:20,959 INFO [optim.py:368] (1/8) 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 23:37:45,821 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=26950.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 23:37:46,400 INFO [train.py:904] (1/8) Epoch 3, batch 6650, loss[loss=0.2685, simple_loss=0.3437, pruned_loss=0.09667, over 16661.00 frames. ], tot_loss[loss=0.2854, simple_loss=0.3529, pruned_loss=0.1089, over 3063671.50 frames. ], batch size: 76, lr: 1.99e-02, grad_scale: 8.0 2023-04-27 23:37:51,224 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=26954.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 23:37:52,605 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.9383, 3.1181, 3.4731, 3.4119, 3.4431, 3.0745, 3.2532, 3.3250], device='cuda:1'), covar=tensor([0.0302, 0.0418, 0.0330, 0.0460, 0.0406, 0.0373, 0.0695, 0.0323], device='cuda:1'), in_proj_covar=tensor([0.0187, 0.0171, 0.0188, 0.0186, 0.0227, 0.0193, 0.0293, 0.0174], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0002], device='cuda:1') 2023-04-27 23:38:09,016 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=26965.0, num_to_drop=1, layers_to_drop={2} 2023-04-27 23:38:51,765 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([1.8679, 3.3115, 3.2209, 1.4542, 3.5135, 3.4716, 2.8290, 2.5713], device='cuda:1'), covar=tensor([0.0886, 0.0104, 0.0131, 0.1293, 0.0065, 0.0057, 0.0272, 0.0453], device='cuda:1'), in_proj_covar=tensor([0.0141, 0.0080, 0.0079, 0.0143, 0.0073, 0.0070, 0.0113, 0.0126], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003, 0.0003], device='cuda:1') 2023-04-27 23:39:03,601 INFO [train.py:904] (1/8) Epoch 3, batch 6700, loss[loss=0.2761, simple_loss=0.3454, pruned_loss=0.1034, over 16265.00 frames. ], tot_loss[loss=0.2839, simple_loss=0.351, pruned_loss=0.1084, over 3078477.06 frames. ], batch size: 165, lr: 1.99e-02, grad_scale: 8.0 2023-04-27 23:39:18,496 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=27011.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 23:39:51,185 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.9498, 4.6847, 4.9272, 5.2144, 5.3408, 4.7314, 5.3549, 5.2787], device='cuda:1'), covar=tensor([0.0663, 0.0576, 0.0897, 0.0353, 0.0321, 0.0432, 0.0324, 0.0262], device='cuda:1'), in_proj_covar=tensor([0.0282, 0.0346, 0.0450, 0.0345, 0.0258, 0.0246, 0.0283, 0.0282], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-27 23:39:57,110 INFO [optim.py:368] (1/8) 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 23:40:21,183 INFO [train.py:904] (1/8) Epoch 3, batch 6750, loss[loss=0.2771, simple_loss=0.3438, pruned_loss=0.1052, over 15235.00 frames. ], tot_loss[loss=0.2823, simple_loss=0.3495, pruned_loss=0.1075, over 3088455.04 frames. ], batch size: 190, lr: 1.99e-02, grad_scale: 8.0 2023-04-27 23:40:26,099 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([1.7583, 3.1697, 3.2776, 2.0626, 2.9449, 3.1174, 3.1050, 1.7362], device='cuda:1'), covar=tensor([0.0364, 0.0028, 0.0031, 0.0247, 0.0041, 0.0075, 0.0030, 0.0301], device='cuda:1'), in_proj_covar=tensor([0.0109, 0.0048, 0.0056, 0.0108, 0.0052, 0.0060, 0.0057, 0.0104], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0001, 0.0002, 0.0003, 0.0001, 0.0002, 0.0002, 0.0003], device='cuda:1') 2023-04-27 23:40:33,269 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.5823, 3.6100, 1.5823, 3.7046, 2.3649, 3.6882, 1.7993, 2.6087], device='cuda:1'), covar=tensor([0.0060, 0.0221, 0.1624, 0.0037, 0.0751, 0.0370, 0.1438, 0.0595], device='cuda:1'), in_proj_covar=tensor([0.0087, 0.0130, 0.0173, 0.0081, 0.0161, 0.0161, 0.0181, 0.0158], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-04-27 23:40:34,714 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.49 vs. limit=2.0 2023-04-27 23:40:51,198 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=27070.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 23:41:37,190 INFO [train.py:904] (1/8) Epoch 3, batch 6800, loss[loss=0.3126, simple_loss=0.3662, pruned_loss=0.1295, over 11538.00 frames. ], tot_loss[loss=0.2801, simple_loss=0.3481, pruned_loss=0.1061, over 3092781.08 frames. ], batch size: 246, lr: 1.98e-02, grad_scale: 8.0 2023-04-27 23:41:43,277 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.8172, 5.1208, 4.7779, 4.8907, 4.4769, 4.2999, 4.6949, 5.1852], device='cuda:1'), covar=tensor([0.0526, 0.0600, 0.0892, 0.0382, 0.0505, 0.0674, 0.0470, 0.0597], device='cuda:1'), in_proj_covar=tensor([0.0276, 0.0380, 0.0340, 0.0244, 0.0244, 0.0244, 0.0302, 0.0268], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-27 23:42:03,929 INFO [zipformer.py:625] (1/8) 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] (1/8) 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 23:42:51,178 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=27148.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 23:42:55,393 INFO [train.py:904] (1/8) Epoch 3, batch 6850, loss[loss=0.2653, simple_loss=0.3541, pruned_loss=0.08821, over 16673.00 frames. ], tot_loss[loss=0.2824, simple_loss=0.3502, pruned_loss=0.1073, over 3076694.02 frames. ], batch size: 134, lr: 1.98e-02, grad_scale: 8.0 2023-04-27 23:43:20,566 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.2882, 4.3932, 4.4452, 4.5167, 4.4725, 5.0022, 4.6245, 4.3141], device='cuda:1'), covar=tensor([0.1154, 0.1457, 0.1392, 0.1619, 0.2271, 0.0932, 0.1187, 0.2420], device='cuda:1'), in_proj_covar=tensor([0.0238, 0.0320, 0.0301, 0.0275, 0.0362, 0.0325, 0.0263, 0.0377], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003], device='cuda:1') 2023-04-27 23:43:46,933 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=2.00 vs. limit=2.0 2023-04-27 23:44:10,167 INFO [train.py:904] (1/8) Epoch 3, batch 6900, loss[loss=0.2596, simple_loss=0.3361, pruned_loss=0.09152, over 16930.00 frames. ], tot_loss[loss=0.2818, simple_loss=0.3517, pruned_loss=0.106, over 3085067.41 frames. ], batch size: 109, lr: 1.98e-02, grad_scale: 8.0 2023-04-27 23:44:11,594 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.48 vs. limit=2.0 2023-04-27 23:44:22,759 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=27209.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 23:44:41,157 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([1.7568, 3.4495, 3.6774, 2.3654, 3.3450, 3.4571, 3.4807, 1.6834], device='cuda:1'), covar=tensor([0.0352, 0.0024, 0.0028, 0.0235, 0.0032, 0.0070, 0.0026, 0.0339], device='cuda:1'), in_proj_covar=tensor([0.0111, 0.0051, 0.0057, 0.0111, 0.0052, 0.0062, 0.0058, 0.0106], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0001, 0.0002, 0.0003, 0.0001, 0.0002, 0.0002, 0.0003], device='cuda:1') 2023-04-27 23:45:02,576 INFO [optim.py:368] (1/8) 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 23:45:15,463 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.2496, 3.1089, 3.2022, 3.3665, 3.4149, 3.0756, 3.3753, 3.4516], device='cuda:1'), covar=tensor([0.0546, 0.0552, 0.0995, 0.0420, 0.0471, 0.1726, 0.0598, 0.0439], device='cuda:1'), in_proj_covar=tensor([0.0288, 0.0358, 0.0460, 0.0351, 0.0269, 0.0255, 0.0289, 0.0288], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-27 23:45:22,908 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=4.81 vs. limit=5.0 2023-04-27 23:45:26,966 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.5650, 3.5342, 4.0074, 3.9821, 3.9897, 3.6283, 3.7143, 3.7503], device='cuda:1'), covar=tensor([0.0244, 0.0337, 0.0331, 0.0366, 0.0372, 0.0302, 0.0701, 0.0290], device='cuda:1'), in_proj_covar=tensor([0.0189, 0.0177, 0.0192, 0.0186, 0.0232, 0.0196, 0.0299, 0.0174], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0002], device='cuda:1') 2023-04-27 23:45:28,455 INFO [train.py:904] (1/8) Epoch 3, batch 6950, loss[loss=0.3135, simple_loss=0.3798, pruned_loss=0.1236, over 15146.00 frames. ], tot_loss[loss=0.2859, simple_loss=0.3543, pruned_loss=0.1088, over 3086473.01 frames. ], batch size: 190, lr: 1.98e-02, grad_scale: 8.0 2023-04-27 23:45:41,656 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=27260.0, num_to_drop=1, layers_to_drop={2} 2023-04-27 23:46:43,369 INFO [train.py:904] (1/8) Epoch 3, batch 7000, loss[loss=0.2569, simple_loss=0.3448, pruned_loss=0.08449, over 16710.00 frames. ], tot_loss[loss=0.2835, simple_loss=0.3537, pruned_loss=0.1067, over 3101872.42 frames. ], batch size: 124, lr: 1.98e-02, grad_scale: 8.0 2023-04-27 23:46:50,999 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=27306.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 23:47:35,868 INFO [optim.py:368] (1/8) 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 23:48:01,108 INFO [train.py:904] (1/8) Epoch 3, batch 7050, loss[loss=0.2579, simple_loss=0.3391, pruned_loss=0.08831, over 16203.00 frames. ], tot_loss[loss=0.2849, simple_loss=0.3553, pruned_loss=0.1072, over 3109572.74 frames. ], batch size: 35, lr: 1.98e-02, grad_scale: 8.0 2023-04-27 23:48:32,950 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.9922, 2.1695, 1.7137, 1.9322, 2.6971, 2.5343, 3.1763, 2.9859], device='cuda:1'), covar=tensor([0.0016, 0.0150, 0.0189, 0.0186, 0.0089, 0.0131, 0.0037, 0.0065], device='cuda:1'), in_proj_covar=tensor([0.0054, 0.0121, 0.0122, 0.0121, 0.0115, 0.0125, 0.0077, 0.0101], device='cuda:1'), out_proj_covar=tensor([7.2089e-05, 1.7168e-04, 1.6767e-04, 1.7160e-04, 1.6655e-04, 1.7973e-04, 1.0906e-04, 1.4622e-04], device='cuda:1') 2023-04-27 23:49:19,631 INFO [train.py:904] (1/8) Epoch 3, batch 7100, loss[loss=0.3634, simple_loss=0.3926, pruned_loss=0.1671, over 11093.00 frames. ], tot_loss[loss=0.2843, simple_loss=0.3539, pruned_loss=0.1074, over 3083658.30 frames. ], batch size: 248, lr: 1.97e-02, grad_scale: 8.0 2023-04-27 23:49:46,792 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=2.67 vs. limit=5.0 2023-04-27 23:50:12,101 INFO [optim.py:368] (1/8) Clipping_scale=2.0, grad-norm quartiles 3.055e+02 4.788e+02 6.087e+02 7.690e+02 2.114e+03, threshold=1.217e+03, percent-clipped=3.0 2023-04-27 23:50:36,340 INFO [train.py:904] (1/8) Epoch 3, batch 7150, loss[loss=0.2372, simple_loss=0.3143, pruned_loss=0.08002, over 16565.00 frames. ], tot_loss[loss=0.2816, simple_loss=0.3508, pruned_loss=0.1062, over 3090119.09 frames. ], batch size: 62, lr: 1.97e-02, grad_scale: 8.0 2023-04-27 23:50:50,445 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.41 vs. limit=2.0 2023-04-27 23:51:51,150 INFO [train.py:904] (1/8) Epoch 3, batch 7200, loss[loss=0.3028, simple_loss=0.3601, pruned_loss=0.1228, over 11994.00 frames. ], tot_loss[loss=0.2786, simple_loss=0.3483, pruned_loss=0.1045, over 3065898.60 frames. ], batch size: 248, lr: 1.97e-02, grad_scale: 8.0 2023-04-27 23:51:55,800 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=27504.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 23:51:58,280 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.5068, 1.6519, 1.3950, 1.6354, 2.0796, 2.0588, 2.3655, 2.4598], device='cuda:1'), covar=tensor([0.0018, 0.0162, 0.0188, 0.0185, 0.0089, 0.0153, 0.0048, 0.0070], device='cuda:1'), in_proj_covar=tensor([0.0054, 0.0120, 0.0122, 0.0122, 0.0114, 0.0125, 0.0078, 0.0099], device='cuda:1'), out_proj_covar=tensor([7.2461e-05, 1.7112e-04, 1.6720e-04, 1.7319e-04, 1.6455e-04, 1.7900e-04, 1.0932e-04, 1.4383e-04], device='cuda:1') 2023-04-27 23:52:45,517 INFO [optim.py:368] (1/8) 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 23:52:46,026 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.4007, 3.3884, 3.2201, 3.3376, 2.9939, 3.3586, 3.1494, 3.2537], device='cuda:1'), covar=tensor([0.0384, 0.0235, 0.0182, 0.0140, 0.0499, 0.0214, 0.1001, 0.0264], device='cuda:1'), in_proj_covar=tensor([0.0142, 0.0126, 0.0166, 0.0138, 0.0196, 0.0158, 0.0123, 0.0170], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-04-27 23:53:12,423 INFO [train.py:904] (1/8) Epoch 3, batch 7250, loss[loss=0.2852, simple_loss=0.3383, pruned_loss=0.1161, over 11517.00 frames. ], tot_loss[loss=0.2749, simple_loss=0.345, pruned_loss=0.1024, over 3066245.15 frames. ], batch size: 248, lr: 1.97e-02, grad_scale: 8.0 2023-04-27 23:53:26,541 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=27560.0, num_to_drop=1, layers_to_drop={2} 2023-04-27 23:54:26,523 INFO [train.py:904] (1/8) Epoch 3, batch 7300, loss[loss=0.2661, simple_loss=0.3589, pruned_loss=0.08661, over 16687.00 frames. ], tot_loss[loss=0.2746, simple_loss=0.3441, pruned_loss=0.1025, over 3047939.07 frames. ], batch size: 89, lr: 1.97e-02, grad_scale: 8.0 2023-04-27 23:54:35,450 INFO [zipformer.py:625] (1/8) 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] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=27608.0, num_to_drop=1, layers_to_drop={0} 2023-04-27 23:54:42,342 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([1.8507, 1.5077, 1.3455, 1.4245, 1.6887, 1.5917, 1.7625, 1.8751], device='cuda:1'), covar=tensor([0.0022, 0.0125, 0.0157, 0.0136, 0.0083, 0.0115, 0.0047, 0.0069], device='cuda:1'), in_proj_covar=tensor([0.0054, 0.0121, 0.0123, 0.0122, 0.0114, 0.0125, 0.0078, 0.0100], device='cuda:1'), out_proj_covar=tensor([7.3241e-05, 1.7227e-04, 1.6818e-04, 1.7263e-04, 1.6412e-04, 1.7945e-04, 1.0951e-04, 1.4557e-04], device='cuda:1') 2023-04-27 23:55:17,368 INFO [optim.py:368] (1/8) Clipping_scale=2.0, grad-norm quartiles 2.622e+02 4.400e+02 5.943e+02 7.946e+02 1.369e+03, threshold=1.189e+03, percent-clipped=13.0 2023-04-27 23:55:40,973 INFO [train.py:904] (1/8) Epoch 3, batch 7350, loss[loss=0.2336, simple_loss=0.3145, pruned_loss=0.07634, over 16631.00 frames. ], tot_loss[loss=0.2749, simple_loss=0.3439, pruned_loss=0.103, over 3028153.72 frames. ], batch size: 62, lr: 1.97e-02, grad_scale: 8.0 2023-04-27 23:55:46,127 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=27654.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 23:56:12,822 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.47 vs. limit=2.0 2023-04-27 23:56:25,284 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.9362, 4.8707, 4.7585, 3.9808, 4.7482, 1.9228, 4.4959, 4.7493], device='cuda:1'), covar=tensor([0.0049, 0.0037, 0.0047, 0.0255, 0.0043, 0.1319, 0.0061, 0.0080], device='cuda:1'), in_proj_covar=tensor([0.0073, 0.0061, 0.0093, 0.0109, 0.0069, 0.0119, 0.0081, 0.0092], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0001, 0.0002, 0.0002, 0.0001, 0.0003, 0.0002, 0.0002], device='cuda:1') 2023-04-27 23:56:59,424 INFO [train.py:904] (1/8) Epoch 3, batch 7400, loss[loss=0.2709, simple_loss=0.3452, pruned_loss=0.09826, over 16912.00 frames. ], tot_loss[loss=0.2773, simple_loss=0.3458, pruned_loss=0.1044, over 3026800.75 frames. ], batch size: 109, lr: 1.96e-02, grad_scale: 4.0 2023-04-27 23:57:20,386 INFO [zipformer.py:625] (1/8) 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] (1/8) 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 23:58:05,886 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.6046, 4.4411, 4.3815, 4.3377, 3.8360, 4.5149, 4.4344, 4.1868], device='cuda:1'), covar=tensor([0.0330, 0.0188, 0.0160, 0.0134, 0.0734, 0.0192, 0.0208, 0.0274], device='cuda:1'), in_proj_covar=tensor([0.0144, 0.0130, 0.0168, 0.0143, 0.0200, 0.0160, 0.0127, 0.0172], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-04-27 23:58:18,284 INFO [train.py:904] (1/8) Epoch 3, batch 7450, loss[loss=0.279, simple_loss=0.3516, pruned_loss=0.1032, over 16927.00 frames. ], tot_loss[loss=0.2793, simple_loss=0.3471, pruned_loss=0.1058, over 3036887.05 frames. ], batch size: 116, lr: 1.96e-02, grad_scale: 4.0 2023-04-27 23:58:42,543 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=2.65 vs. limit=5.0 2023-04-27 23:58:50,128 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.71 vs. limit=2.0 2023-04-27 23:58:58,274 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=27775.0, num_to_drop=1, layers_to_drop={1} 2023-04-27 23:59:39,384 INFO [train.py:904] (1/8) Epoch 3, batch 7500, loss[loss=0.3488, simple_loss=0.3922, pruned_loss=0.1527, over 11616.00 frames. ], tot_loss[loss=0.2799, simple_loss=0.3483, pruned_loss=0.1057, over 3039904.11 frames. ], batch size: 247, lr: 1.96e-02, grad_scale: 4.0 2023-04-27 23:59:44,136 INFO [zipformer.py:625] (1/8) 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] (1/8) 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 00:00:47,671 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.0583, 3.9075, 3.8694, 3.9522, 3.5214, 3.9877, 3.8254, 3.7091], device='cuda:1'), covar=tensor([0.0331, 0.0228, 0.0175, 0.0135, 0.0553, 0.0208, 0.0397, 0.0278], device='cuda:1'), in_proj_covar=tensor([0.0139, 0.0129, 0.0166, 0.0139, 0.0193, 0.0156, 0.0123, 0.0169], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0002], device='cuda:1') 2023-04-28 00:00:55,938 INFO [train.py:904] (1/8) Epoch 3, batch 7550, loss[loss=0.2399, simple_loss=0.3169, pruned_loss=0.08145, over 16906.00 frames. ], tot_loss[loss=0.2795, simple_loss=0.3477, pruned_loss=0.1057, over 3043379.29 frames. ], batch size: 83, lr: 1.96e-02, grad_scale: 4.0 2023-04-28 00:00:58,221 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=27852.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 00:01:09,420 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.8504, 3.7534, 1.6080, 3.9656, 2.4507, 3.9889, 1.8448, 2.7394], device='cuda:1'), covar=tensor([0.0054, 0.0210, 0.1680, 0.0032, 0.0810, 0.0254, 0.1501, 0.0605], device='cuda:1'), in_proj_covar=tensor([0.0088, 0.0131, 0.0170, 0.0079, 0.0160, 0.0159, 0.0180, 0.0159], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-04-28 00:01:26,702 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([1.7927, 3.5682, 3.6450, 2.3520, 3.2311, 3.4300, 3.6143, 1.7738], device='cuda:1'), covar=tensor([0.0334, 0.0024, 0.0027, 0.0223, 0.0042, 0.0065, 0.0022, 0.0309], device='cuda:1'), in_proj_covar=tensor([0.0111, 0.0052, 0.0057, 0.0110, 0.0053, 0.0063, 0.0057, 0.0106], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0001, 0.0002, 0.0003, 0.0001, 0.0002, 0.0002, 0.0003], device='cuda:1') 2023-04-28 00:02:13,359 INFO [train.py:904] (1/8) Epoch 3, batch 7600, loss[loss=0.2721, simple_loss=0.346, pruned_loss=0.09903, over 17197.00 frames. ], tot_loss[loss=0.2794, simple_loss=0.3472, pruned_loss=0.1058, over 3045794.35 frames. ], batch size: 46, lr: 1.96e-02, grad_scale: 8.0 2023-04-28 00:02:59,295 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=2.07 vs. limit=2.0 2023-04-28 00:03:08,896 INFO [optim.py:368] (1/8) 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] (1/8) Epoch 3, batch 7650, loss[loss=0.3477, simple_loss=0.3832, pruned_loss=0.1561, over 11313.00 frames. ], tot_loss[loss=0.2809, simple_loss=0.3483, pruned_loss=0.1068, over 3045290.93 frames. ], batch size: 247, lr: 1.96e-02, grad_scale: 8.0 2023-04-28 00:04:33,124 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([5.3081, 5.5131, 5.2483, 5.3389, 4.8568, 4.7271, 5.0889, 5.6014], device='cuda:1'), covar=tensor([0.0422, 0.0503, 0.0699, 0.0336, 0.0438, 0.0455, 0.0409, 0.0480], device='cuda:1'), in_proj_covar=tensor([0.0272, 0.0380, 0.0338, 0.0245, 0.0243, 0.0252, 0.0305, 0.0262], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-28 00:04:52,685 INFO [train.py:904] (1/8) Epoch 3, batch 7700, loss[loss=0.2494, simple_loss=0.3281, pruned_loss=0.08533, over 15271.00 frames. ], tot_loss[loss=0.2827, simple_loss=0.349, pruned_loss=0.1082, over 3045636.33 frames. ], batch size: 191, lr: 1.95e-02, grad_scale: 8.0 2023-04-28 00:05:46,979 INFO [optim.py:368] (1/8) 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] (1/8) Epoch 3, batch 7750, loss[loss=0.222, simple_loss=0.3087, pruned_loss=0.06771, over 16906.00 frames. ], tot_loss[loss=0.2812, simple_loss=0.3485, pruned_loss=0.107, over 3063640.45 frames. ], batch size: 96, lr: 1.95e-02, grad_scale: 8.0 2023-04-28 00:06:18,123 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.46 vs. limit=2.0 2023-04-28 00:06:18,146 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.38 vs. limit=2.0 2023-04-28 00:06:40,930 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=28070.0, num_to_drop=1, layers_to_drop={3} 2023-04-28 00:07:28,347 INFO [train.py:904] (1/8) Epoch 3, batch 7800, loss[loss=0.2354, simple_loss=0.3234, pruned_loss=0.07364, over 16529.00 frames. ], tot_loss[loss=0.284, simple_loss=0.3501, pruned_loss=0.1089, over 3035177.07 frames. ], batch size: 68, lr: 1.95e-02, grad_scale: 8.0 2023-04-28 00:08:22,845 INFO [optim.py:368] (1/8) 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 00:08:45,059 INFO [train.py:904] (1/8) Epoch 3, batch 7850, loss[loss=0.2839, simple_loss=0.3579, pruned_loss=0.105, over 16357.00 frames. ], tot_loss[loss=0.2833, simple_loss=0.3507, pruned_loss=0.108, over 3040491.48 frames. ], batch size: 146, lr: 1.95e-02, grad_scale: 8.0 2023-04-28 00:10:00,834 INFO [train.py:904] (1/8) Epoch 3, batch 7900, loss[loss=0.3284, simple_loss=0.3683, pruned_loss=0.1442, over 11713.00 frames. ], tot_loss[loss=0.2818, simple_loss=0.3497, pruned_loss=0.1069, over 3043833.28 frames. ], batch size: 248, lr: 1.95e-02, grad_scale: 8.0 2023-04-28 00:10:45,223 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=28229.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 00:10:55,797 INFO [optim.py:368] (1/8) 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 00:11:18,514 INFO [train.py:904] (1/8) Epoch 3, batch 7950, loss[loss=0.368, simple_loss=0.3971, pruned_loss=0.1695, over 11395.00 frames. ], tot_loss[loss=0.2807, simple_loss=0.3486, pruned_loss=0.1064, over 3051913.90 frames. ], batch size: 246, lr: 1.94e-02, grad_scale: 8.0 2023-04-28 00:11:26,026 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=28255.0, num_to_drop=1, layers_to_drop={0} 2023-04-28 00:12:13,496 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=28287.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 00:12:18,162 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=28290.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 00:12:33,513 INFO [train.py:904] (1/8) Epoch 3, batch 8000, loss[loss=0.2686, simple_loss=0.341, pruned_loss=0.09808, over 16679.00 frames. ], tot_loss[loss=0.2806, simple_loss=0.3485, pruned_loss=0.1063, over 3070434.77 frames. ], batch size: 134, lr: 1.94e-02, grad_scale: 8.0 2023-04-28 00:12:56,318 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=28316.0, num_to_drop=1, layers_to_drop={2} 2023-04-28 00:13:27,063 INFO [optim.py:368] (1/8) 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 00:13:45,843 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=28348.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 00:13:49,311 INFO [train.py:904] (1/8) Epoch 3, batch 8050, loss[loss=0.2689, simple_loss=0.3394, pruned_loss=0.09926, over 16288.00 frames. ], tot_loss[loss=0.2799, simple_loss=0.3476, pruned_loss=0.1061, over 3057352.10 frames. ], batch size: 165, lr: 1.94e-02, grad_scale: 8.0 2023-04-28 00:14:18,824 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=28370.0, num_to_drop=1, layers_to_drop={0} 2023-04-28 00:15:05,406 INFO [train.py:904] (1/8) Epoch 3, batch 8100, loss[loss=0.2648, simple_loss=0.3434, pruned_loss=0.09305, over 16593.00 frames. ], tot_loss[loss=0.279, simple_loss=0.3474, pruned_loss=0.1053, over 3065714.37 frames. ], batch size: 68, lr: 1.94e-02, grad_scale: 8.0 2023-04-28 00:15:30,570 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=28418.0, num_to_drop=1, layers_to_drop={1} 2023-04-28 00:15:57,041 INFO [optim.py:368] (1/8) 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] (1/8) Epoch 3, batch 8150, loss[loss=0.3203, simple_loss=0.3657, pruned_loss=0.1374, over 11791.00 frames. ], tot_loss[loss=0.2753, simple_loss=0.3439, pruned_loss=0.1033, over 3090194.29 frames. ], batch size: 247, lr: 1.94e-02, grad_scale: 8.0 2023-04-28 00:16:30,992 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.7242, 1.4552, 1.9041, 2.5785, 2.4570, 2.7614, 1.6818, 2.7071], device='cuda:1'), covar=tensor([0.0040, 0.0220, 0.0136, 0.0093, 0.0073, 0.0062, 0.0182, 0.0051], device='cuda:1'), in_proj_covar=tensor([0.0085, 0.0120, 0.0105, 0.0097, 0.0093, 0.0070, 0.0116, 0.0063], device='cuda:1'), out_proj_covar=tensor([1.2850e-04, 1.8684e-04, 1.6716e-04, 1.5351e-04, 1.4305e-04, 1.0642e-04, 1.7616e-04, 9.4468e-05], device='cuda:1') 2023-04-28 00:17:06,172 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.4488, 2.8525, 2.5058, 2.3795, 2.2705, 2.0521, 2.7677, 3.0025], device='cuda:1'), covar=tensor([0.1330, 0.0543, 0.0971, 0.0917, 0.1480, 0.1143, 0.0374, 0.0417], device='cuda:1'), in_proj_covar=tensor([0.0270, 0.0243, 0.0256, 0.0218, 0.0308, 0.0193, 0.0226, 0.0217], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-28 00:17:14,585 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=28486.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 00:17:23,194 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 3, batch 8200, loss[loss=0.2572, simple_loss=0.3287, pruned_loss=0.09289, over 16558.00 frames. ], tot_loss[loss=0.2721, simple_loss=0.3404, pruned_loss=0.1019, over 3098389.83 frames. ], batch size: 62, lr: 1.94e-02, grad_scale: 8.0 2023-04-28 00:18:31,949 INFO [optim.py:368] (1/8) 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 00:18:50,509 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 3, batch 8250, loss[loss=0.2285, simple_loss=0.3174, pruned_loss=0.06978, over 16476.00 frames. ], tot_loss[loss=0.2711, simple_loss=0.3401, pruned_loss=0.101, over 3041983.23 frames. ], batch size: 68, lr: 1.94e-02, grad_scale: 8.0 2023-04-28 00:19:00,189 INFO [zipformer.py:625] (1/8) 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] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=28585.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 00:20:16,982 INFO [train.py:904] (1/8) Epoch 3, batch 8300, loss[loss=0.223, simple_loss=0.3088, pruned_loss=0.06858, over 16295.00 frames. ], tot_loss[loss=0.2642, simple_loss=0.3358, pruned_loss=0.09627, over 3043478.30 frames. ], batch size: 165, lr: 1.93e-02, grad_scale: 8.0 2023-04-28 00:20:21,231 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=2.01 vs. limit=2.0 2023-04-28 00:20:23,369 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.58 vs. limit=2.0 2023-04-28 00:20:33,938 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=28611.0, num_to_drop=1, layers_to_drop={1} 2023-04-28 00:20:47,225 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.39 vs. limit=2.0 2023-04-28 00:21:14,649 INFO [optim.py:368] (1/8) 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] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=28643.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 00:21:39,704 INFO [train.py:904] (1/8) Epoch 3, batch 8350, loss[loss=0.234, simple_loss=0.3194, pruned_loss=0.07436, over 16954.00 frames. ], tot_loss[loss=0.2594, simple_loss=0.3333, pruned_loss=0.09281, over 3038991.32 frames. ], batch size: 109, lr: 1.93e-02, grad_scale: 8.0 2023-04-28 00:23:00,530 INFO [train.py:904] (1/8) Epoch 3, batch 8400, loss[loss=0.2032, simple_loss=0.2975, pruned_loss=0.05451, over 16767.00 frames. ], tot_loss[loss=0.2545, simple_loss=0.3299, pruned_loss=0.08954, over 3042771.88 frames. ], batch size: 102, lr: 1.93e-02, grad_scale: 8.0 2023-04-28 00:23:16,014 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.51 vs. limit=2.0 2023-04-28 00:23:58,229 INFO [optim.py:368] (1/8) 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 00:24:20,121 INFO [train.py:904] (1/8) Epoch 3, batch 8450, loss[loss=0.2143, simple_loss=0.3048, pruned_loss=0.06188, over 16830.00 frames. ], tot_loss[loss=0.2505, simple_loss=0.3273, pruned_loss=0.08685, over 3057041.56 frames. ], batch size: 102, lr: 1.93e-02, grad_scale: 4.0 2023-04-28 00:24:33,910 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.0173, 3.3460, 3.3519, 2.3638, 3.2062, 3.3509, 3.1846, 1.7802], device='cuda:1'), covar=tensor([0.0325, 0.0020, 0.0028, 0.0217, 0.0030, 0.0038, 0.0032, 0.0341], device='cuda:1'), in_proj_covar=tensor([0.0112, 0.0054, 0.0057, 0.0111, 0.0052, 0.0063, 0.0058, 0.0109], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0001, 0.0002, 0.0003, 0.0001, 0.0002, 0.0002, 0.0003], device='cuda:1') 2023-04-28 00:24:35,160 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.3990, 1.8079, 1.7121, 1.5525, 2.1763, 2.0135, 2.4456, 2.3435], device='cuda:1'), covar=tensor([0.0019, 0.0127, 0.0134, 0.0169, 0.0067, 0.0113, 0.0039, 0.0065], device='cuda:1'), in_proj_covar=tensor([0.0055, 0.0120, 0.0122, 0.0124, 0.0112, 0.0120, 0.0076, 0.0097], device='cuda:1'), out_proj_covar=tensor([7.2900e-05, 1.6676e-04, 1.6632e-04, 1.7295e-04, 1.5921e-04, 1.6880e-04, 1.0511e-04, 1.3767e-04], device='cuda:1') 2023-04-28 00:25:39,200 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([1.9826, 2.7637, 2.7318, 1.7431, 2.8992, 2.8727, 2.5588, 2.4701], device='cuda:1'), covar=tensor([0.0648, 0.0129, 0.0140, 0.0911, 0.0079, 0.0082, 0.0256, 0.0309], device='cuda:1'), in_proj_covar=tensor([0.0133, 0.0079, 0.0076, 0.0140, 0.0069, 0.0070, 0.0108, 0.0116], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003, 0.0003], device='cuda:1') 2023-04-28 00:25:42,011 INFO [train.py:904] (1/8) Epoch 3, batch 8500, loss[loss=0.2258, simple_loss=0.3053, pruned_loss=0.07317, over 16431.00 frames. ], tot_loss[loss=0.2437, simple_loss=0.3223, pruned_loss=0.0826, over 3065355.86 frames. ], batch size: 146, lr: 1.93e-02, grad_scale: 4.0 2023-04-28 00:25:54,397 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.2043, 4.2325, 4.2909, 4.2489, 4.2164, 4.6892, 4.5144, 4.1921], device='cuda:1'), covar=tensor([0.0945, 0.1373, 0.1059, 0.1429, 0.2075, 0.0828, 0.0837, 0.2083], device='cuda:1'), in_proj_covar=tensor([0.0225, 0.0313, 0.0289, 0.0268, 0.0346, 0.0322, 0.0256, 0.0360], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-28 00:26:08,034 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.3880, 4.2322, 3.9238, 1.9430, 3.0622, 2.5045, 3.5477, 4.1380], device='cuda:1'), covar=tensor([0.0252, 0.0472, 0.0402, 0.1640, 0.0712, 0.0929, 0.0714, 0.0590], device='cuda:1'), in_proj_covar=tensor([0.0135, 0.0114, 0.0151, 0.0145, 0.0133, 0.0129, 0.0141, 0.0118], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:1') 2023-04-28 00:26:13,861 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.0391, 3.7299, 3.9888, 4.1679, 4.2651, 3.8164, 4.3062, 4.2173], device='cuda:1'), covar=tensor([0.0519, 0.0671, 0.1033, 0.0469, 0.0354, 0.0692, 0.0323, 0.0389], device='cuda:1'), in_proj_covar=tensor([0.0283, 0.0352, 0.0448, 0.0344, 0.0261, 0.0248, 0.0280, 0.0286], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-28 00:26:40,692 INFO [optim.py:368] (1/8) Clipping_scale=2.0, grad-norm quartiles 2.012e+02 3.668e+02 4.634e+02 5.862e+02 2.485e+03, threshold=9.268e+02, percent-clipped=6.0 2023-04-28 00:26:50,000 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=28842.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 00:27:00,697 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=28848.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 00:27:05,699 INFO [train.py:904] (1/8) Epoch 3, batch 8550, loss[loss=0.2156, simple_loss=0.311, pruned_loss=0.06007, over 16733.00 frames. ], tot_loss[loss=0.2407, simple_loss=0.3194, pruned_loss=0.08097, over 3065699.03 frames. ], batch size: 89, lr: 1.93e-02, grad_scale: 4.0 2023-04-28 00:27:25,996 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=28861.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 00:28:12,794 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=28885.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 00:28:43,306 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.9785, 4.1063, 4.4846, 4.4231, 4.4261, 4.0590, 4.1470, 3.9865], device='cuda:1'), covar=tensor([0.0268, 0.0259, 0.0294, 0.0382, 0.0377, 0.0248, 0.0648, 0.0297], device='cuda:1'), in_proj_covar=tensor([0.0177, 0.0169, 0.0180, 0.0184, 0.0215, 0.0188, 0.0274, 0.0164], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003, 0.0002], device='cuda:1') 2023-04-28 00:28:46,470 INFO [train.py:904] (1/8) Epoch 3, batch 8600, loss[loss=0.2465, simple_loss=0.3124, pruned_loss=0.09033, over 12455.00 frames. ], tot_loss[loss=0.2401, simple_loss=0.3198, pruned_loss=0.08024, over 3043501.06 frames. ], batch size: 247, lr: 1.92e-02, grad_scale: 4.0 2023-04-28 00:29:07,415 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=28911.0, num_to_drop=1, layers_to_drop={1} 2023-04-28 00:29:12,331 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.8276, 2.7541, 1.7683, 2.7729, 2.1986, 2.7942, 1.9721, 2.4657], device='cuda:1'), covar=tensor([0.0098, 0.0277, 0.1149, 0.0072, 0.0644, 0.0427, 0.1062, 0.0477], device='cuda:1'), in_proj_covar=tensor([0.0088, 0.0130, 0.0173, 0.0079, 0.0158, 0.0155, 0.0184, 0.0159], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-04-28 00:29:30,918 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=28922.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 00:29:50,787 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=28933.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 00:29:56,755 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.65 vs. limit=2.0 2023-04-28 00:29:57,554 INFO [optim.py:368] (1/8) 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 00:30:01,061 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.9219, 3.8608, 3.7064, 3.7298, 3.3739, 3.8111, 3.6227, 3.5871], device='cuda:1'), covar=tensor([0.0326, 0.0218, 0.0208, 0.0156, 0.0674, 0.0309, 0.0502, 0.0320], device='cuda:1'), in_proj_covar=tensor([0.0138, 0.0124, 0.0164, 0.0133, 0.0188, 0.0154, 0.0117, 0.0168], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0002], device='cuda:1') 2023-04-28 00:30:11,236 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=28943.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 00:30:25,449 INFO [train.py:904] (1/8) Epoch 3, batch 8650, loss[loss=0.2188, simple_loss=0.2964, pruned_loss=0.07057, over 12031.00 frames. ], tot_loss[loss=0.2367, simple_loss=0.3175, pruned_loss=0.07791, over 3044125.63 frames. ], batch size: 248, lr: 1.92e-02, grad_scale: 4.0 2023-04-28 00:30:45,514 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=28959.0, num_to_drop=1, layers_to_drop={0} 2023-04-28 00:30:55,863 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=4.95 vs. limit=5.0 2023-04-28 00:31:51,196 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.1338, 4.4018, 4.4028, 2.0150, 4.7097, 4.6125, 3.3727, 3.5645], device='cuda:1'), covar=tensor([0.0535, 0.0088, 0.0148, 0.1267, 0.0031, 0.0031, 0.0261, 0.0260], device='cuda:1'), in_proj_covar=tensor([0.0132, 0.0080, 0.0076, 0.0143, 0.0068, 0.0068, 0.0107, 0.0118], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003, 0.0003], device='cuda:1') 2023-04-28 00:31:52,872 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=28991.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 00:32:12,023 INFO [train.py:904] (1/8) Epoch 3, batch 8700, loss[loss=0.2338, simple_loss=0.3166, pruned_loss=0.07553, over 16288.00 frames. ], tot_loss[loss=0.2327, simple_loss=0.3136, pruned_loss=0.07588, over 3054536.56 frames. ], batch size: 165, lr: 1.92e-02, grad_scale: 4.0 2023-04-28 00:32:15,179 INFO [zipformer.py:625] (1/8) 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] (1/8) 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] (1/8) Epoch 3, batch 8750, loss[loss=0.2667, simple_loss=0.3462, pruned_loss=0.09363, over 15555.00 frames. ], tot_loss[loss=0.2318, simple_loss=0.3134, pruned_loss=0.07514, over 3055639.43 frames. ], batch size: 192, lr: 1.92e-02, grad_scale: 4.0 2023-04-28 00:34:18,777 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=29063.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 00:34:27,330 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([1.9830, 2.7501, 2.3279, 3.8803, 3.6917, 3.6328, 1.7134, 2.8407], device='cuda:1'), covar=tensor([0.1140, 0.0446, 0.1006, 0.0056, 0.0172, 0.0311, 0.1088, 0.0615], device='cuda:1'), in_proj_covar=tensor([0.0145, 0.0135, 0.0162, 0.0071, 0.0133, 0.0145, 0.0155, 0.0158], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0004, 0.0003, 0.0004], device='cuda:1') 2023-04-28 00:35:38,457 INFO [train.py:904] (1/8) Epoch 3, batch 8800, loss[loss=0.252, simple_loss=0.3318, pruned_loss=0.08613, over 16674.00 frames. ], tot_loss[loss=0.2286, simple_loss=0.3107, pruned_loss=0.07324, over 3054542.54 frames. ], batch size: 134, lr: 1.92e-02, grad_scale: 8.0 2023-04-28 00:36:38,523 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.4987, 3.6758, 2.9881, 2.1076, 2.5810, 2.0218, 3.6602, 3.6870], device='cuda:1'), covar=tensor([0.1973, 0.0603, 0.1014, 0.1269, 0.1757, 0.1305, 0.0408, 0.0430], device='cuda:1'), in_proj_covar=tensor([0.0255, 0.0234, 0.0245, 0.0208, 0.0238, 0.0187, 0.0209, 0.0194], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-28 00:36:52,045 INFO [optim.py:368] (1/8) 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 00:37:03,534 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=29142.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 00:37:16,920 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 3, batch 8850, loss[loss=0.222, simple_loss=0.2988, pruned_loss=0.07265, over 12486.00 frames. ], tot_loss[loss=0.2286, simple_loss=0.3124, pruned_loss=0.07237, over 3049842.84 frames. ], batch size: 250, lr: 1.92e-02, grad_scale: 8.0 2023-04-28 00:38:45,268 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=29190.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 00:38:57,573 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 3, batch 8900, loss[loss=0.2258, simple_loss=0.3038, pruned_loss=0.07388, over 12759.00 frames. ], tot_loss[loss=0.2283, simple_loss=0.3127, pruned_loss=0.072, over 3031369.16 frames. ], batch size: 248, lr: 1.91e-02, grad_scale: 8.0 2023-04-28 00:39:19,745 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.9326, 5.2106, 4.9770, 4.9260, 4.6155, 4.3885, 4.7293, 5.2896], device='cuda:1'), covar=tensor([0.0485, 0.0540, 0.0610, 0.0317, 0.0462, 0.0548, 0.0438, 0.0504], device='cuda:1'), in_proj_covar=tensor([0.0261, 0.0366, 0.0305, 0.0234, 0.0232, 0.0240, 0.0291, 0.0252], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-28 00:39:40,211 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=29217.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 00:39:57,725 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.8902, 3.1452, 3.4264, 3.3733, 3.3558, 3.1725, 2.9185, 3.2529], device='cuda:1'), covar=tensor([0.0473, 0.0553, 0.0467, 0.0594, 0.0648, 0.0459, 0.1153, 0.0399], device='cuda:1'), in_proj_covar=tensor([0.0172, 0.0162, 0.0172, 0.0178, 0.0207, 0.0182, 0.0268, 0.0161], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0002], device='cuda:1') 2023-04-28 00:40:16,969 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.57 vs. limit=2.0 2023-04-28 00:40:26,646 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.8977, 1.5329, 2.0737, 2.6295, 2.5298, 2.6912, 1.6446, 2.7645], device='cuda:1'), covar=tensor([0.0034, 0.0203, 0.0126, 0.0075, 0.0064, 0.0061, 0.0172, 0.0041], device='cuda:1'), in_proj_covar=tensor([0.0087, 0.0123, 0.0110, 0.0099, 0.0096, 0.0071, 0.0117, 0.0064], device='cuda:1'), out_proj_covar=tensor([1.2962e-04, 1.8897e-04, 1.7350e-04, 1.5381e-04, 1.4636e-04, 1.0601e-04, 1.7799e-04, 9.5032e-05], device='cuda:1') 2023-04-28 00:40:35,266 INFO [zipformer.py:625] (1/8) 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] (1/8) 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 00:40:46,555 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([1.7636, 3.0126, 2.3846, 4.0247, 3.8548, 3.8278, 1.4819, 2.8966], device='cuda:1'), covar=tensor([0.1370, 0.0461, 0.1205, 0.0079, 0.0173, 0.0294, 0.1336, 0.0701], device='cuda:1'), in_proj_covar=tensor([0.0139, 0.0132, 0.0159, 0.0068, 0.0128, 0.0139, 0.0152, 0.0153], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:1') 2023-04-28 00:41:11,852 INFO [train.py:904] (1/8) Epoch 3, batch 8950, loss[loss=0.2261, simple_loss=0.302, pruned_loss=0.07514, over 12681.00 frames. ], tot_loss[loss=0.2296, simple_loss=0.3132, pruned_loss=0.07306, over 3027682.91 frames. ], batch size: 248, lr: 1.91e-02, grad_scale: 8.0 2023-04-28 00:41:39,746 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.57 vs. limit=2.0 2023-04-28 00:42:53,649 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=29297.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 00:43:00,671 INFO [train.py:904] (1/8) Epoch 3, batch 9000, loss[loss=0.2396, simple_loss=0.3058, pruned_loss=0.08669, over 11797.00 frames. ], tot_loss[loss=0.2252, simple_loss=0.3089, pruned_loss=0.07077, over 3024533.97 frames. ], batch size: 247, lr: 1.91e-02, grad_scale: 8.0 2023-04-28 00:43:00,671 INFO [train.py:929] (1/8) Computing validation loss 2023-04-28 00:43:11,837 INFO [train.py:938] (1/8) Epoch 3, validation: loss=0.1886, simple_loss=0.2904, pruned_loss=0.04341, over 944034.00 frames. 2023-04-28 00:43:11,839 INFO [train.py:939] (1/8) Maximum memory allocated so far is 17676MB 2023-04-28 00:44:26,933 INFO [optim.py:368] (1/8) 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] (1/8) Epoch 3, batch 9050, loss[loss=0.213, simple_loss=0.2921, pruned_loss=0.06696, over 15480.00 frames. ], tot_loss[loss=0.2276, simple_loss=0.311, pruned_loss=0.07208, over 3024621.99 frames. ], batch size: 191, lr: 1.91e-02, grad_scale: 8.0 2023-04-28 00:45:12,123 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=29358.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 00:46:41,627 INFO [train.py:904] (1/8) Epoch 3, batch 9100, loss[loss=0.2337, simple_loss=0.3209, pruned_loss=0.07321, over 16366.00 frames. ], tot_loss[loss=0.2286, simple_loss=0.3114, pruned_loss=0.07292, over 3045943.62 frames. ], batch size: 146, lr: 1.91e-02, grad_scale: 8.0 2023-04-28 00:48:08,504 INFO [optim.py:368] (1/8) Clipping_scale=2.0, grad-norm quartiles 3.097e+02 4.371e+02 5.293e+02 6.986e+02 1.215e+03, threshold=1.059e+03, percent-clipped=12.0 2023-04-28 00:48:40,640 INFO [train.py:904] (1/8) Epoch 3, batch 9150, loss[loss=0.2101, simple_loss=0.2976, pruned_loss=0.06131, over 16662.00 frames. ], tot_loss[loss=0.2276, simple_loss=0.311, pruned_loss=0.0721, over 3044339.58 frames. ], batch size: 57, lr: 1.91e-02, grad_scale: 8.0 2023-04-28 00:50:22,366 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.73 vs. limit=2.0 2023-04-28 00:50:24,879 INFO [train.py:904] (1/8) Epoch 3, batch 9200, loss[loss=0.231, simple_loss=0.3158, pruned_loss=0.07315, over 15512.00 frames. ], tot_loss[loss=0.2233, simple_loss=0.3061, pruned_loss=0.07027, over 3051015.29 frames. ], batch size: 190, lr: 1.90e-02, grad_scale: 8.0 2023-04-28 00:50:54,899 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=29517.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 00:51:01,322 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.37 vs. limit=2.0 2023-04-28 00:51:23,413 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.45 vs. limit=2.0 2023-04-28 00:51:32,421 INFO [optim.py:368] (1/8) 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 00:52:01,391 INFO [train.py:904] (1/8) Epoch 3, batch 9250, loss[loss=0.2187, simple_loss=0.3034, pruned_loss=0.067, over 16713.00 frames. ], tot_loss[loss=0.223, simple_loss=0.3055, pruned_loss=0.07029, over 3053288.91 frames. ], batch size: 124, lr: 1.90e-02, grad_scale: 8.0 2023-04-28 00:52:02,490 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=29551.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 00:52:29,198 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=29565.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 00:53:31,859 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=29592.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 00:53:50,315 INFO [train.py:904] (1/8) Epoch 3, batch 9300, loss[loss=0.2093, simple_loss=0.2945, pruned_loss=0.06205, over 16208.00 frames. ], tot_loss[loss=0.2211, simple_loss=0.3036, pruned_loss=0.06927, over 3051206.17 frames. ], batch size: 165, lr: 1.90e-02, grad_scale: 4.0 2023-04-28 00:54:02,348 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.9788, 2.1791, 2.1495, 3.1669, 1.9024, 2.9805, 2.2524, 1.8658], device='cuda:1'), covar=tensor([0.0363, 0.1002, 0.0502, 0.0253, 0.1975, 0.0343, 0.1055, 0.1756], device='cuda:1'), in_proj_covar=tensor([0.0263, 0.0247, 0.0206, 0.0267, 0.0315, 0.0215, 0.0235, 0.0293], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-28 00:54:15,949 INFO [zipformer.py:625] (1/8) 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] (1/8) 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 00:55:12,352 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.58 vs. limit=2.0 2023-04-28 00:55:35,074 INFO [train.py:904] (1/8) Epoch 3, batch 9350, loss[loss=0.22, simple_loss=0.3094, pruned_loss=0.06535, over 16721.00 frames. ], tot_loss[loss=0.2215, simple_loss=0.3043, pruned_loss=0.06935, over 3049888.30 frames. ], batch size: 89, lr: 1.90e-02, grad_scale: 4.0 2023-04-28 00:55:51,525 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=29658.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 00:56:00,140 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=29662.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 00:57:07,657 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.1234, 3.9090, 3.9553, 2.9196, 3.6729, 3.8267, 3.9273, 2.1788], device='cuda:1'), covar=tensor([0.0380, 0.0026, 0.0033, 0.0220, 0.0062, 0.0061, 0.0051, 0.0377], device='cuda:1'), in_proj_covar=tensor([0.0113, 0.0052, 0.0056, 0.0111, 0.0054, 0.0061, 0.0057, 0.0107], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0001, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:1') 2023-04-28 00:57:16,961 INFO [train.py:904] (1/8) Epoch 3, batch 9400, loss[loss=0.2057, simple_loss=0.2822, pruned_loss=0.06466, over 12558.00 frames. ], tot_loss[loss=0.2204, simple_loss=0.3035, pruned_loss=0.06861, over 3051946.94 frames. ], batch size: 248, lr: 1.90e-02, grad_scale: 4.0 2023-04-28 00:57:27,902 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=29706.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 00:57:28,140 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.7391, 2.0645, 2.0075, 2.9521, 1.7880, 2.5898, 1.9883, 1.6622], device='cuda:1'), covar=tensor([0.0526, 0.1354, 0.0691, 0.0368, 0.2529, 0.0530, 0.1497, 0.2404], device='cuda:1'), in_proj_covar=tensor([0.0262, 0.0247, 0.0207, 0.0268, 0.0315, 0.0217, 0.0235, 0.0291], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-28 00:58:01,510 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=29723.0, num_to_drop=1, layers_to_drop={2} 2023-04-28 00:58:32,710 INFO [optim.py:368] (1/8) 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 00:58:58,180 INFO [train.py:904] (1/8) Epoch 3, batch 9450, loss[loss=0.2293, simple_loss=0.3002, pruned_loss=0.07915, over 12429.00 frames. ], tot_loss[loss=0.2218, simple_loss=0.305, pruned_loss=0.06933, over 3022918.58 frames. ], batch size: 248, lr: 1.90e-02, grad_scale: 4.0 2023-04-28 01:00:37,285 INFO [train.py:904] (1/8) Epoch 3, batch 9500, loss[loss=0.2158, simple_loss=0.3039, pruned_loss=0.06384, over 16248.00 frames. ], tot_loss[loss=0.22, simple_loss=0.3036, pruned_loss=0.06817, over 3042139.00 frames. ], batch size: 165, lr: 1.90e-02, grad_scale: 4.0 2023-04-28 01:00:41,154 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=29802.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 01:01:17,672 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.55 vs. limit=2.0 2023-04-28 01:01:50,890 INFO [optim.py:368] (1/8) Clipping_scale=2.0, grad-norm quartiles 2.435e+02 3.682e+02 4.369e+02 5.544e+02 1.182e+03, threshold=8.738e+02, percent-clipped=2.0 2023-04-28 01:02:22,322 INFO [train.py:904] (1/8) Epoch 3, batch 9550, loss[loss=0.2653, simple_loss=0.3476, pruned_loss=0.09155, over 16197.00 frames. ], tot_loss[loss=0.2203, simple_loss=0.3035, pruned_loss=0.06853, over 3052094.08 frames. ], batch size: 165, lr: 1.89e-02, grad_scale: 4.0 2023-04-28 01:02:49,222 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=29863.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 01:03:48,942 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=29892.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 01:04:04,156 INFO [train.py:904] (1/8) Epoch 3, batch 9600, loss[loss=0.261, simple_loss=0.3457, pruned_loss=0.08812, over 16158.00 frames. ], tot_loss[loss=0.2233, simple_loss=0.3055, pruned_loss=0.0705, over 3017031.33 frames. ], batch size: 165, lr: 1.89e-02, grad_scale: 8.0 2023-04-28 01:04:16,286 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=29907.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 01:05:18,577 INFO [optim.py:368] (1/8) 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 01:05:23,703 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=29940.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 01:05:25,991 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.1794, 3.7708, 3.7782, 2.7243, 3.6843, 3.8875, 3.7249, 2.2862], device='cuda:1'), covar=tensor([0.0343, 0.0021, 0.0033, 0.0229, 0.0024, 0.0033, 0.0025, 0.0290], device='cuda:1'), in_proj_covar=tensor([0.0110, 0.0052, 0.0055, 0.0108, 0.0053, 0.0060, 0.0055, 0.0105], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0001, 0.0002, 0.0003, 0.0001, 0.0002, 0.0002, 0.0003], device='cuda:1') 2023-04-28 01:05:51,207 INFO [train.py:904] (1/8) Epoch 3, batch 9650, loss[loss=0.2322, simple_loss=0.3183, pruned_loss=0.07308, over 15325.00 frames. ], tot_loss[loss=0.225, simple_loss=0.3083, pruned_loss=0.07091, over 3033964.70 frames. ], batch size: 191, lr: 1.89e-02, grad_scale: 8.0 2023-04-28 01:06:07,311 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([1.8427, 3.0153, 2.5829, 4.5701, 4.4167, 4.2422, 1.8300, 3.1681], device='cuda:1'), covar=tensor([0.1278, 0.0528, 0.1047, 0.0052, 0.0136, 0.0273, 0.1273, 0.0605], device='cuda:1'), in_proj_covar=tensor([0.0142, 0.0134, 0.0162, 0.0067, 0.0129, 0.0144, 0.0154, 0.0158], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003, 0.0003, 0.0004], device='cuda:1') 2023-04-28 01:06:20,290 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.1556, 2.9891, 2.7177, 1.9811, 2.5492, 2.0813, 2.6570, 2.9575], device='cuda:1'), covar=tensor([0.0264, 0.0438, 0.0451, 0.1343, 0.0623, 0.0925, 0.0595, 0.0437], device='cuda:1'), in_proj_covar=tensor([0.0131, 0.0107, 0.0149, 0.0143, 0.0133, 0.0128, 0.0137, 0.0115], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:1') 2023-04-28 01:07:41,654 INFO [train.py:904] (1/8) Epoch 3, batch 9700, loss[loss=0.2074, simple_loss=0.2978, pruned_loss=0.05855, over 16587.00 frames. ], tot_loss[loss=0.223, simple_loss=0.3064, pruned_loss=0.06979, over 3035553.87 frames. ], batch size: 68, lr: 1.89e-02, grad_scale: 8.0 2023-04-28 01:08:16,002 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=30018.0, num_to_drop=1, layers_to_drop={2} 2023-04-28 01:08:59,980 INFO [optim.py:368] (1/8) Clipping_scale=2.0, grad-norm quartiles 1.947e+02 3.446e+02 4.381e+02 5.437e+02 9.929e+02, threshold=8.763e+02, percent-clipped=3.0 2023-04-28 01:09:24,292 INFO [train.py:904] (1/8) Epoch 3, batch 9750, loss[loss=0.2343, simple_loss=0.3031, pruned_loss=0.08273, over 12364.00 frames. ], tot_loss[loss=0.2226, simple_loss=0.3052, pruned_loss=0.07006, over 3025735.25 frames. ], batch size: 248, lr: 1.89e-02, grad_scale: 8.0 2023-04-28 01:09:25,055 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.6845, 2.7065, 1.6475, 2.7833, 2.0940, 2.7826, 1.9131, 2.4345], device='cuda:1'), covar=tensor([0.0112, 0.0270, 0.1180, 0.0061, 0.0628, 0.0468, 0.1167, 0.0530], device='cuda:1'), in_proj_covar=tensor([0.0090, 0.0127, 0.0169, 0.0076, 0.0151, 0.0150, 0.0174, 0.0153], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003, 0.0004, 0.0003], device='cuda:1') 2023-04-28 01:11:02,931 INFO [train.py:904] (1/8) Epoch 3, batch 9800, loss[loss=0.1912, simple_loss=0.2678, pruned_loss=0.05729, over 12027.00 frames. ], tot_loss[loss=0.2215, simple_loss=0.3054, pruned_loss=0.06881, over 3046225.30 frames. ], batch size: 247, lr: 1.89e-02, grad_scale: 8.0 2023-04-28 01:11:48,328 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=30125.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 01:12:14,772 INFO [optim.py:368] (1/8) 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] (1/8) Epoch 3, batch 9850, loss[loss=0.2275, simple_loss=0.3019, pruned_loss=0.07661, over 12301.00 frames. ], tot_loss[loss=0.2216, simple_loss=0.3064, pruned_loss=0.06839, over 3048095.98 frames. ], batch size: 248, lr: 1.88e-02, grad_scale: 8.0 2023-04-28 01:13:02,424 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=30158.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 01:14:06,709 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=30186.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 01:14:23,277 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.52 vs. limit=2.0 2023-04-28 01:14:39,079 INFO [train.py:904] (1/8) Epoch 3, batch 9900, loss[loss=0.2027, simple_loss=0.2994, pruned_loss=0.05297, over 16792.00 frames. ], tot_loss[loss=0.222, simple_loss=0.3066, pruned_loss=0.06866, over 3025983.46 frames. ], batch size: 83, lr: 1.88e-02, grad_scale: 8.0 2023-04-28 01:14:54,192 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=30207.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 01:14:59,697 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=30209.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 01:15:08,998 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([1.7189, 3.0868, 3.0239, 2.1196, 2.9877, 2.9847, 3.0454, 1.6274], device='cuda:1'), covar=tensor([0.0349, 0.0021, 0.0035, 0.0206, 0.0033, 0.0039, 0.0026, 0.0331], device='cuda:1'), in_proj_covar=tensor([0.0111, 0.0052, 0.0056, 0.0107, 0.0054, 0.0060, 0.0055, 0.0105], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0001, 0.0002, 0.0003, 0.0001, 0.0002, 0.0002, 0.0003], device='cuda:1') 2023-04-28 01:16:05,826 INFO [optim.py:368] (1/8) 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 01:16:35,562 INFO [train.py:904] (1/8) Epoch 3, batch 9950, loss[loss=0.2269, simple_loss=0.3126, pruned_loss=0.07062, over 17061.00 frames. ], tot_loss[loss=0.2229, simple_loss=0.3086, pruned_loss=0.06856, over 3047853.47 frames. ], batch size: 53, lr: 1.88e-02, grad_scale: 8.0 2023-04-28 01:16:46,522 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=30255.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 01:16:48,670 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.9227, 5.3226, 4.9535, 5.1099, 4.5846, 4.5176, 4.6910, 5.3400], device='cuda:1'), covar=tensor([0.0568, 0.0557, 0.0704, 0.0329, 0.0570, 0.0583, 0.0486, 0.0645], device='cuda:1'), in_proj_covar=tensor([0.0257, 0.0358, 0.0306, 0.0236, 0.0236, 0.0239, 0.0288, 0.0257], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-28 01:17:23,122 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=30270.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 01:18:13,850 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.5912, 3.4490, 3.0154, 1.7062, 2.5531, 2.0994, 2.9431, 3.2307], device='cuda:1'), covar=tensor([0.0239, 0.0351, 0.0489, 0.1503, 0.0670, 0.0927, 0.0683, 0.0578], device='cuda:1'), in_proj_covar=tensor([0.0128, 0.0106, 0.0149, 0.0142, 0.0131, 0.0126, 0.0133, 0.0114], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0004, 0.0003, 0.0003, 0.0004, 0.0003], device='cuda:1') 2023-04-28 01:18:20,706 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.9444, 5.2765, 4.9852, 4.9350, 4.5884, 4.4821, 4.7448, 5.2867], device='cuda:1'), covar=tensor([0.0424, 0.0521, 0.0678, 0.0383, 0.0503, 0.0578, 0.0405, 0.0562], device='cuda:1'), in_proj_covar=tensor([0.0261, 0.0361, 0.0312, 0.0241, 0.0240, 0.0242, 0.0292, 0.0259], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-28 01:18:37,070 INFO [train.py:904] (1/8) Epoch 3, batch 10000, loss[loss=0.174, simple_loss=0.2631, pruned_loss=0.04249, over 16257.00 frames. ], tot_loss[loss=0.2205, simple_loss=0.3063, pruned_loss=0.06733, over 3066832.63 frames. ], batch size: 35, lr: 1.88e-02, grad_scale: 8.0 2023-04-28 01:19:11,829 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=30318.0, num_to_drop=1, layers_to_drop={2} 2023-04-28 01:19:24,228 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.25 vs. limit=2.0 2023-04-28 01:19:55,679 INFO [optim.py:368] (1/8) 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 01:20:19,868 INFO [train.py:904] (1/8) Epoch 3, batch 10050, loss[loss=0.2373, simple_loss=0.3156, pruned_loss=0.07954, over 12388.00 frames. ], tot_loss[loss=0.2207, simple_loss=0.3067, pruned_loss=0.06738, over 3069373.77 frames. ], batch size: 250, lr: 1.88e-02, grad_scale: 8.0 2023-04-28 01:20:27,941 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.4173, 4.1762, 4.4657, 4.6845, 4.7350, 4.1631, 4.8033, 4.7357], device='cuda:1'), covar=tensor([0.0517, 0.0560, 0.0907, 0.0364, 0.0418, 0.0598, 0.0346, 0.0301], device='cuda:1'), in_proj_covar=tensor([0.0277, 0.0338, 0.0428, 0.0334, 0.0261, 0.0238, 0.0271, 0.0277], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-28 01:20:38,333 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=30360.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 01:20:41,970 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.4651, 2.2760, 2.0698, 3.7727, 1.7046, 3.4340, 2.1302, 2.0210], device='cuda:1'), covar=tensor([0.0358, 0.1126, 0.0649, 0.0219, 0.2278, 0.0334, 0.1193, 0.1817], device='cuda:1'), in_proj_covar=tensor([0.0263, 0.0251, 0.0207, 0.0265, 0.0314, 0.0220, 0.0234, 0.0294], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-28 01:20:48,231 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.5802, 5.0021, 4.8840, 4.9293, 4.9793, 5.4587, 5.0450, 4.5999], device='cuda:1'), covar=tensor([0.0759, 0.0935, 0.1064, 0.1166, 0.1469, 0.0668, 0.0841, 0.1586], device='cuda:1'), in_proj_covar=tensor([0.0215, 0.0305, 0.0291, 0.0266, 0.0341, 0.0313, 0.0243, 0.0353], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-28 01:20:50,751 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=30366.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 01:21:54,400 INFO [train.py:904] (1/8) Epoch 3, batch 10100, loss[loss=0.1955, simple_loss=0.277, pruned_loss=0.05704, over 12691.00 frames. ], tot_loss[loss=0.2225, simple_loss=0.308, pruned_loss=0.06848, over 3061135.13 frames. ], batch size: 250, lr: 1.88e-02, grad_scale: 8.0 2023-04-28 01:22:36,902 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=30421.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 01:23:00,424 INFO [optim.py:368] (1/8) 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] (1/8) Epoch 4, batch 0, loss[loss=0.2842, simple_loss=0.3624, pruned_loss=0.103, over 17258.00 frames. ], tot_loss[loss=0.2842, simple_loss=0.3624, pruned_loss=0.103, over 17258.00 frames. ], batch size: 52, lr: 1.75e-02, grad_scale: 8.0 2023-04-28 01:23:38,627 INFO [train.py:929] (1/8) Computing validation loss 2023-04-28 01:23:46,513 INFO [train.py:938] (1/8) Epoch 4, validation: loss=0.188, simple_loss=0.2904, pruned_loss=0.04284, over 944034.00 frames. 2023-04-28 01:23:46,514 INFO [train.py:939] (1/8) Maximum memory allocated so far is 17676MB 2023-04-28 01:23:56,880 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.9965, 4.7145, 4.9589, 5.2621, 5.3515, 4.7096, 5.4084, 5.2122], device='cuda:1'), covar=tensor([0.0667, 0.0599, 0.1090, 0.0433, 0.0386, 0.0401, 0.0379, 0.0366], device='cuda:1'), in_proj_covar=tensor([0.0280, 0.0343, 0.0436, 0.0335, 0.0263, 0.0242, 0.0275, 0.0278], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-28 01:23:57,935 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=30458.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 01:24:09,380 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.3557, 3.8942, 3.9940, 2.8041, 3.7904, 3.7901, 3.9092, 2.3089], device='cuda:1'), covar=tensor([0.0293, 0.0015, 0.0027, 0.0200, 0.0029, 0.0038, 0.0022, 0.0278], device='cuda:1'), in_proj_covar=tensor([0.0111, 0.0052, 0.0056, 0.0109, 0.0055, 0.0060, 0.0055, 0.0105], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0001, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:1') 2023-04-28 01:24:29,163 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=30481.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 01:24:56,020 INFO [train.py:904] (1/8) Epoch 4, batch 50, loss[loss=0.2621, simple_loss=0.339, pruned_loss=0.09261, over 16775.00 frames. ], tot_loss[loss=0.2726, simple_loss=0.336, pruned_loss=0.1046, over 753989.10 frames. ], batch size: 57, lr: 1.75e-02, grad_scale: 1.0 2023-04-28 01:25:02,246 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=30506.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 01:25:07,045 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.72 vs. limit=2.0 2023-04-28 01:25:31,734 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.70 vs. limit=2.0 2023-04-28 01:25:49,876 INFO [optim.py:368] (1/8) 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 01:26:02,161 INFO [train.py:904] (1/8) Epoch 4, batch 100, loss[loss=0.2893, simple_loss=0.3367, pruned_loss=0.121, over 16848.00 frames. ], tot_loss[loss=0.2617, simple_loss=0.3273, pruned_loss=0.09809, over 1321456.17 frames. ], batch size: 102, lr: 1.75e-02, grad_scale: 1.0 2023-04-28 01:26:23,714 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=30565.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 01:27:11,704 INFO [train.py:904] (1/8) Epoch 4, batch 150, loss[loss=0.263, simple_loss=0.3422, pruned_loss=0.09186, over 17044.00 frames. ], tot_loss[loss=0.2557, simple_loss=0.323, pruned_loss=0.09417, over 1773486.26 frames. ], batch size: 53, lr: 1.75e-02, grad_scale: 1.0 2023-04-28 01:27:50,835 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.90 vs. limit=2.0 2023-04-28 01:28:04,890 INFO [optim.py:368] (1/8) 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 01:28:18,575 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.4861, 2.3221, 2.3266, 3.8903, 1.8943, 3.4029, 2.1690, 2.2530], device='cuda:1'), covar=tensor([0.0387, 0.1017, 0.0552, 0.0234, 0.1906, 0.0353, 0.1273, 0.1419], device='cuda:1'), in_proj_covar=tensor([0.0272, 0.0261, 0.0215, 0.0277, 0.0324, 0.0229, 0.0243, 0.0318], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-28 01:28:19,172 INFO [train.py:904] (1/8) Epoch 4, batch 200, loss[loss=0.2665, simple_loss=0.32, pruned_loss=0.1065, over 16712.00 frames. ], tot_loss[loss=0.2541, simple_loss=0.3219, pruned_loss=0.09314, over 2116223.74 frames. ], batch size: 134, lr: 1.75e-02, grad_scale: 1.0 2023-04-28 01:28:24,049 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=30655.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 01:28:24,086 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.3418, 1.7933, 2.1207, 3.0381, 2.9416, 3.4502, 1.6116, 3.1227], device='cuda:1'), covar=tensor([0.0040, 0.0180, 0.0155, 0.0083, 0.0080, 0.0053, 0.0175, 0.0053], device='cuda:1'), in_proj_covar=tensor([0.0095, 0.0128, 0.0113, 0.0105, 0.0102, 0.0077, 0.0118, 0.0065], device='cuda:1'), out_proj_covar=tensor([1.4153e-04, 1.9377e-04, 1.7549e-04, 1.6138e-04, 1.5309e-04, 1.1326e-04, 1.7651e-04, 9.7396e-05], device='cuda:1') 2023-04-28 01:28:27,744 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.89 vs. limit=2.0 2023-04-28 01:29:19,522 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.8267, 2.6271, 2.4432, 4.4530, 1.9351, 3.9159, 2.4698, 2.4929], device='cuda:1'), covar=tensor([0.0373, 0.1031, 0.0605, 0.0181, 0.2101, 0.0350, 0.1116, 0.1604], device='cuda:1'), in_proj_covar=tensor([0.0275, 0.0264, 0.0218, 0.0279, 0.0326, 0.0231, 0.0245, 0.0324], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-28 01:29:26,979 INFO [train.py:904] (1/8) Epoch 4, batch 250, loss[loss=0.1985, simple_loss=0.2688, pruned_loss=0.06411, over 15955.00 frames. ], tot_loss[loss=0.2508, simple_loss=0.3186, pruned_loss=0.09147, over 2385573.75 frames. ], batch size: 35, lr: 1.75e-02, grad_scale: 1.0 2023-04-28 01:29:48,579 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=30716.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 01:29:48,757 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=30716.0, num_to_drop=1, layers_to_drop={3} 2023-04-28 01:30:21,642 INFO [optim.py:368] (1/8) 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] (1/8) Epoch 4, batch 300, loss[loss=0.2476, simple_loss=0.3097, pruned_loss=0.09279, over 16684.00 frames. ], tot_loss[loss=0.2448, simple_loss=0.3136, pruned_loss=0.08797, over 2595613.90 frames. ], batch size: 134, lr: 1.75e-02, grad_scale: 1.0 2023-04-28 01:30:47,898 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.9503, 5.3796, 5.0450, 5.1651, 4.6923, 4.5134, 4.8390, 5.3669], device='cuda:1'), covar=tensor([0.0534, 0.0545, 0.0708, 0.0367, 0.0537, 0.0658, 0.0503, 0.0604], device='cuda:1'), in_proj_covar=tensor([0.0298, 0.0421, 0.0362, 0.0272, 0.0274, 0.0273, 0.0334, 0.0294], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-28 01:31:16,606 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=30781.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 01:31:43,559 INFO [train.py:904] (1/8) Epoch 4, batch 350, loss[loss=0.2594, simple_loss=0.3045, pruned_loss=0.1072, over 16907.00 frames. ], tot_loss[loss=0.2399, simple_loss=0.3095, pruned_loss=0.08513, over 2752114.44 frames. ], batch size: 116, lr: 1.74e-02, grad_scale: 1.0 2023-04-28 01:32:03,614 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.3974, 4.5443, 2.2209, 4.9782, 3.1474, 4.9163, 2.5368, 3.5690], device='cuda:1'), covar=tensor([0.0070, 0.0211, 0.1451, 0.0031, 0.0688, 0.0189, 0.1179, 0.0478], device='cuda:1'), in_proj_covar=tensor([0.0097, 0.0140, 0.0169, 0.0080, 0.0157, 0.0164, 0.0176, 0.0154], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0004, 0.0003], device='cuda:1') 2023-04-28 01:32:20,666 INFO [zipformer.py:625] (1/8) 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] (1/8) 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] (1/8) Epoch 4, batch 400, loss[loss=0.2326, simple_loss=0.2999, pruned_loss=0.08264, over 16828.00 frames. ], tot_loss[loss=0.2387, simple_loss=0.3077, pruned_loss=0.0849, over 2860182.98 frames. ], batch size: 83, lr: 1.74e-02, grad_scale: 2.0 2023-04-28 01:33:11,900 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=30865.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 01:33:55,460 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.45 vs. limit=2.0 2023-04-28 01:34:01,538 INFO [train.py:904] (1/8) Epoch 4, batch 450, loss[loss=0.2283, simple_loss=0.2862, pruned_loss=0.08517, over 16893.00 frames. ], tot_loss[loss=0.2356, simple_loss=0.3047, pruned_loss=0.0832, over 2967595.20 frames. ], batch size: 109, lr: 1.74e-02, grad_scale: 2.0 2023-04-28 01:34:05,097 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=30903.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 01:34:17,421 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=30913.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 01:34:35,324 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=30925.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 01:34:45,847 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=30934.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 01:34:56,469 INFO [optim.py:368] (1/8) 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 01:35:09,303 INFO [train.py:904] (1/8) Epoch 4, batch 500, loss[loss=0.2568, simple_loss=0.3087, pruned_loss=0.1024, over 16762.00 frames. ], tot_loss[loss=0.2321, simple_loss=0.3019, pruned_loss=0.08117, over 3052983.78 frames. ], batch size: 124, lr: 1.74e-02, grad_scale: 2.0 2023-04-28 01:35:28,319 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=30964.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 01:35:58,390 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=30986.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 01:36:06,827 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.9410, 4.5218, 4.3722, 5.0912, 5.1885, 4.5817, 5.2270, 5.1376], device='cuda:1'), covar=tensor([0.0635, 0.0766, 0.2149, 0.0781, 0.0664, 0.0526, 0.0634, 0.0546], device='cuda:1'), in_proj_covar=tensor([0.0332, 0.0405, 0.0537, 0.0414, 0.0309, 0.0288, 0.0320, 0.0332], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-28 01:36:09,048 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=30995.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 01:36:17,824 INFO [train.py:904] (1/8) Epoch 4, batch 550, loss[loss=0.1956, simple_loss=0.279, pruned_loss=0.0561, over 17044.00 frames. ], tot_loss[loss=0.2317, simple_loss=0.3015, pruned_loss=0.08097, over 3115144.01 frames. ], batch size: 50, lr: 1.74e-02, grad_scale: 2.0 2023-04-28 01:36:33,159 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=31011.0, num_to_drop=1, layers_to_drop={2} 2023-04-28 01:36:33,253 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=31011.0, num_to_drop=1, layers_to_drop={1} 2023-04-28 01:36:39,294 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=31016.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 01:37:13,481 INFO [optim.py:368] (1/8) Clipping_scale=2.0, grad-norm quartiles 2.455e+02 3.696e+02 4.486e+02 5.421e+02 1.042e+03, threshold=8.971e+02, percent-clipped=3.0 2023-04-28 01:37:28,709 INFO [train.py:904] (1/8) Epoch 4, batch 600, loss[loss=0.2168, simple_loss=0.2738, pruned_loss=0.07996, over 16916.00 frames. ], tot_loss[loss=0.2326, simple_loss=0.3016, pruned_loss=0.08175, over 3164106.26 frames. ], batch size: 96, lr: 1.74e-02, grad_scale: 2.0 2023-04-28 01:37:46,200 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=31064.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 01:37:54,821 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.4767, 4.2270, 3.6400, 1.9570, 3.0045, 2.4768, 3.7735, 3.9069], device='cuda:1'), covar=tensor([0.0260, 0.0533, 0.0451, 0.1545, 0.0679, 0.0879, 0.0518, 0.0716], device='cuda:1'), in_proj_covar=tensor([0.0138, 0.0126, 0.0155, 0.0145, 0.0136, 0.0129, 0.0140, 0.0128], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003], device='cuda:1') 2023-04-28 01:37:58,182 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=31072.0, num_to_drop=1, layers_to_drop={2} 2023-04-28 01:38:19,005 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.4321, 3.2163, 3.9103, 2.9757, 3.7523, 3.8672, 3.8308, 2.2814], device='cuda:1'), covar=tensor([0.0294, 0.0086, 0.0028, 0.0187, 0.0031, 0.0055, 0.0028, 0.0276], device='cuda:1'), in_proj_covar=tensor([0.0110, 0.0057, 0.0058, 0.0110, 0.0057, 0.0063, 0.0056, 0.0103], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:1') 2023-04-28 01:38:36,814 INFO [train.py:904] (1/8) Epoch 4, batch 650, loss[loss=0.2308, simple_loss=0.287, pruned_loss=0.08731, over 15428.00 frames. ], tot_loss[loss=0.23, simple_loss=0.2996, pruned_loss=0.08023, over 3205833.68 frames. ], batch size: 190, lr: 1.74e-02, grad_scale: 2.0 2023-04-28 01:38:40,683 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.1728, 4.4640, 3.6088, 2.4714, 3.1872, 2.3279, 4.7195, 4.4909], device='cuda:1'), covar=tensor([0.1807, 0.0433, 0.0893, 0.1193, 0.2199, 0.1398, 0.0241, 0.0381], device='cuda:1'), in_proj_covar=tensor([0.0273, 0.0249, 0.0262, 0.0222, 0.0289, 0.0203, 0.0223, 0.0233], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0003, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-28 01:39:30,534 INFO [optim.py:368] (1/8) 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 01:39:45,332 INFO [train.py:904] (1/8) Epoch 4, batch 700, loss[loss=0.1814, simple_loss=0.2615, pruned_loss=0.05061, over 15931.00 frames. ], tot_loss[loss=0.2295, simple_loss=0.2993, pruned_loss=0.07988, over 3232799.85 frames. ], batch size: 35, lr: 1.73e-02, grad_scale: 2.0 2023-04-28 01:39:45,709 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.0549, 3.6251, 4.0561, 2.8225, 3.9275, 3.9948, 3.9954, 2.1444], device='cuda:1'), covar=tensor([0.0337, 0.0044, 0.0036, 0.0223, 0.0034, 0.0064, 0.0030, 0.0306], device='cuda:1'), in_proj_covar=tensor([0.0111, 0.0058, 0.0059, 0.0112, 0.0058, 0.0063, 0.0057, 0.0105], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:1') 2023-04-28 01:40:11,204 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.59 vs. limit=2.0 2023-04-28 01:40:53,738 INFO [train.py:904] (1/8) Epoch 4, batch 750, loss[loss=0.2607, simple_loss=0.3169, pruned_loss=0.1022, over 16907.00 frames. ], tot_loss[loss=0.2299, simple_loss=0.3, pruned_loss=0.07996, over 3240765.89 frames. ], batch size: 109, lr: 1.73e-02, grad_scale: 2.0 2023-04-28 01:40:55,451 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.5009, 4.3867, 3.7548, 1.7949, 2.8134, 2.4039, 3.6363, 4.2592], device='cuda:1'), covar=tensor([0.0240, 0.0425, 0.0491, 0.1636, 0.0763, 0.0959, 0.0567, 0.0575], device='cuda:1'), in_proj_covar=tensor([0.0137, 0.0126, 0.0156, 0.0144, 0.0136, 0.0128, 0.0139, 0.0127], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:1') 2023-04-28 01:41:13,987 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.2505, 1.6691, 2.2736, 2.8995, 2.8212, 3.3234, 1.9447, 3.0952], device='cuda:1'), covar=tensor([0.0062, 0.0212, 0.0131, 0.0089, 0.0088, 0.0059, 0.0176, 0.0060], device='cuda:1'), in_proj_covar=tensor([0.0105, 0.0133, 0.0119, 0.0113, 0.0108, 0.0083, 0.0126, 0.0070], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001], device='cuda:1') 2023-04-28 01:41:48,308 INFO [optim.py:368] (1/8) 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] (1/8) Epoch 4, batch 800, loss[loss=0.2755, simple_loss=0.3235, pruned_loss=0.1137, over 16392.00 frames. ], tot_loss[loss=0.2275, simple_loss=0.2983, pruned_loss=0.07838, over 3263471.72 frames. ], batch size: 146, lr: 1.73e-02, grad_scale: 4.0 2023-04-28 01:42:14,288 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=31259.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 01:42:40,319 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=3.77 vs. limit=5.0 2023-04-28 01:42:44,446 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=31281.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 01:42:55,438 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 4, batch 850, loss[loss=0.2542, simple_loss=0.307, pruned_loss=0.1007, over 12517.00 frames. ], tot_loss[loss=0.2285, simple_loss=0.2991, pruned_loss=0.07891, over 3277032.56 frames. ], batch size: 246, lr: 1.73e-02, grad_scale: 4.0 2023-04-28 01:43:24,613 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=31311.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 01:43:37,435 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.9164, 4.1627, 3.3776, 2.4587, 3.0880, 2.3989, 4.3569, 4.2997], device='cuda:1'), covar=tensor([0.2041, 0.0555, 0.1102, 0.1315, 0.2182, 0.1398, 0.0368, 0.0505], device='cuda:1'), in_proj_covar=tensor([0.0269, 0.0244, 0.0258, 0.0222, 0.0288, 0.0198, 0.0220, 0.0233], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-28 01:44:07,361 INFO [optim.py:368] (1/8) 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 01:44:17,670 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.65 vs. limit=2.0 2023-04-28 01:44:19,701 INFO [train.py:904] (1/8) Epoch 4, batch 900, loss[loss=0.2276, simple_loss=0.2956, pruned_loss=0.07979, over 12542.00 frames. ], tot_loss[loss=0.226, simple_loss=0.2976, pruned_loss=0.07719, over 3290630.84 frames. ], batch size: 246, lr: 1.73e-02, grad_scale: 4.0 2023-04-28 01:44:32,186 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=31359.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 01:44:42,742 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=31367.0, num_to_drop=1, layers_to_drop={2} 2023-04-28 01:45:31,310 INFO [train.py:904] (1/8) Epoch 4, batch 950, loss[loss=0.2063, simple_loss=0.2802, pruned_loss=0.06616, over 16660.00 frames. ], tot_loss[loss=0.2245, simple_loss=0.2967, pruned_loss=0.07615, over 3301891.19 frames. ], batch size: 89, lr: 1.73e-02, grad_scale: 4.0 2023-04-28 01:45:41,886 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=31409.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 01:46:26,070 INFO [optim.py:368] (1/8) 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 01:46:38,671 INFO [train.py:904] (1/8) Epoch 4, batch 1000, loss[loss=0.1894, simple_loss=0.2672, pruned_loss=0.05576, over 16857.00 frames. ], tot_loss[loss=0.2244, simple_loss=0.2961, pruned_loss=0.07633, over 3297883.35 frames. ], batch size: 42, lr: 1.73e-02, grad_scale: 4.0 2023-04-28 01:47:06,252 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=31470.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 01:47:11,583 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.89 vs. limit=2.0 2023-04-28 01:47:36,231 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=31491.0, num_to_drop=1, layers_to_drop={0} 2023-04-28 01:47:48,996 INFO [train.py:904] (1/8) Epoch 4, batch 1050, loss[loss=0.2692, simple_loss=0.3191, pruned_loss=0.1097, over 16814.00 frames. ], tot_loss[loss=0.2238, simple_loss=0.2957, pruned_loss=0.076, over 3307196.43 frames. ], batch size: 102, lr: 1.72e-02, grad_scale: 4.0 2023-04-28 01:48:45,339 INFO [optim.py:368] (1/8) 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 01:48:51,894 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=4.76 vs. limit=5.0 2023-04-28 01:49:00,844 INFO [train.py:904] (1/8) Epoch 4, batch 1100, loss[loss=0.2324, simple_loss=0.2998, pruned_loss=0.08247, over 17192.00 frames. ], tot_loss[loss=0.222, simple_loss=0.2945, pruned_loss=0.07479, over 3322451.56 frames. ], batch size: 44, lr: 1.72e-02, grad_scale: 4.0 2023-04-28 01:49:02,394 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=31552.0, num_to_drop=1, layers_to_drop={2} 2023-04-28 01:49:10,975 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=31559.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 01:49:39,142 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.4531, 3.8682, 4.3354, 3.1047, 3.8159, 4.2325, 4.0637, 2.1325], device='cuda:1'), covar=tensor([0.0318, 0.0031, 0.0023, 0.0190, 0.0033, 0.0032, 0.0021, 0.0298], device='cuda:1'), in_proj_covar=tensor([0.0108, 0.0056, 0.0057, 0.0106, 0.0057, 0.0061, 0.0055, 0.0101], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:1') 2023-04-28 01:49:41,937 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=31581.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 01:49:55,806 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 4, batch 1150, loss[loss=0.1924, simple_loss=0.2824, pruned_loss=0.05121, over 17260.00 frames. ], tot_loss[loss=0.223, simple_loss=0.2954, pruned_loss=0.07534, over 3319557.83 frames. ], batch size: 52, lr: 1.72e-02, grad_scale: 4.0 2023-04-28 01:50:09,987 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.1579, 4.0016, 3.8450, 1.6095, 4.1211, 4.0294, 3.2603, 2.8249], device='cuda:1'), covar=tensor([0.1039, 0.0074, 0.0151, 0.1374, 0.0070, 0.0082, 0.0272, 0.0519], device='cuda:1'), in_proj_covar=tensor([0.0138, 0.0080, 0.0079, 0.0141, 0.0074, 0.0074, 0.0111, 0.0123], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003, 0.0003], device='cuda:1') 2023-04-28 01:50:18,982 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=31607.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 01:50:48,414 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=31629.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 01:51:00,549 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=31638.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 01:51:04,270 INFO [optim.py:368] (1/8) 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] (1/8) Epoch 4, batch 1200, loss[loss=0.2175, simple_loss=0.289, pruned_loss=0.07298, over 16781.00 frames. ], tot_loss[loss=0.222, simple_loss=0.2948, pruned_loss=0.07458, over 3322598.69 frames. ], batch size: 39, lr: 1.72e-02, grad_scale: 8.0 2023-04-28 01:51:41,040 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=31667.0, num_to_drop=1, layers_to_drop={0} 2023-04-28 01:52:27,377 INFO [train.py:904] (1/8) Epoch 4, batch 1250, loss[loss=0.24, simple_loss=0.303, pruned_loss=0.08849, over 16601.00 frames. ], tot_loss[loss=0.2233, simple_loss=0.2954, pruned_loss=0.07557, over 3319765.04 frames. ], batch size: 62, lr: 1.72e-02, grad_scale: 8.0 2023-04-28 01:52:31,860 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=31704.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 01:52:47,963 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=31715.0, num_to_drop=1, layers_to_drop={0} 2023-04-28 01:52:54,427 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.1928, 1.7173, 2.3585, 2.9820, 2.8106, 3.2879, 2.0915, 3.1438], device='cuda:1'), covar=tensor([0.0050, 0.0199, 0.0115, 0.0091, 0.0079, 0.0062, 0.0163, 0.0055], device='cuda:1'), in_proj_covar=tensor([0.0101, 0.0129, 0.0115, 0.0109, 0.0106, 0.0081, 0.0123, 0.0068], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001], device='cuda:1') 2023-04-28 01:53:22,524 INFO [optim.py:368] (1/8) 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 01:53:23,234 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.28 vs. limit=2.0 2023-04-28 01:53:35,321 INFO [train.py:904] (1/8) Epoch 4, batch 1300, loss[loss=0.2496, simple_loss=0.3278, pruned_loss=0.08572, over 17050.00 frames. ], tot_loss[loss=0.2229, simple_loss=0.2951, pruned_loss=0.07538, over 3325735.04 frames. ], batch size: 50, lr: 1.72e-02, grad_scale: 8.0 2023-04-28 01:53:47,379 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.3182, 1.4250, 1.8931, 2.2084, 2.3260, 2.3279, 1.6329, 2.3800], device='cuda:1'), covar=tensor([0.0063, 0.0187, 0.0112, 0.0101, 0.0083, 0.0071, 0.0153, 0.0043], device='cuda:1'), in_proj_covar=tensor([0.0101, 0.0129, 0.0116, 0.0111, 0.0107, 0.0081, 0.0124, 0.0069], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001], device='cuda:1') 2023-04-28 01:53:54,700 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=31765.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 01:53:54,887 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=31765.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 01:54:26,634 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 4, batch 1350, loss[loss=0.2106, simple_loss=0.2945, pruned_loss=0.06338, over 17107.00 frames. ], tot_loss[loss=0.2225, simple_loss=0.2949, pruned_loss=0.07501, over 3328836.11 frames. ], batch size: 55, lr: 1.72e-02, grad_scale: 8.0 2023-04-28 01:55:14,625 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=31823.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 01:55:39,667 INFO [optim.py:368] (1/8) Clipping_scale=2.0, grad-norm quartiles 2.154e+02 3.282e+02 4.037e+02 5.139e+02 9.544e+02, threshold=8.075e+02, percent-clipped=2.0 2023-04-28 01:55:48,228 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=31847.0, num_to_drop=1, layers_to_drop={2} 2023-04-28 01:55:50,804 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=31849.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 01:55:52,773 INFO [train.py:904] (1/8) Epoch 4, batch 1400, loss[loss=0.2338, simple_loss=0.3161, pruned_loss=0.0758, over 17063.00 frames. ], tot_loss[loss=0.2225, simple_loss=0.2948, pruned_loss=0.07512, over 3329709.49 frames. ], batch size: 53, lr: 1.72e-02, grad_scale: 8.0 2023-04-28 01:56:32,163 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=31878.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 01:56:40,460 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=31884.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 01:57:02,740 INFO [train.py:904] (1/8) Epoch 4, batch 1450, loss[loss=0.2857, simple_loss=0.3223, pruned_loss=0.1245, over 16887.00 frames. ], tot_loss[loss=0.2214, simple_loss=0.2937, pruned_loss=0.07452, over 3328772.81 frames. ], batch size: 96, lr: 1.71e-02, grad_scale: 4.0 2023-04-28 01:57:22,419 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=31915.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 01:57:56,650 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=31939.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 01:58:00,354 INFO [optim.py:368] (1/8) 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 01:58:12,682 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=31950.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 01:58:13,457 INFO [train.py:904] (1/8) Epoch 4, batch 1500, loss[loss=0.2427, simple_loss=0.3142, pruned_loss=0.08554, over 16720.00 frames. ], tot_loss[loss=0.2208, simple_loss=0.2935, pruned_loss=0.07406, over 3336195.72 frames. ], batch size: 62, lr: 1.71e-02, grad_scale: 4.0 2023-04-28 01:58:47,436 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=31976.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 01:59:26,618 INFO [train.py:904] (1/8) Epoch 4, batch 1550, loss[loss=0.2617, simple_loss=0.3109, pruned_loss=0.1063, over 16757.00 frames. ], tot_loss[loss=0.2231, simple_loss=0.2949, pruned_loss=0.07559, over 3331800.45 frames. ], batch size: 83, lr: 1.71e-02, grad_scale: 4.0 2023-04-28 01:59:40,575 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=32011.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 02:00:21,778 INFO [optim.py:368] (1/8) 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 02:00:30,589 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.3839, 4.2726, 3.9109, 2.0536, 2.7943, 2.4459, 3.7628, 4.2185], device='cuda:1'), covar=tensor([0.0342, 0.0439, 0.0425, 0.1447, 0.0750, 0.0939, 0.0604, 0.0601], device='cuda:1'), in_proj_covar=tensor([0.0143, 0.0131, 0.0153, 0.0143, 0.0135, 0.0127, 0.0140, 0.0132], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:1') 2023-04-28 02:00:34,386 INFO [train.py:904] (1/8) Epoch 4, batch 1600, loss[loss=0.2181, simple_loss=0.2996, pruned_loss=0.06832, over 17071.00 frames. ], tot_loss[loss=0.2248, simple_loss=0.296, pruned_loss=0.07684, over 3331951.28 frames. ], batch size: 55, lr: 1.71e-02, grad_scale: 8.0 2023-04-28 02:00:46,072 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=32060.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 02:00:47,394 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.0455, 1.9489, 2.4031, 2.8312, 2.5915, 3.2702, 2.0975, 3.2924], device='cuda:1'), covar=tensor([0.0064, 0.0165, 0.0123, 0.0097, 0.0090, 0.0058, 0.0157, 0.0035], device='cuda:1'), in_proj_covar=tensor([0.0103, 0.0128, 0.0116, 0.0113, 0.0108, 0.0081, 0.0124, 0.0070], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001], device='cuda:1') 2023-04-28 02:00:52,782 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 4, batch 1650, loss[loss=0.2302, simple_loss=0.3021, pruned_loss=0.07917, over 16485.00 frames. ], tot_loss[loss=0.2267, simple_loss=0.2982, pruned_loss=0.07766, over 3332333.37 frames. ], batch size: 68, lr: 1.71e-02, grad_scale: 8.0 2023-04-28 02:01:51,320 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.8226, 3.8673, 4.2727, 4.2540, 4.2701, 3.8845, 3.9825, 3.9115], device='cuda:1'), covar=tensor([0.0265, 0.0397, 0.0312, 0.0354, 0.0314, 0.0303, 0.0618, 0.0350], device='cuda:1'), in_proj_covar=tensor([0.0215, 0.0211, 0.0220, 0.0216, 0.0262, 0.0234, 0.0330, 0.0190], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:1') 2023-04-28 02:01:58,771 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=32113.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 02:02:16,903 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.6809, 4.5107, 4.3948, 1.8186, 4.8117, 4.7864, 3.3603, 3.6471], device='cuda:1'), covar=tensor([0.0742, 0.0083, 0.0184, 0.1121, 0.0039, 0.0031, 0.0270, 0.0327], device='cuda:1'), in_proj_covar=tensor([0.0140, 0.0082, 0.0081, 0.0143, 0.0073, 0.0077, 0.0111, 0.0126], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003, 0.0003], device='cuda:1') 2023-04-28 02:02:37,463 INFO [optim.py:368] (1/8) 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 02:02:40,717 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=32144.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 02:02:44,912 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=32147.0, num_to_drop=1, layers_to_drop={1} 2023-04-28 02:02:50,055 INFO [train.py:904] (1/8) Epoch 4, batch 1700, loss[loss=0.279, simple_loss=0.3471, pruned_loss=0.1055, over 15497.00 frames. ], tot_loss[loss=0.2302, simple_loss=0.3017, pruned_loss=0.07939, over 3324871.29 frames. ], batch size: 190, lr: 1.71e-02, grad_scale: 8.0 2023-04-28 02:03:29,666 INFO [zipformer.py:625] (1/8) 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] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=32195.0, num_to_drop=1, layers_to_drop={1} 2023-04-28 02:04:01,145 INFO [train.py:904] (1/8) Epoch 4, batch 1750, loss[loss=0.2529, simple_loss=0.319, pruned_loss=0.09346, over 16811.00 frames. ], tot_loss[loss=0.2311, simple_loss=0.303, pruned_loss=0.07965, over 3323160.46 frames. ], batch size: 124, lr: 1.71e-02, grad_scale: 8.0 2023-04-28 02:04:20,369 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.55 vs. limit=2.0 2023-04-28 02:04:46,382 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.47 vs. limit=2.0 2023-04-28 02:04:47,678 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=32234.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 02:04:58,763 INFO [optim.py:368] (1/8) 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 02:05:11,464 INFO [train.py:904] (1/8) Epoch 4, batch 1800, loss[loss=0.2398, simple_loss=0.31, pruned_loss=0.08483, over 16252.00 frames. ], tot_loss[loss=0.2309, simple_loss=0.3039, pruned_loss=0.07892, over 3324547.51 frames. ], batch size: 165, lr: 1.71e-02, grad_scale: 8.0 2023-04-28 02:05:39,579 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=32271.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 02:06:18,944 INFO [train.py:904] (1/8) Epoch 4, batch 1850, loss[loss=0.2042, simple_loss=0.291, pruned_loss=0.05868, over 17137.00 frames. ], tot_loss[loss=0.2307, simple_loss=0.3045, pruned_loss=0.07842, over 3319211.56 frames. ], batch size: 49, lr: 1.70e-02, grad_scale: 4.0 2023-04-28 02:06:26,592 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=32306.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 02:06:28,974 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.7838, 4.7499, 5.3102, 5.3214, 5.3273, 4.8232, 4.8554, 4.6075], device='cuda:1'), covar=tensor([0.0208, 0.0319, 0.0308, 0.0353, 0.0315, 0.0238, 0.0736, 0.0275], device='cuda:1'), in_proj_covar=tensor([0.0221, 0.0217, 0.0225, 0.0224, 0.0270, 0.0243, 0.0344, 0.0200], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:1') 2023-04-28 02:07:18,152 INFO [optim.py:368] (1/8) 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] (1/8) Epoch 4, batch 1900, loss[loss=0.2291, simple_loss=0.3146, pruned_loss=0.07183, over 17143.00 frames. ], tot_loss[loss=0.2294, simple_loss=0.3039, pruned_loss=0.07746, over 3316809.80 frames. ], batch size: 48, lr: 1.70e-02, grad_scale: 4.0 2023-04-28 02:07:43,049 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=32360.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 02:08:41,340 INFO [train.py:904] (1/8) Epoch 4, batch 1950, loss[loss=0.1901, simple_loss=0.2691, pruned_loss=0.05555, over 17012.00 frames. ], tot_loss[loss=0.2272, simple_loss=0.3026, pruned_loss=0.07593, over 3330631.72 frames. ], batch size: 41, lr: 1.70e-02, grad_scale: 4.0 2023-04-28 02:08:44,234 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.7261, 4.4992, 4.6913, 5.0277, 5.1752, 4.3921, 5.1414, 5.0651], device='cuda:1'), covar=tensor([0.0772, 0.0676, 0.1387, 0.0470, 0.0424, 0.0614, 0.0416, 0.0400], device='cuda:1'), in_proj_covar=tensor([0.0355, 0.0427, 0.0572, 0.0437, 0.0325, 0.0319, 0.0345, 0.0357], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-28 02:08:45,541 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.7426, 4.4742, 4.7634, 5.0604, 5.1834, 4.3776, 5.1779, 5.0593], device='cuda:1'), covar=tensor([0.0754, 0.0693, 0.1314, 0.0460, 0.0417, 0.0630, 0.0393, 0.0416], device='cuda:1'), in_proj_covar=tensor([0.0356, 0.0427, 0.0572, 0.0438, 0.0325, 0.0319, 0.0346, 0.0357], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-28 02:08:50,798 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=32408.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 02:09:19,900 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.6400, 4.6512, 4.5578, 3.9715, 4.5449, 1.9877, 4.3847, 4.5754], device='cuda:1'), covar=tensor([0.0070, 0.0051, 0.0090, 0.0285, 0.0067, 0.1285, 0.0076, 0.0104], device='cuda:1'), in_proj_covar=tensor([0.0086, 0.0077, 0.0115, 0.0127, 0.0085, 0.0129, 0.0104, 0.0116], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-04-28 02:09:37,980 INFO [optim.py:368] (1/8) 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 02:09:39,388 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 4, batch 2000, loss[loss=0.2228, simple_loss=0.3064, pruned_loss=0.06958, over 17124.00 frames. ], tot_loss[loss=0.2275, simple_loss=0.3025, pruned_loss=0.07628, over 3334669.53 frames. ], batch size: 47, lr: 1.70e-02, grad_scale: 8.0 2023-04-28 02:10:28,207 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=32479.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 02:10:45,039 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=32492.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 02:10:58,284 INFO [train.py:904] (1/8) Epoch 4, batch 2050, loss[loss=0.2407, simple_loss=0.3214, pruned_loss=0.07998, over 16625.00 frames. ], tot_loss[loss=0.2284, simple_loss=0.3026, pruned_loss=0.07709, over 3327675.89 frames. ], batch size: 57, lr: 1.70e-02, grad_scale: 8.0 2023-04-28 02:11:05,603 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.9311, 4.9445, 5.5506, 5.5630, 5.5086, 5.1336, 5.0678, 4.8640], device='cuda:1'), covar=tensor([0.0206, 0.0269, 0.0249, 0.0288, 0.0302, 0.0238, 0.0576, 0.0275], device='cuda:1'), in_proj_covar=tensor([0.0214, 0.0210, 0.0220, 0.0218, 0.0263, 0.0235, 0.0333, 0.0194], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:1') 2023-04-28 02:11:22,254 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.36 vs. limit=2.0 2023-04-28 02:11:34,834 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=32527.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 02:11:44,271 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=32534.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 02:11:57,876 INFO [optim.py:368] (1/8) 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] (1/8) Epoch 4, batch 2100, loss[loss=0.2067, simple_loss=0.2897, pruned_loss=0.0619, over 16884.00 frames. ], tot_loss[loss=0.2314, simple_loss=0.3045, pruned_loss=0.07917, over 3325019.97 frames. ], batch size: 42, lr: 1.70e-02, grad_scale: 8.0 2023-04-28 02:12:36,512 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=32571.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 02:12:50,998 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=32582.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 02:13:08,186 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.0548, 3.7170, 3.0157, 1.8542, 2.5352, 2.0368, 3.5022, 3.5862], device='cuda:1'), covar=tensor([0.0200, 0.0437, 0.0563, 0.1449, 0.0736, 0.0988, 0.0453, 0.0565], device='cuda:1'), in_proj_covar=tensor([0.0138, 0.0129, 0.0153, 0.0141, 0.0133, 0.0126, 0.0138, 0.0130], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:1') 2023-04-28 02:13:17,653 INFO [train.py:904] (1/8) Epoch 4, batch 2150, loss[loss=0.3314, simple_loss=0.3715, pruned_loss=0.1457, over 11888.00 frames. ], tot_loss[loss=0.232, simple_loss=0.3052, pruned_loss=0.07944, over 3327610.00 frames. ], batch size: 248, lr: 1.70e-02, grad_scale: 8.0 2023-04-28 02:13:24,695 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=32606.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 02:13:37,705 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([1.7768, 3.2399, 2.4024, 4.4803, 4.3114, 4.2220, 1.6348, 2.9829], device='cuda:1'), covar=tensor([0.1332, 0.0408, 0.1146, 0.0077, 0.0266, 0.0254, 0.1256, 0.0679], device='cuda:1'), in_proj_covar=tensor([0.0140, 0.0136, 0.0162, 0.0078, 0.0170, 0.0156, 0.0154, 0.0157], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0003, 0.0004], device='cuda:1') 2023-04-28 02:13:42,051 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=32619.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 02:14:05,100 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([1.9332, 3.3797, 2.6530, 4.6106, 4.4253, 4.2014, 1.6356, 3.1714], device='cuda:1'), covar=tensor([0.1214, 0.0372, 0.0954, 0.0063, 0.0225, 0.0267, 0.1239, 0.0577], device='cuda:1'), in_proj_covar=tensor([0.0139, 0.0135, 0.0161, 0.0078, 0.0169, 0.0155, 0.0153, 0.0155], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0003, 0.0004], device='cuda:1') 2023-04-28 02:14:15,065 INFO [optim.py:368] (1/8) 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 02:14:26,821 INFO [train.py:904] (1/8) Epoch 4, batch 2200, loss[loss=0.2586, simple_loss=0.3398, pruned_loss=0.08867, over 16570.00 frames. ], tot_loss[loss=0.2317, simple_loss=0.3049, pruned_loss=0.0792, over 3329561.00 frames. ], batch size: 68, lr: 1.70e-02, grad_scale: 8.0 2023-04-28 02:14:30,432 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=32654.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 02:14:48,270 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([5.1625, 5.1850, 5.1062, 3.9849, 4.9860, 2.0482, 4.7513, 5.1483], device='cuda:1'), covar=tensor([0.0064, 0.0052, 0.0078, 0.0420, 0.0058, 0.1530, 0.0087, 0.0114], device='cuda:1'), in_proj_covar=tensor([0.0087, 0.0078, 0.0117, 0.0126, 0.0085, 0.0128, 0.0103, 0.0117], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-04-28 02:14:53,036 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.3334, 4.3224, 4.2807, 3.6975, 4.2948, 1.8099, 4.0704, 4.2199], device='cuda:1'), covar=tensor([0.0077, 0.0060, 0.0083, 0.0311, 0.0057, 0.1500, 0.0081, 0.0123], device='cuda:1'), in_proj_covar=tensor([0.0087, 0.0077, 0.0117, 0.0126, 0.0085, 0.0128, 0.0103, 0.0117], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-04-28 02:15:24,586 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.1475, 4.6296, 4.7398, 3.4012, 4.4275, 4.7569, 4.4789, 2.7217], device='cuda:1'), covar=tensor([0.0229, 0.0017, 0.0019, 0.0178, 0.0021, 0.0032, 0.0015, 0.0255], device='cuda:1'), in_proj_covar=tensor([0.0111, 0.0055, 0.0059, 0.0109, 0.0059, 0.0065, 0.0059, 0.0105], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:1') 2023-04-28 02:15:34,030 INFO [train.py:904] (1/8) Epoch 4, batch 2250, loss[loss=0.2631, simple_loss=0.3266, pruned_loss=0.09976, over 16849.00 frames. ], tot_loss[loss=0.2322, simple_loss=0.3055, pruned_loss=0.07949, over 3320182.50 frames. ], batch size: 109, lr: 1.69e-02, grad_scale: 8.0 2023-04-28 02:16:32,021 INFO [optim.py:368] (1/8) Clipping_scale=2.0, grad-norm quartiles 2.152e+02 3.558e+02 4.554e+02 5.446e+02 8.366e+02, threshold=9.108e+02, percent-clipped=0.0 2023-04-28 02:16:42,208 INFO [train.py:904] (1/8) Epoch 4, batch 2300, loss[loss=0.2875, simple_loss=0.3453, pruned_loss=0.1149, over 15281.00 frames. ], tot_loss[loss=0.2329, simple_loss=0.3067, pruned_loss=0.07961, over 3320969.55 frames. ], batch size: 190, lr: 1.69e-02, grad_scale: 4.0 2023-04-28 02:16:43,210 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.54 vs. limit=2.0 2023-04-28 02:17:51,370 INFO [train.py:904] (1/8) Epoch 4, batch 2350, loss[loss=0.3267, simple_loss=0.3765, pruned_loss=0.1385, over 11765.00 frames. ], tot_loss[loss=0.2342, simple_loss=0.3072, pruned_loss=0.08061, over 3321663.02 frames. ], batch size: 246, lr: 1.69e-02, grad_scale: 4.0 2023-04-28 02:17:53,988 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([5.2817, 5.1187, 5.0076, 5.0016, 4.5054, 5.1068, 5.0978, 4.7072], device='cuda:1'), covar=tensor([0.0375, 0.0216, 0.0210, 0.0173, 0.0966, 0.0246, 0.0200, 0.0435], device='cuda:1'), in_proj_covar=tensor([0.0176, 0.0166, 0.0214, 0.0179, 0.0251, 0.0200, 0.0153, 0.0214], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-04-28 02:18:48,983 INFO [optim.py:368] (1/8) 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 02:18:57,193 INFO [train.py:904] (1/8) Epoch 4, batch 2400, loss[loss=0.2078, simple_loss=0.2878, pruned_loss=0.06395, over 16847.00 frames. ], tot_loss[loss=0.2354, simple_loss=0.3082, pruned_loss=0.08129, over 3309401.93 frames. ], batch size: 42, lr: 1.69e-02, grad_scale: 8.0 2023-04-28 02:20:06,099 INFO [train.py:904] (1/8) Epoch 4, batch 2450, loss[loss=0.21, simple_loss=0.2978, pruned_loss=0.0611, over 17261.00 frames. ], tot_loss[loss=0.2355, simple_loss=0.309, pruned_loss=0.08103, over 3306347.50 frames. ], batch size: 52, lr: 1.69e-02, grad_scale: 8.0 2023-04-28 02:21:03,694 INFO [optim.py:368] (1/8) Clipping_scale=2.0, grad-norm quartiles 2.564e+02 3.644e+02 4.364e+02 5.421e+02 9.049e+02, threshold=8.728e+02, percent-clipped=5.0 2023-04-28 02:21:13,283 INFO [train.py:904] (1/8) Epoch 4, batch 2500, loss[loss=0.2382, simple_loss=0.3045, pruned_loss=0.08596, over 16850.00 frames. ], tot_loss[loss=0.2342, simple_loss=0.3077, pruned_loss=0.08033, over 3309870.90 frames. ], batch size: 90, lr: 1.69e-02, grad_scale: 8.0 2023-04-28 02:21:34,391 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=32966.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 02:22:21,148 INFO [train.py:904] (1/8) Epoch 4, batch 2550, loss[loss=0.2023, simple_loss=0.2835, pruned_loss=0.06058, over 17218.00 frames. ], tot_loss[loss=0.233, simple_loss=0.3064, pruned_loss=0.07977, over 3321142.35 frames. ], batch size: 44, lr: 1.69e-02, grad_scale: 8.0 2023-04-28 02:22:57,646 INFO [zipformer.py:625] (1/8) 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] (1/8) 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 02:23:30,980 INFO [train.py:904] (1/8) Epoch 4, batch 2600, loss[loss=0.2742, simple_loss=0.3334, pruned_loss=0.1075, over 12126.00 frames. ], tot_loss[loss=0.2318, simple_loss=0.3057, pruned_loss=0.07895, over 3325307.87 frames. ], batch size: 246, lr: 1.69e-02, grad_scale: 8.0 2023-04-28 02:23:55,338 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 4, batch 2650, loss[loss=0.2387, simple_loss=0.322, pruned_loss=0.07772, over 17053.00 frames. ], tot_loss[loss=0.2322, simple_loss=0.3065, pruned_loss=0.07894, over 3323951.81 frames. ], batch size: 50, lr: 1.68e-02, grad_scale: 8.0 2023-04-28 02:25:19,502 INFO [zipformer.py:625] (1/8) 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] (1/8) 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 02:25:46,693 INFO [train.py:904] (1/8) Epoch 4, batch 2700, loss[loss=0.254, simple_loss=0.3193, pruned_loss=0.09438, over 12486.00 frames. ], tot_loss[loss=0.2321, simple_loss=0.307, pruned_loss=0.07864, over 3319757.33 frames. ], batch size: 246, lr: 1.68e-02, grad_scale: 8.0 2023-04-28 02:25:54,282 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 4, batch 2750, loss[loss=0.2032, simple_loss=0.284, pruned_loss=0.06123, over 17222.00 frames. ], tot_loss[loss=0.231, simple_loss=0.3063, pruned_loss=0.07779, over 3322734.17 frames. ], batch size: 44, lr: 1.68e-02, grad_scale: 8.0 2023-04-28 02:27:16,334 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=33216.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 02:27:54,082 INFO [optim.py:368] (1/8) 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 02:27:55,621 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([5.4313, 5.7695, 5.3962, 5.6141, 5.0227, 4.7120, 5.2195, 5.8292], device='cuda:1'), covar=tensor([0.0556, 0.0665, 0.0988, 0.0418, 0.0550, 0.0582, 0.0640, 0.0644], device='cuda:1'), in_proj_covar=tensor([0.0332, 0.0461, 0.0392, 0.0291, 0.0291, 0.0289, 0.0360, 0.0318], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-28 02:28:04,360 INFO [train.py:904] (1/8) Epoch 4, batch 2800, loss[loss=0.2404, simple_loss=0.3129, pruned_loss=0.08396, over 15419.00 frames. ], tot_loss[loss=0.2307, simple_loss=0.3059, pruned_loss=0.0777, over 3319189.57 frames. ], batch size: 191, lr: 1.68e-02, grad_scale: 8.0 2023-04-28 02:28:13,074 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.2563, 4.2738, 4.6766, 4.7369, 4.6931, 4.2974, 4.3639, 4.2086], device='cuda:1'), covar=tensor([0.0253, 0.0309, 0.0297, 0.0362, 0.0389, 0.0268, 0.0666, 0.0402], device='cuda:1'), in_proj_covar=tensor([0.0223, 0.0221, 0.0227, 0.0230, 0.0278, 0.0237, 0.0346, 0.0207], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:1') 2023-04-28 02:28:27,991 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.6748, 3.7257, 4.1145, 4.1127, 4.1130, 3.7399, 3.8244, 3.8259], device='cuda:1'), covar=tensor([0.0346, 0.0392, 0.0337, 0.0472, 0.0488, 0.0371, 0.0759, 0.0399], device='cuda:1'), in_proj_covar=tensor([0.0224, 0.0222, 0.0227, 0.0231, 0.0279, 0.0237, 0.0348, 0.0207], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:1') 2023-04-28 02:29:10,606 INFO [train.py:904] (1/8) Epoch 4, batch 2850, loss[loss=0.1958, simple_loss=0.2724, pruned_loss=0.05962, over 17001.00 frames. ], tot_loss[loss=0.2301, simple_loss=0.305, pruned_loss=0.0776, over 3312656.65 frames. ], batch size: 41, lr: 1.68e-02, grad_scale: 8.0 2023-04-28 02:29:39,597 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=33322.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 02:30:09,811 INFO [optim.py:368] (1/8) 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] (1/8) Epoch 4, batch 2900, loss[loss=0.3357, simple_loss=0.3635, pruned_loss=0.1539, over 11676.00 frames. ], tot_loss[loss=0.2295, simple_loss=0.3038, pruned_loss=0.07754, over 3320569.38 frames. ], batch size: 246, lr: 1.68e-02, grad_scale: 8.0 2023-04-28 02:31:29,005 INFO [train.py:904] (1/8) Epoch 4, batch 2950, loss[loss=0.3311, simple_loss=0.3765, pruned_loss=0.1428, over 11841.00 frames. ], tot_loss[loss=0.2309, simple_loss=0.3041, pruned_loss=0.07882, over 3305696.69 frames. ], batch size: 246, lr: 1.68e-02, grad_scale: 8.0 2023-04-28 02:32:01,325 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=33425.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 02:32:16,257 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.4396, 2.8011, 2.7317, 4.8718, 1.9992, 4.4904, 2.4946, 2.6669], device='cuda:1'), covar=tensor([0.0334, 0.1186, 0.0634, 0.0173, 0.2417, 0.0343, 0.1341, 0.1955], device='cuda:1'), in_proj_covar=tensor([0.0292, 0.0271, 0.0227, 0.0290, 0.0337, 0.0258, 0.0254, 0.0341], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-28 02:32:26,523 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.3247, 4.2673, 3.7844, 1.8632, 2.8440, 2.5188, 3.4992, 3.9834], device='cuda:1'), covar=tensor([0.0316, 0.0439, 0.0422, 0.1558, 0.0726, 0.0893, 0.0852, 0.0743], device='cuda:1'), in_proj_covar=tensor([0.0142, 0.0129, 0.0153, 0.0141, 0.0131, 0.0127, 0.0143, 0.0133], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:1') 2023-04-28 02:32:27,159 INFO [optim.py:368] (1/8) Clipping_scale=2.0, grad-norm quartiles 2.112e+02 3.646e+02 4.580e+02 6.032e+02 1.054e+03, threshold=9.160e+02, percent-clipped=6.0 2023-04-28 02:32:35,895 INFO [train.py:904] (1/8) Epoch 4, batch 3000, loss[loss=0.3114, simple_loss=0.3653, pruned_loss=0.1287, over 12200.00 frames. ], tot_loss[loss=0.233, simple_loss=0.3052, pruned_loss=0.08033, over 3300710.19 frames. ], batch size: 246, lr: 1.68e-02, grad_scale: 8.0 2023-04-28 02:32:35,895 INFO [train.py:929] (1/8) Computing validation loss 2023-04-28 02:32:41,687 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.9831, 4.9962, 4.8080, 4.6413, 4.2993, 4.9080, 5.0270, 4.6611], device='cuda:1'), covar=tensor([0.0389, 0.0153, 0.0171, 0.0170, 0.0913, 0.0198, 0.0142, 0.0425], device='cuda:1'), in_proj_covar=tensor([0.0179, 0.0169, 0.0215, 0.0181, 0.0254, 0.0202, 0.0157, 0.0218], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-04-28 02:32:45,553 INFO [train.py:938] (1/8) Epoch 4, validation: loss=0.1627, simple_loss=0.2694, pruned_loss=0.02796, over 944034.00 frames. 2023-04-28 02:32:45,553 INFO [train.py:939] (1/8) Maximum memory allocated so far is 17676MB 2023-04-28 02:33:12,893 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=33472.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 02:33:20,913 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.0613, 3.3128, 3.5815, 3.5173, 3.4911, 3.2235, 3.2281, 3.3449], device='cuda:1'), covar=tensor([0.0337, 0.0377, 0.0292, 0.0464, 0.0392, 0.0379, 0.0784, 0.0384], device='cuda:1'), in_proj_covar=tensor([0.0222, 0.0218, 0.0223, 0.0229, 0.0272, 0.0235, 0.0344, 0.0202], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:1') 2023-04-28 02:33:54,600 INFO [train.py:904] (1/8) Epoch 4, batch 3050, loss[loss=0.2319, simple_loss=0.298, pruned_loss=0.08289, over 16850.00 frames. ], tot_loss[loss=0.2329, simple_loss=0.3051, pruned_loss=0.08036, over 3296736.59 frames. ], batch size: 96, lr: 1.67e-02, grad_scale: 4.0 2023-04-28 02:34:07,860 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=33511.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 02:34:38,009 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=33533.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 02:34:54,165 INFO [optim.py:368] (1/8) Clipping_scale=2.0, grad-norm quartiles 1.952e+02 3.585e+02 4.279e+02 5.215e+02 1.682e+03, threshold=8.559e+02, percent-clipped=1.0 2023-04-28 02:35:02,724 INFO [train.py:904] (1/8) Epoch 4, batch 3100, loss[loss=0.1976, simple_loss=0.2775, pruned_loss=0.05887, over 16860.00 frames. ], tot_loss[loss=0.232, simple_loss=0.3048, pruned_loss=0.07965, over 3302978.00 frames. ], batch size: 42, lr: 1.67e-02, grad_scale: 4.0 2023-04-28 02:35:09,974 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.21 vs. limit=2.0 2023-04-28 02:35:11,475 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.7585, 4.4816, 4.6951, 5.0611, 5.1619, 4.5879, 5.0810, 5.1263], device='cuda:1'), covar=tensor([0.0800, 0.0722, 0.1306, 0.0482, 0.0420, 0.0483, 0.0477, 0.0375], device='cuda:1'), in_proj_covar=tensor([0.0354, 0.0431, 0.0578, 0.0454, 0.0338, 0.0318, 0.0353, 0.0366], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-28 02:35:18,969 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.5112, 4.3965, 4.3830, 4.3795, 3.9591, 4.4215, 4.2792, 4.0974], device='cuda:1'), covar=tensor([0.0374, 0.0272, 0.0178, 0.0142, 0.0799, 0.0221, 0.0339, 0.0358], device='cuda:1'), in_proj_covar=tensor([0.0179, 0.0170, 0.0215, 0.0182, 0.0254, 0.0205, 0.0155, 0.0219], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-04-28 02:36:10,677 INFO [train.py:904] (1/8) Epoch 4, batch 3150, loss[loss=0.1947, simple_loss=0.274, pruned_loss=0.05766, over 15906.00 frames. ], tot_loss[loss=0.2319, simple_loss=0.3048, pruned_loss=0.07947, over 3301483.21 frames. ], batch size: 35, lr: 1.67e-02, grad_scale: 4.0 2023-04-28 02:36:35,013 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.61 vs. limit=2.0 2023-04-28 02:36:39,473 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=33622.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 02:37:09,908 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.7615, 1.6441, 2.2004, 2.6009, 2.7061, 2.5631, 1.5371, 2.6597], device='cuda:1'), covar=tensor([0.0050, 0.0197, 0.0128, 0.0095, 0.0062, 0.0097, 0.0200, 0.0056], device='cuda:1'), in_proj_covar=tensor([0.0108, 0.0134, 0.0121, 0.0115, 0.0114, 0.0083, 0.0129, 0.0074], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001], device='cuda:1') 2023-04-28 02:37:11,328 INFO [optim.py:368] (1/8) 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 02:37:18,476 INFO [train.py:904] (1/8) Epoch 4, batch 3200, loss[loss=0.2162, simple_loss=0.3018, pruned_loss=0.06525, over 17121.00 frames. ], tot_loss[loss=0.2297, simple_loss=0.303, pruned_loss=0.07822, over 3310481.78 frames. ], batch size: 49, lr: 1.67e-02, grad_scale: 8.0 2023-04-28 02:37:44,850 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=33670.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 02:37:54,519 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 4, batch 3250, loss[loss=0.2296, simple_loss=0.31, pruned_loss=0.07466, over 16826.00 frames. ], tot_loss[loss=0.2295, simple_loss=0.3028, pruned_loss=0.07813, over 3318790.53 frames. ], batch size: 39, lr: 1.67e-02, grad_scale: 8.0 2023-04-28 02:38:58,386 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=33725.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 02:39:16,377 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=33738.0, num_to_drop=1, layers_to_drop={1} 2023-04-28 02:39:26,584 INFO [optim.py:368] (1/8) 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] (1/8) Epoch 4, batch 3300, loss[loss=0.1959, simple_loss=0.2893, pruned_loss=0.05121, over 17244.00 frames. ], tot_loss[loss=0.2308, simple_loss=0.304, pruned_loss=0.0788, over 3309552.20 frames. ], batch size: 52, lr: 1.67e-02, grad_scale: 8.0 2023-04-28 02:39:43,040 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.89 vs. limit=2.0 2023-04-28 02:39:46,431 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.37 vs. limit=2.0 2023-04-28 02:40:07,409 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=33773.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 02:40:33,097 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.6942, 4.5956, 4.5435, 4.0123, 4.5723, 2.0334, 4.3874, 4.5919], device='cuda:1'), covar=tensor([0.0053, 0.0050, 0.0071, 0.0242, 0.0049, 0.1244, 0.0070, 0.0086], device='cuda:1'), in_proj_covar=tensor([0.0090, 0.0081, 0.0122, 0.0130, 0.0090, 0.0129, 0.0106, 0.0120], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-04-28 02:40:45,511 INFO [train.py:904] (1/8) Epoch 4, batch 3350, loss[loss=0.1683, simple_loss=0.2462, pruned_loss=0.04514, over 16963.00 frames. ], tot_loss[loss=0.2312, simple_loss=0.3044, pruned_loss=0.07895, over 3307327.35 frames. ], batch size: 41, lr: 1.67e-02, grad_scale: 8.0 2023-04-28 02:40:59,029 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=33811.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 02:41:10,781 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.4026, 4.7616, 4.4572, 4.5607, 4.2027, 4.1250, 4.2552, 4.7987], device='cuda:1'), covar=tensor([0.0701, 0.0634, 0.0916, 0.0441, 0.0573, 0.0885, 0.0611, 0.0672], device='cuda:1'), in_proj_covar=tensor([0.0341, 0.0468, 0.0399, 0.0293, 0.0297, 0.0292, 0.0369, 0.0323], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-28 02:41:21,435 INFO [zipformer.py:625] (1/8) 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] (1/8) 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] (1/8) Epoch 4, batch 3400, loss[loss=0.2184, simple_loss=0.2985, pruned_loss=0.0691, over 16455.00 frames. ], tot_loss[loss=0.23, simple_loss=0.3039, pruned_loss=0.07808, over 3302234.34 frames. ], batch size: 68, lr: 1.67e-02, grad_scale: 8.0 2023-04-28 02:42:04,788 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=33859.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 02:42:17,697 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.7957, 4.1860, 3.3372, 2.5112, 3.2092, 2.2914, 4.4367, 4.4626], device='cuda:1'), covar=tensor([0.1951, 0.0567, 0.0955, 0.1274, 0.1898, 0.1259, 0.0298, 0.0381], device='cuda:1'), in_proj_covar=tensor([0.0266, 0.0245, 0.0259, 0.0228, 0.0300, 0.0194, 0.0228, 0.0244], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-28 02:43:03,518 INFO [train.py:904] (1/8) Epoch 4, batch 3450, loss[loss=0.1895, simple_loss=0.2674, pruned_loss=0.05577, over 16990.00 frames. ], tot_loss[loss=0.2272, simple_loss=0.3018, pruned_loss=0.07629, over 3306883.68 frames. ], batch size: 41, lr: 1.66e-02, grad_scale: 8.0 2023-04-28 02:43:57,745 INFO [zipformer.py:625] (1/8) 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] (1/8) 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] (1/8) Epoch 4, batch 3500, loss[loss=0.2347, simple_loss=0.3229, pruned_loss=0.0732, over 17068.00 frames. ], tot_loss[loss=0.2257, simple_loss=0.3003, pruned_loss=0.07559, over 3312685.78 frames. ], batch size: 55, lr: 1.66e-02, grad_scale: 8.0 2023-04-28 02:44:41,348 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.85 vs. limit=2.0 2023-04-28 02:45:13,997 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.3938, 2.3650, 2.0633, 2.2723, 2.7528, 2.7482, 3.6056, 3.1138], device='cuda:1'), covar=tensor([0.0024, 0.0140, 0.0156, 0.0150, 0.0099, 0.0130, 0.0052, 0.0075], device='cuda:1'), in_proj_covar=tensor([0.0077, 0.0139, 0.0134, 0.0135, 0.0131, 0.0139, 0.0106, 0.0118], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002], device='cuda:1') 2023-04-28 02:45:22,689 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.90 vs. limit=2.0 2023-04-28 02:45:26,917 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=34000.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 02:45:27,575 INFO [train.py:904] (1/8) Epoch 4, batch 3550, loss[loss=0.1864, simple_loss=0.2779, pruned_loss=0.04745, over 17198.00 frames. ], tot_loss[loss=0.2248, simple_loss=0.2994, pruned_loss=0.0751, over 3318660.13 frames. ], batch size: 46, lr: 1.66e-02, grad_scale: 8.0 2023-04-28 02:45:32,769 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.8402, 2.5729, 2.5133, 4.4236, 1.9887, 3.8993, 2.4625, 2.5158], device='cuda:1'), covar=tensor([0.0413, 0.1122, 0.0650, 0.0211, 0.2267, 0.0418, 0.1299, 0.1544], device='cuda:1'), in_proj_covar=tensor([0.0298, 0.0276, 0.0231, 0.0292, 0.0340, 0.0260, 0.0256, 0.0342], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-28 02:46:11,008 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=34033.0, num_to_drop=1, layers_to_drop={0} 2023-04-28 02:46:27,190 INFO [zipformer.py:625] (1/8) 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] (1/8) Clipping_scale=2.0, grad-norm quartiles 2.116e+02 3.324e+02 4.237e+02 5.025e+02 1.251e+03, threshold=8.474e+02, percent-clipped=3.0 2023-04-28 02:46:35,602 INFO [train.py:904] (1/8) Epoch 4, batch 3600, loss[loss=0.2013, simple_loss=0.2833, pruned_loss=0.05969, over 17200.00 frames. ], tot_loss[loss=0.2236, simple_loss=0.2982, pruned_loss=0.07456, over 3310208.34 frames. ], batch size: 46, lr: 1.66e-02, grad_scale: 8.0 2023-04-28 02:47:03,192 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([5.2902, 5.5956, 5.3088, 5.3959, 4.9842, 4.7523, 5.1437, 5.6829], device='cuda:1'), covar=tensor([0.0580, 0.0646, 0.0772, 0.0361, 0.0520, 0.0593, 0.0468, 0.0608], device='cuda:1'), in_proj_covar=tensor([0.0341, 0.0464, 0.0392, 0.0290, 0.0296, 0.0288, 0.0364, 0.0319], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-28 02:47:48,227 INFO [train.py:904] (1/8) Epoch 4, batch 3650, loss[loss=0.238, simple_loss=0.2917, pruned_loss=0.09218, over 16804.00 frames. ], tot_loss[loss=0.2233, simple_loss=0.2968, pruned_loss=0.07492, over 3299270.64 frames. ], batch size: 102, lr: 1.66e-02, grad_scale: 8.0 2023-04-28 02:47:54,685 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=34105.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 02:48:11,147 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.4122, 3.6905, 3.6060, 1.7117, 3.7795, 3.8587, 3.0433, 2.7137], device='cuda:1'), covar=tensor([0.0655, 0.0091, 0.0143, 0.1050, 0.0066, 0.0054, 0.0365, 0.0375], device='cuda:1'), in_proj_covar=tensor([0.0136, 0.0082, 0.0082, 0.0140, 0.0073, 0.0075, 0.0113, 0.0126], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003, 0.0003], device='cuda:1') 2023-04-28 02:48:29,198 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=34128.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 02:48:53,739 INFO [optim.py:368] (1/8) 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 02:48:54,257 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 4, batch 3700, loss[loss=0.2221, simple_loss=0.2839, pruned_loss=0.08013, over 16712.00 frames. ], tot_loss[loss=0.2254, simple_loss=0.2964, pruned_loss=0.07716, over 3283397.87 frames. ], batch size: 134, lr: 1.66e-02, grad_scale: 8.0 2023-04-28 02:49:41,417 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=34176.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 02:50:17,552 INFO [train.py:904] (1/8) Epoch 4, batch 3750, loss[loss=0.2752, simple_loss=0.3337, pruned_loss=0.1083, over 11402.00 frames. ], tot_loss[loss=0.2273, simple_loss=0.2974, pruned_loss=0.07857, over 3282777.75 frames. ], batch size: 247, lr: 1.66e-02, grad_scale: 8.0 2023-04-28 02:50:25,300 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=34206.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 02:50:51,530 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.5243, 3.5111, 3.9766, 3.9811, 4.0086, 3.5866, 3.6995, 3.6361], device='cuda:1'), covar=tensor([0.0299, 0.0483, 0.0323, 0.0368, 0.0344, 0.0361, 0.0721, 0.0441], device='cuda:1'), in_proj_covar=tensor([0.0220, 0.0212, 0.0221, 0.0221, 0.0264, 0.0231, 0.0334, 0.0195], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003], device='cuda:1') 2023-04-28 02:51:21,419 INFO [optim.py:368] (1/8) 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] (1/8) Epoch 4, batch 3800, loss[loss=0.2397, simple_loss=0.3008, pruned_loss=0.08924, over 16756.00 frames. ], tot_loss[loss=0.2294, simple_loss=0.2987, pruned_loss=0.0801, over 3276884.38 frames. ], batch size: 134, lr: 1.66e-02, grad_scale: 8.0 2023-04-28 02:52:35,700 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 4, batch 3850, loss[loss=0.2153, simple_loss=0.2835, pruned_loss=0.07353, over 16868.00 frames. ], tot_loss[loss=0.2291, simple_loss=0.2978, pruned_loss=0.08021, over 3285854.86 frames. ], batch size: 124, lr: 1.65e-02, grad_scale: 8.0 2023-04-28 02:53:31,075 INFO [zipformer.py:625] (1/8) 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] (1/8) 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] (1/8) Epoch 4, batch 3900, loss[loss=0.2345, simple_loss=0.2988, pruned_loss=0.08512, over 16892.00 frames. ], tot_loss[loss=0.2295, simple_loss=0.2973, pruned_loss=0.08091, over 3278347.72 frames. ], batch size: 109, lr: 1.65e-02, grad_scale: 4.0 2023-04-28 02:54:40,913 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=34381.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 02:54:57,886 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.56 vs. limit=2.0 2023-04-28 02:55:09,051 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=34400.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 02:55:10,262 INFO [train.py:904] (1/8) Epoch 4, batch 3950, loss[loss=0.2327, simple_loss=0.2827, pruned_loss=0.0913, over 16711.00 frames. ], tot_loss[loss=0.2297, simple_loss=0.2965, pruned_loss=0.0814, over 3265740.98 frames. ], batch size: 83, lr: 1.65e-02, grad_scale: 4.0 2023-04-28 02:55:17,748 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.9551, 4.0037, 4.4655, 4.4320, 4.4060, 4.0212, 4.1280, 3.9498], device='cuda:1'), covar=tensor([0.0217, 0.0375, 0.0230, 0.0283, 0.0323, 0.0279, 0.0612, 0.0359], device='cuda:1'), in_proj_covar=tensor([0.0216, 0.0212, 0.0218, 0.0221, 0.0264, 0.0227, 0.0329, 0.0194], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003], device='cuda:1') 2023-04-28 02:55:40,262 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.9873, 4.9256, 4.8158, 4.7849, 4.3775, 4.8253, 4.7117, 4.5740], device='cuda:1'), covar=tensor([0.0346, 0.0223, 0.0171, 0.0136, 0.0708, 0.0222, 0.0269, 0.0289], device='cuda:1'), in_proj_covar=tensor([0.0172, 0.0163, 0.0206, 0.0173, 0.0240, 0.0195, 0.0149, 0.0211], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-04-28 02:56:07,022 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([5.2240, 5.5248, 5.1558, 5.2961, 4.7864, 4.6892, 4.9669, 5.5267], device='cuda:1'), covar=tensor([0.0563, 0.0549, 0.0919, 0.0381, 0.0652, 0.0597, 0.0538, 0.0586], device='cuda:1'), in_proj_covar=tensor([0.0331, 0.0443, 0.0372, 0.0282, 0.0287, 0.0280, 0.0354, 0.0310], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-28 02:56:16,757 INFO [optim.py:368] (1/8) 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] (1/8) Epoch 4, batch 4000, loss[loss=0.2331, simple_loss=0.3041, pruned_loss=0.08108, over 16881.00 frames. ], tot_loss[loss=0.2284, simple_loss=0.2952, pruned_loss=0.08081, over 3273042.90 frames. ], batch size: 116, lr: 1.65e-02, grad_scale: 8.0 2023-04-28 02:57:02,966 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.9581, 3.8586, 3.8731, 3.3806, 3.9714, 1.8335, 3.7294, 3.7226], device='cuda:1'), covar=tensor([0.0067, 0.0049, 0.0086, 0.0264, 0.0051, 0.1452, 0.0083, 0.0115], device='cuda:1'), in_proj_covar=tensor([0.0089, 0.0078, 0.0116, 0.0126, 0.0087, 0.0128, 0.0103, 0.0115], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-04-28 02:57:27,821 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=4.94 vs. limit=5.0 2023-04-28 02:57:32,767 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.8786, 4.3102, 3.4309, 2.5255, 3.2920, 2.7152, 4.6084, 4.8490], device='cuda:1'), covar=tensor([0.1963, 0.0557, 0.1035, 0.1222, 0.1833, 0.1118, 0.0297, 0.0249], device='cuda:1'), in_proj_covar=tensor([0.0275, 0.0249, 0.0267, 0.0237, 0.0316, 0.0202, 0.0234, 0.0249], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0003, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-28 02:57:36,428 INFO [train.py:904] (1/8) Epoch 4, batch 4050, loss[loss=0.203, simple_loss=0.2848, pruned_loss=0.06061, over 16744.00 frames. ], tot_loss[loss=0.2261, simple_loss=0.2947, pruned_loss=0.07874, over 3274431.61 frames. ], batch size: 124, lr: 1.65e-02, grad_scale: 8.0 2023-04-28 02:57:36,748 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=34501.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 02:58:41,514 INFO [optim.py:368] (1/8) 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 02:58:48,880 INFO [train.py:904] (1/8) Epoch 4, batch 4100, loss[loss=0.223, simple_loss=0.2992, pruned_loss=0.07335, over 16501.00 frames. ], tot_loss[loss=0.2241, simple_loss=0.2946, pruned_loss=0.07676, over 3267681.57 frames. ], batch size: 68, lr: 1.65e-02, grad_scale: 8.0 2023-04-28 02:59:53,944 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=34593.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 02:59:56,525 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 4, batch 4150, loss[loss=0.228, simple_loss=0.3145, pruned_loss=0.07074, over 16844.00 frames. ], tot_loss[loss=0.2329, simple_loss=0.3038, pruned_loss=0.08096, over 3245351.40 frames. ], batch size: 109, lr: 1.65e-02, grad_scale: 8.0 2023-04-28 03:00:15,213 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.46 vs. limit=2.0 2023-04-28 03:00:17,342 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.6724, 2.4882, 1.8158, 2.1570, 3.0167, 2.6350, 3.7441, 3.2724], device='cuda:1'), covar=tensor([0.0014, 0.0117, 0.0175, 0.0155, 0.0070, 0.0123, 0.0032, 0.0059], device='cuda:1'), in_proj_covar=tensor([0.0074, 0.0137, 0.0137, 0.0135, 0.0129, 0.0139, 0.0104, 0.0117], device='cuda:1'), out_proj_covar=tensor([9.6843e-05, 1.8159e-04, 1.7672e-04, 1.7658e-04, 1.7240e-04, 1.8678e-04, 1.3779e-04, 1.5764e-04], device='cuda:1') 2023-04-28 03:00:38,853 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=34622.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 03:00:44,927 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=34626.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 03:01:10,642 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=34643.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 03:01:15,082 INFO [optim.py:368] (1/8) 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 03:01:22,987 INFO [train.py:904] (1/8) Epoch 4, batch 4200, loss[loss=0.2628, simple_loss=0.3434, pruned_loss=0.09112, over 16576.00 frames. ], tot_loss[loss=0.24, simple_loss=0.3122, pruned_loss=0.08386, over 3234867.81 frames. ], batch size: 62, lr: 1.65e-02, grad_scale: 8.0 2023-04-28 03:01:28,638 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=34654.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 03:02:13,246 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=34683.0, num_to_drop=1, layers_to_drop={3} 2023-04-28 03:02:19,296 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=34687.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 03:02:39,113 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 4, batch 4250, loss[loss=0.2146, simple_loss=0.3038, pruned_loss=0.06268, over 15476.00 frames. ], tot_loss[loss=0.241, simple_loss=0.3149, pruned_loss=0.08354, over 3215971.70 frames. ], batch size: 190, lr: 1.65e-02, grad_scale: 8.0 2023-04-28 03:02:53,643 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=34710.0, num_to_drop=1, layers_to_drop={1} 2023-04-28 03:03:05,173 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([5.1612, 5.1350, 4.7939, 4.1992, 5.0164, 1.7888, 4.7314, 4.8161], device='cuda:1'), covar=tensor([0.0038, 0.0033, 0.0062, 0.0261, 0.0032, 0.1471, 0.0058, 0.0086], device='cuda:1'), in_proj_covar=tensor([0.0082, 0.0073, 0.0109, 0.0118, 0.0081, 0.0123, 0.0097, 0.0108], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002], device='cuda:1') 2023-04-28 03:03:14,434 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.2119, 3.0519, 2.7841, 1.8987, 2.5435, 2.0778, 2.7071, 2.9501], device='cuda:1'), covar=tensor([0.0307, 0.0445, 0.0464, 0.1467, 0.0622, 0.0841, 0.0629, 0.0530], device='cuda:1'), in_proj_covar=tensor([0.0135, 0.0125, 0.0153, 0.0141, 0.0134, 0.0125, 0.0140, 0.0128], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:1') 2023-04-28 03:03:22,011 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.48 vs. limit=2.0 2023-04-28 03:03:38,624 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.1061, 4.2501, 4.2617, 4.2246, 4.1120, 4.6643, 4.3570, 4.0471], device='cuda:1'), covar=tensor([0.1235, 0.1274, 0.1167, 0.1443, 0.2375, 0.0875, 0.0985, 0.2178], device='cuda:1'), in_proj_covar=tensor([0.0250, 0.0346, 0.0322, 0.0296, 0.0390, 0.0356, 0.0280, 0.0403], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003], device='cuda:1') 2023-04-28 03:03:47,977 INFO [optim.py:368] (1/8) 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] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=34748.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 03:03:56,215 INFO [train.py:904] (1/8) Epoch 4, batch 4300, loss[loss=0.2609, simple_loss=0.3392, pruned_loss=0.09127, over 17231.00 frames. ], tot_loss[loss=0.2397, simple_loss=0.3152, pruned_loss=0.08208, over 3199718.41 frames. ], batch size: 44, lr: 1.64e-02, grad_scale: 8.0 2023-04-28 03:04:08,929 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.4960, 2.3154, 2.0690, 3.9941, 1.7607, 3.4739, 2.3069, 2.2797], device='cuda:1'), covar=tensor([0.0446, 0.1216, 0.0736, 0.0243, 0.2232, 0.0408, 0.1222, 0.1756], device='cuda:1'), in_proj_covar=tensor([0.0291, 0.0277, 0.0228, 0.0288, 0.0339, 0.0254, 0.0253, 0.0344], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-28 03:04:27,286 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=34771.0, num_to_drop=1, layers_to_drop={3} 2023-04-28 03:05:10,252 INFO [train.py:904] (1/8) Epoch 4, batch 4350, loss[loss=0.2752, simple_loss=0.3479, pruned_loss=0.1013, over 16907.00 frames. ], tot_loss[loss=0.2435, simple_loss=0.3194, pruned_loss=0.08376, over 3193095.22 frames. ], batch size: 116, lr: 1.64e-02, grad_scale: 8.0 2023-04-28 03:05:10,618 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=34801.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 03:06:03,047 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.3487, 3.9208, 3.8507, 1.5783, 4.2691, 4.2808, 3.1116, 3.1419], device='cuda:1'), covar=tensor([0.0778, 0.0115, 0.0220, 0.1255, 0.0023, 0.0037, 0.0304, 0.0368], device='cuda:1'), in_proj_covar=tensor([0.0137, 0.0082, 0.0081, 0.0139, 0.0069, 0.0071, 0.0111, 0.0125], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003, 0.0003], device='cuda:1') 2023-04-28 03:06:15,520 INFO [optim.py:368] (1/8) 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 03:06:19,648 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=34849.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 03:06:22,168 INFO [train.py:904] (1/8) Epoch 4, batch 4400, loss[loss=0.3012, simple_loss=0.3569, pruned_loss=0.1227, over 11642.00 frames. ], tot_loss[loss=0.2456, simple_loss=0.3215, pruned_loss=0.08483, over 3181117.19 frames. ], batch size: 247, lr: 1.64e-02, grad_scale: 8.0 2023-04-28 03:06:41,265 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=2.00 vs. limit=2.0 2023-04-28 03:07:32,100 INFO [train.py:904] (1/8) Epoch 4, batch 4450, loss[loss=0.2439, simple_loss=0.3286, pruned_loss=0.07956, over 15252.00 frames. ], tot_loss[loss=0.247, simple_loss=0.3243, pruned_loss=0.08481, over 3194497.94 frames. ], batch size: 190, lr: 1.64e-02, grad_scale: 8.0 2023-04-28 03:08:36,347 INFO [optim.py:368] (1/8) Clipping_scale=2.0, grad-norm quartiles 1.991e+02 2.921e+02 3.471e+02 4.143e+02 7.527e+02, threshold=6.942e+02, percent-clipped=0.0 2023-04-28 03:08:40,976 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=34949.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 03:08:43,011 INFO [train.py:904] (1/8) Epoch 4, batch 4500, loss[loss=0.2188, simple_loss=0.3002, pruned_loss=0.06874, over 16831.00 frames. ], tot_loss[loss=0.2464, simple_loss=0.3239, pruned_loss=0.08452, over 3208215.22 frames. ], batch size: 83, lr: 1.64e-02, grad_scale: 8.0 2023-04-28 03:08:54,197 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.8188, 2.4892, 2.4903, 4.6082, 1.8870, 3.8418, 2.4949, 2.4821], device='cuda:1'), covar=tensor([0.0407, 0.1252, 0.0674, 0.0193, 0.2520, 0.0391, 0.1173, 0.1866], device='cuda:1'), in_proj_covar=tensor([0.0291, 0.0279, 0.0231, 0.0290, 0.0347, 0.0253, 0.0253, 0.0350], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-28 03:09:23,018 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=34978.0, num_to_drop=1, layers_to_drop={3} 2023-04-28 03:09:28,650 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=34982.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 03:09:54,753 INFO [train.py:904] (1/8) Epoch 4, batch 4550, loss[loss=0.2739, simple_loss=0.3399, pruned_loss=0.1039, over 17044.00 frames. ], tot_loss[loss=0.2463, simple_loss=0.3238, pruned_loss=0.08442, over 3213515.36 frames. ], batch size: 55, lr: 1.64e-02, grad_scale: 8.0 2023-04-28 03:10:48,293 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.83 vs. limit=2.0 2023-04-28 03:10:54,643 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.71 vs. limit=2.0 2023-04-28 03:10:57,621 INFO [optim.py:368] (1/8) 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] (1/8) Epoch 4, batch 4600, loss[loss=0.2063, simple_loss=0.2973, pruned_loss=0.05759, over 16853.00 frames. ], tot_loss[loss=0.2443, simple_loss=0.3226, pruned_loss=0.08299, over 3235112.52 frames. ], batch size: 102, lr: 1.64e-02, grad_scale: 8.0 2023-04-28 03:11:25,507 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=35066.0, num_to_drop=1, layers_to_drop={0} 2023-04-28 03:12:01,882 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=35092.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 03:12:15,719 INFO [train.py:904] (1/8) Epoch 4, batch 4650, loss[loss=0.246, simple_loss=0.3158, pruned_loss=0.08815, over 16605.00 frames. ], tot_loss[loss=0.2438, simple_loss=0.3214, pruned_loss=0.08307, over 3223339.19 frames. ], batch size: 62, lr: 1.64e-02, grad_scale: 8.0 2023-04-28 03:12:42,206 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.41 vs. limit=2.0 2023-04-28 03:13:20,502 INFO [optim.py:368] (1/8) 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 03:13:28,268 INFO [train.py:904] (1/8) Epoch 4, batch 4700, loss[loss=0.2171, simple_loss=0.2964, pruned_loss=0.06895, over 17038.00 frames. ], tot_loss[loss=0.2413, simple_loss=0.3188, pruned_loss=0.08186, over 3221665.98 frames. ], batch size: 55, lr: 1.63e-02, grad_scale: 8.0 2023-04-28 03:13:32,487 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=35153.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 03:14:04,358 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([5.2465, 5.1644, 5.0930, 4.3538, 5.1539, 2.2213, 4.8690, 5.1944], device='cuda:1'), covar=tensor([0.0048, 0.0042, 0.0043, 0.0300, 0.0037, 0.1176, 0.0062, 0.0081], device='cuda:1'), in_proj_covar=tensor([0.0078, 0.0070, 0.0103, 0.0116, 0.0078, 0.0120, 0.0092, 0.0105], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-28 03:14:41,629 INFO [train.py:904] (1/8) Epoch 4, batch 4750, loss[loss=0.206, simple_loss=0.2896, pruned_loss=0.06122, over 16814.00 frames. ], tot_loss[loss=0.2379, simple_loss=0.3153, pruned_loss=0.08028, over 3222492.76 frames. ], batch size: 89, lr: 1.63e-02, grad_scale: 8.0 2023-04-28 03:15:11,589 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=4.04 vs. limit=5.0 2023-04-28 03:15:12,653 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([5.0723, 4.9735, 5.5203, 5.5794, 5.6565, 5.1434, 5.1543, 4.8065], device='cuda:1'), covar=tensor([0.0182, 0.0270, 0.0278, 0.0282, 0.0294, 0.0203, 0.0610, 0.0272], device='cuda:1'), in_proj_covar=tensor([0.0203, 0.0200, 0.0205, 0.0209, 0.0251, 0.0215, 0.0310, 0.0183], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0002], device='cuda:1') 2023-04-28 03:15:23,294 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.35 vs. limit=2.0 2023-04-28 03:15:45,876 INFO [optim.py:368] (1/8) 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 03:15:51,446 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=35249.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 03:15:53,449 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.30 vs. limit=2.0 2023-04-28 03:15:53,750 INFO [train.py:904] (1/8) Epoch 4, batch 4800, loss[loss=0.2433, simple_loss=0.3278, pruned_loss=0.07943, over 16889.00 frames. ], tot_loss[loss=0.2339, simple_loss=0.3116, pruned_loss=0.07813, over 3222799.48 frames. ], batch size: 96, lr: 1.63e-02, grad_scale: 8.0 2023-04-28 03:16:14,794 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=3.00 vs. limit=5.0 2023-04-28 03:16:32,901 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=35278.0, num_to_drop=1, layers_to_drop={1} 2023-04-28 03:16:38,631 INFO [zipformer.py:625] (1/8) 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] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=35297.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 03:17:05,718 INFO [train.py:904] (1/8) Epoch 4, batch 4850, loss[loss=0.2408, simple_loss=0.3301, pruned_loss=0.07575, over 16339.00 frames. ], tot_loss[loss=0.235, simple_loss=0.313, pruned_loss=0.07845, over 3192340.45 frames. ], batch size: 146, lr: 1.63e-02, grad_scale: 8.0 2023-04-28 03:17:41,749 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=35326.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 03:17:47,990 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=35330.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 03:17:50,726 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.4707, 1.8898, 1.6111, 1.5649, 2.1419, 1.9798, 2.1004, 2.3013], device='cuda:1'), covar=tensor([0.0026, 0.0159, 0.0189, 0.0189, 0.0089, 0.0137, 0.0055, 0.0100], device='cuda:1'), in_proj_covar=tensor([0.0067, 0.0134, 0.0138, 0.0136, 0.0129, 0.0141, 0.0098, 0.0115], device='cuda:1'), out_proj_covar=tensor([8.7206e-05, 1.7748e-04, 1.7685e-04, 1.7654e-04, 1.7252e-04, 1.8745e-04, 1.2789e-04, 1.5422e-04], device='cuda:1') 2023-04-28 03:18:11,918 INFO [zipformer.py:625] (1/8) 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] (1/8) 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 03:18:19,084 INFO [train.py:904] (1/8) Epoch 4, batch 4900, loss[loss=0.2561, simple_loss=0.3263, pruned_loss=0.09301, over 11816.00 frames. ], tot_loss[loss=0.2332, simple_loss=0.3125, pruned_loss=0.07697, over 3186038.32 frames. ], batch size: 247, lr: 1.63e-02, grad_scale: 8.0 2023-04-28 03:18:36,746 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=2.78 vs. limit=5.0 2023-04-28 03:18:42,136 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=35366.0, num_to_drop=1, layers_to_drop={0} 2023-04-28 03:18:53,743 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=35374.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 03:18:55,922 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=3.44 vs. limit=5.0 2023-04-28 03:19:32,911 INFO [train.py:904] (1/8) Epoch 4, batch 4950, loss[loss=0.2402, simple_loss=0.333, pruned_loss=0.07372, over 16871.00 frames. ], tot_loss[loss=0.2338, simple_loss=0.313, pruned_loss=0.07727, over 3189489.47 frames. ], batch size: 102, lr: 1.63e-02, grad_scale: 8.0 2023-04-28 03:19:41,844 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=35406.0, num_to_drop=1, layers_to_drop={2} 2023-04-28 03:19:53,435 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=35414.0, num_to_drop=1, layers_to_drop={0} 2023-04-28 03:19:59,858 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.6428, 4.6268, 5.1020, 5.0971, 5.0828, 4.6213, 4.6770, 4.3987], device='cuda:1'), covar=tensor([0.0194, 0.0256, 0.0228, 0.0285, 0.0335, 0.0226, 0.0584, 0.0294], device='cuda:1'), in_proj_covar=tensor([0.0203, 0.0198, 0.0207, 0.0207, 0.0249, 0.0215, 0.0311, 0.0184], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0002], device='cuda:1') 2023-04-28 03:20:22,750 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=35435.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 03:20:24,347 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.3712, 4.1062, 4.0154, 2.7961, 3.5988, 4.0458, 3.7819, 1.9295], device='cuda:1'), covar=tensor([0.0260, 0.0010, 0.0014, 0.0180, 0.0031, 0.0026, 0.0027, 0.0277], device='cuda:1'), in_proj_covar=tensor([0.0109, 0.0048, 0.0053, 0.0107, 0.0056, 0.0061, 0.0056, 0.0101], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0001, 0.0001, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:1') 2023-04-28 03:20:37,429 INFO [optim.py:368] (1/8) 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 03:20:41,763 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=35448.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 03:20:45,295 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=3.24 vs. limit=5.0 2023-04-28 03:20:45,623 INFO [train.py:904] (1/8) Epoch 4, batch 5000, loss[loss=0.2248, simple_loss=0.3152, pruned_loss=0.06715, over 16792.00 frames. ], tot_loss[loss=0.2343, simple_loss=0.3142, pruned_loss=0.07717, over 3201671.59 frames. ], batch size: 102, lr: 1.63e-02, grad_scale: 8.0 2023-04-28 03:21:53,273 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.59 vs. limit=2.0 2023-04-28 03:21:57,693 INFO [train.py:904] (1/8) Epoch 4, batch 5050, loss[loss=0.2376, simple_loss=0.3233, pruned_loss=0.0759, over 16248.00 frames. ], tot_loss[loss=0.2347, simple_loss=0.3147, pruned_loss=0.0774, over 3201017.43 frames. ], batch size: 165, lr: 1.63e-02, grad_scale: 8.0 2023-04-28 03:22:34,516 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.8939, 2.2235, 2.1061, 3.1827, 1.9331, 2.7973, 2.2258, 1.9710], device='cuda:1'), covar=tensor([0.0444, 0.1115, 0.0704, 0.0306, 0.2062, 0.0515, 0.1281, 0.1633], device='cuda:1'), in_proj_covar=tensor([0.0295, 0.0277, 0.0234, 0.0293, 0.0344, 0.0250, 0.0256, 0.0344], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-28 03:22:39,977 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=35530.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 03:23:03,493 INFO [optim.py:368] (1/8) 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 03:23:10,553 INFO [train.py:904] (1/8) Epoch 4, batch 5100, loss[loss=0.2212, simple_loss=0.3033, pruned_loss=0.06954, over 16624.00 frames. ], tot_loss[loss=0.2329, simple_loss=0.3129, pruned_loss=0.07649, over 3200975.29 frames. ], batch size: 75, lr: 1.63e-02, grad_scale: 8.0 2023-04-28 03:24:08,072 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=35591.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 03:24:22,825 INFO [train.py:904] (1/8) Epoch 4, batch 5150, loss[loss=0.2277, simple_loss=0.3008, pruned_loss=0.07728, over 16211.00 frames. ], tot_loss[loss=0.2323, simple_loss=0.3131, pruned_loss=0.07579, over 3204716.81 frames. ], batch size: 35, lr: 1.62e-02, grad_scale: 8.0 2023-04-28 03:25:29,063 INFO [optim.py:368] (1/8) 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] (1/8) Epoch 4, batch 5200, loss[loss=0.2041, simple_loss=0.2896, pruned_loss=0.0593, over 16758.00 frames. ], tot_loss[loss=0.2314, simple_loss=0.3123, pruned_loss=0.07524, over 3207144.42 frames. ], batch size: 83, lr: 1.62e-02, grad_scale: 8.0 2023-04-28 03:26:37,499 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([1.7937, 2.9143, 2.5601, 4.4481, 3.9673, 4.0074, 1.3588, 3.0482], device='cuda:1'), covar=tensor([0.1238, 0.0472, 0.1017, 0.0050, 0.0146, 0.0224, 0.1362, 0.0648], device='cuda:1'), in_proj_covar=tensor([0.0141, 0.0133, 0.0162, 0.0074, 0.0148, 0.0154, 0.0154, 0.0158], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0004, 0.0003, 0.0004], device='cuda:1') 2023-04-28 03:26:46,333 INFO [train.py:904] (1/8) Epoch 4, batch 5250, loss[loss=0.2261, simple_loss=0.3108, pruned_loss=0.0707, over 15428.00 frames. ], tot_loss[loss=0.2303, simple_loss=0.3102, pruned_loss=0.07522, over 3206043.62 frames. ], batch size: 190, lr: 1.62e-02, grad_scale: 8.0 2023-04-28 03:26:46,853 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=35701.0, num_to_drop=1, layers_to_drop={0} 2023-04-28 03:27:28,537 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=35730.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 03:27:28,798 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.5210, 3.6412, 3.0665, 2.3161, 2.7907, 2.2593, 3.6668, 3.9359], device='cuda:1'), covar=tensor([0.1923, 0.0643, 0.0960, 0.1289, 0.1718, 0.1186, 0.0418, 0.0462], device='cuda:1'), in_proj_covar=tensor([0.0267, 0.0241, 0.0253, 0.0226, 0.0291, 0.0192, 0.0228, 0.0229], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-28 03:27:52,066 INFO [optim.py:368] (1/8) 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 03:27:55,701 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 4, batch 5300, loss[loss=0.1803, simple_loss=0.2552, pruned_loss=0.0527, over 17007.00 frames. ], tot_loss[loss=0.2265, simple_loss=0.3057, pruned_loss=0.07366, over 3209475.84 frames. ], batch size: 41, lr: 1.62e-02, grad_scale: 8.0 2023-04-28 03:28:17,079 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.8865, 1.4470, 2.1276, 2.9233, 2.5829, 3.0551, 1.6991, 3.0048], device='cuda:1'), covar=tensor([0.0047, 0.0228, 0.0126, 0.0079, 0.0093, 0.0057, 0.0199, 0.0037], device='cuda:1'), in_proj_covar=tensor([0.0100, 0.0127, 0.0118, 0.0111, 0.0113, 0.0080, 0.0129, 0.0074], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001], device='cuda:1') 2023-04-28 03:28:28,813 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=35771.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 03:28:55,242 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.7721, 3.9850, 3.2308, 2.4379, 3.1845, 2.5067, 4.0950, 4.3193], device='cuda:1'), covar=tensor([0.1930, 0.0629, 0.0996, 0.1242, 0.1594, 0.1074, 0.0366, 0.0418], device='cuda:1'), in_proj_covar=tensor([0.0266, 0.0238, 0.0252, 0.0227, 0.0294, 0.0191, 0.0226, 0.0228], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-28 03:29:00,665 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=2.01 vs. limit=2.0 2023-04-28 03:29:02,750 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 4, batch 5350, loss[loss=0.2255, simple_loss=0.2986, pruned_loss=0.07621, over 12328.00 frames. ], tot_loss[loss=0.2252, simple_loss=0.3042, pruned_loss=0.07305, over 3191944.59 frames. ], batch size: 246, lr: 1.62e-02, grad_scale: 4.0 2023-04-28 03:29:24,884 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=35811.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 03:29:47,549 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.0168, 3.2421, 3.5174, 3.5046, 3.4482, 3.1956, 3.2478, 3.3162], device='cuda:1'), covar=tensor([0.0288, 0.0461, 0.0306, 0.0334, 0.0422, 0.0326, 0.0700, 0.0388], device='cuda:1'), in_proj_covar=tensor([0.0209, 0.0203, 0.0209, 0.0217, 0.0257, 0.0223, 0.0317, 0.0189], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0002], device='cuda:1') 2023-04-28 03:29:56,968 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=35832.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 03:30:17,961 INFO [optim.py:368] (1/8) 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] (1/8) Epoch 4, batch 5400, loss[loss=0.2673, simple_loss=0.3428, pruned_loss=0.09585, over 16740.00 frames. ], tot_loss[loss=0.2282, simple_loss=0.3078, pruned_loss=0.07432, over 3188072.20 frames. ], batch size: 124, lr: 1.62e-02, grad_scale: 4.0 2023-04-28 03:30:53,681 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=35872.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 03:31:13,840 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=35886.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 03:31:27,246 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([1.9296, 2.7485, 2.6917, 1.7289, 2.8896, 2.8803, 2.5253, 2.2986], device='cuda:1'), covar=tensor([0.0720, 0.0120, 0.0138, 0.0933, 0.0076, 0.0084, 0.0302, 0.0419], device='cuda:1'), in_proj_covar=tensor([0.0137, 0.0082, 0.0077, 0.0137, 0.0067, 0.0072, 0.0110, 0.0125], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003, 0.0003], device='cuda:1') 2023-04-28 03:31:38,572 INFO [train.py:904] (1/8) Epoch 4, batch 5450, loss[loss=0.2882, simple_loss=0.3515, pruned_loss=0.1125, over 16243.00 frames. ], tot_loss[loss=0.2339, simple_loss=0.3124, pruned_loss=0.07772, over 3161757.82 frames. ], batch size: 165, lr: 1.62e-02, grad_scale: 4.0 2023-04-28 03:32:17,295 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.3218, 4.2030, 4.1058, 4.0774, 3.7149, 4.1901, 4.0124, 3.8454], device='cuda:1'), covar=tensor([0.0422, 0.0389, 0.0202, 0.0173, 0.0772, 0.0335, 0.0415, 0.0429], device='cuda:1'), in_proj_covar=tensor([0.0170, 0.0162, 0.0203, 0.0169, 0.0233, 0.0194, 0.0147, 0.0210], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-04-28 03:32:50,253 INFO [optim.py:368] (1/8) 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] (1/8) Epoch 4, batch 5500, loss[loss=0.2539, simple_loss=0.333, pruned_loss=0.08745, over 17104.00 frames. ], tot_loss[loss=0.2459, simple_loss=0.3221, pruned_loss=0.08486, over 3151120.49 frames. ], batch size: 47, lr: 1.62e-02, grad_scale: 4.0 2023-04-28 03:33:15,851 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.75 vs. limit=2.0 2023-04-28 03:34:16,680 INFO [train.py:904] (1/8) Epoch 4, batch 5550, loss[loss=0.2402, simple_loss=0.3132, pruned_loss=0.0836, over 16851.00 frames. ], tot_loss[loss=0.2566, simple_loss=0.3305, pruned_loss=0.09138, over 3159133.64 frames. ], batch size: 42, lr: 1.62e-02, grad_scale: 4.0 2023-04-28 03:34:17,149 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=36001.0, num_to_drop=1, layers_to_drop={0} 2023-04-28 03:34:42,838 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.1041, 3.1588, 1.4935, 3.3134, 2.2787, 3.3247, 1.8671, 2.5386], device='cuda:1'), covar=tensor([0.0136, 0.0249, 0.1550, 0.0047, 0.0680, 0.0352, 0.1282, 0.0578], device='cuda:1'), in_proj_covar=tensor([0.0109, 0.0143, 0.0175, 0.0078, 0.0157, 0.0173, 0.0181, 0.0161], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-04-28 03:34:58,459 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.7279, 3.6733, 3.7979, 3.6948, 3.7825, 4.1613, 3.9765, 3.7516], device='cuda:1'), covar=tensor([0.1538, 0.1679, 0.1316, 0.2155, 0.2593, 0.1169, 0.1129, 0.2349], device='cuda:1'), in_proj_covar=tensor([0.0252, 0.0348, 0.0329, 0.0309, 0.0402, 0.0359, 0.0285, 0.0411], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003], device='cuda:1') 2023-04-28 03:35:02,230 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=36030.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 03:35:29,646 INFO [optim.py:368] (1/8) 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 03:35:33,008 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=36049.0, num_to_drop=1, layers_to_drop={1} 2023-04-28 03:35:35,559 INFO [train.py:904] (1/8) Epoch 4, batch 5600, loss[loss=0.2997, simple_loss=0.3605, pruned_loss=0.1194, over 15323.00 frames. ], tot_loss[loss=0.2661, simple_loss=0.337, pruned_loss=0.09758, over 3124468.56 frames. ], batch size: 190, lr: 1.61e-02, grad_scale: 8.0 2023-04-28 03:36:20,418 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=36078.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 03:36:57,478 INFO [train.py:904] (1/8) Epoch 4, batch 5650, loss[loss=0.4034, simple_loss=0.4209, pruned_loss=0.193, over 11660.00 frames. ], tot_loss[loss=0.2759, simple_loss=0.3441, pruned_loss=0.1038, over 3084518.14 frames. ], batch size: 247, lr: 1.61e-02, grad_scale: 8.0 2023-04-28 03:37:41,229 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=36127.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 03:37:42,907 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.65 vs. limit=2.0 2023-04-28 03:38:09,921 INFO [optim.py:368] (1/8) 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 03:38:10,552 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([1.8112, 2.7559, 2.6708, 4.2391, 3.9222, 3.9241, 1.5291, 3.0160], device='cuda:1'), covar=tensor([0.1217, 0.0497, 0.0947, 0.0082, 0.0173, 0.0280, 0.1260, 0.0636], device='cuda:1'), in_proj_covar=tensor([0.0139, 0.0134, 0.0160, 0.0072, 0.0149, 0.0154, 0.0152, 0.0157], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0004, 0.0003, 0.0004], device='cuda:1') 2023-04-28 03:38:17,750 INFO [train.py:904] (1/8) Epoch 4, batch 5700, loss[loss=0.2609, simple_loss=0.3329, pruned_loss=0.09443, over 16614.00 frames. ], tot_loss[loss=0.2788, simple_loss=0.3462, pruned_loss=0.1058, over 3068722.78 frames. ], batch size: 68, lr: 1.61e-02, grad_scale: 8.0 2023-04-28 03:38:42,662 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=36167.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 03:39:05,023 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=36181.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 03:39:13,383 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=36186.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 03:39:36,193 INFO [train.py:904] (1/8) Epoch 4, batch 5750, loss[loss=0.3306, simple_loss=0.3701, pruned_loss=0.1455, over 11244.00 frames. ], tot_loss[loss=0.2834, simple_loss=0.35, pruned_loss=0.1084, over 3024974.08 frames. ], batch size: 248, lr: 1.61e-02, grad_scale: 8.0 2023-04-28 03:39:51,199 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.1965, 3.9400, 3.9387, 1.8666, 4.1973, 4.2134, 3.0727, 3.0390], device='cuda:1'), covar=tensor([0.0921, 0.0075, 0.0141, 0.1132, 0.0036, 0.0038, 0.0307, 0.0433], device='cuda:1'), in_proj_covar=tensor([0.0138, 0.0081, 0.0077, 0.0137, 0.0066, 0.0071, 0.0112, 0.0125], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003, 0.0003], device='cuda:1') 2023-04-28 03:40:14,890 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=3.34 vs. limit=5.0 2023-04-28 03:40:29,639 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=36234.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 03:40:42,480 INFO [zipformer.py:625] (1/8) 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] (1/8) 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 03:40:55,811 INFO [train.py:904] (1/8) Epoch 4, batch 5800, loss[loss=0.2706, simple_loss=0.339, pruned_loss=0.101, over 17022.00 frames. ], tot_loss[loss=0.2802, simple_loss=0.3486, pruned_loss=0.1059, over 3042593.74 frames. ], batch size: 53, lr: 1.61e-02, grad_scale: 8.0 2023-04-28 03:41:53,410 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.4234, 4.3228, 4.2879, 3.6404, 4.3027, 1.6646, 4.0421, 4.2758], device='cuda:1'), covar=tensor([0.0065, 0.0062, 0.0079, 0.0318, 0.0058, 0.1595, 0.0087, 0.0117], device='cuda:1'), in_proj_covar=tensor([0.0079, 0.0069, 0.0106, 0.0117, 0.0079, 0.0127, 0.0092, 0.0105], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0001, 0.0002, 0.0002, 0.0001, 0.0003, 0.0002, 0.0002], device='cuda:1') 2023-04-28 03:42:02,863 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.9507, 3.1241, 3.4687, 3.4531, 3.4090, 3.1667, 3.2624, 3.3338], device='cuda:1'), covar=tensor([0.0347, 0.0519, 0.0419, 0.0506, 0.0486, 0.0432, 0.0813, 0.0430], device='cuda:1'), in_proj_covar=tensor([0.0204, 0.0193, 0.0204, 0.0211, 0.0245, 0.0214, 0.0309, 0.0186], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0002], device='cuda:1') 2023-04-28 03:42:12,358 INFO [train.py:904] (1/8) Epoch 4, batch 5850, loss[loss=0.2852, simple_loss=0.3438, pruned_loss=0.1133, over 11857.00 frames. ], tot_loss[loss=0.2756, simple_loss=0.3451, pruned_loss=0.103, over 3042474.09 frames. ], batch size: 247, lr: 1.61e-02, grad_scale: 8.0 2023-04-28 03:43:17,820 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=4.55 vs. limit=5.0 2023-04-28 03:43:25,070 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.6045, 3.5751, 3.0185, 2.2002, 2.6231, 2.1437, 3.6230, 3.8072], device='cuda:1'), covar=tensor([0.1876, 0.0599, 0.1031, 0.1382, 0.1732, 0.1284, 0.0379, 0.0410], device='cuda:1'), in_proj_covar=tensor([0.0276, 0.0245, 0.0263, 0.0232, 0.0306, 0.0197, 0.0231, 0.0233], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-28 03:43:29,064 INFO [optim.py:368] (1/8) 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 03:43:33,995 INFO [train.py:904] (1/8) Epoch 4, batch 5900, loss[loss=0.2358, simple_loss=0.3261, pruned_loss=0.0728, over 16755.00 frames. ], tot_loss[loss=0.2746, simple_loss=0.3442, pruned_loss=0.1025, over 3036250.16 frames. ], batch size: 83, lr: 1.61e-02, grad_scale: 4.0 2023-04-28 03:44:55,801 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.93 vs. limit=2.0 2023-04-28 03:44:56,221 INFO [train.py:904] (1/8) Epoch 4, batch 5950, loss[loss=0.2627, simple_loss=0.3443, pruned_loss=0.09052, over 16811.00 frames. ], tot_loss[loss=0.2724, simple_loss=0.3443, pruned_loss=0.1003, over 3055496.74 frames. ], batch size: 102, lr: 1.61e-02, grad_scale: 4.0 2023-04-28 03:45:37,407 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=36427.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 03:45:43,846 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=2.03 vs. limit=2.0 2023-04-28 03:46:10,128 INFO [optim.py:368] (1/8) 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 03:46:14,570 INFO [train.py:904] (1/8) Epoch 4, batch 6000, loss[loss=0.2227, simple_loss=0.3111, pruned_loss=0.0672, over 16854.00 frames. ], tot_loss[loss=0.2698, simple_loss=0.3422, pruned_loss=0.09869, over 3072253.25 frames. ], batch size: 102, lr: 1.61e-02, grad_scale: 8.0 2023-04-28 03:46:14,570 INFO [train.py:929] (1/8) Computing validation loss 2023-04-28 03:46:25,217 INFO [train.py:938] (1/8) Epoch 4, validation: loss=0.1966, simple_loss=0.3069, pruned_loss=0.04309, over 944034.00 frames. 2023-04-28 03:46:25,217 INFO [train.py:939] (1/8) Maximum memory allocated so far is 17676MB 2023-04-28 03:46:49,514 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.5919, 3.6326, 4.1494, 4.1068, 4.0743, 3.7156, 3.8315, 3.7033], device='cuda:1'), covar=tensor([0.0309, 0.0390, 0.0338, 0.0414, 0.0440, 0.0374, 0.0740, 0.0437], device='cuda:1'), in_proj_covar=tensor([0.0213, 0.0199, 0.0207, 0.0214, 0.0254, 0.0221, 0.0317, 0.0191], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0002], device='cuda:1') 2023-04-28 03:46:49,540 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=36467.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 03:47:01,318 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=36475.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 03:47:44,054 INFO [train.py:904] (1/8) Epoch 4, batch 6050, loss[loss=0.231, simple_loss=0.3278, pruned_loss=0.06706, over 16898.00 frames. ], tot_loss[loss=0.2678, simple_loss=0.3403, pruned_loss=0.09769, over 3084231.75 frames. ], batch size: 96, lr: 1.61e-02, grad_scale: 8.0 2023-04-28 03:47:58,032 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.0931, 1.2988, 1.7921, 2.0263, 2.2310, 2.2628, 1.3746, 2.1319], device='cuda:1'), covar=tensor([0.0066, 0.0186, 0.0108, 0.0092, 0.0082, 0.0054, 0.0179, 0.0049], device='cuda:1'), in_proj_covar=tensor([0.0100, 0.0132, 0.0117, 0.0110, 0.0116, 0.0082, 0.0129, 0.0075], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001], device='cuda:1') 2023-04-28 03:48:06,425 INFO [zipformer.py:625] (1/8) 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] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=36537.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 03:48:58,314 INFO [optim.py:368] (1/8) 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 03:49:03,036 INFO [train.py:904] (1/8) Epoch 4, batch 6100, loss[loss=0.3029, simple_loss=0.3628, pruned_loss=0.1215, over 15398.00 frames. ], tot_loss[loss=0.2658, simple_loss=0.3394, pruned_loss=0.09613, over 3102245.57 frames. ], batch size: 191, lr: 1.60e-02, grad_scale: 8.0 2023-04-28 03:49:16,169 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=4.32 vs. limit=5.0 2023-04-28 03:50:22,075 INFO [train.py:904] (1/8) Epoch 4, batch 6150, loss[loss=0.216, simple_loss=0.3048, pruned_loss=0.06364, over 16934.00 frames. ], tot_loss[loss=0.2637, simple_loss=0.337, pruned_loss=0.09515, over 3108089.73 frames. ], batch size: 96, lr: 1.60e-02, grad_scale: 8.0 2023-04-28 03:50:29,657 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=36605.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 03:50:35,066 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=36608.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 03:50:41,772 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=36612.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 03:51:38,879 INFO [optim.py:368] (1/8) 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 03:51:44,319 INFO [train.py:904] (1/8) Epoch 4, batch 6200, loss[loss=0.3037, simple_loss=0.3626, pruned_loss=0.1224, over 15472.00 frames. ], tot_loss[loss=0.2622, simple_loss=0.3351, pruned_loss=0.09464, over 3099415.69 frames. ], batch size: 191, lr: 1.60e-02, grad_scale: 8.0 2023-04-28 03:52:07,194 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=36666.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 03:52:11,521 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=36669.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 03:52:17,482 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=36673.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 03:52:57,798 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=36700.0, num_to_drop=1, layers_to_drop={0} 2023-04-28 03:52:59,118 INFO [train.py:904] (1/8) Epoch 4, batch 6250, loss[loss=0.2654, simple_loss=0.3417, pruned_loss=0.09449, over 16885.00 frames. ], tot_loss[loss=0.2628, simple_loss=0.335, pruned_loss=0.09527, over 3089010.51 frames. ], batch size: 116, lr: 1.60e-02, grad_scale: 8.0 2023-04-28 03:53:23,994 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.0741, 3.0170, 2.7114, 1.9879, 2.5256, 2.0676, 2.7459, 2.9724], device='cuda:1'), covar=tensor([0.0297, 0.0432, 0.0443, 0.1309, 0.0642, 0.0847, 0.0561, 0.0430], device='cuda:1'), in_proj_covar=tensor([0.0140, 0.0124, 0.0155, 0.0143, 0.0138, 0.0130, 0.0144, 0.0128], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-04-28 03:54:09,137 INFO [optim.py:368] (1/8) 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] (1/8) Epoch 4, batch 6300, loss[loss=0.2825, simple_loss=0.3361, pruned_loss=0.1145, over 11707.00 frames. ], tot_loss[loss=0.2619, simple_loss=0.3348, pruned_loss=0.09452, over 3092807.71 frames. ], batch size: 247, lr: 1.60e-02, grad_scale: 8.0 2023-04-28 03:54:30,618 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=36761.0, num_to_drop=1, layers_to_drop={1} 2023-04-28 03:55:32,004 INFO [train.py:904] (1/8) Epoch 4, batch 6350, loss[loss=0.2921, simple_loss=0.3579, pruned_loss=0.1132, over 15387.00 frames. ], tot_loss[loss=0.2645, simple_loss=0.3362, pruned_loss=0.09642, over 3075686.44 frames. ], batch size: 190, lr: 1.60e-02, grad_scale: 8.0 2023-04-28 03:55:38,772 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=36805.0, num_to_drop=1, layers_to_drop={0} 2023-04-28 03:56:28,072 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=36837.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 03:56:43,684 INFO [optim.py:368] (1/8) 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 03:56:48,092 INFO [train.py:904] (1/8) Epoch 4, batch 6400, loss[loss=0.2796, simple_loss=0.3382, pruned_loss=0.1105, over 16700.00 frames. ], tot_loss[loss=0.2665, simple_loss=0.337, pruned_loss=0.09801, over 3066487.39 frames. ], batch size: 62, lr: 1.60e-02, grad_scale: 8.0 2023-04-28 03:56:51,216 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=2.59 vs. limit=5.0 2023-04-28 03:57:10,864 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=36866.0, num_to_drop=1, layers_to_drop={2} 2023-04-28 03:57:14,290 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.6859, 5.0221, 4.7271, 4.6831, 4.3680, 4.2571, 4.5123, 5.0440], device='cuda:1'), covar=tensor([0.0490, 0.0588, 0.0775, 0.0485, 0.0537, 0.0792, 0.0490, 0.0634], device='cuda:1'), in_proj_covar=tensor([0.0325, 0.0439, 0.0382, 0.0289, 0.0279, 0.0287, 0.0355, 0.0307], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-28 03:57:27,178 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.30 vs. limit=2.0 2023-04-28 03:57:34,233 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.1806, 1.3370, 1.8367, 2.0331, 2.2220, 2.2987, 1.4566, 2.1839], device='cuda:1'), covar=tensor([0.0061, 0.0166, 0.0102, 0.0083, 0.0070, 0.0047, 0.0148, 0.0047], device='cuda:1'), in_proj_covar=tensor([0.0099, 0.0131, 0.0118, 0.0112, 0.0117, 0.0082, 0.0130, 0.0075], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001], device='cuda:1') 2023-04-28 03:57:39,964 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=36885.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 03:57:56,696 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.0102, 3.9018, 3.9701, 4.2040, 4.2770, 3.8316, 4.2485, 4.2829], device='cuda:1'), covar=tensor([0.0813, 0.0666, 0.1122, 0.0507, 0.0488, 0.0930, 0.0524, 0.0422], device='cuda:1'), in_proj_covar=tensor([0.0340, 0.0420, 0.0536, 0.0421, 0.0318, 0.0309, 0.0345, 0.0348], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-28 03:58:04,376 INFO [train.py:904] (1/8) Epoch 4, batch 6450, loss[loss=0.2276, simple_loss=0.3124, pruned_loss=0.07135, over 16656.00 frames. ], tot_loss[loss=0.2662, simple_loss=0.3372, pruned_loss=0.09757, over 3039298.86 frames. ], batch size: 62, lr: 1.60e-02, grad_scale: 8.0 2023-04-28 03:58:20,083 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=36911.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 03:59:18,067 INFO [optim.py:368] (1/8) 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 03:59:23,782 INFO [train.py:904] (1/8) Epoch 4, batch 6500, loss[loss=0.2586, simple_loss=0.333, pruned_loss=0.09213, over 16867.00 frames. ], tot_loss[loss=0.2629, simple_loss=0.3341, pruned_loss=0.09588, over 3054230.18 frames. ], batch size: 102, lr: 1.60e-02, grad_scale: 8.0 2023-04-28 03:59:34,543 INFO [zipformer.py:625] (1/8) 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] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=36961.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 03:59:44,060 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=36964.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 03:59:49,605 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=36968.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 03:59:56,654 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=36972.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 04:00:44,412 INFO [train.py:904] (1/8) Epoch 4, batch 6550, loss[loss=0.2465, simple_loss=0.3371, pruned_loss=0.07797, over 16419.00 frames. ], tot_loss[loss=0.2657, simple_loss=0.3381, pruned_loss=0.09666, over 3079276.79 frames. ], batch size: 146, lr: 1.59e-02, grad_scale: 8.0 2023-04-28 04:00:50,592 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=37004.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 04:01:13,054 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=37019.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 04:01:38,703 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([5.3241, 5.1259, 5.1649, 5.0685, 4.6895, 5.1537, 5.0996, 4.8140], device='cuda:1'), covar=tensor([0.0376, 0.0236, 0.0160, 0.0132, 0.0683, 0.0261, 0.0157, 0.0413], device='cuda:1'), in_proj_covar=tensor([0.0159, 0.0155, 0.0192, 0.0159, 0.0221, 0.0184, 0.0143, 0.0205], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-04-28 04:01:47,992 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.2497, 2.2094, 2.1562, 3.6912, 1.7905, 3.1263, 2.2206, 2.0357], device='cuda:1'), covar=tensor([0.0514, 0.1324, 0.0769, 0.0253, 0.2485, 0.0536, 0.1418, 0.1847], device='cuda:1'), in_proj_covar=tensor([0.0292, 0.0279, 0.0232, 0.0291, 0.0346, 0.0254, 0.0256, 0.0343], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-28 04:01:56,564 INFO [optim.py:368] (1/8) 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] (1/8) Epoch 4, batch 6600, loss[loss=0.2639, simple_loss=0.3333, pruned_loss=0.09724, over 17024.00 frames. ], tot_loss[loss=0.2687, simple_loss=0.341, pruned_loss=0.09825, over 3079539.20 frames. ], batch size: 55, lr: 1.59e-02, grad_scale: 8.0 2023-04-28 04:02:09,098 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=37056.0, num_to_drop=1, layers_to_drop={1} 2023-04-28 04:02:12,298 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.2754, 3.9816, 4.2683, 1.4794, 4.4630, 4.5999, 3.0990, 3.1986], device='cuda:1'), covar=tensor([0.0867, 0.0095, 0.0134, 0.1395, 0.0041, 0.0039, 0.0290, 0.0459], device='cuda:1'), in_proj_covar=tensor([0.0139, 0.0084, 0.0079, 0.0140, 0.0067, 0.0071, 0.0112, 0.0127], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003, 0.0003], device='cuda:1') 2023-04-28 04:02:22,644 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=37065.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 04:03:19,632 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.32 vs. limit=2.0 2023-04-28 04:03:19,954 INFO [train.py:904] (1/8) Epoch 4, batch 6650, loss[loss=0.2492, simple_loss=0.331, pruned_loss=0.08368, over 16750.00 frames. ], tot_loss[loss=0.2686, simple_loss=0.3403, pruned_loss=0.09842, over 3089602.73 frames. ], batch size: 83, lr: 1.59e-02, grad_scale: 8.0 2023-04-28 04:03:34,591 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.3523, 4.1379, 4.3240, 4.5748, 4.6831, 4.1783, 4.6200, 4.5988], device='cuda:1'), covar=tensor([0.0678, 0.0705, 0.1076, 0.0397, 0.0376, 0.0703, 0.0407, 0.0424], device='cuda:1'), in_proj_covar=tensor([0.0333, 0.0414, 0.0526, 0.0412, 0.0314, 0.0306, 0.0342, 0.0346], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-28 04:04:33,021 INFO [optim.py:368] (1/8) 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] (1/8) Epoch 4, batch 6700, loss[loss=0.2454, simple_loss=0.326, pruned_loss=0.08236, over 17056.00 frames. ], tot_loss[loss=0.2674, simple_loss=0.3388, pruned_loss=0.09803, over 3083808.51 frames. ], batch size: 55, lr: 1.59e-02, grad_scale: 8.0 2023-04-28 04:04:53,079 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=37161.0, num_to_drop=1, layers_to_drop={1} 2023-04-28 04:05:53,504 INFO [train.py:904] (1/8) Epoch 4, batch 6750, loss[loss=0.3305, simple_loss=0.3756, pruned_loss=0.1427, over 11606.00 frames. ], tot_loss[loss=0.2664, simple_loss=0.3372, pruned_loss=0.09782, over 3087493.45 frames. ], batch size: 246, lr: 1.59e-02, grad_scale: 8.0 2023-04-28 04:05:58,905 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.5838, 3.5234, 2.8292, 2.2872, 2.6196, 2.0426, 3.6251, 3.7287], device='cuda:1'), covar=tensor([0.2048, 0.0679, 0.1281, 0.1362, 0.1985, 0.1419, 0.0394, 0.0506], device='cuda:1'), in_proj_covar=tensor([0.0281, 0.0248, 0.0263, 0.0233, 0.0312, 0.0199, 0.0233, 0.0240], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-28 04:06:00,147 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.1424, 3.8917, 4.0957, 4.3326, 4.4319, 3.9924, 4.3859, 4.3745], device='cuda:1'), covar=tensor([0.0736, 0.0709, 0.1179, 0.0436, 0.0343, 0.0749, 0.0416, 0.0404], device='cuda:1'), in_proj_covar=tensor([0.0332, 0.0412, 0.0531, 0.0413, 0.0313, 0.0303, 0.0343, 0.0343], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-28 04:07:06,101 INFO [optim.py:368] (1/8) 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] (1/8) Epoch 4, batch 6800, loss[loss=0.2706, simple_loss=0.349, pruned_loss=0.09613, over 16734.00 frames. ], tot_loss[loss=0.2649, simple_loss=0.3363, pruned_loss=0.09672, over 3112119.62 frames. ], batch size: 124, lr: 1.59e-02, grad_scale: 8.0 2023-04-28 04:07:27,139 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=37261.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 04:07:31,495 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=37264.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 04:07:35,281 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=37267.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 04:07:36,733 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=37268.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 04:07:36,752 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=37268.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 04:08:27,075 INFO [train.py:904] (1/8) Epoch 4, batch 6850, loss[loss=0.3279, simple_loss=0.3688, pruned_loss=0.1435, over 11921.00 frames. ], tot_loss[loss=0.2667, simple_loss=0.3377, pruned_loss=0.09785, over 3106770.37 frames. ], batch size: 248, lr: 1.59e-02, grad_scale: 4.0 2023-04-28 04:08:38,764 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=37309.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 04:08:43,093 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=37312.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 04:08:46,843 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=37314.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 04:08:49,156 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=37316.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 04:09:08,794 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=37329.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 04:09:18,960 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.54 vs. limit=2.0 2023-04-28 04:09:36,160 INFO [optim.py:368] (1/8) 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 04:09:39,830 INFO [train.py:904] (1/8) Epoch 4, batch 6900, loss[loss=0.3553, simple_loss=0.3893, pruned_loss=0.1606, over 11457.00 frames. ], tot_loss[loss=0.2691, simple_loss=0.3409, pruned_loss=0.09866, over 3101825.21 frames. ], batch size: 247, lr: 1.59e-02, grad_scale: 4.0 2023-04-28 04:09:42,018 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.38 vs. limit=2.0 2023-04-28 04:09:48,200 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=37356.0, num_to_drop=1, layers_to_drop={0} 2023-04-28 04:09:52,851 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=37360.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 04:09:59,904 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=37364.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 04:10:13,746 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.0175, 2.4657, 2.3316, 3.2549, 2.7839, 3.2540, 1.7677, 2.7433], device='cuda:1'), covar=tensor([0.1073, 0.0415, 0.0877, 0.0097, 0.0237, 0.0307, 0.1129, 0.0615], device='cuda:1'), in_proj_covar=tensor([0.0141, 0.0136, 0.0164, 0.0076, 0.0159, 0.0160, 0.0157, 0.0159], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0003, 0.0004], device='cuda:1') 2023-04-28 04:10:48,889 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.98 vs. limit=2.0 2023-04-28 04:10:55,352 INFO [train.py:904] (1/8) Epoch 4, batch 6950, loss[loss=0.2686, simple_loss=0.3466, pruned_loss=0.09533, over 16676.00 frames. ], tot_loss[loss=0.2729, simple_loss=0.3435, pruned_loss=0.1011, over 3088185.43 frames. ], batch size: 134, lr: 1.59e-02, grad_scale: 4.0 2023-04-28 04:11:00,665 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=37404.0, num_to_drop=1, layers_to_drop={0} 2023-04-28 04:11:32,920 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.4659, 3.4518, 3.2433, 3.2694, 3.0460, 3.3762, 3.1755, 3.1872], device='cuda:1'), covar=tensor([0.0431, 0.0261, 0.0185, 0.0162, 0.0587, 0.0253, 0.0919, 0.0367], device='cuda:1'), in_proj_covar=tensor([0.0161, 0.0154, 0.0190, 0.0158, 0.0223, 0.0185, 0.0144, 0.0206], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-04-28 04:11:33,006 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=37425.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 04:12:10,356 INFO [optim.py:368] (1/8) 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] (1/8) Epoch 4, batch 7000, loss[loss=0.24, simple_loss=0.3384, pruned_loss=0.07081, over 16638.00 frames. ], tot_loss[loss=0.2717, simple_loss=0.3434, pruned_loss=0.1, over 3080925.27 frames. ], batch size: 76, lr: 1.58e-02, grad_scale: 2.0 2023-04-28 04:12:28,175 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=37461.0, num_to_drop=1, layers_to_drop={0} 2023-04-28 04:12:39,343 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.68 vs. limit=2.0 2023-04-28 04:12:58,548 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.7685, 3.6208, 3.1620, 1.8072, 2.6720, 2.1446, 3.1614, 3.4304], device='cuda:1'), covar=tensor([0.0285, 0.0463, 0.0555, 0.1677, 0.0798, 0.0934, 0.0708, 0.0680], device='cuda:1'), in_proj_covar=tensor([0.0136, 0.0121, 0.0153, 0.0141, 0.0136, 0.0126, 0.0143, 0.0128], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-04-28 04:13:29,622 INFO [train.py:904] (1/8) Epoch 4, batch 7050, loss[loss=0.256, simple_loss=0.3378, pruned_loss=0.08714, over 16178.00 frames. ], tot_loss[loss=0.2715, simple_loss=0.344, pruned_loss=0.09944, over 3084378.85 frames. ], batch size: 165, lr: 1.58e-02, grad_scale: 2.0 2023-04-28 04:13:43,150 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=37509.0, num_to_drop=1, layers_to_drop={1} 2023-04-28 04:14:28,294 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([1.7766, 2.6150, 2.2629, 3.7253, 3.2528, 3.5734, 1.5028, 2.8089], device='cuda:1'), covar=tensor([0.1280, 0.0514, 0.1113, 0.0074, 0.0251, 0.0316, 0.1280, 0.0665], device='cuda:1'), in_proj_covar=tensor([0.0141, 0.0137, 0.0165, 0.0074, 0.0158, 0.0158, 0.0153, 0.0159], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0003, 0.0004], device='cuda:1') 2023-04-28 04:14:45,407 INFO [optim.py:368] (1/8) 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 04:14:46,715 INFO [train.py:904] (1/8) Epoch 4, batch 7100, loss[loss=0.2618, simple_loss=0.3376, pruned_loss=0.09299, over 16156.00 frames. ], tot_loss[loss=0.2727, simple_loss=0.3437, pruned_loss=0.1009, over 3046884.20 frames. ], batch size: 165, lr: 1.58e-02, grad_scale: 2.0 2023-04-28 04:15:13,100 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=37567.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 04:16:02,665 INFO [train.py:904] (1/8) Epoch 4, batch 7150, loss[loss=0.2855, simple_loss=0.3469, pruned_loss=0.112, over 15441.00 frames. ], tot_loss[loss=0.2696, simple_loss=0.3404, pruned_loss=0.09935, over 3057455.87 frames. ], batch size: 190, lr: 1.58e-02, grad_scale: 2.0 2023-04-28 04:16:22,604 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=37614.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 04:16:24,832 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=37615.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 04:16:38,273 INFO [zipformer.py:625] (1/8) 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] (1/8) 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] (1/8) Epoch 4, batch 7200, loss[loss=0.2315, simple_loss=0.3247, pruned_loss=0.06918, over 16773.00 frames. ], tot_loss[loss=0.2662, simple_loss=0.3381, pruned_loss=0.09719, over 3064903.20 frames. ], batch size: 102, lr: 1.58e-02, grad_scale: 4.0 2023-04-28 04:17:33,268 INFO [zipformer.py:625] (1/8) 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] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=37662.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 04:17:39,803 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.96 vs. limit=2.0 2023-04-28 04:17:58,934 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.57 vs. limit=2.0 2023-04-28 04:18:40,008 INFO [train.py:904] (1/8) Epoch 4, batch 7250, loss[loss=0.277, simple_loss=0.3321, pruned_loss=0.1109, over 11342.00 frames. ], tot_loss[loss=0.2614, simple_loss=0.3342, pruned_loss=0.09429, over 3073807.64 frames. ], batch size: 247, lr: 1.58e-02, grad_scale: 4.0 2023-04-28 04:18:52,051 INFO [zipformer.py:625] (1/8) 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] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=37720.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 04:19:55,472 INFO [optim.py:368] (1/8) 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 04:19:57,423 INFO [train.py:904] (1/8) Epoch 4, batch 7300, loss[loss=0.2184, simple_loss=0.3048, pruned_loss=0.06595, over 16878.00 frames. ], tot_loss[loss=0.2608, simple_loss=0.3333, pruned_loss=0.09417, over 3066908.87 frames. ], batch size: 42, lr: 1.58e-02, grad_scale: 4.0 2023-04-28 04:20:05,521 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.0617, 3.8368, 4.0322, 4.2356, 4.3069, 3.8490, 4.2758, 4.2659], device='cuda:1'), covar=tensor([0.0712, 0.0733, 0.1091, 0.0465, 0.0380, 0.0933, 0.0428, 0.0392], device='cuda:1'), in_proj_covar=tensor([0.0339, 0.0416, 0.0532, 0.0426, 0.0314, 0.0310, 0.0340, 0.0345], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-28 04:20:23,057 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([1.9210, 3.5073, 3.5383, 2.3557, 3.3171, 3.5304, 3.4021, 2.0984], device='cuda:1'), covar=tensor([0.0329, 0.0016, 0.0032, 0.0220, 0.0031, 0.0041, 0.0026, 0.0244], device='cuda:1'), in_proj_covar=tensor([0.0113, 0.0051, 0.0055, 0.0111, 0.0057, 0.0064, 0.0059, 0.0105], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0001, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:1') 2023-04-28 04:21:14,972 INFO [train.py:904] (1/8) Epoch 4, batch 7350, loss[loss=0.3082, simple_loss=0.3532, pruned_loss=0.1316, over 11085.00 frames. ], tot_loss[loss=0.261, simple_loss=0.3334, pruned_loss=0.0943, over 3051785.52 frames. ], batch size: 248, lr: 1.58e-02, grad_scale: 4.0 2023-04-28 04:22:29,890 INFO [optim.py:368] (1/8) 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] (1/8) Epoch 4, batch 7400, loss[loss=0.2552, simple_loss=0.3383, pruned_loss=0.08604, over 16903.00 frames. ], tot_loss[loss=0.2632, simple_loss=0.3354, pruned_loss=0.09553, over 3066185.56 frames. ], batch size: 109, lr: 1.58e-02, grad_scale: 4.0 2023-04-28 04:23:49,002 INFO [train.py:904] (1/8) Epoch 4, batch 7450, loss[loss=0.285, simple_loss=0.3402, pruned_loss=0.1149, over 11523.00 frames. ], tot_loss[loss=0.2657, simple_loss=0.337, pruned_loss=0.09719, over 3052202.79 frames. ], batch size: 248, lr: 1.58e-02, grad_scale: 4.0 2023-04-28 04:24:26,774 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.0788, 4.0872, 3.9625, 3.9129, 3.5912, 4.0382, 3.8200, 3.7663], device='cuda:1'), covar=tensor([0.0427, 0.0260, 0.0200, 0.0165, 0.0710, 0.0236, 0.0495, 0.0401], device='cuda:1'), in_proj_covar=tensor([0.0160, 0.0149, 0.0187, 0.0155, 0.0215, 0.0180, 0.0140, 0.0200], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0002], device='cuda:1') 2023-04-28 04:24:26,785 INFO [zipformer.py:625] (1/8) 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] (1/8) 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 04:25:07,723 INFO [train.py:904] (1/8) Epoch 4, batch 7500, loss[loss=0.2514, simple_loss=0.3351, pruned_loss=0.0838, over 16443.00 frames. ], tot_loss[loss=0.2648, simple_loss=0.3371, pruned_loss=0.09629, over 3057316.38 frames. ], batch size: 68, lr: 1.57e-02, grad_scale: 4.0 2023-04-28 04:25:40,562 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=37972.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 04:25:50,230 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=37979.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 04:26:00,429 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.1081, 4.1273, 4.6300, 4.6254, 4.5789, 4.1675, 4.2662, 4.1656], device='cuda:1'), covar=tensor([0.0297, 0.0349, 0.0332, 0.0429, 0.0423, 0.0274, 0.0822, 0.0359], device='cuda:1'), in_proj_covar=tensor([0.0215, 0.0208, 0.0215, 0.0221, 0.0257, 0.0226, 0.0329, 0.0196], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0002], device='cuda:1') 2023-04-28 04:26:27,336 INFO [train.py:904] (1/8) Epoch 4, batch 7550, loss[loss=0.2282, simple_loss=0.3003, pruned_loss=0.07807, over 16609.00 frames. ], tot_loss[loss=0.2647, simple_loss=0.3366, pruned_loss=0.09638, over 3047963.95 frames. ], batch size: 62, lr: 1.57e-02, grad_scale: 4.0 2023-04-28 04:26:54,330 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=38020.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 04:27:23,988 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=38040.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 04:27:38,139 INFO [optim.py:368] (1/8) 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 04:27:40,071 INFO [train.py:904] (1/8) Epoch 4, batch 7600, loss[loss=0.3094, simple_loss=0.3551, pruned_loss=0.1318, over 11512.00 frames. ], tot_loss[loss=0.2655, simple_loss=0.3365, pruned_loss=0.09728, over 3035670.90 frames. ], batch size: 246, lr: 1.57e-02, grad_scale: 8.0 2023-04-28 04:28:05,542 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=38068.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 04:28:37,904 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.7783, 5.1785, 5.2244, 5.3681, 5.2023, 5.7668, 5.3782, 5.0952], device='cuda:1'), covar=tensor([0.0818, 0.1304, 0.1043, 0.1286, 0.2035, 0.0839, 0.1008, 0.1938], device='cuda:1'), in_proj_covar=tensor([0.0253, 0.0357, 0.0339, 0.0307, 0.0406, 0.0370, 0.0283, 0.0416], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003], device='cuda:1') 2023-04-28 04:28:55,064 INFO [train.py:904] (1/8) Epoch 4, batch 7650, loss[loss=0.3132, simple_loss=0.359, pruned_loss=0.1337, over 11056.00 frames. ], tot_loss[loss=0.2652, simple_loss=0.3359, pruned_loss=0.09728, over 3044679.52 frames. ], batch size: 246, lr: 1.57e-02, grad_scale: 8.0 2023-04-28 04:30:08,813 INFO [optim.py:368] (1/8) 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 04:30:09,971 INFO [train.py:904] (1/8) Epoch 4, batch 7700, loss[loss=0.2413, simple_loss=0.3266, pruned_loss=0.07796, over 16712.00 frames. ], tot_loss[loss=0.2664, simple_loss=0.3365, pruned_loss=0.09813, over 3049802.99 frames. ], batch size: 76, lr: 1.57e-02, grad_scale: 8.0 2023-04-28 04:31:26,750 INFO [train.py:904] (1/8) Epoch 4, batch 7750, loss[loss=0.2408, simple_loss=0.3254, pruned_loss=0.07809, over 16837.00 frames. ], tot_loss[loss=0.2661, simple_loss=0.3366, pruned_loss=0.09781, over 3056859.82 frames. ], batch size: 102, lr: 1.57e-02, grad_scale: 8.0 2023-04-28 04:32:40,399 INFO [optim.py:368] (1/8) 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] (1/8) Epoch 4, batch 7800, loss[loss=0.2695, simple_loss=0.3424, pruned_loss=0.09827, over 16452.00 frames. ], tot_loss[loss=0.2689, simple_loss=0.3385, pruned_loss=0.09967, over 3039302.96 frames. ], batch size: 68, lr: 1.57e-02, grad_scale: 8.0 2023-04-28 04:33:04,357 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.0988, 2.2450, 1.5227, 1.7000, 2.7821, 2.5081, 3.1763, 3.0278], device='cuda:1'), covar=tensor([0.0026, 0.0164, 0.0256, 0.0222, 0.0082, 0.0151, 0.0069, 0.0076], device='cuda:1'), in_proj_covar=tensor([0.0069, 0.0140, 0.0145, 0.0141, 0.0133, 0.0144, 0.0105, 0.0120], device='cuda:1'), out_proj_covar=tensor([8.8748e-05, 1.8115e-04, 1.8302e-04, 1.7800e-04, 1.7202e-04, 1.8645e-04, 1.3140e-04, 1.5692e-04], device='cuda:1') 2023-04-28 04:33:12,151 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=2.66 vs. limit=5.0 2023-04-28 04:33:58,981 INFO [train.py:904] (1/8) Epoch 4, batch 7850, loss[loss=0.3075, simple_loss=0.3587, pruned_loss=0.1282, over 11586.00 frames. ], tot_loss[loss=0.2684, simple_loss=0.339, pruned_loss=0.09891, over 3050490.53 frames. ], batch size: 248, lr: 1.57e-02, grad_scale: 8.0 2023-04-28 04:34:50,602 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=38335.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 04:35:10,362 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([5.3698, 5.7300, 5.3543, 5.4418, 5.0189, 4.7590, 5.2099, 5.7689], device='cuda:1'), covar=tensor([0.0512, 0.0546, 0.0802, 0.0399, 0.0524, 0.0591, 0.0500, 0.0535], device='cuda:1'), in_proj_covar=tensor([0.0328, 0.0432, 0.0380, 0.0282, 0.0280, 0.0295, 0.0358, 0.0310], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-28 04:35:12,442 INFO [optim.py:368] (1/8) 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 04:35:14,727 INFO [train.py:904] (1/8) Epoch 4, batch 7900, loss[loss=0.237, simple_loss=0.3135, pruned_loss=0.08029, over 16391.00 frames. ], tot_loss[loss=0.2659, simple_loss=0.3373, pruned_loss=0.0972, over 3067446.61 frames. ], batch size: 35, lr: 1.57e-02, grad_scale: 8.0 2023-04-28 04:35:26,042 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.3487, 4.1783, 4.3401, 4.5999, 4.6751, 4.1661, 4.6691, 4.6437], device='cuda:1'), covar=tensor([0.0834, 0.0738, 0.1088, 0.0400, 0.0404, 0.0653, 0.0371, 0.0431], device='cuda:1'), in_proj_covar=tensor([0.0345, 0.0416, 0.0538, 0.0429, 0.0321, 0.0309, 0.0345, 0.0352], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-28 04:35:28,606 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=38360.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 04:36:34,671 INFO [train.py:904] (1/8) Epoch 4, batch 7950, loss[loss=0.2714, simple_loss=0.3386, pruned_loss=0.1021, over 15496.00 frames. ], tot_loss[loss=0.2671, simple_loss=0.3379, pruned_loss=0.09815, over 3046805.91 frames. ], batch size: 191, lr: 1.57e-02, grad_scale: 8.0 2023-04-28 04:37:05,062 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=38421.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 04:37:37,232 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.8727, 2.5937, 2.4546, 4.5887, 1.9692, 3.6711, 2.3562, 2.5019], device='cuda:1'), covar=tensor([0.0485, 0.1467, 0.0839, 0.0213, 0.2688, 0.0617, 0.1472, 0.2011], device='cuda:1'), in_proj_covar=tensor([0.0293, 0.0281, 0.0235, 0.0292, 0.0351, 0.0258, 0.0257, 0.0343], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-28 04:37:49,210 INFO [optim.py:368] (1/8) 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 04:37:50,954 INFO [train.py:904] (1/8) Epoch 4, batch 8000, loss[loss=0.2458, simple_loss=0.3308, pruned_loss=0.08039, over 16747.00 frames. ], tot_loss[loss=0.267, simple_loss=0.3377, pruned_loss=0.09818, over 3045482.19 frames. ], batch size: 89, lr: 1.56e-02, grad_scale: 8.0 2023-04-28 04:38:05,611 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.7910, 3.7261, 4.1892, 4.1810, 4.1483, 3.7763, 3.8676, 3.8302], device='cuda:1'), covar=tensor([0.0243, 0.0373, 0.0292, 0.0371, 0.0390, 0.0304, 0.0751, 0.0354], device='cuda:1'), in_proj_covar=tensor([0.0211, 0.0207, 0.0213, 0.0214, 0.0255, 0.0220, 0.0322, 0.0192], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0002], device='cuda:1') 2023-04-28 04:38:14,684 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.6890, 3.5880, 3.7216, 3.6430, 3.7369, 4.1107, 3.8457, 3.5576], device='cuda:1'), covar=tensor([0.1698, 0.1611, 0.1316, 0.1775, 0.2181, 0.1302, 0.1082, 0.1990], device='cuda:1'), in_proj_covar=tensor([0.0253, 0.0355, 0.0342, 0.0310, 0.0407, 0.0376, 0.0289, 0.0417], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-04-28 04:39:04,864 INFO [train.py:904] (1/8) Epoch 4, batch 8050, loss[loss=0.2737, simple_loss=0.3466, pruned_loss=0.1004, over 16386.00 frames. ], tot_loss[loss=0.2673, simple_loss=0.338, pruned_loss=0.09832, over 3042674.10 frames. ], batch size: 146, lr: 1.56e-02, grad_scale: 8.0 2023-04-28 04:39:11,767 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.3418, 3.6416, 3.7097, 2.6060, 3.4040, 3.6638, 3.4766, 1.8100], device='cuda:1'), covar=tensor([0.0264, 0.0026, 0.0024, 0.0186, 0.0039, 0.0044, 0.0034, 0.0295], device='cuda:1'), in_proj_covar=tensor([0.0111, 0.0048, 0.0055, 0.0109, 0.0056, 0.0064, 0.0059, 0.0102], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0001, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:1') 2023-04-28 04:40:21,961 INFO [optim.py:368] (1/8) 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] (1/8) Epoch 4, batch 8100, loss[loss=0.2373, simple_loss=0.3162, pruned_loss=0.07914, over 16583.00 frames. ], tot_loss[loss=0.2667, simple_loss=0.3377, pruned_loss=0.09784, over 3036669.71 frames. ], batch size: 57, lr: 1.56e-02, grad_scale: 8.0 2023-04-28 04:41:41,533 INFO [train.py:904] (1/8) Epoch 4, batch 8150, loss[loss=0.2544, simple_loss=0.3272, pruned_loss=0.09083, over 16689.00 frames. ], tot_loss[loss=0.2638, simple_loss=0.3349, pruned_loss=0.09637, over 3036064.88 frames. ], batch size: 89, lr: 1.56e-02, grad_scale: 8.0 2023-04-28 04:42:04,253 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.4736, 4.3320, 4.2603, 4.2376, 3.7952, 4.3312, 4.1841, 4.0508], device='cuda:1'), covar=tensor([0.0384, 0.0259, 0.0185, 0.0150, 0.0867, 0.0237, 0.0338, 0.0403], device='cuda:1'), in_proj_covar=tensor([0.0164, 0.0162, 0.0194, 0.0160, 0.0224, 0.0187, 0.0146, 0.0208], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-04-28 04:42:34,011 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=38635.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 04:42:37,637 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.43 vs. limit=2.0 2023-04-28 04:42:57,099 INFO [optim.py:368] (1/8) 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 04:42:59,071 INFO [train.py:904] (1/8) Epoch 4, batch 8200, loss[loss=0.2648, simple_loss=0.3423, pruned_loss=0.09364, over 15390.00 frames. ], tot_loss[loss=0.2614, simple_loss=0.3322, pruned_loss=0.09524, over 3053556.87 frames. ], batch size: 190, lr: 1.56e-02, grad_scale: 8.0 2023-04-28 04:43:33,416 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.2768, 1.7531, 2.2985, 2.9537, 2.8315, 3.2921, 1.5993, 3.4319], device='cuda:1'), covar=tensor([0.0055, 0.0206, 0.0144, 0.0095, 0.0085, 0.0069, 0.0226, 0.0039], device='cuda:1'), in_proj_covar=tensor([0.0102, 0.0132, 0.0119, 0.0112, 0.0116, 0.0083, 0.0131, 0.0075], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001], device='cuda:1') 2023-04-28 04:43:53,409 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=38683.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 04:44:23,316 INFO [train.py:904] (1/8) Epoch 4, batch 8250, loss[loss=0.2368, simple_loss=0.3205, pruned_loss=0.07653, over 15383.00 frames. ], tot_loss[loss=0.2583, simple_loss=0.3313, pruned_loss=0.09266, over 3055068.61 frames. ], batch size: 191, lr: 1.56e-02, grad_scale: 8.0 2023-04-28 04:44:48,444 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=38716.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 04:44:56,944 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.1497, 2.9966, 2.8261, 1.9782, 2.6272, 1.9907, 2.8103, 2.9385], device='cuda:1'), covar=tensor([0.0247, 0.0389, 0.0380, 0.1331, 0.0601, 0.0979, 0.0573, 0.0505], device='cuda:1'), in_proj_covar=tensor([0.0135, 0.0118, 0.0152, 0.0137, 0.0131, 0.0125, 0.0138, 0.0125], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:1') 2023-04-28 04:45:43,965 INFO [optim.py:368] (1/8) 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 04:45:45,839 INFO [train.py:904] (1/8) Epoch 4, batch 8300, loss[loss=0.2372, simple_loss=0.3065, pruned_loss=0.08397, over 12007.00 frames. ], tot_loss[loss=0.2526, simple_loss=0.3272, pruned_loss=0.08899, over 3032530.85 frames. ], batch size: 247, lr: 1.56e-02, grad_scale: 8.0 2023-04-28 04:45:49,979 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.90 vs. limit=2.0 2023-04-28 04:45:56,416 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.6587, 3.8304, 1.6706, 3.9547, 2.4552, 4.0013, 2.0188, 2.9511], device='cuda:1'), covar=tensor([0.0101, 0.0155, 0.1539, 0.0039, 0.0835, 0.0250, 0.1442, 0.0541], device='cuda:1'), in_proj_covar=tensor([0.0108, 0.0142, 0.0173, 0.0077, 0.0157, 0.0169, 0.0182, 0.0159], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-04-28 04:46:03,041 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.2012, 3.9535, 3.7904, 4.3527, 4.4618, 4.0349, 4.4658, 4.4544], device='cuda:1'), covar=tensor([0.0817, 0.0733, 0.1876, 0.0711, 0.0641, 0.0810, 0.0627, 0.0545], device='cuda:1'), in_proj_covar=tensor([0.0340, 0.0410, 0.0527, 0.0423, 0.0315, 0.0301, 0.0339, 0.0346], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-28 04:46:33,826 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([1.8467, 3.0540, 2.5083, 4.4626, 3.9258, 4.1561, 1.4898, 3.2588], device='cuda:1'), covar=tensor([0.1348, 0.0448, 0.1057, 0.0051, 0.0162, 0.0245, 0.1410, 0.0597], device='cuda:1'), in_proj_covar=tensor([0.0142, 0.0135, 0.0164, 0.0077, 0.0154, 0.0160, 0.0155, 0.0156], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0003, 0.0004], device='cuda:1') 2023-04-28 04:47:07,347 INFO [train.py:904] (1/8) Epoch 4, batch 8350, loss[loss=0.239, simple_loss=0.3266, pruned_loss=0.07567, over 16868.00 frames. ], tot_loss[loss=0.2479, simple_loss=0.3249, pruned_loss=0.08545, over 3028256.48 frames. ], batch size: 96, lr: 1.56e-02, grad_scale: 8.0 2023-04-28 04:47:44,538 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=38823.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 04:48:19,563 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.8045, 3.6328, 3.7842, 3.7910, 3.9012, 4.2408, 3.9808, 3.6384], device='cuda:1'), covar=tensor([0.1483, 0.1650, 0.1309, 0.1674, 0.2002, 0.1134, 0.1012, 0.1905], device='cuda:1'), in_proj_covar=tensor([0.0238, 0.0333, 0.0325, 0.0290, 0.0386, 0.0359, 0.0275, 0.0396], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-28 04:48:29,398 INFO [optim.py:368] (1/8) 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 04:48:30,623 INFO [train.py:904] (1/8) Epoch 4, batch 8400, loss[loss=0.2129, simple_loss=0.3008, pruned_loss=0.06251, over 16566.00 frames. ], tot_loss[loss=0.243, simple_loss=0.3209, pruned_loss=0.08249, over 3015229.08 frames. ], batch size: 75, lr: 1.56e-02, grad_scale: 8.0 2023-04-28 04:49:26,951 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 4, batch 8450, loss[loss=0.2102, simple_loss=0.299, pruned_loss=0.06068, over 16730.00 frames. ], tot_loss[loss=0.238, simple_loss=0.3175, pruned_loss=0.07927, over 3022831.67 frames. ], batch size: 89, lr: 1.56e-02, grad_scale: 4.0 2023-04-28 04:50:03,711 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=3.43 vs. limit=5.0 2023-04-28 04:50:26,469 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.8749, 4.1820, 3.9784, 4.0287, 3.6115, 3.6663, 3.8904, 4.1476], device='cuda:1'), covar=tensor([0.0651, 0.0704, 0.0780, 0.0447, 0.0620, 0.1117, 0.0546, 0.0757], device='cuda:1'), in_proj_covar=tensor([0.0311, 0.0417, 0.0364, 0.0270, 0.0268, 0.0286, 0.0338, 0.0298], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-28 04:50:36,369 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.6223, 4.7985, 4.8736, 5.0312, 4.9522, 5.4175, 5.0781, 4.8611], device='cuda:1'), covar=tensor([0.0796, 0.1459, 0.1065, 0.1453, 0.2079, 0.0840, 0.1013, 0.2085], device='cuda:1'), in_proj_covar=tensor([0.0241, 0.0335, 0.0328, 0.0291, 0.0390, 0.0356, 0.0279, 0.0395], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-28 04:51:13,821 INFO [optim.py:368] (1/8) 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] (1/8) Epoch 4, batch 8500, loss[loss=0.2029, simple_loss=0.2755, pruned_loss=0.06517, over 11745.00 frames. ], tot_loss[loss=0.2323, simple_loss=0.3128, pruned_loss=0.07586, over 3026681.70 frames. ], batch size: 248, lr: 1.55e-02, grad_scale: 4.0 2023-04-28 04:51:54,907 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([1.9542, 1.7643, 1.4704, 1.4634, 1.8506, 1.6674, 1.8807, 1.8827], device='cuda:1'), covar=tensor([0.0026, 0.0112, 0.0161, 0.0142, 0.0077, 0.0114, 0.0062, 0.0085], device='cuda:1'), in_proj_covar=tensor([0.0068, 0.0139, 0.0141, 0.0139, 0.0132, 0.0141, 0.0101, 0.0118], device='cuda:1'), out_proj_covar=tensor([8.6672e-05, 1.7875e-04, 1.7632e-04, 1.7431e-04, 1.7156e-04, 1.8083e-04, 1.2556e-04, 1.5343e-04], device='cuda:1') 2023-04-28 04:52:39,867 INFO [train.py:904] (1/8) Epoch 4, batch 8550, loss[loss=0.2453, simple_loss=0.3268, pruned_loss=0.08184, over 15370.00 frames. ], tot_loss[loss=0.2288, simple_loss=0.3096, pruned_loss=0.07402, over 3015769.35 frames. ], batch size: 190, lr: 1.55e-02, grad_scale: 4.0 2023-04-28 04:53:09,952 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=39016.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 04:53:26,416 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.6197, 1.5261, 1.7266, 2.4801, 2.4309, 2.6431, 1.5466, 2.6879], device='cuda:1'), covar=tensor([0.0054, 0.0217, 0.0150, 0.0114, 0.0092, 0.0075, 0.0221, 0.0054], device='cuda:1'), in_proj_covar=tensor([0.0103, 0.0130, 0.0117, 0.0112, 0.0117, 0.0080, 0.0130, 0.0073], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001], device='cuda:1') 2023-04-28 04:53:29,732 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=4.36 vs. limit=5.0 2023-04-28 04:54:11,240 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([5.0938, 5.3885, 5.0831, 5.0620, 4.7590, 4.6154, 4.9261, 5.4245], device='cuda:1'), covar=tensor([0.0535, 0.0681, 0.0907, 0.0468, 0.0630, 0.0600, 0.0553, 0.0631], device='cuda:1'), in_proj_covar=tensor([0.0309, 0.0415, 0.0361, 0.0270, 0.0269, 0.0284, 0.0336, 0.0300], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-28 04:54:21,290 INFO [optim.py:368] (1/8) 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 04:54:21,313 INFO [train.py:904] (1/8) Epoch 4, batch 8600, loss[loss=0.2232, simple_loss=0.3092, pruned_loss=0.06863, over 15220.00 frames. ], tot_loss[loss=0.2275, simple_loss=0.3096, pruned_loss=0.07264, over 3015731.48 frames. ], batch size: 190, lr: 1.55e-02, grad_scale: 4.0 2023-04-28 04:54:45,521 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.9211, 4.1917, 3.9329, 3.9848, 3.6453, 3.7178, 3.9409, 4.1146], device='cuda:1'), covar=tensor([0.0613, 0.0741, 0.0917, 0.0489, 0.0675, 0.1185, 0.0527, 0.0904], device='cuda:1'), in_proj_covar=tensor([0.0306, 0.0412, 0.0359, 0.0269, 0.0267, 0.0282, 0.0333, 0.0300], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-28 04:54:50,961 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=39064.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 04:55:04,177 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.5995, 2.2387, 2.2109, 4.0726, 1.7616, 3.3652, 2.2684, 2.0335], device='cuda:1'), covar=tensor([0.0476, 0.1665, 0.0952, 0.0232, 0.2982, 0.0597, 0.1581, 0.2257], device='cuda:1'), in_proj_covar=tensor([0.0286, 0.0280, 0.0231, 0.0284, 0.0346, 0.0254, 0.0255, 0.0331], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-28 04:55:58,610 INFO [train.py:904] (1/8) Epoch 4, batch 8650, loss[loss=0.2, simple_loss=0.2897, pruned_loss=0.05516, over 16634.00 frames. ], tot_loss[loss=0.2243, simple_loss=0.3076, pruned_loss=0.07057, over 3027402.18 frames. ], batch size: 134, lr: 1.55e-02, grad_scale: 4.0 2023-04-28 04:57:44,933 INFO [optim.py:368] (1/8) 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 04:57:44,948 INFO [train.py:904] (1/8) Epoch 4, batch 8700, loss[loss=0.207, simple_loss=0.2975, pruned_loss=0.05822, over 16338.00 frames. ], tot_loss[loss=0.2205, simple_loss=0.3039, pruned_loss=0.06858, over 3045339.74 frames. ], batch size: 146, lr: 1.55e-02, grad_scale: 4.0 2023-04-28 04:58:36,852 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=39179.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 04:58:49,741 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=39185.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 04:59:20,253 INFO [train.py:904] (1/8) Epoch 4, batch 8750, loss[loss=0.2453, simple_loss=0.3341, pruned_loss=0.07819, over 16268.00 frames. ], tot_loss[loss=0.2201, simple_loss=0.3038, pruned_loss=0.06823, over 3048843.19 frames. ], batch size: 165, lr: 1.55e-02, grad_scale: 2.0 2023-04-28 04:59:58,706 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([5.1377, 5.5079, 5.1994, 5.3003, 4.9647, 4.7434, 5.0005, 5.5157], device='cuda:1'), covar=tensor([0.0518, 0.0524, 0.0685, 0.0368, 0.0496, 0.0517, 0.0475, 0.0558], device='cuda:1'), in_proj_covar=tensor([0.0303, 0.0407, 0.0352, 0.0267, 0.0264, 0.0277, 0.0332, 0.0296], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-28 05:00:45,638 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.3711, 2.1692, 2.2681, 3.7185, 1.8623, 3.1092, 2.2712, 2.0934], device='cuda:1'), covar=tensor([0.0433, 0.1521, 0.0776, 0.0266, 0.2504, 0.0618, 0.1396, 0.1969], device='cuda:1'), in_proj_covar=tensor([0.0281, 0.0276, 0.0227, 0.0278, 0.0338, 0.0250, 0.0252, 0.0324], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-28 05:01:05,034 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=39246.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 05:01:14,442 INFO [train.py:904] (1/8) Epoch 4, batch 8800, loss[loss=0.217, simple_loss=0.3022, pruned_loss=0.06586, over 15445.00 frames. ], tot_loss[loss=0.2191, simple_loss=0.3029, pruned_loss=0.0677, over 3051060.31 frames. ], batch size: 191, lr: 1.55e-02, grad_scale: 4.0 2023-04-28 05:01:15,986 INFO [optim.py:368] (1/8) 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 05:01:33,663 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.18 vs. limit=2.0 2023-04-28 05:02:17,133 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.7050, 3.5249, 3.5359, 3.8582, 3.8425, 3.4892, 3.9507, 3.9527], device='cuda:1'), covar=tensor([0.0804, 0.0856, 0.1569, 0.0675, 0.0881, 0.1626, 0.0616, 0.0496], device='cuda:1'), in_proj_covar=tensor([0.0324, 0.0397, 0.0496, 0.0401, 0.0304, 0.0294, 0.0323, 0.0329], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-28 05:02:23,543 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=39284.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 05:02:58,289 INFO [train.py:904] (1/8) Epoch 4, batch 8850, loss[loss=0.2172, simple_loss=0.2902, pruned_loss=0.07213, over 12106.00 frames. ], tot_loss[loss=0.2186, simple_loss=0.3044, pruned_loss=0.06636, over 3041727.40 frames. ], batch size: 246, lr: 1.55e-02, grad_scale: 4.0 2023-04-28 05:04:32,322 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=39345.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 05:04:44,178 INFO [train.py:904] (1/8) Epoch 4, batch 8900, loss[loss=0.2211, simple_loss=0.3199, pruned_loss=0.06118, over 17071.00 frames. ], tot_loss[loss=0.2171, simple_loss=0.304, pruned_loss=0.06514, over 3047091.44 frames. ], batch size: 97, lr: 1.55e-02, grad_scale: 2.0 2023-04-28 05:04:49,516 INFO [optim.py:368] (1/8) 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 05:05:07,315 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=2.60 vs. limit=5.0 2023-04-28 05:06:47,800 INFO [train.py:904] (1/8) Epoch 4, batch 8950, loss[loss=0.2087, simple_loss=0.3042, pruned_loss=0.05662, over 16819.00 frames. ], tot_loss[loss=0.2177, simple_loss=0.3038, pruned_loss=0.06577, over 3048830.76 frames. ], batch size: 124, lr: 1.55e-02, grad_scale: 2.0 2023-04-28 05:06:51,192 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.4912, 3.5306, 3.3469, 3.2493, 3.0999, 3.4528, 3.1988, 3.2954], device='cuda:1'), covar=tensor([0.0437, 0.0319, 0.0198, 0.0163, 0.0484, 0.0275, 0.0867, 0.0361], device='cuda:1'), in_proj_covar=tensor([0.0155, 0.0149, 0.0184, 0.0150, 0.0201, 0.0177, 0.0136, 0.0194], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-28 05:08:35,748 INFO [train.py:904] (1/8) Epoch 4, batch 9000, loss[loss=0.1957, simple_loss=0.2832, pruned_loss=0.05416, over 16671.00 frames. ], tot_loss[loss=0.2148, simple_loss=0.3007, pruned_loss=0.06443, over 3047143.67 frames. ], batch size: 134, lr: 1.54e-02, grad_scale: 2.0 2023-04-28 05:08:35,748 INFO [train.py:929] (1/8) Computing validation loss 2023-04-28 05:08:45,781 INFO [train.py:938] (1/8) 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] (1/8) Maximum memory allocated so far is 17676MB 2023-04-28 05:08:49,857 INFO [optim.py:368] (1/8) 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 05:09:11,616 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=39463.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 05:09:29,879 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([5.1283, 4.8425, 5.0813, 5.3892, 5.4727, 4.8156, 5.4681, 5.3880], device='cuda:1'), covar=tensor([0.0776, 0.0668, 0.1175, 0.0344, 0.0425, 0.0520, 0.0315, 0.0375], device='cuda:1'), in_proj_covar=tensor([0.0321, 0.0394, 0.0497, 0.0396, 0.0300, 0.0290, 0.0323, 0.0324], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-28 05:09:43,576 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=39479.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 05:10:02,029 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.8050, 3.6390, 3.1516, 1.8290, 2.7338, 2.2068, 3.0553, 3.3141], device='cuda:1'), covar=tensor([0.0265, 0.0345, 0.0508, 0.1505, 0.0683, 0.0899, 0.0715, 0.0732], device='cuda:1'), in_proj_covar=tensor([0.0133, 0.0112, 0.0151, 0.0138, 0.0131, 0.0125, 0.0136, 0.0123], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:1') 2023-04-28 05:10:29,875 INFO [train.py:904] (1/8) Epoch 4, batch 9050, loss[loss=0.2036, simple_loss=0.2927, pruned_loss=0.05721, over 15512.00 frames. ], tot_loss[loss=0.2161, simple_loss=0.302, pruned_loss=0.06507, over 3068686.46 frames. ], batch size: 191, lr: 1.54e-02, grad_scale: 2.0 2023-04-28 05:11:17,238 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=39524.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 05:11:21,897 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=39527.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 05:11:54,713 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=39541.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 05:12:14,619 INFO [train.py:904] (1/8) Epoch 4, batch 9100, loss[loss=0.2245, simple_loss=0.3024, pruned_loss=0.07328, over 12618.00 frames. ], tot_loss[loss=0.2168, simple_loss=0.3021, pruned_loss=0.06578, over 3057471.90 frames. ], batch size: 247, lr: 1.54e-02, grad_scale: 2.0 2023-04-28 05:12:18,749 INFO [optim.py:368] (1/8) 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 05:14:15,379 INFO [train.py:904] (1/8) Epoch 4, batch 9150, loss[loss=0.2226, simple_loss=0.3033, pruned_loss=0.07092, over 16934.00 frames. ], tot_loss[loss=0.2164, simple_loss=0.3026, pruned_loss=0.06508, over 3075355.02 frames. ], batch size: 109, lr: 1.54e-02, grad_scale: 2.0 2023-04-28 05:15:40,865 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=39639.0, num_to_drop=1, layers_to_drop={1} 2023-04-28 05:15:43,230 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=39640.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 05:16:00,925 INFO [train.py:904] (1/8) Epoch 4, batch 9200, loss[loss=0.2124, simple_loss=0.2932, pruned_loss=0.06577, over 16808.00 frames. ], tot_loss[loss=0.2127, simple_loss=0.2977, pruned_loss=0.06386, over 3068541.75 frames. ], batch size: 124, lr: 1.54e-02, grad_scale: 4.0 2023-04-28 05:16:04,310 INFO [optim.py:368] (1/8) 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 05:16:43,707 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.2739, 4.6108, 4.3996, 4.4547, 4.0384, 3.9940, 4.1335, 4.5676], device='cuda:1'), covar=tensor([0.0654, 0.0638, 0.0759, 0.0372, 0.0680, 0.1073, 0.0662, 0.0810], device='cuda:1'), in_proj_covar=tensor([0.0305, 0.0418, 0.0350, 0.0266, 0.0269, 0.0284, 0.0334, 0.0304], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-28 05:17:00,465 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.4561, 3.3897, 3.3871, 2.7188, 3.3269, 1.8747, 3.1983, 3.0034], device='cuda:1'), covar=tensor([0.0138, 0.0114, 0.0132, 0.0342, 0.0096, 0.1898, 0.0151, 0.0204], device='cuda:1'), in_proj_covar=tensor([0.0081, 0.0068, 0.0107, 0.0105, 0.0078, 0.0130, 0.0094, 0.0101], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-28 05:17:34,394 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=39700.0, num_to_drop=1, layers_to_drop={1} 2023-04-28 05:17:35,638 INFO [train.py:904] (1/8) Epoch 4, batch 9250, loss[loss=0.2087, simple_loss=0.2949, pruned_loss=0.06125, over 15223.00 frames. ], tot_loss[loss=0.2126, simple_loss=0.2971, pruned_loss=0.06405, over 3062690.76 frames. ], batch size: 190, lr: 1.54e-02, grad_scale: 4.0 2023-04-28 05:19:26,128 INFO [train.py:904] (1/8) Epoch 4, batch 9300, loss[loss=0.187, simple_loss=0.2749, pruned_loss=0.04952, over 16639.00 frames. ], tot_loss[loss=0.2105, simple_loss=0.2951, pruned_loss=0.06294, over 3056782.42 frames. ], batch size: 57, lr: 1.54e-02, grad_scale: 4.0 2023-04-28 05:19:30,018 INFO [optim.py:368] (1/8) Clipping_scale=2.0, grad-norm quartiles 2.165e+02 3.034e+02 3.554e+02 4.321e+02 7.893e+02, threshold=7.107e+02, percent-clipped=0.0 2023-04-28 05:20:06,099 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=39767.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 05:21:11,794 INFO [train.py:904] (1/8) Epoch 4, batch 9350, loss[loss=0.2224, simple_loss=0.301, pruned_loss=0.07185, over 15262.00 frames. ], tot_loss[loss=0.2105, simple_loss=0.2952, pruned_loss=0.06293, over 3064598.31 frames. ], batch size: 190, lr: 1.54e-02, grad_scale: 2.0 2023-04-28 05:21:41,197 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=39815.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 05:21:48,595 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=39819.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 05:22:05,632 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=39828.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 05:22:30,201 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=39841.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 05:22:51,020 INFO [train.py:904] (1/8) Epoch 4, batch 9400, loss[loss=0.2148, simple_loss=0.3115, pruned_loss=0.05905, over 16970.00 frames. ], tot_loss[loss=0.2104, simple_loss=0.2955, pruned_loss=0.06263, over 3066521.45 frames. ], batch size: 116, lr: 1.54e-02, grad_scale: 2.0 2023-04-28 05:22:57,590 INFO [optim.py:368] (1/8) 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 05:23:41,208 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=39876.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 05:24:09,632 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 4, batch 9450, loss[loss=0.2189, simple_loss=0.2968, pruned_loss=0.07047, over 12364.00 frames. ], tot_loss[loss=0.2116, simple_loss=0.2972, pruned_loss=0.06296, over 3069422.77 frames. ], batch size: 248, lr: 1.54e-02, grad_scale: 2.0 2023-04-28 05:25:54,132 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=39940.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 05:26:13,854 INFO [train.py:904] (1/8) Epoch 4, batch 9500, loss[loss=0.1889, simple_loss=0.2787, pruned_loss=0.04959, over 15325.00 frames. ], tot_loss[loss=0.2099, simple_loss=0.2956, pruned_loss=0.06208, over 3055497.47 frames. ], batch size: 190, lr: 1.54e-02, grad_scale: 2.0 2023-04-28 05:26:21,218 INFO [optim.py:368] (1/8) 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 05:27:31,421 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=39988.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 05:27:48,799 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=39995.0, num_to_drop=1, layers_to_drop={3} 2023-04-28 05:28:04,404 INFO [train.py:904] (1/8) Epoch 4, batch 9550, loss[loss=0.2225, simple_loss=0.2962, pruned_loss=0.0744, over 12373.00 frames. ], tot_loss[loss=0.2101, simple_loss=0.2958, pruned_loss=0.06219, over 3079147.98 frames. ], batch size: 248, lr: 1.53e-02, grad_scale: 2.0 2023-04-28 05:29:46,541 INFO [train.py:904] (1/8) Epoch 4, batch 9600, loss[loss=0.2324, simple_loss=0.2993, pruned_loss=0.08275, over 12027.00 frames. ], tot_loss[loss=0.2122, simple_loss=0.2976, pruned_loss=0.06342, over 3071469.87 frames. ], batch size: 248, lr: 1.53e-02, grad_scale: 4.0 2023-04-28 05:29:52,055 INFO [optim.py:368] (1/8) 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 05:30:20,561 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.94 vs. limit=2.0 2023-04-28 05:30:34,742 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.1998, 3.8438, 3.9511, 2.6028, 3.5958, 3.9671, 3.7434, 2.2971], device='cuda:1'), covar=tensor([0.0337, 0.0018, 0.0032, 0.0220, 0.0037, 0.0036, 0.0036, 0.0279], device='cuda:1'), in_proj_covar=tensor([0.0110, 0.0050, 0.0056, 0.0108, 0.0054, 0.0063, 0.0058, 0.0102], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0001, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:1') 2023-04-28 05:31:33,080 INFO [train.py:904] (1/8) Epoch 4, batch 9650, loss[loss=0.212, simple_loss=0.3041, pruned_loss=0.05994, over 16912.00 frames. ], tot_loss[loss=0.2134, simple_loss=0.299, pruned_loss=0.06389, over 3045102.64 frames. ], batch size: 116, lr: 1.53e-02, grad_scale: 4.0 2023-04-28 05:32:17,622 INFO [zipformer.py:625] (1/8) 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] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=40123.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 05:32:54,912 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=4.63 vs. limit=5.0 2023-04-28 05:33:07,581 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.4734, 3.4043, 2.8460, 2.2322, 2.3386, 2.1241, 3.3865, 3.4070], device='cuda:1'), covar=tensor([0.1970, 0.0683, 0.1074, 0.1425, 0.1671, 0.1269, 0.0362, 0.0524], device='cuda:1'), in_proj_covar=tensor([0.0265, 0.0236, 0.0251, 0.0228, 0.0237, 0.0189, 0.0219, 0.0216], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-28 05:33:21,182 INFO [train.py:904] (1/8) Epoch 4, batch 9700, loss[loss=0.213, simple_loss=0.2991, pruned_loss=0.06348, over 16936.00 frames. ], tot_loss[loss=0.2115, simple_loss=0.2971, pruned_loss=0.06294, over 3052711.17 frames. ], batch size: 116, lr: 1.53e-02, grad_scale: 4.0 2023-04-28 05:33:26,546 INFO [optim.py:368] (1/8) 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 05:33:39,442 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.5751, 3.5015, 3.4466, 3.0884, 3.4474, 1.9477, 3.3220, 3.1924], device='cuda:1'), covar=tensor([0.0059, 0.0052, 0.0071, 0.0164, 0.0062, 0.1323, 0.0080, 0.0114], device='cuda:1'), in_proj_covar=tensor([0.0078, 0.0065, 0.0103, 0.0102, 0.0077, 0.0131, 0.0092, 0.0098], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-28 05:33:53,101 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=40167.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 05:34:01,336 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=40171.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 05:34:54,837 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 4, batch 9750, loss[loss=0.2174, simple_loss=0.2865, pruned_loss=0.07421, over 12336.00 frames. ], tot_loss[loss=0.2113, simple_loss=0.296, pruned_loss=0.06328, over 3030250.63 frames. ], batch size: 249, lr: 1.53e-02, grad_scale: 4.0 2023-04-28 05:35:34,922 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.2050, 3.0654, 2.8234, 1.9301, 2.5669, 2.1117, 2.7366, 2.9536], device='cuda:1'), covar=tensor([0.0311, 0.0433, 0.0396, 0.1347, 0.0637, 0.0862, 0.0629, 0.0612], device='cuda:1'), in_proj_covar=tensor([0.0135, 0.0113, 0.0154, 0.0142, 0.0134, 0.0126, 0.0138, 0.0123], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003], device='cuda:1') 2023-04-28 05:35:38,789 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=40218.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 05:36:45,109 INFO [train.py:904] (1/8) Epoch 4, batch 9800, loss[loss=0.2411, simple_loss=0.3375, pruned_loss=0.07231, over 16333.00 frames. ], tot_loss[loss=0.2099, simple_loss=0.2959, pruned_loss=0.06198, over 3047799.62 frames. ], batch size: 146, lr: 1.53e-02, grad_scale: 4.0 2023-04-28 05:36:51,067 INFO [optim.py:368] (1/8) 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 05:36:56,142 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=40256.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 05:37:24,899 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=40272.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 05:37:37,803 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=40279.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 05:38:17,481 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 4, batch 9850, loss[loss=0.1956, simple_loss=0.2954, pruned_loss=0.04788, over 17033.00 frames. ], tot_loss[loss=0.2097, simple_loss=0.297, pruned_loss=0.06119, over 3063832.53 frames. ], batch size: 41, lr: 1.53e-02, grad_scale: 4.0 2023-04-28 05:38:56,854 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.7132, 3.7266, 3.8374, 3.8430, 3.8672, 4.2163, 3.8817, 3.6506], device='cuda:1'), covar=tensor([0.1809, 0.1569, 0.1292, 0.1766, 0.1995, 0.1180, 0.1225, 0.2269], device='cuda:1'), in_proj_covar=tensor([0.0231, 0.0337, 0.0327, 0.0293, 0.0388, 0.0354, 0.0273, 0.0389], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-28 05:39:19,011 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.52 vs. limit=2.0 2023-04-28 05:39:30,913 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=40330.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 05:39:39,168 INFO [zipformer.py:625] (1/8) 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] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=40343.0, num_to_drop=1, layers_to_drop={1} 2023-04-28 05:40:07,956 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([1.6768, 2.6677, 2.5113, 4.0901, 3.7201, 3.8916, 1.2854, 2.9817], device='cuda:1'), covar=tensor([0.1299, 0.0569, 0.0969, 0.0080, 0.0155, 0.0277, 0.1441, 0.0642], device='cuda:1'), in_proj_covar=tensor([0.0141, 0.0134, 0.0162, 0.0073, 0.0139, 0.0157, 0.0157, 0.0159], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0004, 0.0003, 0.0004], device='cuda:1') 2023-04-28 05:40:15,474 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.8576, 5.3649, 5.4464, 5.4207, 5.3921, 5.7864, 5.5566, 5.3412], device='cuda:1'), covar=tensor([0.0670, 0.1260, 0.0940, 0.1521, 0.1898, 0.0792, 0.1002, 0.2101], device='cuda:1'), in_proj_covar=tensor([0.0231, 0.0338, 0.0325, 0.0292, 0.0387, 0.0354, 0.0272, 0.0389], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-28 05:40:19,741 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.2087, 3.9667, 3.6856, 2.0012, 2.9026, 2.6119, 3.3222, 3.6533], device='cuda:1'), covar=tensor([0.0199, 0.0385, 0.0372, 0.1393, 0.0673, 0.0772, 0.0664, 0.0624], device='cuda:1'), in_proj_covar=tensor([0.0134, 0.0112, 0.0154, 0.0142, 0.0133, 0.0126, 0.0137, 0.0122], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003], device='cuda:1') 2023-04-28 05:40:21,695 INFO [train.py:904] (1/8) Epoch 4, batch 9900, loss[loss=0.187, simple_loss=0.2856, pruned_loss=0.04416, over 16689.00 frames. ], tot_loss[loss=0.2102, simple_loss=0.2976, pruned_loss=0.06139, over 3034702.91 frames. ], batch size: 76, lr: 1.53e-02, grad_scale: 4.0 2023-04-28 05:40:27,910 INFO [optim.py:368] (1/8) 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 05:40:29,130 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.5021, 3.9154, 3.9316, 1.5814, 4.1460, 4.2850, 3.3079, 3.0353], device='cuda:1'), covar=tensor([0.0703, 0.0091, 0.0131, 0.1253, 0.0046, 0.0030, 0.0241, 0.0400], device='cuda:1'), in_proj_covar=tensor([0.0136, 0.0083, 0.0076, 0.0140, 0.0068, 0.0072, 0.0109, 0.0125], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003, 0.0003], device='cuda:1') 2023-04-28 05:40:33,046 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.25 vs. limit=2.0 2023-04-28 05:41:56,796 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=40391.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 05:42:12,591 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.89 vs. limit=2.0 2023-04-28 05:42:18,075 INFO [train.py:904] (1/8) Epoch 4, batch 9950, loss[loss=0.2094, simple_loss=0.3019, pruned_loss=0.05852, over 15304.00 frames. ], tot_loss[loss=0.2116, simple_loss=0.2997, pruned_loss=0.06169, over 3043674.00 frames. ], batch size: 191, lr: 1.53e-02, grad_scale: 4.0 2023-04-28 05:42:29,336 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.4491, 3.5194, 3.0100, 2.3391, 2.4578, 2.1539, 3.6591, 3.5359], device='cuda:1'), covar=tensor([0.2022, 0.0577, 0.1078, 0.1467, 0.1484, 0.1262, 0.0339, 0.0478], device='cuda:1'), in_proj_covar=tensor([0.0264, 0.0238, 0.0253, 0.0229, 0.0236, 0.0190, 0.0222, 0.0221], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-28 05:43:13,897 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 4, batch 10000, loss[loss=0.2138, simple_loss=0.3036, pruned_loss=0.062, over 16434.00 frames. ], tot_loss[loss=0.2099, simple_loss=0.2978, pruned_loss=0.06106, over 3050112.13 frames. ], batch size: 147, lr: 1.53e-02, grad_scale: 8.0 2023-04-28 05:44:26,639 INFO [optim.py:368] (1/8) 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 05:44:27,872 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.3925, 3.2383, 3.3343, 3.4974, 3.5228, 3.1533, 3.4766, 3.5019], device='cuda:1'), covar=tensor([0.0656, 0.0650, 0.1165, 0.0536, 0.0537, 0.1890, 0.0773, 0.0575], device='cuda:1'), in_proj_covar=tensor([0.0328, 0.0407, 0.0503, 0.0409, 0.0306, 0.0298, 0.0330, 0.0335], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-28 05:45:01,304 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=40471.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 05:45:01,454 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=40471.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 05:45:46,823 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=40492.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 05:46:02,921 INFO [train.py:904] (1/8) Epoch 4, batch 10050, loss[loss=0.2211, simple_loss=0.3135, pruned_loss=0.06439, over 16414.00 frames. ], tot_loss[loss=0.21, simple_loss=0.298, pruned_loss=0.06098, over 3063351.43 frames. ], batch size: 146, lr: 1.53e-02, grad_scale: 8.0 2023-04-28 05:46:38,832 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=40519.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 05:46:59,046 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.9915, 2.4714, 2.3859, 4.5444, 1.9127, 3.4646, 2.4141, 2.3015], device='cuda:1'), covar=tensor([0.0421, 0.1663, 0.0869, 0.0199, 0.2841, 0.0583, 0.1581, 0.2157], device='cuda:1'), in_proj_covar=tensor([0.0292, 0.0286, 0.0233, 0.0289, 0.0346, 0.0259, 0.0260, 0.0330], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-28 05:47:38,802 INFO [train.py:904] (1/8) Epoch 4, batch 10100, loss[loss=0.1851, simple_loss=0.2718, pruned_loss=0.04921, over 16143.00 frames. ], tot_loss[loss=0.2109, simple_loss=0.2985, pruned_loss=0.06168, over 3065429.20 frames. ], batch size: 165, lr: 1.52e-02, grad_scale: 8.0 2023-04-28 05:47:39,296 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=40551.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 05:47:42,188 INFO [zipformer.py:625] (1/8) 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] (1/8) 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 05:48:27,649 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 5, batch 0, loss[loss=0.3368, simple_loss=0.3688, pruned_loss=0.1524, over 16941.00 frames. ], tot_loss[loss=0.3368, simple_loss=0.3688, pruned_loss=0.1524, over 16941.00 frames. ], batch size: 96, lr: 1.42e-02, grad_scale: 8.0 2023-04-28 05:49:25,172 INFO [train.py:929] (1/8) Computing validation loss 2023-04-28 05:49:32,548 INFO [train.py:938] (1/8) Epoch 5, validation: loss=0.1789, simple_loss=0.2817, pruned_loss=0.03802, over 944034.00 frames. 2023-04-28 05:49:32,549 INFO [train.py:939] (1/8) Maximum memory allocated so far is 17676MB 2023-04-28 05:49:38,532 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.2913, 4.0470, 3.2221, 1.9206, 2.7468, 2.1920, 3.5804, 3.7018], device='cuda:1'), covar=tensor([0.0148, 0.0333, 0.0525, 0.1409, 0.0660, 0.0959, 0.0444, 0.0500], device='cuda:1'), in_proj_covar=tensor([0.0133, 0.0111, 0.0154, 0.0142, 0.0132, 0.0126, 0.0136, 0.0123], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003], device='cuda:1') 2023-04-28 05:49:53,896 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.59 vs. limit=2.0 2023-04-28 05:50:11,888 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=40628.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 05:50:42,807 INFO [train.py:904] (1/8) Epoch 5, batch 50, loss[loss=0.2217, simple_loss=0.3075, pruned_loss=0.068, over 17138.00 frames. ], tot_loss[loss=0.2573, simple_loss=0.324, pruned_loss=0.09527, over 737582.39 frames. ], batch size: 49, lr: 1.42e-02, grad_scale: 1.0 2023-04-28 05:50:43,162 INFO [zipformer.py:625] (1/8) 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] (1/8) Clipping_scale=2.0, grad-norm quartiles 2.398e+02 3.847e+02 4.852e+02 6.011e+02 1.299e+03, threshold=9.705e+02, percent-clipped=3.0 2023-04-28 05:51:29,205 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=40686.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 05:51:50,912 INFO [train.py:904] (1/8) Epoch 5, batch 100, loss[loss=0.2343, simple_loss=0.2972, pruned_loss=0.08566, over 16720.00 frames. ], tot_loss[loss=0.2435, simple_loss=0.3147, pruned_loss=0.08608, over 1307935.19 frames. ], batch size: 83, lr: 1.42e-02, grad_scale: 1.0 2023-04-28 05:52:06,682 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=40712.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 05:52:18,621 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.95 vs. limit=2.0 2023-04-28 05:52:29,352 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.6825, 4.4480, 4.2064, 2.0304, 3.4249, 2.7269, 4.0325, 4.1566], device='cuda:1'), covar=tensor([0.0228, 0.0454, 0.0362, 0.1461, 0.0592, 0.0865, 0.0573, 0.0744], device='cuda:1'), in_proj_covar=tensor([0.0135, 0.0116, 0.0154, 0.0142, 0.0132, 0.0125, 0.0136, 0.0127], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-04-28 05:52:59,301 INFO [train.py:904] (1/8) Epoch 5, batch 150, loss[loss=0.2576, simple_loss=0.3355, pruned_loss=0.0899, over 16722.00 frames. ], tot_loss[loss=0.2408, simple_loss=0.3133, pruned_loss=0.08411, over 1753135.42 frames. ], batch size: 57, lr: 1.42e-02, grad_scale: 1.0 2023-04-28 05:53:08,164 INFO [optim.py:368] (1/8) 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 05:53:54,556 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.5178, 3.3279, 2.5606, 2.2798, 2.4634, 2.1607, 3.4181, 3.4301], device='cuda:1'), covar=tensor([0.1844, 0.0732, 0.1265, 0.1424, 0.1961, 0.1363, 0.0456, 0.0622], device='cuda:1'), in_proj_covar=tensor([0.0273, 0.0249, 0.0267, 0.0239, 0.0270, 0.0199, 0.0232, 0.0235], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-28 05:54:09,222 INFO [train.py:904] (1/8) Epoch 5, batch 200, loss[loss=0.2706, simple_loss=0.3197, pruned_loss=0.1107, over 16752.00 frames. ], tot_loss[loss=0.2387, simple_loss=0.3116, pruned_loss=0.0829, over 2104440.29 frames. ], batch size: 124, lr: 1.42e-02, grad_scale: 1.0 2023-04-28 05:55:06,311 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.3439, 4.0078, 4.2912, 4.5286, 4.5894, 4.0842, 4.4267, 4.6074], device='cuda:1'), covar=tensor([0.0742, 0.0750, 0.1033, 0.0446, 0.0388, 0.0870, 0.0783, 0.0316], device='cuda:1'), in_proj_covar=tensor([0.0370, 0.0459, 0.0580, 0.0456, 0.0341, 0.0333, 0.0364, 0.0380], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-28 05:55:13,917 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=40848.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 05:55:17,642 INFO [train.py:904] (1/8) Epoch 5, batch 250, loss[loss=0.2493, simple_loss=0.3044, pruned_loss=0.09715, over 16786.00 frames. ], tot_loss[loss=0.2365, simple_loss=0.3081, pruned_loss=0.08243, over 2382384.84 frames. ], batch size: 102, lr: 1.41e-02, grad_scale: 1.0 2023-04-28 05:55:17,981 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=40851.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 05:55:25,557 INFO [optim.py:368] (1/8) 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] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=40865.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 05:55:50,402 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=40874.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 05:55:51,724 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.5635, 3.6914, 1.8584, 3.6817, 2.5675, 3.7225, 1.9975, 2.7718], device='cuda:1'), covar=tensor([0.0113, 0.0225, 0.1383, 0.0112, 0.0663, 0.0382, 0.1220, 0.0531], device='cuda:1'), in_proj_covar=tensor([0.0112, 0.0147, 0.0176, 0.0082, 0.0159, 0.0176, 0.0185, 0.0163], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-04-28 05:56:24,969 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=40899.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 05:56:26,383 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=40900.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 05:56:27,682 INFO [train.py:904] (1/8) Epoch 5, batch 300, loss[loss=0.2595, simple_loss=0.3194, pruned_loss=0.09975, over 12454.00 frames. ], tot_loss[loss=0.2317, simple_loss=0.3039, pruned_loss=0.07977, over 2583857.72 frames. ], batch size: 246, lr: 1.41e-02, grad_scale: 1.0 2023-04-28 05:56:58,125 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=40922.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 05:57:04,442 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=40926.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 05:57:07,200 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=40928.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 05:57:21,533 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.44 vs. limit=2.0 2023-04-28 05:57:29,492 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.6305, 2.7382, 1.6174, 2.7387, 2.1312, 2.7866, 1.9475, 2.3490], device='cuda:1'), covar=tensor([0.0159, 0.0290, 0.1309, 0.0103, 0.0607, 0.0419, 0.1097, 0.0637], device='cuda:1'), in_proj_covar=tensor([0.0113, 0.0148, 0.0177, 0.0082, 0.0159, 0.0177, 0.0186, 0.0163], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-04-28 05:57:39,679 INFO [train.py:904] (1/8) Epoch 5, batch 350, loss[loss=0.2467, simple_loss=0.3254, pruned_loss=0.08405, over 16706.00 frames. ], tot_loss[loss=0.2283, simple_loss=0.3007, pruned_loss=0.07801, over 2741382.99 frames. ], batch size: 57, lr: 1.41e-02, grad_scale: 1.0 2023-04-28 05:57:45,330 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.89 vs. limit=2.0 2023-04-28 05:57:48,103 INFO [optim.py:368] (1/8) 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 05:57:54,048 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=40961.0, num_to_drop=1, layers_to_drop={3} 2023-04-28 05:58:05,106 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.38 vs. limit=2.0 2023-04-28 05:58:14,878 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=40976.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 05:58:22,906 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.3843, 4.8774, 5.3796, 5.3076, 5.2232, 4.7885, 4.3205, 4.5473], device='cuda:1'), covar=tensor([0.0562, 0.0398, 0.0330, 0.0467, 0.0637, 0.0465, 0.1416, 0.0463], device='cuda:1'), in_proj_covar=tensor([0.0228, 0.0220, 0.0224, 0.0230, 0.0266, 0.0238, 0.0338, 0.0205], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003], device='cuda:1') 2023-04-28 05:58:28,034 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=40986.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 05:58:44,174 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 5, batch 400, loss[loss=0.2031, simple_loss=0.2844, pruned_loss=0.06086, over 16680.00 frames. ], tot_loss[loss=0.2259, simple_loss=0.2979, pruned_loss=0.07697, over 2873496.16 frames. ], batch size: 62, lr: 1.41e-02, grad_scale: 2.0 2023-04-28 05:58:58,509 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=41007.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 05:59:36,592 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=41034.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 05:59:51,061 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 5, batch 450, loss[loss=0.2424, simple_loss=0.3207, pruned_loss=0.08199, over 17095.00 frames. ], tot_loss[loss=0.2233, simple_loss=0.2962, pruned_loss=0.07521, over 2975951.18 frames. ], batch size: 53, lr: 1.41e-02, grad_scale: 2.0 2023-04-28 06:00:09,409 INFO [optim.py:368] (1/8) 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 06:00:11,802 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=41058.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 06:01:10,604 INFO [train.py:904] (1/8) Epoch 5, batch 500, loss[loss=0.1884, simple_loss=0.267, pruned_loss=0.05495, over 17190.00 frames. ], tot_loss[loss=0.2214, simple_loss=0.2946, pruned_loss=0.07411, over 3063982.34 frames. ], batch size: 46, lr: 1.41e-02, grad_scale: 2.0 2023-04-28 06:01:12,239 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=41102.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 06:01:16,682 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=41106.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 06:01:58,157 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.38 vs. limit=2.0 2023-04-28 06:02:14,428 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 5, batch 550, loss[loss=0.2085, simple_loss=0.2995, pruned_loss=0.05874, over 16670.00 frames. ], tot_loss[loss=0.221, simple_loss=0.2946, pruned_loss=0.07369, over 3117626.72 frames. ], batch size: 57, lr: 1.41e-02, grad_scale: 2.0 2023-04-28 06:02:27,244 INFO [optim.py:368] (1/8) 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 06:02:35,957 INFO [zipformer.py:625] (1/8) 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] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=41196.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 06:03:27,689 INFO [train.py:904] (1/8) Epoch 5, batch 600, loss[loss=0.1899, simple_loss=0.2772, pruned_loss=0.05131, over 17210.00 frames. ], tot_loss[loss=0.2204, simple_loss=0.2936, pruned_loss=0.07363, over 3159203.17 frames. ], batch size: 46, lr: 1.41e-02, grad_scale: 2.0 2023-04-28 06:03:54,891 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=41221.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 06:04:23,215 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.5883, 4.3640, 3.9693, 2.1033, 3.2102, 2.3735, 3.8400, 4.0283], device='cuda:1'), covar=tensor([0.0252, 0.0417, 0.0437, 0.1457, 0.0628, 0.1005, 0.0568, 0.0775], device='cuda:1'), in_proj_covar=tensor([0.0140, 0.0128, 0.0155, 0.0142, 0.0134, 0.0127, 0.0140, 0.0135], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-04-28 06:04:33,941 INFO [train.py:904] (1/8) Epoch 5, batch 650, loss[loss=0.2066, simple_loss=0.2721, pruned_loss=0.07049, over 16854.00 frames. ], tot_loss[loss=0.2178, simple_loss=0.2912, pruned_loss=0.0722, over 3203885.57 frames. ], batch size: 109, lr: 1.41e-02, grad_scale: 2.0 2023-04-28 06:04:40,901 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=41256.0, num_to_drop=1, layers_to_drop={2} 2023-04-28 06:04:42,445 INFO [optim.py:368] (1/8) Clipping_scale=2.0, grad-norm quartiles 2.330e+02 3.128e+02 3.726e+02 4.799e+02 9.270e+02, threshold=7.452e+02, percent-clipped=1.0 2023-04-28 06:04:44,149 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([1.7498, 3.2079, 2.8304, 4.4502, 3.9790, 4.1338, 1.7960, 3.2174], device='cuda:1'), covar=tensor([0.1324, 0.0485, 0.0901, 0.0080, 0.0234, 0.0263, 0.1157, 0.0601], device='cuda:1'), in_proj_covar=tensor([0.0143, 0.0137, 0.0163, 0.0082, 0.0163, 0.0167, 0.0155, 0.0159], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0003, 0.0004], device='cuda:1') 2023-04-28 06:05:32,852 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.90 vs. limit=2.0 2023-04-28 06:05:39,909 INFO [train.py:904] (1/8) Epoch 5, batch 700, loss[loss=0.2224, simple_loss=0.2869, pruned_loss=0.07892, over 16885.00 frames. ], tot_loss[loss=0.218, simple_loss=0.2917, pruned_loss=0.07217, over 3233869.47 frames. ], batch size: 96, lr: 1.41e-02, grad_scale: 2.0 2023-04-28 06:05:49,028 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=41307.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 06:06:36,846 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.39 vs. limit=2.0 2023-04-28 06:06:49,378 INFO [train.py:904] (1/8) Epoch 5, batch 750, loss[loss=0.2046, simple_loss=0.2855, pruned_loss=0.06182, over 17058.00 frames. ], tot_loss[loss=0.2182, simple_loss=0.2917, pruned_loss=0.07237, over 3262894.72 frames. ], batch size: 50, lr: 1.41e-02, grad_scale: 2.0 2023-04-28 06:06:52,746 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=41353.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 06:06:55,088 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=41355.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 06:06:57,955 INFO [optim.py:368] (1/8) 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 06:07:58,821 INFO [train.py:904] (1/8) Epoch 5, batch 800, loss[loss=0.2012, simple_loss=0.2743, pruned_loss=0.06409, over 16526.00 frames. ], tot_loss[loss=0.2176, simple_loss=0.2908, pruned_loss=0.07216, over 3281678.17 frames. ], batch size: 75, lr: 1.40e-02, grad_scale: 4.0 2023-04-28 06:07:59,138 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=41401.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 06:09:05,234 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=4.04 vs. limit=5.0 2023-04-28 06:09:06,198 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.8552, 4.2702, 4.6579, 2.9958, 4.1135, 4.6049, 4.2399, 2.6910], device='cuda:1'), covar=tensor([0.0252, 0.0017, 0.0014, 0.0204, 0.0037, 0.0023, 0.0026, 0.0226], device='cuda:1'), in_proj_covar=tensor([0.0113, 0.0059, 0.0059, 0.0112, 0.0059, 0.0066, 0.0061, 0.0103], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:1') 2023-04-28 06:09:08,623 INFO [train.py:904] (1/8) Epoch 5, batch 850, loss[loss=0.2003, simple_loss=0.2939, pruned_loss=0.05334, over 17056.00 frames. ], tot_loss[loss=0.2157, simple_loss=0.2893, pruned_loss=0.07102, over 3285130.02 frames. ], batch size: 53, lr: 1.40e-02, grad_scale: 4.0 2023-04-28 06:09:16,402 INFO [optim.py:368] (1/8) 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 06:09:17,864 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=41458.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 06:10:16,011 INFO [train.py:904] (1/8) Epoch 5, batch 900, loss[loss=0.2377, simple_loss=0.2984, pruned_loss=0.08848, over 11972.00 frames. ], tot_loss[loss=0.2147, simple_loss=0.2886, pruned_loss=0.0704, over 3277681.32 frames. ], batch size: 248, lr: 1.40e-02, grad_scale: 4.0 2023-04-28 06:10:44,982 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=41521.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 06:11:27,561 INFO [train.py:904] (1/8) Epoch 5, batch 950, loss[loss=0.2135, simple_loss=0.2795, pruned_loss=0.07375, over 16416.00 frames. ], tot_loss[loss=0.2147, simple_loss=0.2889, pruned_loss=0.07029, over 3291249.90 frames. ], batch size: 146, lr: 1.40e-02, grad_scale: 4.0 2023-04-28 06:11:34,465 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=41556.0, num_to_drop=1, layers_to_drop={0} 2023-04-28 06:11:35,283 INFO [optim.py:368] (1/8) 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 06:11:43,352 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.5700, 4.3430, 4.0238, 2.0592, 3.0652, 2.5158, 3.7976, 4.0166], device='cuda:1'), covar=tensor([0.0305, 0.0548, 0.0428, 0.1538, 0.0724, 0.0941, 0.0664, 0.0825], device='cuda:1'), in_proj_covar=tensor([0.0141, 0.0131, 0.0156, 0.0142, 0.0135, 0.0126, 0.0141, 0.0137], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-04-28 06:11:52,918 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=41569.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 06:12:33,339 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.5242, 4.4337, 4.3673, 3.8903, 4.3655, 1.8771, 4.1784, 4.2668], device='cuda:1'), covar=tensor([0.0071, 0.0060, 0.0089, 0.0230, 0.0061, 0.1532, 0.0088, 0.0113], device='cuda:1'), in_proj_covar=tensor([0.0096, 0.0079, 0.0127, 0.0130, 0.0095, 0.0142, 0.0110, 0.0122], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-04-28 06:12:37,216 INFO [train.py:904] (1/8) Epoch 5, batch 1000, loss[loss=0.2273, simple_loss=0.2904, pruned_loss=0.08212, over 11772.00 frames. ], tot_loss[loss=0.2138, simple_loss=0.2878, pruned_loss=0.06994, over 3297697.51 frames. ], batch size: 246, lr: 1.40e-02, grad_scale: 4.0 2023-04-28 06:12:41,515 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=41604.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 06:13:43,777 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.6169, 4.9989, 4.6709, 4.7527, 4.3927, 4.2589, 4.4246, 4.9500], device='cuda:1'), covar=tensor([0.0663, 0.0686, 0.0832, 0.0434, 0.0623, 0.0877, 0.0659, 0.0832], device='cuda:1'), in_proj_covar=tensor([0.0363, 0.0497, 0.0419, 0.0319, 0.0311, 0.0315, 0.0393, 0.0356], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-28 06:13:45,696 INFO [train.py:904] (1/8) Epoch 5, batch 1050, loss[loss=0.2523, simple_loss=0.3035, pruned_loss=0.1006, over 16926.00 frames. ], tot_loss[loss=0.2134, simple_loss=0.2873, pruned_loss=0.06978, over 3303458.20 frames. ], batch size: 109, lr: 1.40e-02, grad_scale: 4.0 2023-04-28 06:13:48,333 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=41653.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 06:13:50,300 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=41654.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 06:13:54,766 INFO [optim.py:368] (1/8) 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 06:14:05,760 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.0277, 3.2148, 1.6948, 3.2011, 2.3596, 3.2872, 1.9704, 2.5956], device='cuda:1'), covar=tensor([0.0143, 0.0267, 0.1425, 0.0139, 0.0699, 0.0428, 0.1202, 0.0571], device='cuda:1'), in_proj_covar=tensor([0.0115, 0.0151, 0.0174, 0.0086, 0.0159, 0.0183, 0.0184, 0.0163], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-04-28 06:14:56,088 INFO [train.py:904] (1/8) Epoch 5, batch 1100, loss[loss=0.1888, simple_loss=0.2824, pruned_loss=0.04758, over 17127.00 frames. ], tot_loss[loss=0.2131, simple_loss=0.2872, pruned_loss=0.06948, over 3292379.95 frames. ], batch size: 47, lr: 1.40e-02, grad_scale: 4.0 2023-04-28 06:14:56,430 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=41701.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 06:14:56,493 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=41701.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 06:15:15,124 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=41715.0, num_to_drop=1, layers_to_drop={0} 2023-04-28 06:15:37,335 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=4.63 vs. limit=5.0 2023-04-28 06:16:02,694 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=41749.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 06:16:05,297 INFO [train.py:904] (1/8) Epoch 5, batch 1150, loss[loss=0.2303, simple_loss=0.2913, pruned_loss=0.08466, over 15472.00 frames. ], tot_loss[loss=0.212, simple_loss=0.2867, pruned_loss=0.06867, over 3287783.39 frames. ], batch size: 191, lr: 1.40e-02, grad_scale: 4.0 2023-04-28 06:16:12,831 INFO [optim.py:368] (1/8) Clipping_scale=2.0, grad-norm quartiles 1.896e+02 2.850e+02 3.603e+02 4.605e+02 7.939e+02, threshold=7.207e+02, percent-clipped=2.0 2023-04-28 06:16:15,044 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=41758.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 06:16:21,160 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([5.2366, 5.1357, 4.9569, 4.2931, 4.9216, 1.9779, 4.7565, 5.0297], device='cuda:1'), covar=tensor([0.0046, 0.0044, 0.0083, 0.0318, 0.0052, 0.1511, 0.0079, 0.0104], device='cuda:1'), in_proj_covar=tensor([0.0097, 0.0080, 0.0128, 0.0133, 0.0095, 0.0143, 0.0110, 0.0124], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-04-28 06:17:14,374 INFO [train.py:904] (1/8) Epoch 5, batch 1200, loss[loss=0.2179, simple_loss=0.2845, pruned_loss=0.07563, over 16511.00 frames. ], tot_loss[loss=0.2103, simple_loss=0.2857, pruned_loss=0.06747, over 3300525.77 frames. ], batch size: 146, lr: 1.40e-02, grad_scale: 8.0 2023-04-28 06:17:21,143 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 5, batch 1250, loss[loss=0.2048, simple_loss=0.2972, pruned_loss=0.0562, over 17230.00 frames. ], tot_loss[loss=0.2125, simple_loss=0.2862, pruned_loss=0.06942, over 3301217.42 frames. ], batch size: 52, lr: 1.40e-02, grad_scale: 8.0 2023-04-28 06:18:26,481 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.2995, 3.8885, 3.9726, 1.9137, 4.1235, 4.0641, 3.2716, 3.1044], device='cuda:1'), covar=tensor([0.0818, 0.0104, 0.0137, 0.1148, 0.0053, 0.0072, 0.0280, 0.0426], device='cuda:1'), in_proj_covar=tensor([0.0140, 0.0086, 0.0085, 0.0141, 0.0071, 0.0080, 0.0113, 0.0130], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003, 0.0003], device='cuda:1') 2023-04-28 06:18:31,520 INFO [optim.py:368] (1/8) 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 06:18:36,738 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.9974, 4.4004, 2.0720, 4.6750, 2.8516, 4.6066, 2.1464, 3.0571], device='cuda:1'), covar=tensor([0.0109, 0.0181, 0.1512, 0.0036, 0.0761, 0.0302, 0.1318, 0.0607], device='cuda:1'), in_proj_covar=tensor([0.0118, 0.0156, 0.0176, 0.0088, 0.0163, 0.0187, 0.0187, 0.0166], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-04-28 06:19:04,825 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.2619, 3.3753, 1.7973, 3.4135, 2.4511, 3.4441, 2.0220, 2.7500], device='cuda:1'), covar=tensor([0.0148, 0.0280, 0.1398, 0.0145, 0.0743, 0.0471, 0.1102, 0.0483], device='cuda:1'), in_proj_covar=tensor([0.0118, 0.0156, 0.0176, 0.0088, 0.0163, 0.0187, 0.0186, 0.0165], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-04-28 06:19:17,627 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.2958, 2.5845, 1.8798, 2.2034, 2.9438, 2.7148, 3.4759, 3.2243], device='cuda:1'), covar=tensor([0.0031, 0.0171, 0.0250, 0.0214, 0.0102, 0.0156, 0.0083, 0.0085], device='cuda:1'), in_proj_covar=tensor([0.0082, 0.0153, 0.0151, 0.0150, 0.0147, 0.0151, 0.0125, 0.0132], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-28 06:19:23,645 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.29 vs. limit=2.0 2023-04-28 06:19:30,797 INFO [train.py:904] (1/8) Epoch 5, batch 1300, loss[loss=0.1848, simple_loss=0.2794, pruned_loss=0.04511, over 17018.00 frames. ], tot_loss[loss=0.2118, simple_loss=0.286, pruned_loss=0.06882, over 3300815.05 frames. ], batch size: 50, lr: 1.40e-02, grad_scale: 8.0 2023-04-28 06:19:32,266 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.7237, 5.0526, 5.1117, 5.1615, 4.9621, 5.6282, 5.2474, 4.9444], device='cuda:1'), covar=tensor([0.0791, 0.1475, 0.1451, 0.1469, 0.2808, 0.0987, 0.0946, 0.1920], device='cuda:1'), in_proj_covar=tensor([0.0268, 0.0396, 0.0378, 0.0336, 0.0452, 0.0408, 0.0313, 0.0454], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-04-28 06:19:38,946 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=41906.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 06:20:01,371 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.85 vs. limit=2.0 2023-04-28 06:20:22,097 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.8229, 1.5003, 2.2454, 2.6382, 2.6998, 2.8362, 1.7603, 2.9829], device='cuda:1'), covar=tensor([0.0067, 0.0219, 0.0150, 0.0116, 0.0095, 0.0095, 0.0188, 0.0032], device='cuda:1'), in_proj_covar=tensor([0.0117, 0.0141, 0.0128, 0.0126, 0.0125, 0.0090, 0.0139, 0.0078], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001], device='cuda:1') 2023-04-28 06:20:34,349 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.6480, 4.4846, 4.0456, 2.1618, 3.0813, 2.6538, 3.6589, 4.1238], device='cuda:1'), covar=tensor([0.0229, 0.0411, 0.0441, 0.1407, 0.0701, 0.0898, 0.0673, 0.0787], device='cuda:1'), in_proj_covar=tensor([0.0138, 0.0132, 0.0152, 0.0140, 0.0132, 0.0124, 0.0141, 0.0137], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-04-28 06:20:38,052 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=41948.0, num_to_drop=1, layers_to_drop={0} 2023-04-28 06:20:42,283 INFO [train.py:904] (1/8) Epoch 5, batch 1350, loss[loss=0.2289, simple_loss=0.3041, pruned_loss=0.07685, over 15613.00 frames. ], tot_loss[loss=0.2117, simple_loss=0.2862, pruned_loss=0.06861, over 3312247.71 frames. ], batch size: 191, lr: 1.40e-02, grad_scale: 4.0 2023-04-28 06:20:51,207 INFO [optim.py:368] (1/8) Clipping_scale=2.0, grad-norm quartiles 2.161e+02 3.407e+02 4.000e+02 4.882e+02 1.065e+03, threshold=8.000e+02, percent-clipped=1.0 2023-04-28 06:21:04,968 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.0698, 2.4764, 1.8586, 2.0338, 2.9770, 2.7313, 3.3121, 3.1066], device='cuda:1'), covar=tensor([0.0040, 0.0174, 0.0227, 0.0221, 0.0094, 0.0148, 0.0087, 0.0091], device='cuda:1'), in_proj_covar=tensor([0.0082, 0.0152, 0.0150, 0.0149, 0.0146, 0.0151, 0.0125, 0.0131], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-28 06:21:04,972 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=41967.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 06:21:55,629 INFO [train.py:904] (1/8) Epoch 5, batch 1400, loss[loss=0.1855, simple_loss=0.2658, pruned_loss=0.05259, over 16848.00 frames. ], tot_loss[loss=0.2124, simple_loss=0.2864, pruned_loss=0.06922, over 3309557.94 frames. ], batch size: 42, lr: 1.40e-02, grad_scale: 4.0 2023-04-28 06:22:08,281 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=42009.0, num_to_drop=1, layers_to_drop={1} 2023-04-28 06:22:09,281 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=42010.0, num_to_drop=1, layers_to_drop={0} 2023-04-28 06:22:30,886 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.8422, 4.9570, 5.0824, 5.0358, 4.9975, 5.5875, 5.1880, 4.8620], device='cuda:1'), covar=tensor([0.0803, 0.1588, 0.1228, 0.1548, 0.2380, 0.0859, 0.1049, 0.1978], device='cuda:1'), in_proj_covar=tensor([0.0269, 0.0398, 0.0379, 0.0340, 0.0454, 0.0408, 0.0312, 0.0454], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-04-28 06:23:03,804 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.8586, 4.4567, 4.1737, 4.9942, 5.0828, 4.4949, 5.0269, 5.1522], device='cuda:1'), covar=tensor([0.0766, 0.0804, 0.2653, 0.0913, 0.0800, 0.0753, 0.1094, 0.0680], device='cuda:1'), in_proj_covar=tensor([0.0405, 0.0500, 0.0636, 0.0505, 0.0381, 0.0371, 0.0401, 0.0416], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-28 06:23:05,347 INFO [train.py:904] (1/8) Epoch 5, batch 1450, loss[loss=0.2021, simple_loss=0.2888, pruned_loss=0.05764, over 17283.00 frames. ], tot_loss[loss=0.2111, simple_loss=0.2851, pruned_loss=0.06852, over 3300165.75 frames. ], batch size: 52, lr: 1.39e-02, grad_scale: 4.0 2023-04-28 06:23:11,603 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.35 vs. limit=2.0 2023-04-28 06:23:15,558 INFO [optim.py:368] (1/8) 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 06:24:13,340 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([5.5243, 5.9214, 5.5752, 5.7668, 5.1116, 5.0815, 5.3683, 6.0069], device='cuda:1'), covar=tensor([0.0697, 0.0685, 0.0985, 0.0435, 0.0626, 0.0521, 0.0612, 0.0704], device='cuda:1'), in_proj_covar=tensor([0.0363, 0.0503, 0.0418, 0.0318, 0.0313, 0.0317, 0.0393, 0.0357], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-28 06:24:14,179 INFO [train.py:904] (1/8) Epoch 5, batch 1500, loss[loss=0.1884, simple_loss=0.2701, pruned_loss=0.05336, over 17205.00 frames. ], tot_loss[loss=0.2104, simple_loss=0.2845, pruned_loss=0.06812, over 3291352.18 frames. ], batch size: 44, lr: 1.39e-02, grad_scale: 4.0 2023-04-28 06:24:34,854 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=2.65 vs. limit=5.0 2023-04-28 06:25:01,169 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.64 vs. limit=2.0 2023-04-28 06:25:21,155 INFO [train.py:904] (1/8) Epoch 5, batch 1550, loss[loss=0.2432, simple_loss=0.2985, pruned_loss=0.09398, over 16880.00 frames. ], tot_loss[loss=0.213, simple_loss=0.2864, pruned_loss=0.06981, over 3300466.99 frames. ], batch size: 96, lr: 1.39e-02, grad_scale: 4.0 2023-04-28 06:25:32,954 INFO [optim.py:368] (1/8) 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 06:26:17,032 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=3.49 vs. limit=5.0 2023-04-28 06:26:32,627 INFO [train.py:904] (1/8) Epoch 5, batch 1600, loss[loss=0.2142, simple_loss=0.3074, pruned_loss=0.06053, over 17247.00 frames. ], tot_loss[loss=0.2135, simple_loss=0.2876, pruned_loss=0.06973, over 3306144.98 frames. ], batch size: 52, lr: 1.39e-02, grad_scale: 8.0 2023-04-28 06:26:53,436 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.89 vs. limit=2.0 2023-04-28 06:27:00,570 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.6806, 3.6566, 3.7086, 3.5679, 3.5664, 4.1401, 3.7963, 3.4714], device='cuda:1'), covar=tensor([0.2110, 0.1717, 0.1602, 0.2447, 0.2960, 0.1598, 0.1383, 0.2777], device='cuda:1'), in_proj_covar=tensor([0.0271, 0.0400, 0.0385, 0.0342, 0.0455, 0.0408, 0.0312, 0.0451], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-04-28 06:27:01,881 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.6988, 2.5428, 1.8464, 2.1958, 2.8942, 2.8075, 3.1188, 3.0534], device='cuda:1'), covar=tensor([0.0061, 0.0128, 0.0199, 0.0172, 0.0074, 0.0119, 0.0079, 0.0080], device='cuda:1'), in_proj_covar=tensor([0.0084, 0.0151, 0.0151, 0.0150, 0.0148, 0.0153, 0.0127, 0.0134], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-28 06:27:32,629 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.3093, 3.2606, 3.9181, 2.5966, 3.6099, 3.8966, 3.7227, 2.0269], device='cuda:1'), covar=tensor([0.0315, 0.0113, 0.0031, 0.0219, 0.0050, 0.0041, 0.0033, 0.0296], device='cuda:1'), in_proj_covar=tensor([0.0115, 0.0061, 0.0060, 0.0113, 0.0060, 0.0066, 0.0062, 0.0104], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:1') 2023-04-28 06:27:39,756 INFO [train.py:904] (1/8) Epoch 5, batch 1650, loss[loss=0.2302, simple_loss=0.2889, pruned_loss=0.0858, over 16720.00 frames. ], tot_loss[loss=0.2148, simple_loss=0.2888, pruned_loss=0.07036, over 3303395.54 frames. ], batch size: 134, lr: 1.39e-02, grad_scale: 8.0 2023-04-28 06:27:49,823 INFO [optim.py:368] (1/8) 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 06:27:56,372 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=42262.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 06:28:41,221 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.9673, 5.3222, 5.4325, 5.3174, 5.3762, 5.9337, 5.5489, 5.2790], device='cuda:1'), covar=tensor([0.0778, 0.1769, 0.1510, 0.1744, 0.2622, 0.0908, 0.0994, 0.2193], device='cuda:1'), in_proj_covar=tensor([0.0273, 0.0405, 0.0382, 0.0344, 0.0459, 0.0409, 0.0312, 0.0456], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-04-28 06:28:42,865 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.50 vs. limit=2.0 2023-04-28 06:28:49,288 INFO [train.py:904] (1/8) Epoch 5, batch 1700, loss[loss=0.2103, simple_loss=0.302, pruned_loss=0.05927, over 17048.00 frames. ], tot_loss[loss=0.2157, simple_loss=0.2901, pruned_loss=0.07068, over 3310046.94 frames. ], batch size: 50, lr: 1.39e-02, grad_scale: 8.0 2023-04-28 06:28:54,199 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=42304.0, num_to_drop=1, layers_to_drop={1} 2023-04-28 06:28:56,016 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.46 vs. limit=2.0 2023-04-28 06:29:01,207 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 5, batch 1750, loss[loss=0.3137, simple_loss=0.3641, pruned_loss=0.1316, over 12179.00 frames. ], tot_loss[loss=0.2171, simple_loss=0.2919, pruned_loss=0.07114, over 3310603.87 frames. ], batch size: 246, lr: 1.39e-02, grad_scale: 8.0 2023-04-28 06:30:05,725 INFO [optim.py:368] (1/8) 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 06:30:05,950 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=42358.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 06:30:35,448 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=42379.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 06:31:06,097 INFO [train.py:904] (1/8) Epoch 5, batch 1800, loss[loss=0.2644, simple_loss=0.3405, pruned_loss=0.0941, over 16594.00 frames. ], tot_loss[loss=0.2169, simple_loss=0.2925, pruned_loss=0.07072, over 3317268.57 frames. ], batch size: 68, lr: 1.39e-02, grad_scale: 8.0 2023-04-28 06:32:01,370 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=42440.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 06:32:16,848 INFO [train.py:904] (1/8) Epoch 5, batch 1850, loss[loss=0.2008, simple_loss=0.2996, pruned_loss=0.05101, over 16665.00 frames. ], tot_loss[loss=0.2178, simple_loss=0.2934, pruned_loss=0.07109, over 3325620.62 frames. ], batch size: 62, lr: 1.39e-02, grad_scale: 8.0 2023-04-28 06:32:26,219 INFO [optim.py:368] (1/8) 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 06:33:25,573 INFO [train.py:904] (1/8) Epoch 5, batch 1900, loss[loss=0.2217, simple_loss=0.2919, pruned_loss=0.0757, over 16874.00 frames. ], tot_loss[loss=0.2167, simple_loss=0.2926, pruned_loss=0.07044, over 3319439.11 frames. ], batch size: 116, lr: 1.39e-02, grad_scale: 8.0 2023-04-28 06:33:32,656 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=42506.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 06:34:35,496 INFO [train.py:904] (1/8) Epoch 5, batch 1950, loss[loss=0.2394, simple_loss=0.3127, pruned_loss=0.083, over 16432.00 frames. ], tot_loss[loss=0.217, simple_loss=0.293, pruned_loss=0.07052, over 3325547.01 frames. ], batch size: 146, lr: 1.39e-02, grad_scale: 8.0 2023-04-28 06:34:47,175 INFO [optim.py:368] (1/8) 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 06:34:52,229 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.9812, 4.8355, 4.9488, 3.5382, 4.8598, 1.7959, 4.6118, 4.9146], device='cuda:1'), covar=tensor([0.0108, 0.0095, 0.0109, 0.0590, 0.0088, 0.1951, 0.0117, 0.0191], device='cuda:1'), in_proj_covar=tensor([0.0098, 0.0081, 0.0128, 0.0134, 0.0097, 0.0141, 0.0111, 0.0124], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-04-28 06:34:52,233 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=42562.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 06:34:52,299 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.5620, 2.6379, 2.1215, 2.3713, 2.8846, 2.8309, 3.6865, 3.1693], device='cuda:1'), covar=tensor([0.0025, 0.0155, 0.0202, 0.0184, 0.0113, 0.0152, 0.0073, 0.0091], device='cuda:1'), in_proj_covar=tensor([0.0084, 0.0152, 0.0151, 0.0149, 0.0147, 0.0155, 0.0127, 0.0132], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-28 06:34:53,292 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([5.1705, 5.0162, 4.9982, 4.3501, 4.9578, 1.9540, 4.8145, 5.0686], device='cuda:1'), covar=tensor([0.0058, 0.0051, 0.0079, 0.0317, 0.0063, 0.1480, 0.0079, 0.0104], device='cuda:1'), in_proj_covar=tensor([0.0098, 0.0081, 0.0128, 0.0134, 0.0097, 0.0141, 0.0111, 0.0124], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-04-28 06:34:59,300 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.4129, 4.2900, 4.4051, 4.4080, 4.3130, 4.8623, 4.5250, 4.2449], device='cuda:1'), covar=tensor([0.1310, 0.1756, 0.1318, 0.1569, 0.2353, 0.0940, 0.1089, 0.2109], device='cuda:1'), in_proj_covar=tensor([0.0274, 0.0403, 0.0377, 0.0340, 0.0457, 0.0409, 0.0312, 0.0454], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-04-28 06:34:59,445 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=42567.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 06:35:48,040 INFO [train.py:904] (1/8) Epoch 5, batch 2000, loss[loss=0.1914, simple_loss=0.2612, pruned_loss=0.06083, over 16790.00 frames. ], tot_loss[loss=0.216, simple_loss=0.2922, pruned_loss=0.0699, over 3330434.59 frames. ], batch size: 102, lr: 1.39e-02, grad_scale: 8.0 2023-04-28 06:35:50,538 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.5034, 3.6934, 3.6484, 1.6476, 3.9129, 3.8147, 3.1785, 3.0095], device='cuda:1'), covar=tensor([0.0596, 0.0085, 0.0146, 0.1089, 0.0059, 0.0078, 0.0300, 0.0321], device='cuda:1'), in_proj_covar=tensor([0.0138, 0.0086, 0.0084, 0.0142, 0.0070, 0.0079, 0.0114, 0.0127], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003, 0.0003], device='cuda:1') 2023-04-28 06:35:51,706 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=42604.0, num_to_drop=1, layers_to_drop={0} 2023-04-28 06:36:00,217 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=42610.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 06:36:20,933 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.0226, 4.4119, 3.5696, 2.5483, 3.2029, 2.4101, 4.5968, 4.4013], device='cuda:1'), covar=tensor([0.2055, 0.0572, 0.1021, 0.1489, 0.2348, 0.1419, 0.0327, 0.0580], device='cuda:1'), in_proj_covar=tensor([0.0277, 0.0254, 0.0271, 0.0243, 0.0304, 0.0203, 0.0240, 0.0265], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003], device='cuda:1') 2023-04-28 06:36:47,457 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.5855, 4.3495, 4.5888, 4.8218, 4.8856, 4.3368, 4.7700, 4.8562], device='cuda:1'), covar=tensor([0.0702, 0.0671, 0.1099, 0.0423, 0.0384, 0.0758, 0.0700, 0.0378], device='cuda:1'), in_proj_covar=tensor([0.0405, 0.0494, 0.0636, 0.0513, 0.0381, 0.0373, 0.0399, 0.0418], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-28 06:36:55,036 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=42649.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 06:36:57,537 INFO [train.py:904] (1/8) Epoch 5, batch 2050, loss[loss=0.2516, simple_loss=0.3086, pruned_loss=0.09734, over 16675.00 frames. ], tot_loss[loss=0.2176, simple_loss=0.2936, pruned_loss=0.07074, over 3318152.53 frames. ], batch size: 134, lr: 1.38e-02, grad_scale: 8.0 2023-04-28 06:36:59,031 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=42652.0, num_to_drop=1, layers_to_drop={0} 2023-04-28 06:37:06,885 INFO [optim.py:368] (1/8) 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 06:37:54,682 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.0113, 1.3826, 2.3334, 2.8060, 2.7890, 3.2085, 1.7516, 3.0996], device='cuda:1'), covar=tensor([0.0088, 0.0262, 0.0144, 0.0129, 0.0109, 0.0072, 0.0201, 0.0054], device='cuda:1'), in_proj_covar=tensor([0.0120, 0.0144, 0.0128, 0.0130, 0.0126, 0.0092, 0.0138, 0.0082], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001], device='cuda:1') 2023-04-28 06:38:05,314 INFO [train.py:904] (1/8) Epoch 5, batch 2100, loss[loss=0.226, simple_loss=0.3027, pruned_loss=0.0746, over 17028.00 frames. ], tot_loss[loss=0.2195, simple_loss=0.2953, pruned_loss=0.07186, over 3315451.61 frames. ], batch size: 55, lr: 1.38e-02, grad_scale: 8.0 2023-04-28 06:38:18,111 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=42710.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 06:38:46,981 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.0874, 1.6122, 2.4411, 2.9076, 2.8573, 3.1191, 1.8369, 3.1584], device='cuda:1'), covar=tensor([0.0066, 0.0226, 0.0118, 0.0108, 0.0091, 0.0093, 0.0190, 0.0046], device='cuda:1'), in_proj_covar=tensor([0.0120, 0.0145, 0.0129, 0.0131, 0.0127, 0.0093, 0.0139, 0.0082], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001], device='cuda:1') 2023-04-28 06:38:52,287 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=42735.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 06:39:15,216 INFO [train.py:904] (1/8) Epoch 5, batch 2150, loss[loss=0.2251, simple_loss=0.304, pruned_loss=0.0731, over 16622.00 frames. ], tot_loss[loss=0.2194, simple_loss=0.2954, pruned_loss=0.07169, over 3320715.54 frames. ], batch size: 62, lr: 1.38e-02, grad_scale: 8.0 2023-04-28 06:39:24,098 INFO [optim.py:368] (1/8) 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 06:40:17,532 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.93 vs. limit=2.0 2023-04-28 06:40:23,577 INFO [train.py:904] (1/8) Epoch 5, batch 2200, loss[loss=0.2344, simple_loss=0.3029, pruned_loss=0.08297, over 16844.00 frames. ], tot_loss[loss=0.2196, simple_loss=0.2955, pruned_loss=0.07187, over 3324279.61 frames. ], batch size: 96, lr: 1.38e-02, grad_scale: 8.0 2023-04-28 06:40:44,859 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.81 vs. limit=2.0 2023-04-28 06:40:52,450 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.0204, 3.9219, 3.0374, 5.3156, 4.9130, 4.7009, 1.6021, 3.5326], device='cuda:1'), covar=tensor([0.1192, 0.0383, 0.0915, 0.0060, 0.0192, 0.0258, 0.1418, 0.0567], device='cuda:1'), in_proj_covar=tensor([0.0142, 0.0141, 0.0166, 0.0085, 0.0177, 0.0169, 0.0157, 0.0161], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0003, 0.0004], device='cuda:1') 2023-04-28 06:41:34,694 INFO [train.py:904] (1/8) Epoch 5, batch 2250, loss[loss=0.1935, simple_loss=0.2738, pruned_loss=0.05656, over 16842.00 frames. ], tot_loss[loss=0.22, simple_loss=0.296, pruned_loss=0.07205, over 3326126.15 frames. ], batch size: 42, lr: 1.38e-02, grad_scale: 8.0 2023-04-28 06:41:43,705 INFO [optim.py:368] (1/8) 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 06:41:45,439 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=42859.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 06:41:49,465 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=42862.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 06:42:14,406 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=42879.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 06:42:32,786 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.7726, 3.5539, 3.7106, 3.5598, 3.6810, 4.1472, 3.9345, 3.6265], device='cuda:1'), covar=tensor([0.1788, 0.2285, 0.1695, 0.2696, 0.3015, 0.1619, 0.1224, 0.2514], device='cuda:1'), in_proj_covar=tensor([0.0280, 0.0408, 0.0386, 0.0348, 0.0469, 0.0414, 0.0320, 0.0459], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-04-28 06:42:44,587 INFO [train.py:904] (1/8) Epoch 5, batch 2300, loss[loss=0.229, simple_loss=0.3023, pruned_loss=0.07779, over 16472.00 frames. ], tot_loss[loss=0.2208, simple_loss=0.2967, pruned_loss=0.07245, over 3311243.80 frames. ], batch size: 146, lr: 1.38e-02, grad_scale: 4.0 2023-04-28 06:42:53,835 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([5.2749, 5.1992, 5.1478, 4.5739, 5.0842, 2.0853, 4.8845, 5.1932], device='cuda:1'), covar=tensor([0.0051, 0.0049, 0.0066, 0.0274, 0.0053, 0.1307, 0.0078, 0.0092], device='cuda:1'), in_proj_covar=tensor([0.0097, 0.0082, 0.0127, 0.0134, 0.0096, 0.0138, 0.0112, 0.0123], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002], device='cuda:1') 2023-04-28 06:43:11,616 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=42920.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 06:43:38,237 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=42940.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 06:43:53,246 INFO [train.py:904] (1/8) Epoch 5, batch 2350, loss[loss=0.2101, simple_loss=0.2973, pruned_loss=0.06148, over 17141.00 frames. ], tot_loss[loss=0.2207, simple_loss=0.2965, pruned_loss=0.07241, over 3311139.40 frames. ], batch size: 48, lr: 1.38e-02, grad_scale: 4.0 2023-04-28 06:44:03,299 INFO [optim.py:368] (1/8) 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 06:44:25,697 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([5.2522, 5.3626, 5.1993, 4.4989, 5.1053, 2.0180, 4.8684, 5.1820], device='cuda:1'), covar=tensor([0.0060, 0.0037, 0.0065, 0.0297, 0.0057, 0.1340, 0.0070, 0.0092], device='cuda:1'), in_proj_covar=tensor([0.0097, 0.0083, 0.0128, 0.0134, 0.0097, 0.0138, 0.0111, 0.0123], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-04-28 06:44:26,873 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.7065, 4.1283, 4.4281, 1.7423, 4.6880, 4.6901, 3.4621, 3.4607], device='cuda:1'), covar=tensor([0.0687, 0.0123, 0.0166, 0.1188, 0.0051, 0.0066, 0.0256, 0.0368], device='cuda:1'), in_proj_covar=tensor([0.0138, 0.0085, 0.0085, 0.0139, 0.0071, 0.0079, 0.0113, 0.0126], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003, 0.0003], device='cuda:1') 2023-04-28 06:45:02,226 INFO [train.py:904] (1/8) Epoch 5, batch 2400, loss[loss=0.2198, simple_loss=0.2958, pruned_loss=0.07185, over 17239.00 frames. ], tot_loss[loss=0.2215, simple_loss=0.2978, pruned_loss=0.07264, over 3318506.48 frames. ], batch size: 44, lr: 1.38e-02, grad_scale: 8.0 2023-04-28 06:45:08,089 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=43005.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 06:45:30,627 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.89 vs. limit=2.0 2023-04-28 06:45:52,173 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 5, batch 2450, loss[loss=0.2817, simple_loss=0.346, pruned_loss=0.1086, over 11924.00 frames. ], tot_loss[loss=0.2221, simple_loss=0.2987, pruned_loss=0.07275, over 3316104.99 frames. ], batch size: 246, lr: 1.38e-02, grad_scale: 4.0 2023-04-28 06:46:26,010 INFO [optim.py:368] (1/8) 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 06:46:57,965 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=43083.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 06:47:23,969 INFO [train.py:904] (1/8) Epoch 5, batch 2500, loss[loss=0.2322, simple_loss=0.3081, pruned_loss=0.07811, over 16897.00 frames. ], tot_loss[loss=0.2216, simple_loss=0.2988, pruned_loss=0.07224, over 3313558.54 frames. ], batch size: 116, lr: 1.38e-02, grad_scale: 4.0 2023-04-28 06:47:35,565 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.94 vs. limit=2.0 2023-04-28 06:48:33,486 INFO [train.py:904] (1/8) Epoch 5, batch 2550, loss[loss=0.2614, simple_loss=0.3243, pruned_loss=0.09925, over 16472.00 frames. ], tot_loss[loss=0.2211, simple_loss=0.2984, pruned_loss=0.07193, over 3318421.05 frames. ], batch size: 146, lr: 1.38e-02, grad_scale: 4.0 2023-04-28 06:48:45,560 INFO [optim.py:368] (1/8) 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 06:48:48,807 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=43162.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 06:49:04,580 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.38 vs. limit=2.0 2023-04-28 06:49:15,313 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.8256, 5.0173, 5.0825, 5.0595, 4.9460, 5.5567, 5.2438, 4.9125], device='cuda:1'), covar=tensor([0.0941, 0.1675, 0.1284, 0.1492, 0.2488, 0.0993, 0.1280, 0.2279], device='cuda:1'), in_proj_covar=tensor([0.0281, 0.0398, 0.0383, 0.0341, 0.0465, 0.0413, 0.0321, 0.0457], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-04-28 06:49:43,652 INFO [train.py:904] (1/8) Epoch 5, batch 2600, loss[loss=0.2392, simple_loss=0.3045, pruned_loss=0.087, over 16848.00 frames. ], tot_loss[loss=0.2215, simple_loss=0.2986, pruned_loss=0.07224, over 3306405.49 frames. ], batch size: 96, lr: 1.38e-02, grad_scale: 4.0 2023-04-28 06:49:55,550 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=43210.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 06:50:01,751 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=43215.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 06:50:31,265 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=43235.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 06:50:53,869 INFO [train.py:904] (1/8) Epoch 5, batch 2650, loss[loss=0.2239, simple_loss=0.2898, pruned_loss=0.07903, over 16834.00 frames. ], tot_loss[loss=0.2207, simple_loss=0.2985, pruned_loss=0.07147, over 3315525.00 frames. ], batch size: 96, lr: 1.37e-02, grad_scale: 4.0 2023-04-28 06:51:05,580 INFO [optim.py:368] (1/8) 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 06:51:16,629 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.9071, 4.1984, 4.3626, 1.9019, 4.6077, 4.6965, 3.3023, 3.7779], device='cuda:1'), covar=tensor([0.0653, 0.0117, 0.0142, 0.1134, 0.0050, 0.0063, 0.0314, 0.0314], device='cuda:1'), in_proj_covar=tensor([0.0141, 0.0086, 0.0088, 0.0143, 0.0073, 0.0080, 0.0116, 0.0129], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003, 0.0003], device='cuda:1') 2023-04-28 06:51:20,754 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.85 vs. limit=2.0 2023-04-28 06:51:43,938 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([1.7849, 3.2061, 2.5229, 4.4450, 4.0841, 4.0672, 1.5888, 2.9349], device='cuda:1'), covar=tensor([0.1232, 0.0412, 0.0912, 0.0063, 0.0267, 0.0274, 0.1186, 0.0678], device='cuda:1'), in_proj_covar=tensor([0.0143, 0.0140, 0.0162, 0.0084, 0.0179, 0.0169, 0.0156, 0.0159], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0003, 0.0004], device='cuda:1') 2023-04-28 06:52:00,845 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([1.7405, 3.2257, 2.6215, 4.6271, 4.0923, 4.1642, 1.8318, 2.9247], device='cuda:1'), covar=tensor([0.1306, 0.0470, 0.0901, 0.0060, 0.0312, 0.0289, 0.1148, 0.0716], device='cuda:1'), in_proj_covar=tensor([0.0144, 0.0142, 0.0164, 0.0085, 0.0181, 0.0171, 0.0158, 0.0161], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0003, 0.0004], device='cuda:1') 2023-04-28 06:52:02,500 INFO [train.py:904] (1/8) Epoch 5, batch 2700, loss[loss=0.217, simple_loss=0.3132, pruned_loss=0.06041, over 17271.00 frames. ], tot_loss[loss=0.2199, simple_loss=0.298, pruned_loss=0.07089, over 3316863.33 frames. ], batch size: 52, lr: 1.37e-02, grad_scale: 4.0 2023-04-28 06:52:09,120 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=43305.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 06:52:15,345 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=3.55 vs. limit=5.0 2023-04-28 06:52:34,198 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.3725, 4.3708, 3.5563, 2.0145, 2.7660, 2.4594, 3.4922, 4.0710], device='cuda:1'), covar=tensor([0.0262, 0.0382, 0.0521, 0.1528, 0.0772, 0.0900, 0.0634, 0.0626], device='cuda:1'), in_proj_covar=tensor([0.0139, 0.0132, 0.0153, 0.0140, 0.0132, 0.0124, 0.0139, 0.0139], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:1') 2023-04-28 06:53:12,573 INFO [train.py:904] (1/8) Epoch 5, batch 2750, loss[loss=0.1946, simple_loss=0.2896, pruned_loss=0.04979, over 17027.00 frames. ], tot_loss[loss=0.2193, simple_loss=0.2976, pruned_loss=0.07046, over 3319723.91 frames. ], batch size: 55, lr: 1.37e-02, grad_scale: 4.0 2023-04-28 06:53:15,874 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=43353.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 06:53:25,425 INFO [optim.py:368] (1/8) 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] (1/8) Epoch 5, batch 2800, loss[loss=0.2051, simple_loss=0.2951, pruned_loss=0.05758, over 17251.00 frames. ], tot_loss[loss=0.2193, simple_loss=0.2976, pruned_loss=0.07054, over 3318412.70 frames. ], batch size: 52, lr: 1.37e-02, grad_scale: 8.0 2023-04-28 06:55:33,338 INFO [train.py:904] (1/8) Epoch 5, batch 2850, loss[loss=0.221, simple_loss=0.3056, pruned_loss=0.06823, over 16770.00 frames. ], tot_loss[loss=0.2191, simple_loss=0.2972, pruned_loss=0.07052, over 3319460.91 frames. ], batch size: 57, lr: 1.37e-02, grad_scale: 8.0 2023-04-28 06:55:45,525 INFO [optim.py:368] (1/8) 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 06:56:41,595 INFO [train.py:904] (1/8) Epoch 5, batch 2900, loss[loss=0.2422, simple_loss=0.2928, pruned_loss=0.09582, over 16840.00 frames. ], tot_loss[loss=0.2184, simple_loss=0.2951, pruned_loss=0.0708, over 3318203.17 frames. ], batch size: 116, lr: 1.37e-02, grad_scale: 8.0 2023-04-28 06:57:00,324 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=43514.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 06:57:01,458 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=43515.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 06:57:15,478 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.2069, 2.3837, 1.7674, 2.0956, 2.9486, 2.6955, 3.3578, 3.1771], device='cuda:1'), covar=tensor([0.0032, 0.0170, 0.0222, 0.0193, 0.0097, 0.0148, 0.0090, 0.0085], device='cuda:1'), in_proj_covar=tensor([0.0084, 0.0154, 0.0152, 0.0149, 0.0150, 0.0157, 0.0133, 0.0135], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-28 06:57:28,500 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 5, batch 2950, loss[loss=0.2463, simple_loss=0.3249, pruned_loss=0.08381, over 17067.00 frames. ], tot_loss[loss=0.2169, simple_loss=0.2935, pruned_loss=0.07016, over 3329712.68 frames. ], batch size: 53, lr: 1.37e-02, grad_scale: 8.0 2023-04-28 06:58:02,027 INFO [optim.py:368] (1/8) 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] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=43563.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 06:58:24,115 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=43575.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 06:58:29,591 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.0752, 1.6985, 2.3186, 2.8059, 2.7244, 3.2569, 1.6840, 3.1538], device='cuda:1'), covar=tensor([0.0090, 0.0216, 0.0134, 0.0123, 0.0111, 0.0074, 0.0197, 0.0060], device='cuda:1'), in_proj_covar=tensor([0.0122, 0.0145, 0.0129, 0.0134, 0.0128, 0.0094, 0.0140, 0.0082], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001], device='cuda:1') 2023-04-28 06:58:34,162 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=43583.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 06:58:46,519 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([1.9130, 1.7084, 1.3305, 1.4654, 1.8762, 1.6775, 1.7101, 1.9564], device='cuda:1'), covar=tensor([0.0051, 0.0117, 0.0173, 0.0152, 0.0083, 0.0136, 0.0100, 0.0086], device='cuda:1'), in_proj_covar=tensor([0.0083, 0.0151, 0.0150, 0.0148, 0.0149, 0.0155, 0.0131, 0.0135], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-28 06:58:59,870 INFO [train.py:904] (1/8) Epoch 5, batch 3000, loss[loss=0.2159, simple_loss=0.2807, pruned_loss=0.0756, over 16944.00 frames. ], tot_loss[loss=0.2185, simple_loss=0.2943, pruned_loss=0.07131, over 3325567.09 frames. ], batch size: 109, lr: 1.37e-02, grad_scale: 8.0 2023-04-28 06:58:59,871 INFO [train.py:929] (1/8) Computing validation loss 2023-04-28 06:59:07,851 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.1565, 3.9666, 3.5011, 1.9341, 3.0009, 2.3009, 3.5030, 3.6197], device='cuda:1'), covar=tensor([0.0302, 0.0444, 0.0447, 0.1555, 0.0625, 0.0964, 0.0588, 0.0774], device='cuda:1'), in_proj_covar=tensor([0.0141, 0.0132, 0.0155, 0.0141, 0.0132, 0.0126, 0.0140, 0.0140], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-04-28 06:59:08,833 INFO [train.py:938] (1/8) Epoch 5, validation: loss=0.1564, simple_loss=0.2632, pruned_loss=0.02477, over 944034.00 frames. 2023-04-28 06:59:08,833 INFO [train.py:939] (1/8) Maximum memory allocated so far is 17676MB 2023-04-28 06:59:26,428 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=3.92 vs. limit=5.0 2023-04-28 07:00:08,010 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.2925, 4.3391, 3.5913, 1.8058, 2.8004, 2.3618, 3.4984, 3.7460], device='cuda:1'), covar=tensor([0.0298, 0.0401, 0.0528, 0.1659, 0.0798, 0.0964, 0.0738, 0.0887], device='cuda:1'), in_proj_covar=tensor([0.0143, 0.0134, 0.0155, 0.0141, 0.0134, 0.0126, 0.0141, 0.0141], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-04-28 07:00:15,865 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.7994, 4.8222, 5.4001, 5.3264, 5.3316, 4.9314, 4.9630, 4.7452], device='cuda:1'), covar=tensor([0.0248, 0.0356, 0.0307, 0.0375, 0.0476, 0.0277, 0.0787, 0.0318], device='cuda:1'), in_proj_covar=tensor([0.0252, 0.0244, 0.0249, 0.0251, 0.0301, 0.0262, 0.0378, 0.0221], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:1') 2023-04-28 07:00:18,524 INFO [train.py:904] (1/8) Epoch 5, batch 3050, loss[loss=0.2435, simple_loss=0.3022, pruned_loss=0.0924, over 16329.00 frames. ], tot_loss[loss=0.2196, simple_loss=0.2952, pruned_loss=0.07202, over 3322951.64 frames. ], batch size: 165, lr: 1.37e-02, grad_scale: 4.0 2023-04-28 07:00:31,425 INFO [optim.py:368] (1/8) 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 07:00:42,532 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.0513, 1.7368, 2.3728, 3.0090, 2.6959, 3.2946, 1.9932, 3.2872], device='cuda:1'), covar=tensor([0.0067, 0.0207, 0.0129, 0.0102, 0.0104, 0.0082, 0.0189, 0.0055], device='cuda:1'), in_proj_covar=tensor([0.0121, 0.0142, 0.0128, 0.0132, 0.0127, 0.0094, 0.0139, 0.0081], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001], device='cuda:1') 2023-04-28 07:01:25,947 INFO [train.py:904] (1/8) Epoch 5, batch 3100, loss[loss=0.2376, simple_loss=0.3013, pruned_loss=0.08694, over 16265.00 frames. ], tot_loss[loss=0.2185, simple_loss=0.2944, pruned_loss=0.07128, over 3321404.41 frames. ], batch size: 165, lr: 1.37e-02, grad_scale: 4.0 2023-04-28 07:01:32,815 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.9445, 4.2145, 2.2496, 4.5456, 2.5934, 4.5096, 2.3863, 3.2010], device='cuda:1'), covar=tensor([0.0134, 0.0245, 0.1333, 0.0043, 0.0779, 0.0285, 0.1297, 0.0533], device='cuda:1'), in_proj_covar=tensor([0.0116, 0.0157, 0.0174, 0.0088, 0.0158, 0.0188, 0.0183, 0.0163], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-04-28 07:01:48,775 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=4.84 vs. limit=5.0 2023-04-28 07:02:25,467 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([5.0593, 5.4045, 5.4636, 5.3836, 5.2384, 5.9376, 5.5591, 5.2899], device='cuda:1'), covar=tensor([0.0797, 0.1466, 0.1461, 0.1580, 0.2619, 0.0896, 0.1129, 0.2134], device='cuda:1'), in_proj_covar=tensor([0.0291, 0.0408, 0.0396, 0.0345, 0.0465, 0.0419, 0.0330, 0.0463], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-04-28 07:02:33,553 INFO [train.py:904] (1/8) Epoch 5, batch 3150, loss[loss=0.1772, simple_loss=0.2517, pruned_loss=0.05132, over 16800.00 frames. ], tot_loss[loss=0.2185, simple_loss=0.2938, pruned_loss=0.07158, over 3322436.78 frames. ], batch size: 39, lr: 1.37e-02, grad_scale: 4.0 2023-04-28 07:02:46,878 INFO [optim.py:368] (1/8) 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 07:03:22,219 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=43787.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 07:03:42,789 INFO [train.py:904] (1/8) Epoch 5, batch 3200, loss[loss=0.2135, simple_loss=0.2934, pruned_loss=0.06682, over 16788.00 frames. ], tot_loss[loss=0.217, simple_loss=0.2928, pruned_loss=0.07057, over 3331542.22 frames. ], batch size: 83, lr: 1.37e-02, grad_scale: 8.0 2023-04-28 07:04:48,216 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=43848.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 07:04:52,115 INFO [train.py:904] (1/8) Epoch 5, batch 3250, loss[loss=0.2634, simple_loss=0.3299, pruned_loss=0.09844, over 12281.00 frames. ], tot_loss[loss=0.2191, simple_loss=0.2947, pruned_loss=0.07168, over 3323881.74 frames. ], batch size: 247, lr: 1.37e-02, grad_scale: 8.0 2023-04-28 07:05:06,669 INFO [optim.py:368] (1/8) 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 07:05:19,446 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=43870.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 07:05:20,744 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.4706, 3.8074, 3.8055, 1.7188, 3.9541, 3.9869, 3.2472, 3.0568], device='cuda:1'), covar=tensor([0.0651, 0.0084, 0.0115, 0.1095, 0.0055, 0.0069, 0.0285, 0.0356], device='cuda:1'), in_proj_covar=tensor([0.0140, 0.0087, 0.0087, 0.0143, 0.0073, 0.0080, 0.0117, 0.0129], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003, 0.0003], device='cuda:1') 2023-04-28 07:06:03,320 INFO [train.py:904] (1/8) Epoch 5, batch 3300, loss[loss=0.2262, simple_loss=0.3029, pruned_loss=0.07472, over 16505.00 frames. ], tot_loss[loss=0.2194, simple_loss=0.2955, pruned_loss=0.0717, over 3331555.49 frames. ], batch size: 68, lr: 1.36e-02, grad_scale: 8.0 2023-04-28 07:06:14,794 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=43909.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 07:07:01,330 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([1.7647, 2.9945, 2.5560, 4.2356, 3.8082, 3.9000, 1.4544, 2.8994], device='cuda:1'), covar=tensor([0.1171, 0.0451, 0.0883, 0.0063, 0.0272, 0.0311, 0.1254, 0.0620], device='cuda:1'), in_proj_covar=tensor([0.0140, 0.0139, 0.0162, 0.0085, 0.0181, 0.0171, 0.0156, 0.0157], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0003, 0.0004], device='cuda:1') 2023-04-28 07:07:12,756 INFO [train.py:904] (1/8) Epoch 5, batch 3350, loss[loss=0.2303, simple_loss=0.2995, pruned_loss=0.08056, over 15447.00 frames. ], tot_loss[loss=0.2185, simple_loss=0.2952, pruned_loss=0.07091, over 3335648.47 frames. ], batch size: 191, lr: 1.36e-02, grad_scale: 8.0 2023-04-28 07:07:24,252 INFO [zipformer.py:625] (1/8) 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] (1/8) 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 07:07:39,216 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=43970.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 07:07:47,917 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.6224, 2.7275, 1.6848, 2.7693, 2.0924, 2.7605, 1.8767, 2.3737], device='cuda:1'), covar=tensor([0.0168, 0.0305, 0.1325, 0.0119, 0.0699, 0.0447, 0.1224, 0.0501], device='cuda:1'), in_proj_covar=tensor([0.0117, 0.0158, 0.0175, 0.0090, 0.0159, 0.0190, 0.0183, 0.0162], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-04-28 07:08:22,251 INFO [train.py:904] (1/8) Epoch 5, batch 3400, loss[loss=0.2369, simple_loss=0.2934, pruned_loss=0.09023, over 16346.00 frames. ], tot_loss[loss=0.2182, simple_loss=0.2948, pruned_loss=0.07081, over 3327497.60 frames. ], batch size: 165, lr: 1.36e-02, grad_scale: 8.0 2023-04-28 07:08:47,617 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 5, batch 3450, loss[loss=0.2086, simple_loss=0.2997, pruned_loss=0.05875, over 17133.00 frames. ], tot_loss[loss=0.216, simple_loss=0.2926, pruned_loss=0.0697, over 3334410.13 frames. ], batch size: 47, lr: 1.36e-02, grad_scale: 8.0 2023-04-28 07:09:44,809 INFO [optim.py:368] (1/8) 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 07:09:52,898 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.32 vs. limit=2.0 2023-04-28 07:10:00,116 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.2233, 2.3289, 2.4817, 4.7895, 1.9293, 3.6598, 2.4998, 2.5638], device='cuda:1'), covar=tensor([0.0459, 0.1984, 0.1003, 0.0218, 0.3117, 0.0743, 0.1810, 0.2483], device='cuda:1'), in_proj_covar=tensor([0.0317, 0.0320, 0.0255, 0.0312, 0.0361, 0.0307, 0.0284, 0.0387], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-28 07:10:39,128 INFO [train.py:904] (1/8) Epoch 5, batch 3500, loss[loss=0.2144, simple_loss=0.2937, pruned_loss=0.06749, over 17216.00 frames. ], tot_loss[loss=0.2142, simple_loss=0.2909, pruned_loss=0.0688, over 3330340.52 frames. ], batch size: 44, lr: 1.36e-02, grad_scale: 8.0 2023-04-28 07:11:37,785 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=44143.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 07:11:49,617 INFO [train.py:904] (1/8) Epoch 5, batch 3550, loss[loss=0.1804, simple_loss=0.2655, pruned_loss=0.04765, over 17215.00 frames. ], tot_loss[loss=0.213, simple_loss=0.2897, pruned_loss=0.06817, over 3327259.62 frames. ], batch size: 45, lr: 1.36e-02, grad_scale: 8.0 2023-04-28 07:12:03,021 INFO [optim.py:368] (1/8) 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 07:12:17,208 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.8785, 3.3336, 3.4676, 3.4597, 3.4225, 3.2328, 3.2328, 3.2881], device='cuda:1'), covar=tensor([0.0438, 0.0418, 0.0375, 0.0437, 0.0445, 0.0371, 0.0642, 0.0467], device='cuda:1'), in_proj_covar=tensor([0.0253, 0.0246, 0.0251, 0.0253, 0.0300, 0.0264, 0.0378, 0.0223], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:1') 2023-04-28 07:12:17,230 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=44170.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 07:12:34,232 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.56 vs. limit=2.0 2023-04-28 07:12:59,203 INFO [train.py:904] (1/8) Epoch 5, batch 3600, loss[loss=0.1954, simple_loss=0.2843, pruned_loss=0.05323, over 17087.00 frames. ], tot_loss[loss=0.213, simple_loss=0.2894, pruned_loss=0.0683, over 3313646.65 frames. ], batch size: 55, lr: 1.36e-02, grad_scale: 8.0 2023-04-28 07:13:01,404 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.89 vs. limit=2.0 2023-04-28 07:13:23,275 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=44218.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 07:13:35,668 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.1412, 3.7490, 3.1170, 1.9488, 2.6437, 2.3310, 3.6183, 3.5861], device='cuda:1'), covar=tensor([0.0204, 0.0476, 0.0605, 0.1427, 0.0689, 0.0809, 0.0477, 0.0636], device='cuda:1'), in_proj_covar=tensor([0.0143, 0.0135, 0.0154, 0.0139, 0.0132, 0.0124, 0.0141, 0.0143], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:1') 2023-04-28 07:13:55,933 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.3162, 3.4008, 3.9776, 2.7283, 3.6277, 4.0144, 3.8266, 2.1236], device='cuda:1'), covar=tensor([0.0285, 0.0130, 0.0027, 0.0187, 0.0040, 0.0041, 0.0034, 0.0287], device='cuda:1'), in_proj_covar=tensor([0.0114, 0.0059, 0.0060, 0.0112, 0.0061, 0.0068, 0.0066, 0.0104], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:1') 2023-04-28 07:14:09,666 INFO [train.py:904] (1/8) Epoch 5, batch 3650, loss[loss=0.2448, simple_loss=0.3025, pruned_loss=0.0935, over 11518.00 frames. ], tot_loss[loss=0.2137, simple_loss=0.2888, pruned_loss=0.06933, over 3296536.91 frames. ], batch size: 246, lr: 1.36e-02, grad_scale: 8.0 2023-04-28 07:14:25,438 INFO [optim.py:368] (1/8) 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 07:14:30,507 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=44265.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 07:15:22,468 INFO [train.py:904] (1/8) Epoch 5, batch 3700, loss[loss=0.2136, simple_loss=0.2821, pruned_loss=0.07255, over 16284.00 frames. ], tot_loss[loss=0.2138, simple_loss=0.2869, pruned_loss=0.07034, over 3290999.29 frames. ], batch size: 165, lr: 1.36e-02, grad_scale: 8.0 2023-04-28 07:15:25,269 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.40 vs. limit=2.0 2023-04-28 07:15:44,547 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=44315.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 07:16:36,544 INFO [train.py:904] (1/8) Epoch 5, batch 3750, loss[loss=0.2094, simple_loss=0.2959, pruned_loss=0.06143, over 17105.00 frames. ], tot_loss[loss=0.2165, simple_loss=0.2887, pruned_loss=0.07217, over 3269955.71 frames. ], batch size: 49, lr: 1.36e-02, grad_scale: 8.0 2023-04-28 07:16:52,716 INFO [optim.py:368] (1/8) 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] (1/8) Epoch 5, batch 3800, loss[loss=0.2048, simple_loss=0.2717, pruned_loss=0.06901, over 16842.00 frames. ], tot_loss[loss=0.2184, simple_loss=0.2899, pruned_loss=0.07345, over 3267950.51 frames. ], batch size: 90, lr: 1.36e-02, grad_scale: 8.0 2023-04-28 07:18:32,718 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=3.36 vs. limit=5.0 2023-04-28 07:18:49,798 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=44443.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 07:19:01,027 INFO [train.py:904] (1/8) Epoch 5, batch 3850, loss[loss=0.1843, simple_loss=0.2505, pruned_loss=0.05907, over 16642.00 frames. ], tot_loss[loss=0.2196, simple_loss=0.2902, pruned_loss=0.07448, over 3262021.27 frames. ], batch size: 76, lr: 1.36e-02, grad_scale: 8.0 2023-04-28 07:19:16,580 INFO [optim.py:368] (1/8) Clipping_scale=2.0, grad-norm quartiles 1.887e+02 2.978e+02 3.515e+02 4.161e+02 7.942e+02, threshold=7.030e+02, percent-clipped=2.0 2023-04-28 07:19:57,670 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=44491.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 07:20:11,847 INFO [train.py:904] (1/8) Epoch 5, batch 3900, loss[loss=0.2181, simple_loss=0.2948, pruned_loss=0.07073, over 16893.00 frames. ], tot_loss[loss=0.2192, simple_loss=0.2891, pruned_loss=0.07467, over 3265419.02 frames. ], batch size: 90, lr: 1.36e-02, grad_scale: 8.0 2023-04-28 07:20:25,869 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.5838, 2.1568, 2.1990, 4.1385, 1.9527, 3.0494, 2.2226, 2.1903], device='cuda:1'), covar=tensor([0.0575, 0.1991, 0.1100, 0.0314, 0.2681, 0.0929, 0.2139, 0.2139], device='cuda:1'), in_proj_covar=tensor([0.0318, 0.0325, 0.0259, 0.0313, 0.0364, 0.0314, 0.0292, 0.0395], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-28 07:20:49,848 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=44527.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 07:21:22,429 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 5, batch 3950, loss[loss=0.2123, simple_loss=0.2744, pruned_loss=0.07513, over 16863.00 frames. ], tot_loss[loss=0.2189, simple_loss=0.2883, pruned_loss=0.07479, over 3273473.57 frames. ], batch size: 96, lr: 1.36e-02, grad_scale: 8.0 2023-04-28 07:21:37,715 INFO [optim.py:368] (1/8) 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 07:21:44,216 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=44565.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 07:22:15,348 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=44588.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 07:22:34,880 INFO [train.py:904] (1/8) Epoch 5, batch 4000, loss[loss=0.2062, simple_loss=0.2833, pruned_loss=0.06456, over 16522.00 frames. ], tot_loss[loss=0.2188, simple_loss=0.2881, pruned_loss=0.07477, over 3271529.86 frames. ], batch size: 68, lr: 1.35e-02, grad_scale: 8.0 2023-04-28 07:22:48,841 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=44611.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 07:22:51,757 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=44613.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 07:22:54,854 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=44615.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 07:22:59,791 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.7889, 4.4404, 3.4314, 1.8044, 2.8342, 2.3824, 3.9004, 4.1404], device='cuda:1'), covar=tensor([0.0200, 0.0300, 0.0659, 0.1931, 0.0888, 0.1035, 0.0624, 0.0672], device='cuda:1'), in_proj_covar=tensor([0.0139, 0.0133, 0.0153, 0.0140, 0.0132, 0.0124, 0.0140, 0.0142], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:1') 2023-04-28 07:23:31,054 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([5.2425, 5.6411, 5.2743, 5.3446, 4.8190, 4.7247, 5.0898, 5.7168], device='cuda:1'), covar=tensor([0.0558, 0.0528, 0.0868, 0.0463, 0.0636, 0.0623, 0.0586, 0.0585], device='cuda:1'), in_proj_covar=tensor([0.0365, 0.0495, 0.0418, 0.0321, 0.0314, 0.0323, 0.0397, 0.0352], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-28 07:23:45,848 INFO [train.py:904] (1/8) Epoch 5, batch 4050, loss[loss=0.1969, simple_loss=0.2753, pruned_loss=0.05924, over 16492.00 frames. ], tot_loss[loss=0.216, simple_loss=0.2869, pruned_loss=0.07261, over 3279436.18 frames. ], batch size: 75, lr: 1.35e-02, grad_scale: 8.0 2023-04-28 07:23:57,932 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([5.3536, 5.2800, 4.9750, 4.1328, 5.1874, 1.9179, 4.8303, 4.9445], device='cuda:1'), covar=tensor([0.0041, 0.0044, 0.0096, 0.0369, 0.0042, 0.1686, 0.0078, 0.0144], device='cuda:1'), in_proj_covar=tensor([0.0096, 0.0086, 0.0126, 0.0134, 0.0097, 0.0139, 0.0113, 0.0126], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-04-28 07:24:02,948 INFO [optim.py:368] (1/8) 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 07:24:04,435 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=44663.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 07:25:01,123 INFO [train.py:904] (1/8) Epoch 5, batch 4100, loss[loss=0.2219, simple_loss=0.3042, pruned_loss=0.06985, over 16958.00 frames. ], tot_loss[loss=0.2148, simple_loss=0.2869, pruned_loss=0.07131, over 3272852.24 frames. ], batch size: 109, lr: 1.35e-02, grad_scale: 8.0 2023-04-28 07:25:42,019 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.4771, 4.6174, 4.9448, 4.9224, 4.9741, 4.4978, 4.2828, 4.3773], device='cuda:1'), covar=tensor([0.0345, 0.0371, 0.0415, 0.0544, 0.0615, 0.0401, 0.1162, 0.0392], device='cuda:1'), in_proj_covar=tensor([0.0244, 0.0236, 0.0240, 0.0244, 0.0293, 0.0254, 0.0366, 0.0213], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:1') 2023-04-28 07:25:51,216 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([1.9363, 3.7814, 3.6722, 2.2826, 3.3330, 3.5278, 3.5216, 1.9400], device='cuda:1'), covar=tensor([0.0330, 0.0018, 0.0024, 0.0251, 0.0045, 0.0055, 0.0031, 0.0311], device='cuda:1'), in_proj_covar=tensor([0.0114, 0.0057, 0.0058, 0.0113, 0.0061, 0.0067, 0.0063, 0.0105], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:1') 2023-04-28 07:26:06,057 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([1.7880, 1.6151, 1.4113, 1.4576, 1.7140, 1.5429, 1.7603, 1.8754], device='cuda:1'), covar=tensor([0.0042, 0.0123, 0.0184, 0.0159, 0.0087, 0.0143, 0.0067, 0.0084], device='cuda:1'), in_proj_covar=tensor([0.0082, 0.0150, 0.0151, 0.0148, 0.0148, 0.0154, 0.0126, 0.0135], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-28 07:26:16,728 INFO [train.py:904] (1/8) Epoch 5, batch 4150, loss[loss=0.2849, simple_loss=0.3436, pruned_loss=0.1131, over 11457.00 frames. ], tot_loss[loss=0.2228, simple_loss=0.2953, pruned_loss=0.07509, over 3245154.24 frames. ], batch size: 246, lr: 1.35e-02, grad_scale: 8.0 2023-04-28 07:26:34,907 INFO [optim.py:368] (1/8) 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 07:27:12,914 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([1.8056, 2.8166, 2.6464, 1.9076, 2.5315, 2.5908, 2.6845, 1.7006], device='cuda:1'), covar=tensor([0.0233, 0.0028, 0.0032, 0.0186, 0.0044, 0.0048, 0.0037, 0.0264], device='cuda:1'), in_proj_covar=tensor([0.0114, 0.0056, 0.0058, 0.0113, 0.0060, 0.0068, 0.0063, 0.0105], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:1') 2023-04-28 07:27:35,372 INFO [train.py:904] (1/8) Epoch 5, batch 4200, loss[loss=0.245, simple_loss=0.3288, pruned_loss=0.08059, over 16373.00 frames. ], tot_loss[loss=0.2286, simple_loss=0.3028, pruned_loss=0.07716, over 3212509.39 frames. ], batch size: 146, lr: 1.35e-02, grad_scale: 8.0 2023-04-28 07:28:36,617 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.6866, 4.5762, 4.3683, 3.8090, 4.4786, 1.7659, 4.2699, 4.3760], device='cuda:1'), covar=tensor([0.0043, 0.0041, 0.0080, 0.0220, 0.0049, 0.1592, 0.0069, 0.0109], device='cuda:1'), in_proj_covar=tensor([0.0092, 0.0082, 0.0120, 0.0129, 0.0093, 0.0136, 0.0107, 0.0120], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-28 07:28:51,233 INFO [train.py:904] (1/8) Epoch 5, batch 4250, loss[loss=0.2123, simple_loss=0.2978, pruned_loss=0.06339, over 16199.00 frames. ], tot_loss[loss=0.2302, simple_loss=0.3058, pruned_loss=0.07732, over 3181236.57 frames. ], batch size: 165, lr: 1.35e-02, grad_scale: 8.0 2023-04-28 07:29:07,183 INFO [optim.py:368] (1/8) 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 07:29:38,861 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 5, batch 4300, loss[loss=0.2319, simple_loss=0.3196, pruned_loss=0.07213, over 17241.00 frames. ], tot_loss[loss=0.2297, simple_loss=0.3072, pruned_loss=0.07612, over 3189577.49 frames. ], batch size: 52, lr: 1.35e-02, grad_scale: 8.0 2023-04-28 07:30:13,092 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=44906.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 07:31:19,635 INFO [train.py:904] (1/8) Epoch 5, batch 4350, loss[loss=0.2447, simple_loss=0.3253, pruned_loss=0.08209, over 16358.00 frames. ], tot_loss[loss=0.2337, simple_loss=0.3113, pruned_loss=0.07809, over 3186326.51 frames. ], batch size: 165, lr: 1.35e-02, grad_scale: 8.0 2023-04-28 07:31:36,571 INFO [optim.py:368] (1/8) 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 07:32:35,452 INFO [train.py:904] (1/8) Epoch 5, batch 4400, loss[loss=0.2385, simple_loss=0.3153, pruned_loss=0.0808, over 16666.00 frames. ], tot_loss[loss=0.2361, simple_loss=0.3135, pruned_loss=0.07937, over 3165961.81 frames. ], batch size: 134, lr: 1.35e-02, grad_scale: 8.0 2023-04-28 07:33:49,609 INFO [train.py:904] (1/8) Epoch 5, batch 4450, loss[loss=0.2334, simple_loss=0.3226, pruned_loss=0.07212, over 16694.00 frames. ], tot_loss[loss=0.2371, simple_loss=0.3161, pruned_loss=0.07903, over 3184677.09 frames. ], batch size: 89, lr: 1.35e-02, grad_scale: 8.0 2023-04-28 07:33:56,451 INFO [zipformer.py:625] (1/8) 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] (1/8) 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 07:34:16,586 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.3316, 4.5669, 4.6170, 2.9310, 3.9752, 4.4143, 4.1179, 2.1503], device='cuda:1'), covar=tensor([0.0296, 0.0011, 0.0019, 0.0213, 0.0038, 0.0056, 0.0026, 0.0314], device='cuda:1'), in_proj_covar=tensor([0.0111, 0.0052, 0.0055, 0.0110, 0.0057, 0.0065, 0.0061, 0.0102], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0001, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:1') 2023-04-28 07:34:33,674 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=45081.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 07:35:01,842 INFO [train.py:904] (1/8) Epoch 5, batch 4500, loss[loss=0.2149, simple_loss=0.3043, pruned_loss=0.06274, over 16797.00 frames. ], tot_loss[loss=0.2361, simple_loss=0.3154, pruned_loss=0.0784, over 3195738.15 frames. ], batch size: 102, lr: 1.35e-02, grad_scale: 8.0 2023-04-28 07:35:02,302 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.8224, 5.3479, 5.5303, 5.2806, 5.3210, 5.9087, 5.4901, 5.2004], device='cuda:1'), covar=tensor([0.0733, 0.1331, 0.1404, 0.1579, 0.2112, 0.0805, 0.0956, 0.2043], device='cuda:1'), in_proj_covar=tensor([0.0279, 0.0383, 0.0371, 0.0329, 0.0441, 0.0394, 0.0310, 0.0452], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-04-28 07:35:26,565 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=45117.0, num_to_drop=1, layers_to_drop={3} 2023-04-28 07:35:52,799 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.0054, 3.9860, 1.9471, 4.6002, 2.9510, 4.4079, 2.2026, 2.9184], device='cuda:1'), covar=tensor([0.0089, 0.0174, 0.1320, 0.0022, 0.0593, 0.0197, 0.1188, 0.0549], device='cuda:1'), in_proj_covar=tensor([0.0117, 0.0150, 0.0172, 0.0080, 0.0158, 0.0180, 0.0183, 0.0160], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-04-28 07:36:02,284 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=45142.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 07:36:05,029 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=2.82 vs. limit=5.0 2023-04-28 07:36:13,664 INFO [train.py:904] (1/8) Epoch 5, batch 4550, loss[loss=0.2512, simple_loss=0.328, pruned_loss=0.08721, over 16447.00 frames. ], tot_loss[loss=0.2367, simple_loss=0.3157, pruned_loss=0.07888, over 3205997.07 frames. ], batch size: 68, lr: 1.35e-02, grad_scale: 8.0 2023-04-28 07:36:30,358 INFO [optim.py:368] (1/8) 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 07:37:00,725 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=45183.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 07:37:27,614 INFO [train.py:904] (1/8) Epoch 5, batch 4600, loss[loss=0.2189, simple_loss=0.309, pruned_loss=0.06443, over 16886.00 frames. ], tot_loss[loss=0.2362, simple_loss=0.3157, pruned_loss=0.07838, over 3220756.95 frames. ], batch size: 96, lr: 1.35e-02, grad_scale: 8.0 2023-04-28 07:37:35,969 INFO [zipformer.py:625] (1/8) 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] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=45231.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 07:38:34,238 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.73 vs. limit=2.0 2023-04-28 07:38:40,729 INFO [train.py:904] (1/8) Epoch 5, batch 4650, loss[loss=0.2311, simple_loss=0.3152, pruned_loss=0.07345, over 15504.00 frames. ], tot_loss[loss=0.235, simple_loss=0.3144, pruned_loss=0.07774, over 3225887.68 frames. ], batch size: 190, lr: 1.34e-02, grad_scale: 8.0 2023-04-28 07:38:45,219 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=45254.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 07:38:45,426 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.0140, 4.2487, 1.7644, 4.8390, 2.9887, 4.4808, 1.9771, 2.8981], device='cuda:1'), covar=tensor([0.0118, 0.0156, 0.1716, 0.0018, 0.0610, 0.0203, 0.1366, 0.0568], device='cuda:1'), in_proj_covar=tensor([0.0119, 0.0151, 0.0176, 0.0081, 0.0160, 0.0182, 0.0185, 0.0163], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-04-28 07:38:57,178 INFO [optim.py:368] (1/8) 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 07:39:32,977 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.23 vs. limit=2.0 2023-04-28 07:39:49,264 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.0845, 4.0596, 4.0580, 2.7240, 3.4178, 3.8842, 3.5852, 2.0174], device='cuda:1'), covar=tensor([0.0340, 0.0013, 0.0019, 0.0202, 0.0045, 0.0053, 0.0062, 0.0306], device='cuda:1'), in_proj_covar=tensor([0.0114, 0.0053, 0.0056, 0.0113, 0.0059, 0.0065, 0.0062, 0.0105], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0001, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:1') 2023-04-28 07:39:55,267 INFO [train.py:904] (1/8) Epoch 5, batch 4700, loss[loss=0.2321, simple_loss=0.304, pruned_loss=0.08012, over 12008.00 frames. ], tot_loss[loss=0.2334, simple_loss=0.3127, pruned_loss=0.07705, over 3204499.69 frames. ], batch size: 248, lr: 1.34e-02, grad_scale: 8.0 2023-04-28 07:40:06,974 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.99 vs. limit=2.0 2023-04-28 07:41:03,081 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=45348.0, num_to_drop=1, layers_to_drop={1} 2023-04-28 07:41:04,157 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([5.2204, 5.1752, 5.0108, 4.8799, 4.4249, 5.0062, 4.9703, 4.7118], device='cuda:1'), covar=tensor([0.0361, 0.0220, 0.0155, 0.0139, 0.0820, 0.0298, 0.0158, 0.0400], device='cuda:1'), in_proj_covar=tensor([0.0171, 0.0176, 0.0207, 0.0181, 0.0239, 0.0206, 0.0151, 0.0221], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0002], device='cuda:1') 2023-04-28 07:41:06,793 INFO [train.py:904] (1/8) Epoch 5, batch 4750, loss[loss=0.2127, simple_loss=0.2902, pruned_loss=0.06757, over 16644.00 frames. ], tot_loss[loss=0.2313, simple_loss=0.31, pruned_loss=0.07634, over 3186808.12 frames. ], batch size: 134, lr: 1.34e-02, grad_scale: 8.0 2023-04-28 07:41:20,663 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.33 vs. limit=2.0 2023-04-28 07:41:22,809 INFO [optim.py:368] (1/8) 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 07:41:27,650 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.0156, 1.5271, 2.1364, 2.8657, 2.8036, 3.1911, 1.6522, 3.0443], device='cuda:1'), covar=tensor([0.0064, 0.0255, 0.0163, 0.0123, 0.0105, 0.0061, 0.0231, 0.0047], device='cuda:1'), in_proj_covar=tensor([0.0115, 0.0141, 0.0127, 0.0125, 0.0125, 0.0090, 0.0138, 0.0082], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001], device='cuda:1') 2023-04-28 07:41:34,884 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.5586, 2.6620, 1.6916, 2.7770, 2.1257, 2.7861, 1.8595, 2.3454], device='cuda:1'), covar=tensor([0.0177, 0.0322, 0.1231, 0.0081, 0.0665, 0.0517, 0.1087, 0.0505], device='cuda:1'), in_proj_covar=tensor([0.0118, 0.0149, 0.0173, 0.0080, 0.0158, 0.0180, 0.0183, 0.0160], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-04-28 07:41:37,759 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.0932, 1.5519, 2.1771, 2.9457, 2.7863, 3.2847, 1.7496, 3.1424], device='cuda:1'), covar=tensor([0.0056, 0.0251, 0.0154, 0.0107, 0.0103, 0.0046, 0.0212, 0.0040], device='cuda:1'), in_proj_covar=tensor([0.0116, 0.0141, 0.0127, 0.0126, 0.0126, 0.0090, 0.0138, 0.0082], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001], device='cuda:1') 2023-04-28 07:42:12,428 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([5.0043, 5.3455, 5.0895, 5.1009, 4.7385, 4.5792, 4.8004, 5.4194], device='cuda:1'), covar=tensor([0.0572, 0.0539, 0.0698, 0.0364, 0.0485, 0.0648, 0.0571, 0.0584], device='cuda:1'), in_proj_covar=tensor([0.0340, 0.0453, 0.0392, 0.0297, 0.0290, 0.0301, 0.0371, 0.0327], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-28 07:42:20,534 INFO [train.py:904] (1/8) Epoch 5, batch 4800, loss[loss=0.2567, simple_loss=0.3388, pruned_loss=0.08728, over 15530.00 frames. ], tot_loss[loss=0.2268, simple_loss=0.3056, pruned_loss=0.07402, over 3190404.22 frames. ], batch size: 191, lr: 1.34e-02, grad_scale: 8.0 2023-04-28 07:42:33,276 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=45409.0, num_to_drop=1, layers_to_drop={0} 2023-04-28 07:42:38,044 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=45412.0, num_to_drop=1, layers_to_drop={0} 2023-04-28 07:43:14,513 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=45437.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 07:43:25,810 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.97 vs. limit=2.0 2023-04-28 07:43:34,382 INFO [train.py:904] (1/8) Epoch 5, batch 4850, loss[loss=0.2056, simple_loss=0.2945, pruned_loss=0.05841, over 16712.00 frames. ], tot_loss[loss=0.227, simple_loss=0.3066, pruned_loss=0.07376, over 3169082.78 frames. ], batch size: 83, lr: 1.34e-02, grad_scale: 8.0 2023-04-28 07:43:50,675 INFO [optim.py:368] (1/8) 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 07:44:47,403 INFO [train.py:904] (1/8) Epoch 5, batch 4900, loss[loss=0.2249, simple_loss=0.3088, pruned_loss=0.07049, over 16472.00 frames. ], tot_loss[loss=0.2258, simple_loss=0.306, pruned_loss=0.07284, over 3157198.47 frames. ], batch size: 75, lr: 1.34e-02, grad_scale: 8.0 2023-04-28 07:45:00,254 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.2041, 1.9643, 2.0735, 3.7568, 1.7634, 2.8582, 2.1480, 2.1881], device='cuda:1'), covar=tensor([0.0602, 0.2025, 0.1069, 0.0317, 0.2878, 0.0870, 0.1837, 0.2031], device='cuda:1'), in_proj_covar=tensor([0.0314, 0.0317, 0.0256, 0.0305, 0.0365, 0.0299, 0.0281, 0.0381], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-28 07:45:53,621 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.7016, 4.4614, 4.6608, 4.9631, 5.0904, 4.5075, 5.0327, 5.0360], device='cuda:1'), covar=tensor([0.0826, 0.0712, 0.1069, 0.0418, 0.0374, 0.0520, 0.0362, 0.0342], device='cuda:1'), in_proj_covar=tensor([0.0364, 0.0445, 0.0563, 0.0454, 0.0340, 0.0344, 0.0354, 0.0371], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-28 07:46:01,017 INFO [train.py:904] (1/8) Epoch 5, batch 4950, loss[loss=0.2341, simple_loss=0.3172, pruned_loss=0.07549, over 17264.00 frames. ], tot_loss[loss=0.2247, simple_loss=0.3056, pruned_loss=0.07187, over 3173372.23 frames. ], batch size: 52, lr: 1.34e-02, grad_scale: 8.0 2023-04-28 07:46:15,482 INFO [optim.py:368] (1/8) Clipping_scale=2.0, grad-norm quartiles 2.470e+02 3.018e+02 3.573e+02 4.428e+02 9.346e+02, threshold=7.147e+02, percent-clipped=9.0 2023-04-28 07:46:49,085 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.5744, 4.5902, 3.7038, 1.9632, 3.0391, 2.7168, 3.7346, 4.2079], device='cuda:1'), covar=tensor([0.0237, 0.0325, 0.0578, 0.1552, 0.0708, 0.0784, 0.0619, 0.0553], device='cuda:1'), in_proj_covar=tensor([0.0138, 0.0127, 0.0151, 0.0139, 0.0132, 0.0125, 0.0138, 0.0135], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-04-28 07:47:13,814 INFO [train.py:904] (1/8) Epoch 5, batch 5000, loss[loss=0.2373, simple_loss=0.3216, pruned_loss=0.0765, over 16753.00 frames. ], tot_loss[loss=0.2253, simple_loss=0.3069, pruned_loss=0.07182, over 3192985.45 frames. ], batch size: 89, lr: 1.34e-02, grad_scale: 8.0 2023-04-28 07:48:01,949 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.88 vs. limit=2.0 2023-04-28 07:48:27,075 INFO [train.py:904] (1/8) Epoch 5, batch 5050, loss[loss=0.2769, simple_loss=0.3411, pruned_loss=0.1063, over 11702.00 frames. ], tot_loss[loss=0.2247, simple_loss=0.3068, pruned_loss=0.07128, over 3199453.63 frames. ], batch size: 247, lr: 1.34e-02, grad_scale: 8.0 2023-04-28 07:48:43,127 INFO [optim.py:368] (1/8) 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 07:49:13,334 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.3359, 4.2536, 4.2747, 3.5382, 4.2327, 1.6914, 4.0401, 4.1286], device='cuda:1'), covar=tensor([0.0071, 0.0070, 0.0068, 0.0342, 0.0056, 0.1558, 0.0081, 0.0112], device='cuda:1'), in_proj_covar=tensor([0.0088, 0.0078, 0.0115, 0.0126, 0.0088, 0.0135, 0.0103, 0.0116], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-28 07:49:39,975 INFO [train.py:904] (1/8) Epoch 5, batch 5100, loss[loss=0.1947, simple_loss=0.2732, pruned_loss=0.0581, over 16457.00 frames. ], tot_loss[loss=0.2233, simple_loss=0.305, pruned_loss=0.07081, over 3188088.86 frames. ], batch size: 68, lr: 1.34e-02, grad_scale: 8.0 2023-04-28 07:49:42,424 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=45702.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 07:49:44,453 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=45704.0, num_to_drop=1, layers_to_drop={3} 2023-04-28 07:49:55,988 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=45712.0, num_to_drop=1, layers_to_drop={2} 2023-04-28 07:50:33,159 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=45737.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 07:50:53,052 INFO [train.py:904] (1/8) Epoch 5, batch 5150, loss[loss=0.2375, simple_loss=0.3304, pruned_loss=0.07233, over 16328.00 frames. ], tot_loss[loss=0.2232, simple_loss=0.3056, pruned_loss=0.07039, over 3184920.91 frames. ], batch size: 165, lr: 1.34e-02, grad_scale: 8.0 2023-04-28 07:51:06,960 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=45760.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 07:51:08,846 INFO [optim.py:368] (1/8) 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 07:51:10,528 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=45763.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 07:51:25,993 INFO [zipformer.py:625] (1/8) 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] (1/8) 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] (1/8) Epoch 5, batch 5200, loss[loss=0.1871, simple_loss=0.2746, pruned_loss=0.04982, over 16817.00 frames. ], tot_loss[loss=0.2232, simple_loss=0.3053, pruned_loss=0.07058, over 3183805.63 frames. ], batch size: 83, lr: 1.34e-02, grad_scale: 8.0 2023-04-28 07:52:30,795 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.6203, 3.9960, 1.6248, 4.2432, 2.5571, 4.1476, 1.7610, 2.7221], device='cuda:1'), covar=tensor([0.0145, 0.0175, 0.1742, 0.0029, 0.0735, 0.0245, 0.1534, 0.0673], device='cuda:1'), in_proj_covar=tensor([0.0120, 0.0149, 0.0176, 0.0080, 0.0161, 0.0182, 0.0188, 0.0162], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-04-28 07:52:34,503 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.94 vs. limit=2.0 2023-04-28 07:52:54,176 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=45835.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 07:53:16,335 INFO [train.py:904] (1/8) Epoch 5, batch 5250, loss[loss=0.2001, simple_loss=0.2882, pruned_loss=0.05604, over 16474.00 frames. ], tot_loss[loss=0.2201, simple_loss=0.3015, pruned_loss=0.06938, over 3200700.96 frames. ], batch size: 75, lr: 1.34e-02, grad_scale: 8.0 2023-04-28 07:53:26,728 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=45859.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 07:53:31,051 INFO [optim.py:368] (1/8) Clipping_scale=2.0, grad-norm quartiles 1.833e+02 2.734e+02 3.218e+02 3.993e+02 9.159e+02, threshold=6.436e+02, percent-clipped=4.0 2023-04-28 07:53:46,428 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.0418, 3.6363, 3.5991, 2.2347, 3.2557, 3.5897, 3.5048, 1.8964], device='cuda:1'), covar=tensor([0.0320, 0.0019, 0.0030, 0.0240, 0.0043, 0.0049, 0.0032, 0.0320], device='cuda:1'), in_proj_covar=tensor([0.0117, 0.0055, 0.0060, 0.0115, 0.0060, 0.0068, 0.0064, 0.0109], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0001, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:1') 2023-04-28 07:54:24,704 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.0149, 1.3699, 1.7732, 1.9624, 2.2014, 2.4076, 1.6570, 2.2896], device='cuda:1'), covar=tensor([0.0067, 0.0208, 0.0108, 0.0138, 0.0092, 0.0052, 0.0165, 0.0042], device='cuda:1'), in_proj_covar=tensor([0.0118, 0.0142, 0.0126, 0.0123, 0.0125, 0.0089, 0.0139, 0.0081], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001], device='cuda:1') 2023-04-28 07:54:26,348 INFO [train.py:904] (1/8) Epoch 5, batch 5300, loss[loss=0.2502, simple_loss=0.316, pruned_loss=0.09223, over 12074.00 frames. ], tot_loss[loss=0.2181, simple_loss=0.2986, pruned_loss=0.06874, over 3189746.04 frames. ], batch size: 246, lr: 1.34e-02, grad_scale: 8.0 2023-04-28 07:54:52,335 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=4.63 vs. limit=5.0 2023-04-28 07:54:53,086 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=45920.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 07:54:57,495 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.5947, 3.6656, 2.9440, 2.1443, 2.6349, 2.2232, 3.8609, 3.8384], device='cuda:1'), covar=tensor([0.2177, 0.0679, 0.1244, 0.1725, 0.1943, 0.1399, 0.0421, 0.0607], device='cuda:1'), in_proj_covar=tensor([0.0278, 0.0251, 0.0266, 0.0243, 0.0292, 0.0203, 0.0243, 0.0253], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-28 07:55:36,415 INFO [train.py:904] (1/8) Epoch 5, batch 5350, loss[loss=0.2228, simple_loss=0.3089, pruned_loss=0.06832, over 16555.00 frames. ], tot_loss[loss=0.2155, simple_loss=0.2963, pruned_loss=0.0673, over 3203692.84 frames. ], batch size: 68, lr: 1.33e-02, grad_scale: 8.0 2023-04-28 07:55:51,091 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([1.8212, 3.1463, 3.0844, 2.0007, 2.8058, 3.0172, 2.9752, 1.6405], device='cuda:1'), covar=tensor([0.0352, 0.0022, 0.0031, 0.0261, 0.0050, 0.0044, 0.0036, 0.0355], device='cuda:1'), in_proj_covar=tensor([0.0116, 0.0054, 0.0060, 0.0115, 0.0060, 0.0068, 0.0063, 0.0108], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0001, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:1') 2023-04-28 07:55:53,055 INFO [optim.py:368] (1/8) 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 07:56:52,488 INFO [train.py:904] (1/8) Epoch 5, batch 5400, loss[loss=0.2145, simple_loss=0.3034, pruned_loss=0.06275, over 16844.00 frames. ], tot_loss[loss=0.2182, simple_loss=0.2997, pruned_loss=0.06834, over 3214323.54 frames. ], batch size: 90, lr: 1.33e-02, grad_scale: 8.0 2023-04-28 07:56:57,453 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=46004.0, num_to_drop=1, layers_to_drop={2} 2023-04-28 07:58:09,282 INFO [train.py:904] (1/8) Epoch 5, batch 5450, loss[loss=0.2494, simple_loss=0.3309, pruned_loss=0.08394, over 16675.00 frames. ], tot_loss[loss=0.2226, simple_loss=0.3039, pruned_loss=0.07064, over 3202025.16 frames. ], batch size: 89, lr: 1.33e-02, grad_scale: 8.0 2023-04-28 07:58:11,175 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=46052.0, num_to_drop=1, layers_to_drop={0} 2023-04-28 07:58:19,237 INFO [zipformer.py:625] (1/8) 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] (1/8) 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] (1/8) Epoch 5, batch 5500, loss[loss=0.3123, simple_loss=0.3552, pruned_loss=0.1347, over 11820.00 frames. ], tot_loss[loss=0.2344, simple_loss=0.3131, pruned_loss=0.07782, over 3170445.29 frames. ], batch size: 250, lr: 1.33e-02, grad_scale: 8.0 2023-04-28 08:00:07,587 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=46130.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 08:00:13,930 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.8712, 2.1128, 2.2685, 3.1623, 2.1018, 2.7451, 2.3433, 2.1304], device='cuda:1'), covar=tensor([0.0469, 0.1535, 0.0787, 0.0331, 0.1970, 0.0668, 0.1437, 0.1787], device='cuda:1'), in_proj_covar=tensor([0.0312, 0.0312, 0.0254, 0.0300, 0.0361, 0.0298, 0.0279, 0.0373], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-28 08:00:39,132 INFO [train.py:904] (1/8) Epoch 5, batch 5550, loss[loss=0.3525, simple_loss=0.4031, pruned_loss=0.151, over 15423.00 frames. ], tot_loss[loss=0.2461, simple_loss=0.3222, pruned_loss=0.08497, over 3148726.69 frames. ], batch size: 191, lr: 1.33e-02, grad_scale: 8.0 2023-04-28 08:00:56,875 INFO [optim.py:368] (1/8) Clipping_scale=2.0, grad-norm quartiles 2.500e+02 4.429e+02 5.470e+02 6.755e+02 1.488e+03, threshold=1.094e+03, percent-clipped=17.0 2023-04-28 08:01:58,727 INFO [train.py:904] (1/8) Epoch 5, batch 5600, loss[loss=0.2801, simple_loss=0.3447, pruned_loss=0.1078, over 15246.00 frames. ], tot_loss[loss=0.256, simple_loss=0.3291, pruned_loss=0.09144, over 3118864.10 frames. ], batch size: 190, lr: 1.33e-02, grad_scale: 16.0 2023-04-28 08:02:11,998 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.59 vs. limit=2.0 2023-04-28 08:02:19,042 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.5631, 3.3728, 2.7325, 1.7624, 2.4923, 2.0914, 3.0663, 3.2081], device='cuda:1'), covar=tensor([0.0322, 0.0440, 0.0575, 0.1617, 0.0796, 0.0963, 0.0610, 0.0663], device='cuda:1'), in_proj_covar=tensor([0.0139, 0.0126, 0.0154, 0.0141, 0.0134, 0.0126, 0.0140, 0.0134], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-04-28 08:02:24,003 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=46215.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 08:03:23,171 INFO [train.py:904] (1/8) Epoch 5, batch 5650, loss[loss=0.2276, simple_loss=0.3113, pruned_loss=0.07189, over 16959.00 frames. ], tot_loss[loss=0.2638, simple_loss=0.3347, pruned_loss=0.09642, over 3077177.29 frames. ], batch size: 96, lr: 1.33e-02, grad_scale: 8.0 2023-04-28 08:03:28,447 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.0052, 3.2555, 3.4832, 3.4552, 3.4552, 3.1494, 3.2177, 3.2912], device='cuda:1'), covar=tensor([0.0370, 0.0461, 0.0405, 0.0453, 0.0457, 0.0458, 0.0855, 0.0477], device='cuda:1'), in_proj_covar=tensor([0.0230, 0.0219, 0.0225, 0.0228, 0.0275, 0.0243, 0.0341, 0.0205], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0002], device='cuda:1') 2023-04-28 08:03:42,413 INFO [optim.py:368] (1/8) 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 08:03:49,240 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.7990, 4.0844, 3.8257, 3.8781, 3.5486, 3.7109, 3.8082, 4.0727], device='cuda:1'), covar=tensor([0.0697, 0.0771, 0.0936, 0.0530, 0.0578, 0.1197, 0.0649, 0.0762], device='cuda:1'), in_proj_covar=tensor([0.0350, 0.0467, 0.0406, 0.0304, 0.0289, 0.0308, 0.0376, 0.0334], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-28 08:04:04,191 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=46276.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 08:04:43,823 INFO [train.py:904] (1/8) Epoch 5, batch 5700, loss[loss=0.3621, simple_loss=0.3857, pruned_loss=0.1693, over 11452.00 frames. ], tot_loss[loss=0.2672, simple_loss=0.3368, pruned_loss=0.09875, over 3047489.53 frames. ], batch size: 247, lr: 1.33e-02, grad_scale: 8.0 2023-04-28 08:05:41,861 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=46337.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 08:06:04,707 INFO [train.py:904] (1/8) Epoch 5, batch 5750, loss[loss=0.246, simple_loss=0.3277, pruned_loss=0.08216, over 16517.00 frames. ], tot_loss[loss=0.2705, simple_loss=0.3397, pruned_loss=0.1007, over 3010142.81 frames. ], batch size: 75, lr: 1.33e-02, grad_scale: 8.0 2023-04-28 08:06:16,562 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=46358.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 08:06:23,499 INFO [optim.py:368] (1/8) Clipping_scale=2.0, grad-norm quartiles 2.622e+02 4.192e+02 4.948e+02 6.377e+02 1.174e+03, threshold=9.897e+02, percent-clipped=0.0 2023-04-28 08:06:36,713 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([1.8568, 1.4208, 1.6222, 1.7514, 1.8664, 1.7801, 1.5405, 1.7284], device='cuda:1'), covar=tensor([0.0088, 0.0139, 0.0086, 0.0094, 0.0082, 0.0056, 0.0146, 0.0040], device='cuda:1'), in_proj_covar=tensor([0.0115, 0.0139, 0.0125, 0.0122, 0.0125, 0.0088, 0.0138, 0.0077], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001], device='cuda:1') 2023-04-28 08:06:38,238 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.4386, 2.0319, 2.1053, 3.9605, 1.7385, 2.9771, 2.1905, 2.1345], device='cuda:1'), covar=tensor([0.0601, 0.2212, 0.1160, 0.0303, 0.3093, 0.0975, 0.2028, 0.2361], device='cuda:1'), in_proj_covar=tensor([0.0310, 0.0313, 0.0252, 0.0301, 0.0363, 0.0299, 0.0279, 0.0373], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-28 08:06:50,119 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.7477, 5.0636, 4.7799, 4.8036, 4.4638, 4.4404, 4.6012, 5.1305], device='cuda:1'), covar=tensor([0.0660, 0.0649, 0.0930, 0.0444, 0.0612, 0.0723, 0.0599, 0.0655], device='cuda:1'), in_proj_covar=tensor([0.0355, 0.0467, 0.0410, 0.0308, 0.0294, 0.0311, 0.0379, 0.0338], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-28 08:07:20,469 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=4.99 vs. limit=5.0 2023-04-28 08:07:26,289 INFO [train.py:904] (1/8) Epoch 5, batch 5800, loss[loss=0.2665, simple_loss=0.3384, pruned_loss=0.09723, over 16309.00 frames. ], tot_loss[loss=0.2697, simple_loss=0.3399, pruned_loss=0.09974, over 3028387.83 frames. ], batch size: 165, lr: 1.33e-02, grad_scale: 8.0 2023-04-28 08:07:35,756 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=46406.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 08:08:13,782 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=46430.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 08:08:46,748 INFO [train.py:904] (1/8) Epoch 5, batch 5850, loss[loss=0.24, simple_loss=0.3221, pruned_loss=0.07892, over 16396.00 frames. ], tot_loss[loss=0.2654, simple_loss=0.3372, pruned_loss=0.09682, over 3048789.90 frames. ], batch size: 68, lr: 1.33e-02, grad_scale: 4.0 2023-04-28 08:09:06,683 INFO [optim.py:368] (1/8) 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 08:09:28,719 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 5, batch 5900, loss[loss=0.2644, simple_loss=0.3426, pruned_loss=0.0931, over 17244.00 frames. ], tot_loss[loss=0.2641, simple_loss=0.3363, pruned_loss=0.09592, over 3059564.61 frames. ], batch size: 44, lr: 1.33e-02, grad_scale: 4.0 2023-04-28 08:10:16,254 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.2363, 2.0000, 2.0897, 3.8371, 1.8056, 2.8655, 2.1936, 2.0924], device='cuda:1'), covar=tensor([0.0653, 0.2051, 0.1122, 0.0303, 0.2973, 0.0946, 0.1849, 0.2146], device='cuda:1'), in_proj_covar=tensor([0.0309, 0.0312, 0.0254, 0.0299, 0.0365, 0.0299, 0.0278, 0.0370], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-28 08:10:36,171 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=46515.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 08:11:30,097 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.5958, 3.6747, 2.8686, 2.1345, 2.6066, 1.9636, 3.6429, 3.8579], device='cuda:1'), covar=tensor([0.2051, 0.0568, 0.1280, 0.1633, 0.1859, 0.1605, 0.0440, 0.0483], device='cuda:1'), in_proj_covar=tensor([0.0276, 0.0244, 0.0264, 0.0240, 0.0288, 0.0199, 0.0239, 0.0246], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-28 08:11:31,433 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.70 vs. limit=2.0 2023-04-28 08:11:31,953 INFO [train.py:904] (1/8) Epoch 5, batch 5950, loss[loss=0.25, simple_loss=0.3452, pruned_loss=0.07745, over 17241.00 frames. ], tot_loss[loss=0.2617, simple_loss=0.3359, pruned_loss=0.0937, over 3063411.43 frames. ], batch size: 52, lr: 1.33e-02, grad_scale: 4.0 2023-04-28 08:11:52,724 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=46563.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 08:11:53,495 INFO [optim.py:368] (1/8) 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 08:12:38,739 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.2099, 4.4191, 4.5289, 3.4016, 3.9573, 4.3589, 4.1138, 2.2915], device='cuda:1'), covar=tensor([0.0320, 0.0015, 0.0020, 0.0165, 0.0038, 0.0048, 0.0028, 0.0268], device='cuda:1'), in_proj_covar=tensor([0.0115, 0.0054, 0.0058, 0.0115, 0.0059, 0.0069, 0.0063, 0.0108], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0001, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:1') 2023-04-28 08:12:43,209 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 5, batch 6000, loss[loss=0.2337, simple_loss=0.3129, pruned_loss=0.07726, over 16909.00 frames. ], tot_loss[loss=0.2605, simple_loss=0.335, pruned_loss=0.09302, over 3076991.21 frames. ], batch size: 116, lr: 1.33e-02, grad_scale: 8.0 2023-04-28 08:12:51,974 INFO [train.py:929] (1/8) Computing validation loss 2023-04-28 08:13:04,051 INFO [train.py:938] (1/8) Epoch 5, validation: loss=0.1879, simple_loss=0.2992, pruned_loss=0.03826, over 944034.00 frames. 2023-04-28 08:13:04,052 INFO [train.py:939] (1/8) Maximum memory allocated so far is 17676MB 2023-04-28 08:13:05,772 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=46602.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 08:13:51,748 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=46632.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 08:14:27,391 INFO [train.py:904] (1/8) Epoch 5, batch 6050, loss[loss=0.2711, simple_loss=0.3308, pruned_loss=0.1057, over 11772.00 frames. ], tot_loss[loss=0.257, simple_loss=0.3321, pruned_loss=0.09095, over 3113204.78 frames. ], batch size: 248, lr: 1.32e-02, grad_scale: 8.0 2023-04-28 08:14:36,891 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=46656.0, num_to_drop=1, layers_to_drop={2} 2023-04-28 08:14:48,250 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=46663.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 08:14:48,996 INFO [optim.py:368] (1/8) 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 08:15:15,927 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.4350, 4.3837, 4.2813, 4.2373, 3.8267, 4.3868, 4.2219, 4.0324], device='cuda:1'), covar=tensor([0.0489, 0.0305, 0.0213, 0.0178, 0.0824, 0.0312, 0.0342, 0.0469], device='cuda:1'), in_proj_covar=tensor([0.0181, 0.0185, 0.0208, 0.0179, 0.0237, 0.0209, 0.0150, 0.0227], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0002], device='cuda:1') 2023-04-28 08:15:17,289 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.8951, 1.6031, 2.1807, 2.8131, 2.6043, 3.1077, 1.6707, 2.9735], device='cuda:1'), covar=tensor([0.0072, 0.0277, 0.0169, 0.0123, 0.0118, 0.0069, 0.0272, 0.0052], device='cuda:1'), in_proj_covar=tensor([0.0115, 0.0139, 0.0124, 0.0121, 0.0123, 0.0087, 0.0136, 0.0077], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001], device='cuda:1') 2023-04-28 08:15:23,798 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.25 vs. limit=2.0 2023-04-28 08:15:45,459 INFO [train.py:904] (1/8) Epoch 5, batch 6100, loss[loss=0.2402, simple_loss=0.3221, pruned_loss=0.07912, over 16319.00 frames. ], tot_loss[loss=0.2558, simple_loss=0.3318, pruned_loss=0.08997, over 3114693.11 frames. ], batch size: 165, lr: 1.32e-02, grad_scale: 8.0 2023-04-28 08:16:09,047 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.8359, 1.5166, 2.1270, 2.7665, 2.6080, 2.9196, 1.6752, 2.8453], device='cuda:1'), covar=tensor([0.0060, 0.0244, 0.0146, 0.0102, 0.0095, 0.0065, 0.0215, 0.0037], device='cuda:1'), in_proj_covar=tensor([0.0114, 0.0138, 0.0123, 0.0119, 0.0122, 0.0086, 0.0135, 0.0076], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001], device='cuda:1') 2023-04-28 08:17:01,187 INFO [train.py:904] (1/8) Epoch 5, batch 6150, loss[loss=0.2579, simple_loss=0.3303, pruned_loss=0.09273, over 16264.00 frames. ], tot_loss[loss=0.2545, simple_loss=0.33, pruned_loss=0.08947, over 3118387.35 frames. ], batch size: 165, lr: 1.32e-02, grad_scale: 8.0 2023-04-28 08:17:22,716 INFO [optim.py:368] (1/8) Clipping_scale=2.0, grad-norm quartiles 2.183e+02 3.664e+02 4.778e+02 7.050e+02 1.596e+03, threshold=9.557e+02, percent-clipped=5.0 2023-04-28 08:17:48,079 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.0346, 2.4082, 2.3738, 3.0582, 2.6641, 3.2612, 1.8143, 2.7176], device='cuda:1'), covar=tensor([0.1071, 0.0451, 0.0999, 0.0098, 0.0225, 0.0379, 0.1197, 0.0679], device='cuda:1'), in_proj_covar=tensor([0.0144, 0.0142, 0.0167, 0.0083, 0.0178, 0.0177, 0.0162, 0.0164], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0003, 0.0004], device='cuda:1') 2023-04-28 08:18:08,960 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.46 vs. limit=2.0 2023-04-28 08:18:21,108 INFO [train.py:904] (1/8) Epoch 5, batch 6200, loss[loss=0.2852, simple_loss=0.336, pruned_loss=0.1172, over 11487.00 frames. ], tot_loss[loss=0.2529, simple_loss=0.3276, pruned_loss=0.08907, over 3100342.90 frames. ], batch size: 246, lr: 1.32e-02, grad_scale: 8.0 2023-04-28 08:18:35,840 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=46810.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 08:19:10,821 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.6068, 4.2651, 4.5286, 4.7791, 4.9298, 4.3717, 4.9359, 4.8634], device='cuda:1'), covar=tensor([0.0890, 0.1007, 0.1409, 0.0553, 0.0431, 0.0626, 0.0452, 0.0464], device='cuda:1'), in_proj_covar=tensor([0.0373, 0.0463, 0.0577, 0.0471, 0.0352, 0.0352, 0.0374, 0.0394], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-28 08:19:37,501 INFO [train.py:904] (1/8) Epoch 5, batch 6250, loss[loss=0.232, simple_loss=0.3199, pruned_loss=0.0721, over 16595.00 frames. ], tot_loss[loss=0.2531, simple_loss=0.3276, pruned_loss=0.08932, over 3078889.89 frames. ], batch size: 68, lr: 1.32e-02, grad_scale: 8.0 2023-04-28 08:19:57,339 INFO [optim.py:368] (1/8) 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 08:20:09,465 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=46871.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 08:20:21,850 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=4.70 vs. limit=5.0 2023-04-28 08:20:55,067 INFO [train.py:904] (1/8) Epoch 5, batch 6300, loss[loss=0.2893, simple_loss=0.3656, pruned_loss=0.1065, over 16245.00 frames. ], tot_loss[loss=0.2508, simple_loss=0.3264, pruned_loss=0.08763, over 3102805.53 frames. ], batch size: 165, lr: 1.32e-02, grad_scale: 8.0 2023-04-28 08:21:00,961 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.4327, 4.5413, 4.6037, 4.6443, 4.6693, 5.2211, 4.8680, 4.5830], device='cuda:1'), covar=tensor([0.1118, 0.1773, 0.1698, 0.1893, 0.2566, 0.1027, 0.1273, 0.2455], device='cuda:1'), in_proj_covar=tensor([0.0276, 0.0378, 0.0381, 0.0332, 0.0445, 0.0397, 0.0313, 0.0461], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-04-28 08:21:43,828 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=46932.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 08:22:12,093 INFO [train.py:904] (1/8) Epoch 5, batch 6350, loss[loss=0.2209, simple_loss=0.2973, pruned_loss=0.07223, over 16613.00 frames. ], tot_loss[loss=0.2544, simple_loss=0.3286, pruned_loss=0.09007, over 3087122.80 frames. ], batch size: 57, lr: 1.32e-02, grad_scale: 8.0 2023-04-28 08:22:13,433 INFO [zipformer.py:625] (1/8) 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] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=46958.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 08:22:24,215 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([1.8158, 3.5330, 3.4695, 2.2866, 3.1451, 3.4675, 3.4309, 1.9710], device='cuda:1'), covar=tensor([0.0345, 0.0021, 0.0028, 0.0220, 0.0042, 0.0056, 0.0029, 0.0267], device='cuda:1'), in_proj_covar=tensor([0.0114, 0.0052, 0.0057, 0.0113, 0.0059, 0.0068, 0.0062, 0.0106], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0001, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:1') 2023-04-28 08:22:31,677 INFO [optim.py:368] (1/8) 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 08:22:56,443 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=46980.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 08:23:28,011 INFO [train.py:904] (1/8) Epoch 5, batch 6400, loss[loss=0.2315, simple_loss=0.3041, pruned_loss=0.07942, over 16826.00 frames. ], tot_loss[loss=0.2555, simple_loss=0.3288, pruned_loss=0.09114, over 3081381.50 frames. ], batch size: 83, lr: 1.32e-02, grad_scale: 8.0 2023-04-28 08:23:28,553 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([1.6458, 2.9408, 2.4582, 4.6882, 3.7629, 4.1994, 1.7972, 2.9597], device='cuda:1'), covar=tensor([0.1585, 0.0654, 0.1307, 0.0111, 0.0355, 0.0325, 0.1498, 0.0824], device='cuda:1'), in_proj_covar=tensor([0.0146, 0.0144, 0.0168, 0.0085, 0.0179, 0.0179, 0.0163, 0.0166], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-04-28 08:24:42,540 INFO [train.py:904] (1/8) Epoch 5, batch 6450, loss[loss=0.2541, simple_loss=0.3131, pruned_loss=0.0976, over 11741.00 frames. ], tot_loss[loss=0.2532, simple_loss=0.3271, pruned_loss=0.08966, over 3066361.16 frames. ], batch size: 246, lr: 1.32e-02, grad_scale: 8.0 2023-04-28 08:24:49,092 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.45 vs. limit=2.0 2023-04-28 08:24:58,496 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.4214, 4.6438, 4.7772, 4.8313, 4.8226, 5.3340, 4.8681, 4.6506], device='cuda:1'), covar=tensor([0.1020, 0.1504, 0.1458, 0.1394, 0.2017, 0.0842, 0.1212, 0.2034], device='cuda:1'), in_proj_covar=tensor([0.0275, 0.0374, 0.0380, 0.0330, 0.0437, 0.0399, 0.0310, 0.0456], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-04-28 08:25:02,353 INFO [optim.py:368] (1/8) 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 08:25:03,987 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.9714, 4.0424, 4.4578, 4.3984, 4.4404, 4.0785, 4.0714, 3.9637], device='cuda:1'), covar=tensor([0.0267, 0.0380, 0.0336, 0.0385, 0.0323, 0.0273, 0.0780, 0.0361], device='cuda:1'), in_proj_covar=tensor([0.0236, 0.0226, 0.0234, 0.0234, 0.0277, 0.0245, 0.0356, 0.0208], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0002], device='cuda:1') 2023-04-28 08:25:14,756 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.52 vs. limit=2.0 2023-04-28 08:25:59,692 INFO [train.py:904] (1/8) Epoch 5, batch 6500, loss[loss=0.2011, simple_loss=0.2877, pruned_loss=0.05726, over 17038.00 frames. ], tot_loss[loss=0.2504, simple_loss=0.3244, pruned_loss=0.08815, over 3087245.27 frames. ], batch size: 50, lr: 1.32e-02, grad_scale: 8.0 2023-04-28 08:26:20,872 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=4.74 vs. limit=5.0 2023-04-28 08:27:08,591 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.0296, 3.1602, 1.5109, 3.2594, 2.2644, 3.2652, 1.7472, 2.4577], device='cuda:1'), covar=tensor([0.0154, 0.0267, 0.1553, 0.0076, 0.0724, 0.0459, 0.1385, 0.0660], device='cuda:1'), in_proj_covar=tensor([0.0120, 0.0148, 0.0180, 0.0081, 0.0162, 0.0185, 0.0188, 0.0166], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-04-28 08:27:10,505 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.7421, 5.3146, 5.3660, 5.3814, 5.4106, 5.9195, 5.4468, 5.2107], device='cuda:1'), covar=tensor([0.0790, 0.1303, 0.1557, 0.1600, 0.2317, 0.0811, 0.1114, 0.2332], device='cuda:1'), in_proj_covar=tensor([0.0274, 0.0371, 0.0379, 0.0327, 0.0433, 0.0396, 0.0307, 0.0451], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-04-28 08:27:18,006 INFO [train.py:904] (1/8) Epoch 5, batch 6550, loss[loss=0.266, simple_loss=0.3518, pruned_loss=0.09012, over 16659.00 frames. ], tot_loss[loss=0.2537, simple_loss=0.3282, pruned_loss=0.08963, over 3079896.27 frames. ], batch size: 134, lr: 1.32e-02, grad_scale: 8.0 2023-04-28 08:27:37,105 INFO [optim.py:368] (1/8) 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] (1/8) 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] (1/8) Epoch 5, batch 6600, loss[loss=0.3024, simple_loss=0.3697, pruned_loss=0.1176, over 16342.00 frames. ], tot_loss[loss=0.2563, simple_loss=0.3313, pruned_loss=0.09064, over 3084263.19 frames. ], batch size: 146, lr: 1.32e-02, grad_scale: 8.0 2023-04-28 08:29:21,770 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.89 vs. limit=2.0 2023-04-28 08:29:35,118 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=47241.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 08:29:51,328 INFO [train.py:904] (1/8) Epoch 5, batch 6650, loss[loss=0.2447, simple_loss=0.318, pruned_loss=0.08573, over 16825.00 frames. ], tot_loss[loss=0.2595, simple_loss=0.333, pruned_loss=0.09299, over 3066706.08 frames. ], batch size: 116, lr: 1.32e-02, grad_scale: 8.0 2023-04-28 08:29:51,880 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=47251.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 08:30:02,696 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=47258.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 08:30:11,433 INFO [optim.py:368] (1/8) 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 08:31:04,889 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=47299.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 08:31:06,905 INFO [train.py:904] (1/8) Epoch 5, batch 6700, loss[loss=0.256, simple_loss=0.3305, pruned_loss=0.09069, over 16939.00 frames. ], tot_loss[loss=0.2586, simple_loss=0.3315, pruned_loss=0.09282, over 3067822.32 frames. ], batch size: 109, lr: 1.32e-02, grad_scale: 8.0 2023-04-28 08:31:09,288 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=47302.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 08:31:16,568 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=47306.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 08:32:25,988 INFO [train.py:904] (1/8) Epoch 5, batch 6750, loss[loss=0.2476, simple_loss=0.3237, pruned_loss=0.08573, over 16046.00 frames. ], tot_loss[loss=0.258, simple_loss=0.3307, pruned_loss=0.09262, over 3070810.10 frames. ], batch size: 35, lr: 1.31e-02, grad_scale: 8.0 2023-04-28 08:32:45,840 INFO [optim.py:368] (1/8) 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 08:33:08,524 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=47379.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 08:33:40,654 INFO [train.py:904] (1/8) Epoch 5, batch 6800, loss[loss=0.2671, simple_loss=0.3452, pruned_loss=0.09443, over 16233.00 frames. ], tot_loss[loss=0.2554, simple_loss=0.3293, pruned_loss=0.09078, over 3098426.96 frames. ], batch size: 165, lr: 1.31e-02, grad_scale: 8.0 2023-04-28 08:33:53,147 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([5.3584, 5.6513, 5.2907, 5.4150, 4.9001, 4.7285, 5.2044, 5.7530], device='cuda:1'), covar=tensor([0.0633, 0.0693, 0.0958, 0.0480, 0.0614, 0.0671, 0.0573, 0.0620], device='cuda:1'), in_proj_covar=tensor([0.0369, 0.0476, 0.0415, 0.0311, 0.0301, 0.0323, 0.0391, 0.0344], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-28 08:33:54,645 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.7349, 2.3180, 2.2295, 4.3946, 2.0161, 3.1317, 2.3766, 2.4489], device='cuda:1'), covar=tensor([0.0581, 0.2147, 0.1227, 0.0299, 0.3091, 0.1016, 0.1833, 0.2355], device='cuda:1'), in_proj_covar=tensor([0.0318, 0.0321, 0.0264, 0.0311, 0.0377, 0.0310, 0.0288, 0.0385], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-28 08:34:40,935 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=47440.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 08:34:57,635 INFO [train.py:904] (1/8) Epoch 5, batch 6850, loss[loss=0.2571, simple_loss=0.3452, pruned_loss=0.08452, over 16461.00 frames. ], tot_loss[loss=0.2546, simple_loss=0.3294, pruned_loss=0.08986, over 3117819.89 frames. ], batch size: 146, lr: 1.31e-02, grad_scale: 8.0 2023-04-28 08:35:17,575 INFO [optim.py:368] (1/8) 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 08:35:21,014 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=47466.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 08:35:45,599 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.4283, 4.2647, 4.4317, 3.1361, 4.2005, 4.4354, 4.3585, 2.3037], device='cuda:1'), covar=tensor([0.0319, 0.0017, 0.0018, 0.0180, 0.0021, 0.0035, 0.0018, 0.0300], device='cuda:1'), in_proj_covar=tensor([0.0115, 0.0053, 0.0058, 0.0113, 0.0059, 0.0069, 0.0062, 0.0107], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0001, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:1') 2023-04-28 08:36:12,140 INFO [train.py:904] (1/8) Epoch 5, batch 6900, loss[loss=0.3417, simple_loss=0.3766, pruned_loss=0.1534, over 11597.00 frames. ], tot_loss[loss=0.2553, simple_loss=0.3319, pruned_loss=0.0894, over 3126420.21 frames. ], batch size: 248, lr: 1.31e-02, grad_scale: 8.0 2023-04-28 08:36:33,878 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=47514.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 08:36:44,448 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.4043, 2.9238, 2.6491, 2.3966, 2.2733, 2.1748, 2.8845, 3.0414], device='cuda:1'), covar=tensor([0.1507, 0.0535, 0.0886, 0.1073, 0.1565, 0.1189, 0.0378, 0.0535], device='cuda:1'), in_proj_covar=tensor([0.0273, 0.0242, 0.0262, 0.0238, 0.0284, 0.0195, 0.0235, 0.0243], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-28 08:37:08,389 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.9819, 4.9605, 4.7432, 3.8804, 4.8572, 1.5739, 4.5613, 4.7122], device='cuda:1'), covar=tensor([0.0063, 0.0047, 0.0085, 0.0381, 0.0053, 0.1903, 0.0086, 0.0134], device='cuda:1'), in_proj_covar=tensor([0.0089, 0.0076, 0.0118, 0.0122, 0.0087, 0.0138, 0.0103, 0.0114], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-28 08:37:30,502 INFO [train.py:904] (1/8) Epoch 5, batch 6950, loss[loss=0.3357, simple_loss=0.3733, pruned_loss=0.1491, over 11185.00 frames. ], tot_loss[loss=0.2601, simple_loss=0.3349, pruned_loss=0.09269, over 3104895.98 frames. ], batch size: 248, lr: 1.31e-02, grad_scale: 8.0 2023-04-28 08:37:50,871 INFO [optim.py:368] (1/8) 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 08:38:05,135 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.32 vs. limit=2.0 2023-04-28 08:38:16,022 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.3992, 1.3859, 1.7811, 2.2360, 2.2948, 2.4194, 1.5517, 2.3561], device='cuda:1'), covar=tensor([0.0072, 0.0243, 0.0142, 0.0118, 0.0106, 0.0087, 0.0229, 0.0053], device='cuda:1'), in_proj_covar=tensor([0.0117, 0.0143, 0.0128, 0.0122, 0.0127, 0.0092, 0.0139, 0.0080], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001], device='cuda:1') 2023-04-28 08:38:20,382 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.3954, 3.2972, 2.6663, 2.0360, 2.3504, 2.0446, 3.3626, 3.3685], device='cuda:1'), covar=tensor([0.2160, 0.0641, 0.1282, 0.1745, 0.1907, 0.1496, 0.0464, 0.0693], device='cuda:1'), in_proj_covar=tensor([0.0277, 0.0247, 0.0267, 0.0241, 0.0287, 0.0199, 0.0239, 0.0245], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-28 08:38:42,823 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 5, batch 7000, loss[loss=0.2381, simple_loss=0.3227, pruned_loss=0.07672, over 17054.00 frames. ], tot_loss[loss=0.2589, simple_loss=0.3346, pruned_loss=0.09162, over 3105130.05 frames. ], batch size: 53, lr: 1.31e-02, grad_scale: 8.0 2023-04-28 08:38:52,132 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.9800, 1.4772, 2.0396, 2.6797, 2.5642, 2.9651, 1.8281, 2.9395], device='cuda:1'), covar=tensor([0.0059, 0.0271, 0.0178, 0.0113, 0.0123, 0.0079, 0.0233, 0.0051], device='cuda:1'), in_proj_covar=tensor([0.0116, 0.0142, 0.0127, 0.0121, 0.0125, 0.0091, 0.0138, 0.0079], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001], device='cuda:1') 2023-04-28 08:39:01,236 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.2940, 4.2020, 4.7630, 4.7340, 4.6874, 4.3013, 4.3539, 4.1331], device='cuda:1'), covar=tensor([0.0213, 0.0374, 0.0332, 0.0323, 0.0367, 0.0272, 0.0739, 0.0396], device='cuda:1'), in_proj_covar=tensor([0.0232, 0.0224, 0.0231, 0.0232, 0.0279, 0.0244, 0.0346, 0.0205], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0002], device='cuda:1') 2023-04-28 08:39:29,382 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.4392, 1.9562, 2.1322, 4.0112, 1.8731, 2.9166, 2.2123, 2.1581], device='cuda:1'), covar=tensor([0.0644, 0.2279, 0.1206, 0.0293, 0.3115, 0.1045, 0.1937, 0.2344], device='cuda:1'), in_proj_covar=tensor([0.0317, 0.0316, 0.0260, 0.0306, 0.0373, 0.0309, 0.0286, 0.0379], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-28 08:39:35,824 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.3289, 4.3158, 4.9304, 4.8439, 4.8069, 4.3353, 4.4477, 4.1729], device='cuda:1'), covar=tensor([0.0249, 0.0380, 0.0286, 0.0313, 0.0411, 0.0324, 0.0826, 0.0424], device='cuda:1'), in_proj_covar=tensor([0.0234, 0.0226, 0.0232, 0.0233, 0.0280, 0.0247, 0.0350, 0.0206], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0002], device='cuda:1') 2023-04-28 08:40:06,954 INFO [train.py:904] (1/8) Epoch 5, batch 7050, loss[loss=0.2254, simple_loss=0.3083, pruned_loss=0.07123, over 16542.00 frames. ], tot_loss[loss=0.258, simple_loss=0.3344, pruned_loss=0.09075, over 3117094.18 frames. ], batch size: 35, lr: 1.31e-02, grad_scale: 8.0 2023-04-28 08:40:26,888 INFO [optim.py:368] (1/8) 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 08:40:29,555 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.60 vs. limit=2.0 2023-04-28 08:41:26,209 INFO [train.py:904] (1/8) Epoch 5, batch 7100, loss[loss=0.2951, simple_loss=0.3534, pruned_loss=0.1183, over 15591.00 frames. ], tot_loss[loss=0.2579, simple_loss=0.3335, pruned_loss=0.09117, over 3116151.54 frames. ], batch size: 191, lr: 1.31e-02, grad_scale: 8.0 2023-04-28 08:41:28,286 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.33 vs. limit=2.0 2023-04-28 08:42:09,581 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.46 vs. limit=2.0 2023-04-28 08:42:19,694 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=47735.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 08:42:24,135 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=47738.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 08:42:44,679 INFO [train.py:904] (1/8) Epoch 5, batch 7150, loss[loss=0.2578, simple_loss=0.3283, pruned_loss=0.09362, over 15302.00 frames. ], tot_loss[loss=0.256, simple_loss=0.3311, pruned_loss=0.09044, over 3117580.79 frames. ], batch size: 190, lr: 1.31e-02, grad_scale: 8.0 2023-04-28 08:42:57,072 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([1.9712, 3.5089, 3.4516, 2.2453, 3.1941, 3.4126, 3.3869, 1.8067], device='cuda:1'), covar=tensor([0.0298, 0.0020, 0.0027, 0.0232, 0.0042, 0.0051, 0.0028, 0.0300], device='cuda:1'), in_proj_covar=tensor([0.0116, 0.0053, 0.0059, 0.0116, 0.0060, 0.0069, 0.0062, 0.0108], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0001, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:1') 2023-04-28 08:43:00,431 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.0444, 2.9457, 2.6737, 2.0458, 2.5563, 1.9715, 2.7799, 2.8616], device='cuda:1'), covar=tensor([0.0242, 0.0429, 0.0413, 0.1240, 0.0592, 0.0859, 0.0460, 0.0520], device='cuda:1'), in_proj_covar=tensor([0.0136, 0.0124, 0.0152, 0.0139, 0.0131, 0.0123, 0.0138, 0.0133], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:1') 2023-04-28 08:43:03,967 INFO [optim.py:368] (1/8) 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 08:43:11,285 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([1.8943, 3.3536, 3.3168, 2.1529, 3.0275, 3.2769, 3.2342, 1.8309], device='cuda:1'), covar=tensor([0.0316, 0.0021, 0.0031, 0.0233, 0.0053, 0.0053, 0.0034, 0.0317], device='cuda:1'), in_proj_covar=tensor([0.0117, 0.0053, 0.0059, 0.0116, 0.0060, 0.0070, 0.0063, 0.0109], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0001, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:1') 2023-04-28 08:43:40,021 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=47788.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 08:43:57,119 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=47799.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 08:44:00,011 INFO [train.py:904] (1/8) Epoch 5, batch 7200, loss[loss=0.2545, simple_loss=0.331, pruned_loss=0.08893, over 16378.00 frames. ], tot_loss[loss=0.2527, simple_loss=0.3285, pruned_loss=0.08842, over 3103856.57 frames. ], batch size: 35, lr: 1.31e-02, grad_scale: 8.0 2023-04-28 08:44:13,986 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=47809.0, num_to_drop=1, layers_to_drop={0} 2023-04-28 08:44:29,890 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.9997, 4.4726, 2.3046, 5.0284, 2.7866, 4.8382, 2.5079, 3.1606], device='cuda:1'), covar=tensor([0.0127, 0.0165, 0.1586, 0.0016, 0.0792, 0.0215, 0.1237, 0.0616], device='cuda:1'), in_proj_covar=tensor([0.0118, 0.0147, 0.0177, 0.0079, 0.0159, 0.0179, 0.0187, 0.0164], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-04-28 08:45:22,038 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=47849.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 08:45:24,154 INFO [train.py:904] (1/8) Epoch 5, batch 7250, loss[loss=0.2156, simple_loss=0.2895, pruned_loss=0.07083, over 16480.00 frames. ], tot_loss[loss=0.2493, simple_loss=0.3253, pruned_loss=0.08659, over 3107641.75 frames. ], batch size: 75, lr: 1.31e-02, grad_scale: 8.0 2023-04-28 08:45:43,493 INFO [optim.py:368] (1/8) Clipping_scale=2.0, grad-norm quartiles 2.043e+02 3.368e+02 4.070e+02 5.805e+02 1.027e+03, threshold=8.141e+02, percent-clipped=1.0 2023-04-28 08:45:53,842 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=47870.0, num_to_drop=1, layers_to_drop={0} 2023-04-28 08:45:59,343 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.41 vs. limit=2.0 2023-04-28 08:46:00,302 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([1.7056, 2.6943, 2.2560, 3.6234, 2.8380, 3.6046, 1.4984, 2.7972], device='cuda:1'), covar=tensor([0.1266, 0.0528, 0.1129, 0.0111, 0.0250, 0.0372, 0.1368, 0.0737], device='cuda:1'), in_proj_covar=tensor([0.0144, 0.0143, 0.0168, 0.0086, 0.0181, 0.0178, 0.0160, 0.0164], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0003, 0.0004], device='cuda:1') 2023-04-28 08:46:34,226 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=47897.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 08:46:40,680 INFO [train.py:904] (1/8) Epoch 5, batch 7300, loss[loss=0.2114, simple_loss=0.3082, pruned_loss=0.05734, over 16880.00 frames. ], tot_loss[loss=0.2482, simple_loss=0.3243, pruned_loss=0.08607, over 3107477.59 frames. ], batch size: 96, lr: 1.31e-02, grad_scale: 8.0 2023-04-28 08:47:36,467 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.6828, 3.6526, 3.7649, 3.7209, 3.7977, 4.1162, 3.8510, 3.5824], device='cuda:1'), covar=tensor([0.1742, 0.1698, 0.1377, 0.1847, 0.2039, 0.1207, 0.1258, 0.2404], device='cuda:1'), in_proj_covar=tensor([0.0274, 0.0369, 0.0377, 0.0328, 0.0427, 0.0397, 0.0308, 0.0448], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-04-28 08:47:49,458 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 5, batch 7350, loss[loss=0.2732, simple_loss=0.3444, pruned_loss=0.101, over 15295.00 frames. ], tot_loss[loss=0.2479, simple_loss=0.3239, pruned_loss=0.08596, over 3101494.77 frames. ], batch size: 190, lr: 1.31e-02, grad_scale: 8.0 2023-04-28 08:48:17,667 INFO [optim.py:368] (1/8) 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 08:49:11,423 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.8483, 4.8808, 4.6344, 3.9730, 4.7107, 1.8088, 4.4598, 4.6041], device='cuda:1'), covar=tensor([0.0050, 0.0043, 0.0079, 0.0294, 0.0048, 0.1810, 0.0072, 0.0109], device='cuda:1'), in_proj_covar=tensor([0.0089, 0.0077, 0.0119, 0.0124, 0.0089, 0.0142, 0.0105, 0.0116], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-28 08:49:18,342 INFO [train.py:904] (1/8) Epoch 5, batch 7400, loss[loss=0.2139, simple_loss=0.3051, pruned_loss=0.06135, over 16854.00 frames. ], tot_loss[loss=0.2497, simple_loss=0.3253, pruned_loss=0.0871, over 3088449.60 frames. ], batch size: 90, lr: 1.31e-02, grad_scale: 8.0 2023-04-28 08:50:13,269 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=48035.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 08:50:38,165 INFO [train.py:904] (1/8) Epoch 5, batch 7450, loss[loss=0.2518, simple_loss=0.3357, pruned_loss=0.08393, over 16879.00 frames. ], tot_loss[loss=0.2532, simple_loss=0.3275, pruned_loss=0.08945, over 3071383.25 frames. ], batch size: 116, lr: 1.31e-02, grad_scale: 8.0 2023-04-28 08:51:00,810 INFO [optim.py:368] (1/8) 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 08:51:14,541 INFO [zipformer.py:625] (1/8) 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] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=48083.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 08:51:49,935 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=48094.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 08:51:56,545 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.42 vs. limit=2.0 2023-04-28 08:52:00,146 INFO [train.py:904] (1/8) Epoch 5, batch 7500, loss[loss=0.2316, simple_loss=0.3102, pruned_loss=0.07647, over 16758.00 frames. ], tot_loss[loss=0.2526, simple_loss=0.3278, pruned_loss=0.08867, over 3075705.12 frames. ], batch size: 124, lr: 1.30e-02, grad_scale: 8.0 2023-04-28 08:52:50,866 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=48133.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 08:53:08,103 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 5, batch 7550, loss[loss=0.2268, simple_loss=0.301, pruned_loss=0.0763, over 16596.00 frames. ], tot_loss[loss=0.254, simple_loss=0.3279, pruned_loss=0.09008, over 3050829.50 frames. ], batch size: 62, lr: 1.30e-02, grad_scale: 8.0 2023-04-28 08:53:38,522 INFO [optim.py:368] (1/8) 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 08:53:40,654 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=48165.0, num_to_drop=1, layers_to_drop={1} 2023-04-28 08:54:34,587 INFO [train.py:904] (1/8) Epoch 5, batch 7600, loss[loss=0.2156, simple_loss=0.3046, pruned_loss=0.06328, over 16804.00 frames. ], tot_loss[loss=0.2537, simple_loss=0.3271, pruned_loss=0.09017, over 3051478.44 frames. ], batch size: 102, lr: 1.30e-02, grad_scale: 8.0 2023-04-28 08:54:46,218 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=4.63 vs. limit=5.0 2023-04-28 08:54:51,942 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=48212.0, num_to_drop=1, layers_to_drop={0} 2023-04-28 08:55:07,771 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.9691, 3.0757, 1.5560, 3.2351, 2.2867, 3.2725, 1.8081, 2.5260], device='cuda:1'), covar=tensor([0.0142, 0.0322, 0.1448, 0.0057, 0.0707, 0.0408, 0.1330, 0.0557], device='cuda:1'), in_proj_covar=tensor([0.0119, 0.0151, 0.0178, 0.0080, 0.0161, 0.0180, 0.0187, 0.0164], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-04-28 08:55:50,855 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.90 vs. limit=2.0 2023-04-28 08:55:51,252 INFO [train.py:904] (1/8) Epoch 5, batch 7650, loss[loss=0.3026, simple_loss=0.3505, pruned_loss=0.1274, over 11513.00 frames. ], tot_loss[loss=0.2536, simple_loss=0.3272, pruned_loss=0.08999, over 3068056.79 frames. ], batch size: 248, lr: 1.30e-02, grad_scale: 8.0 2023-04-28 08:56:12,527 INFO [optim.py:368] (1/8) 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 08:56:25,997 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=48273.0, num_to_drop=1, layers_to_drop={1} 2023-04-28 08:57:08,646 INFO [train.py:904] (1/8) Epoch 5, batch 7700, loss[loss=0.2615, simple_loss=0.3318, pruned_loss=0.09558, over 15302.00 frames. ], tot_loss[loss=0.254, simple_loss=0.3276, pruned_loss=0.09023, over 3089518.20 frames. ], batch size: 190, lr: 1.30e-02, grad_scale: 4.0 2023-04-28 08:57:28,110 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([5.2026, 5.1019, 4.9616, 3.4768, 5.0917, 1.5961, 4.6607, 4.7678], device='cuda:1'), covar=tensor([0.0113, 0.0082, 0.0110, 0.0634, 0.0078, 0.2390, 0.0094, 0.0251], device='cuda:1'), in_proj_covar=tensor([0.0091, 0.0077, 0.0119, 0.0124, 0.0090, 0.0142, 0.0104, 0.0117], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-28 08:57:44,147 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 5, batch 7750, loss[loss=0.2402, simple_loss=0.3263, pruned_loss=0.07705, over 16473.00 frames. ], tot_loss[loss=0.2539, simple_loss=0.3278, pruned_loss=0.08995, over 3098774.50 frames. ], batch size: 68, lr: 1.30e-02, grad_scale: 4.0 2023-04-28 08:58:47,829 INFO [optim.py:368] (1/8) 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 08:59:20,047 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=48385.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 08:59:34,662 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=48394.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 08:59:39,516 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=48397.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 08:59:45,047 INFO [train.py:904] (1/8) Epoch 5, batch 7800, loss[loss=0.2226, simple_loss=0.3134, pruned_loss=0.06592, over 16844.00 frames. ], tot_loss[loss=0.2547, simple_loss=0.3285, pruned_loss=0.0904, over 3105836.68 frames. ], batch size: 96, lr: 1.30e-02, grad_scale: 4.0 2023-04-28 08:59:45,683 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.3882, 2.9606, 2.5209, 2.3343, 2.2234, 2.1323, 2.9139, 2.9714], device='cuda:1'), covar=tensor([0.1517, 0.0459, 0.0977, 0.1141, 0.1714, 0.1211, 0.0399, 0.0617], device='cuda:1'), in_proj_covar=tensor([0.0282, 0.0245, 0.0268, 0.0242, 0.0288, 0.0200, 0.0240, 0.0253], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-28 09:00:20,535 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=48423.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 09:00:28,136 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=48428.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 09:00:49,115 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=48442.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 09:00:52,196 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 5, batch 7850, loss[loss=0.2955, simple_loss=0.344, pruned_loss=0.1234, over 11838.00 frames. ], tot_loss[loss=0.2551, simple_loss=0.3294, pruned_loss=0.09042, over 3096115.46 frames. ], batch size: 247, lr: 1.30e-02, grad_scale: 4.0 2023-04-28 09:01:13,092 INFO [zipformer.py:625] (1/8) 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] (1/8) Clipping_scale=2.0, grad-norm quartiles 2.309e+02 3.753e+02 4.508e+02 5.540e+02 1.129e+03, threshold=9.017e+02, percent-clipped=6.0 2023-04-28 09:01:24,036 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=48465.0, num_to_drop=1, layers_to_drop={1} 2023-04-28 09:01:52,505 INFO [zipformer.py:625] (1/8) 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] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=48492.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 09:02:16,622 INFO [train.py:904] (1/8) Epoch 5, batch 7900, loss[loss=0.2195, simple_loss=0.3012, pruned_loss=0.06885, over 16752.00 frames. ], tot_loss[loss=0.2532, simple_loss=0.3278, pruned_loss=0.0893, over 3096390.80 frames. ], batch size: 39, lr: 1.30e-02, grad_scale: 4.0 2023-04-28 09:02:22,922 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.8260, 4.1745, 4.2787, 1.9284, 4.4755, 4.4651, 3.3005, 3.3398], device='cuda:1'), covar=tensor([0.0661, 0.0090, 0.0110, 0.1134, 0.0044, 0.0055, 0.0276, 0.0341], device='cuda:1'), in_proj_covar=tensor([0.0143, 0.0087, 0.0081, 0.0142, 0.0071, 0.0078, 0.0116, 0.0125], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0004, 0.0002, 0.0002, 0.0003, 0.0003], device='cuda:1') 2023-04-28 09:02:35,255 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=48513.0, num_to_drop=1, layers_to_drop={0} 2023-04-28 09:03:12,650 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.3658, 4.2191, 4.4217, 4.6392, 4.7185, 4.2582, 4.7208, 4.6627], device='cuda:1'), covar=tensor([0.0882, 0.0708, 0.1054, 0.0454, 0.0397, 0.0715, 0.0407, 0.0396], device='cuda:1'), in_proj_covar=tensor([0.0368, 0.0463, 0.0582, 0.0480, 0.0356, 0.0346, 0.0374, 0.0399], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-28 09:03:36,614 INFO [train.py:904] (1/8) Epoch 5, batch 7950, loss[loss=0.2286, simple_loss=0.306, pruned_loss=0.07557, over 16697.00 frames. ], tot_loss[loss=0.2532, simple_loss=0.3276, pruned_loss=0.08941, over 3099895.38 frames. ], batch size: 89, lr: 1.30e-02, grad_scale: 4.0 2023-04-28 09:03:56,964 INFO [optim.py:368] (1/8) 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 09:04:02,856 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=48568.0, num_to_drop=1, layers_to_drop={3} 2023-04-28 09:04:50,720 INFO [train.py:904] (1/8) Epoch 5, batch 8000, loss[loss=0.3423, simple_loss=0.3773, pruned_loss=0.1537, over 11664.00 frames. ], tot_loss[loss=0.2564, simple_loss=0.3295, pruned_loss=0.09171, over 3062173.33 frames. ], batch size: 247, lr: 1.30e-02, grad_scale: 8.0 2023-04-28 09:04:53,447 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.26 vs. limit=2.0 2023-04-28 09:05:26,699 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.8488, 1.5445, 2.1611, 2.6166, 2.6289, 3.0500, 1.4845, 2.9254], device='cuda:1'), covar=tensor([0.0066, 0.0230, 0.0132, 0.0116, 0.0091, 0.0055, 0.0240, 0.0039], device='cuda:1'), in_proj_covar=tensor([0.0118, 0.0144, 0.0126, 0.0124, 0.0130, 0.0094, 0.0142, 0.0080], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001], device='cuda:1') 2023-04-28 09:06:06,082 INFO [train.py:904] (1/8) Epoch 5, batch 8050, loss[loss=0.2242, simple_loss=0.309, pruned_loss=0.0697, over 16536.00 frames. ], tot_loss[loss=0.2556, simple_loss=0.3294, pruned_loss=0.09091, over 3073587.31 frames. ], batch size: 68, lr: 1.30e-02, grad_scale: 8.0 2023-04-28 09:06:26,963 INFO [optim.py:368] (1/8) 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 09:06:49,657 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=48680.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 09:06:56,126 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([5.1263, 5.0764, 4.9291, 4.1723, 4.8699, 1.8550, 4.6386, 4.8054], device='cuda:1'), covar=tensor([0.0049, 0.0038, 0.0073, 0.0316, 0.0056, 0.1696, 0.0086, 0.0120], device='cuda:1'), in_proj_covar=tensor([0.0091, 0.0077, 0.0121, 0.0124, 0.0090, 0.0142, 0.0105, 0.0117], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-28 09:07:21,588 INFO [train.py:904] (1/8) Epoch 5, batch 8100, loss[loss=0.2323, simple_loss=0.3098, pruned_loss=0.07742, over 17014.00 frames. ], tot_loss[loss=0.2537, simple_loss=0.328, pruned_loss=0.08973, over 3069651.65 frames. ], batch size: 41, lr: 1.30e-02, grad_scale: 8.0 2023-04-28 09:08:05,528 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=48728.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 09:08:40,601 INFO [train.py:904] (1/8) Epoch 5, batch 8150, loss[loss=0.2517, simple_loss=0.3104, pruned_loss=0.09654, over 15381.00 frames. ], tot_loss[loss=0.2503, simple_loss=0.3249, pruned_loss=0.08783, over 3100343.50 frames. ], batch size: 190, lr: 1.30e-02, grad_scale: 8.0 2023-04-28 09:08:44,092 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=48753.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 09:09:01,945 INFO [optim.py:368] (1/8) Clipping_scale=2.0, grad-norm quartiles 2.147e+02 3.868e+02 4.692e+02 5.916e+02 1.218e+03, threshold=9.385e+02, percent-clipped=3.0 2023-04-28 09:09:20,848 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=48776.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 09:09:24,938 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=48779.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 09:09:43,461 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([1.7001, 2.9318, 2.2355, 4.6087, 3.6128, 3.9448, 1.7465, 2.5610], device='cuda:1'), covar=tensor([0.1416, 0.0658, 0.1376, 0.0073, 0.0429, 0.0356, 0.1372, 0.1021], device='cuda:1'), in_proj_covar=tensor([0.0145, 0.0143, 0.0169, 0.0086, 0.0184, 0.0178, 0.0161, 0.0165], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0003, 0.0004], device='cuda:1') 2023-04-28 09:09:43,653 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.23 vs. limit=2.0 2023-04-28 09:10:00,529 INFO [train.py:904] (1/8) Epoch 5, batch 8200, loss[loss=0.2676, simple_loss=0.3413, pruned_loss=0.09699, over 16179.00 frames. ], tot_loss[loss=0.2482, simple_loss=0.3226, pruned_loss=0.08692, over 3105081.57 frames. ], batch size: 165, lr: 1.30e-02, grad_scale: 8.0 2023-04-28 09:10:32,560 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=48821.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 09:10:45,694 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([1.8721, 2.5165, 2.5739, 1.9516, 2.5956, 2.6013, 2.6319, 1.8075], device='cuda:1'), covar=tensor([0.0275, 0.0030, 0.0047, 0.0209, 0.0040, 0.0050, 0.0039, 0.0301], device='cuda:1'), in_proj_covar=tensor([0.0117, 0.0053, 0.0059, 0.0115, 0.0059, 0.0069, 0.0063, 0.0108], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0001, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:1') 2023-04-28 09:10:47,061 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.9560, 1.9188, 2.2065, 3.2296, 2.0280, 2.5875, 2.2713, 1.9434], device='cuda:1'), covar=tensor([0.0531, 0.2198, 0.1043, 0.0360, 0.2808, 0.1030, 0.1769, 0.2459], device='cuda:1'), in_proj_covar=tensor([0.0310, 0.0318, 0.0258, 0.0305, 0.0370, 0.0307, 0.0286, 0.0374], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-28 09:11:00,870 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 5, batch 8250, loss[loss=0.2381, simple_loss=0.322, pruned_loss=0.07707, over 16317.00 frames. ], tot_loss[loss=0.2455, simple_loss=0.3215, pruned_loss=0.08474, over 3093843.97 frames. ], batch size: 165, lr: 1.29e-02, grad_scale: 8.0 2023-04-28 09:11:34,919 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.0047, 1.9021, 2.1683, 3.1329, 1.9621, 2.5211, 2.2144, 1.8983], device='cuda:1'), covar=tensor([0.0526, 0.2282, 0.1100, 0.0366, 0.3243, 0.1084, 0.1982, 0.2541], device='cuda:1'), in_proj_covar=tensor([0.0309, 0.0317, 0.0259, 0.0304, 0.0371, 0.0307, 0.0287, 0.0375], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-28 09:11:44,535 INFO [optim.py:368] (1/8) 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 09:11:50,743 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=48868.0, num_to_drop=1, layers_to_drop={2} 2023-04-28 09:12:13,768 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=48882.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 09:12:40,551 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=48899.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 09:12:43,300 INFO [train.py:904] (1/8) Epoch 5, batch 8300, loss[loss=0.2095, simple_loss=0.2957, pruned_loss=0.06158, over 16541.00 frames. ], tot_loss[loss=0.2398, simple_loss=0.3178, pruned_loss=0.08096, over 3096091.43 frames. ], batch size: 57, lr: 1.29e-02, grad_scale: 8.0 2023-04-28 09:13:08,659 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=48916.0, num_to_drop=1, layers_to_drop={0} 2023-04-28 09:13:46,487 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.2249, 2.2970, 2.2922, 4.7360, 2.0499, 3.3602, 2.4264, 2.4556], device='cuda:1'), covar=tensor([0.0345, 0.1984, 0.1161, 0.0201, 0.3043, 0.0920, 0.1824, 0.2413], device='cuda:1'), in_proj_covar=tensor([0.0306, 0.0314, 0.0257, 0.0299, 0.0368, 0.0305, 0.0285, 0.0370], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-28 09:14:06,483 INFO [train.py:904] (1/8) Epoch 5, batch 8350, loss[loss=0.2567, simple_loss=0.3172, pruned_loss=0.09813, over 12063.00 frames. ], tot_loss[loss=0.2363, simple_loss=0.316, pruned_loss=0.07833, over 3086658.26 frames. ], batch size: 246, lr: 1.29e-02, grad_scale: 8.0 2023-04-28 09:14:30,518 INFO [optim.py:368] (1/8) 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 09:14:47,273 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([1.7774, 2.6535, 2.5197, 3.6486, 2.7951, 3.7199, 1.5651, 2.8342], device='cuda:1'), covar=tensor([0.1325, 0.0511, 0.0892, 0.0099, 0.0154, 0.0360, 0.1351, 0.0671], device='cuda:1'), in_proj_covar=tensor([0.0143, 0.0139, 0.0164, 0.0085, 0.0176, 0.0175, 0.0160, 0.0162], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0003, 0.0004], device='cuda:1') 2023-04-28 09:14:54,849 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=48980.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 09:15:30,220 INFO [train.py:904] (1/8) Epoch 5, batch 8400, loss[loss=0.2036, simple_loss=0.2789, pruned_loss=0.06417, over 12341.00 frames. ], tot_loss[loss=0.2318, simple_loss=0.3125, pruned_loss=0.07558, over 3083452.47 frames. ], batch size: 248, lr: 1.29e-02, grad_scale: 8.0 2023-04-28 09:16:14,934 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=49028.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 09:16:51,001 INFO [train.py:904] (1/8) Epoch 5, batch 8450, loss[loss=0.2141, simple_loss=0.2866, pruned_loss=0.07081, over 12201.00 frames. ], tot_loss[loss=0.2283, simple_loss=0.3104, pruned_loss=0.07305, over 3081943.78 frames. ], batch size: 246, lr: 1.29e-02, grad_scale: 8.0 2023-04-28 09:16:54,837 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=49053.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 09:17:13,777 INFO [optim.py:368] (1/8) 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 09:17:38,528 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=49079.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 09:18:13,189 INFO [train.py:904] (1/8) Epoch 5, batch 8500, loss[loss=0.2174, simple_loss=0.2997, pruned_loss=0.06761, over 16663.00 frames. ], tot_loss[loss=0.2227, simple_loss=0.3055, pruned_loss=0.06992, over 3077630.99 frames. ], batch size: 134, lr: 1.29e-02, grad_scale: 8.0 2023-04-28 09:18:13,691 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=49101.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 09:18:14,256 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.83 vs. limit=2.0 2023-04-28 09:18:42,879 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.7683, 1.4769, 2.1520, 2.8124, 2.5750, 2.8656, 1.7291, 2.9089], device='cuda:1'), covar=tensor([0.0071, 0.0289, 0.0151, 0.0113, 0.0108, 0.0129, 0.0235, 0.0075], device='cuda:1'), in_proj_covar=tensor([0.0121, 0.0144, 0.0128, 0.0124, 0.0132, 0.0093, 0.0142, 0.0079], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001], device='cuda:1') 2023-04-28 09:18:57,904 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 5, batch 8550, loss[loss=0.2121, simple_loss=0.286, pruned_loss=0.06906, over 11908.00 frames. ], tot_loss[loss=0.219, simple_loss=0.3019, pruned_loss=0.06807, over 3051832.99 frames. ], batch size: 246, lr: 1.29e-02, grad_scale: 8.0 2023-04-28 09:20:04,106 INFO [optim.py:368] (1/8) 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 09:20:29,014 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=49177.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 09:21:02,471 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.2059, 4.2429, 4.0377, 3.9020, 3.7064, 4.1604, 4.0228, 3.8587], device='cuda:1'), covar=tensor([0.0451, 0.0278, 0.0247, 0.0200, 0.0696, 0.0335, 0.0387, 0.0465], device='cuda:1'), in_proj_covar=tensor([0.0179, 0.0179, 0.0206, 0.0176, 0.0229, 0.0203, 0.0146, 0.0231], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-28 09:21:04,487 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 5, batch 8600, loss[loss=0.2196, simple_loss=0.3119, pruned_loss=0.06369, over 15317.00 frames. ], tot_loss[loss=0.2183, simple_loss=0.3022, pruned_loss=0.06719, over 3038842.63 frames. ], batch size: 191, lr: 1.29e-02, grad_scale: 8.0 2023-04-28 09:22:28,791 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=49236.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 09:22:58,657 INFO [train.py:904] (1/8) Epoch 5, batch 8650, loss[loss=0.2058, simple_loss=0.2904, pruned_loss=0.06064, over 11916.00 frames. ], tot_loss[loss=0.2161, simple_loss=0.3004, pruned_loss=0.06594, over 3017484.32 frames. ], batch size: 248, lr: 1.29e-02, grad_scale: 4.0 2023-04-28 09:23:04,547 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([1.9653, 2.7343, 2.5699, 1.9596, 2.5253, 2.5315, 2.5768, 1.7820], device='cuda:1'), covar=tensor([0.0231, 0.0024, 0.0037, 0.0178, 0.0049, 0.0039, 0.0039, 0.0275], device='cuda:1'), in_proj_covar=tensor([0.0111, 0.0051, 0.0056, 0.0110, 0.0057, 0.0063, 0.0061, 0.0105], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0001, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:1') 2023-04-28 09:23:33,948 INFO [optim.py:368] (1/8) 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 09:24:11,567 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.9376, 4.2196, 3.9588, 4.0335, 3.6442, 3.7332, 3.8489, 4.0807], device='cuda:1'), covar=tensor([0.0785, 0.0782, 0.1001, 0.0491, 0.0692, 0.1214, 0.0680, 0.1072], device='cuda:1'), in_proj_covar=tensor([0.0347, 0.0462, 0.0393, 0.0300, 0.0288, 0.0311, 0.0376, 0.0334], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-28 09:24:38,773 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=49297.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 09:24:47,018 INFO [train.py:904] (1/8) Epoch 5, batch 8700, loss[loss=0.1837, simple_loss=0.2731, pruned_loss=0.04717, over 16691.00 frames. ], tot_loss[loss=0.2123, simple_loss=0.2966, pruned_loss=0.06397, over 3014376.98 frames. ], batch size: 83, lr: 1.29e-02, grad_scale: 4.0 2023-04-28 09:25:56,471 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=3.41 vs. limit=5.0 2023-04-28 09:26:24,151 INFO [train.py:904] (1/8) Epoch 5, batch 8750, loss[loss=0.1962, simple_loss=0.2745, pruned_loss=0.05895, over 12332.00 frames. ], tot_loss[loss=0.2115, simple_loss=0.2964, pruned_loss=0.06333, over 3030323.35 frames. ], batch size: 248, lr: 1.29e-02, grad_scale: 4.0 2023-04-28 09:27:05,487 INFO [optim.py:368] (1/8) 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] (1/8) Epoch 5, batch 8800, loss[loss=0.2466, simple_loss=0.3127, pruned_loss=0.09022, over 12444.00 frames. ], tot_loss[loss=0.209, simple_loss=0.2944, pruned_loss=0.0618, over 3036239.06 frames. ], batch size: 247, lr: 1.29e-02, grad_scale: 4.0 2023-04-28 09:28:38,045 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.39 vs. limit=2.0 2023-04-28 09:29:02,001 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=49422.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 09:29:06,844 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([1.9660, 1.2398, 1.6507, 2.1038, 2.0059, 2.1357, 1.3770, 2.1560], device='cuda:1'), covar=tensor([0.0144, 0.0277, 0.0166, 0.0143, 0.0142, 0.0130, 0.0266, 0.0065], device='cuda:1'), in_proj_covar=tensor([0.0121, 0.0145, 0.0130, 0.0124, 0.0131, 0.0093, 0.0144, 0.0078], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001], device='cuda:1') 2023-04-28 09:29:24,421 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=49432.0, num_to_drop=1, layers_to_drop={0} 2023-04-28 09:30:04,519 INFO [train.py:904] (1/8) Epoch 5, batch 8850, loss[loss=0.2105, simple_loss=0.2858, pruned_loss=0.06756, over 12264.00 frames. ], tot_loss[loss=0.2096, simple_loss=0.2967, pruned_loss=0.06122, over 3029603.67 frames. ], batch size: 246, lr: 1.29e-02, grad_scale: 4.0 2023-04-28 09:30:38,663 INFO [optim.py:368] (1/8) 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 09:30:41,886 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.8319, 1.5370, 2.1872, 2.8113, 2.6446, 2.9133, 1.6148, 2.9530], device='cuda:1'), covar=tensor([0.0069, 0.0245, 0.0154, 0.0115, 0.0104, 0.0087, 0.0246, 0.0050], device='cuda:1'), in_proj_covar=tensor([0.0119, 0.0143, 0.0128, 0.0124, 0.0130, 0.0091, 0.0141, 0.0077], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001], device='cuda:1') 2023-04-28 09:31:02,575 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=49477.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 09:31:16,984 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=49483.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 09:31:36,662 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=49493.0, num_to_drop=1, layers_to_drop={3} 2023-04-28 09:31:39,638 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=49494.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 09:31:54,236 INFO [train.py:904] (1/8) Epoch 5, batch 8900, loss[loss=0.2147, simple_loss=0.3045, pruned_loss=0.0624, over 17015.00 frames. ], tot_loss[loss=0.2091, simple_loss=0.2973, pruned_loss=0.06052, over 3035819.88 frames. ], batch size: 109, lr: 1.29e-02, grad_scale: 4.0 2023-04-28 09:32:48,133 INFO [zipformer.py:625] (1/8) 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] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=49542.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 09:33:59,804 INFO [train.py:904] (1/8) Epoch 5, batch 8950, loss[loss=0.177, simple_loss=0.2667, pruned_loss=0.04364, over 16339.00 frames. ], tot_loss[loss=0.2096, simple_loss=0.2973, pruned_loss=0.06092, over 3062124.46 frames. ], batch size: 146, lr: 1.29e-02, grad_scale: 4.0 2023-04-28 09:34:13,225 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.31 vs. limit=2.0 2023-04-28 09:34:35,628 INFO [optim.py:368] (1/8) 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 09:35:10,793 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.5403, 3.4935, 3.5055, 3.0583, 3.4665, 1.9814, 3.2499, 2.9950], device='cuda:1'), covar=tensor([0.0074, 0.0067, 0.0081, 0.0180, 0.0062, 0.1576, 0.0091, 0.0128], device='cuda:1'), in_proj_covar=tensor([0.0090, 0.0076, 0.0120, 0.0114, 0.0089, 0.0141, 0.0103, 0.0113], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-28 09:35:33,166 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 5, batch 9000, loss[loss=0.1892, simple_loss=0.2722, pruned_loss=0.05308, over 11851.00 frames. ], tot_loss[loss=0.2066, simple_loss=0.2943, pruned_loss=0.05944, over 3069421.54 frames. ], batch size: 248, lr: 1.28e-02, grad_scale: 4.0 2023-04-28 09:35:51,807 INFO [train.py:929] (1/8) Computing validation loss 2023-04-28 09:36:02,102 INFO [train.py:938] (1/8) Epoch 5, validation: loss=0.1735, simple_loss=0.2766, pruned_loss=0.0352, over 944034.00 frames. 2023-04-28 09:36:02,103 INFO [train.py:939] (1/8) Maximum memory allocated so far is 17676MB 2023-04-28 09:37:44,738 INFO [train.py:904] (1/8) Epoch 5, batch 9050, loss[loss=0.2291, simple_loss=0.3034, pruned_loss=0.07742, over 12821.00 frames. ], tot_loss[loss=0.208, simple_loss=0.2957, pruned_loss=0.06009, over 3085851.88 frames. ], batch size: 247, lr: 1.28e-02, grad_scale: 4.0 2023-04-28 09:38:18,678 INFO [optim.py:368] (1/8) 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 09:39:29,263 INFO [train.py:904] (1/8) Epoch 5, batch 9100, loss[loss=0.2184, simple_loss=0.3077, pruned_loss=0.06459, over 16812.00 frames. ], tot_loss[loss=0.2083, simple_loss=0.2953, pruned_loss=0.0607, over 3072936.53 frames. ], batch size: 124, lr: 1.28e-02, grad_scale: 4.0 2023-04-28 09:41:26,340 INFO [train.py:904] (1/8) Epoch 5, batch 9150, loss[loss=0.1833, simple_loss=0.2714, pruned_loss=0.04763, over 16979.00 frames. ], tot_loss[loss=0.208, simple_loss=0.2954, pruned_loss=0.06032, over 3062030.47 frames. ], batch size: 109, lr: 1.28e-02, grad_scale: 4.0 2023-04-28 09:42:00,558 INFO [zipformer.py:625] (1/8) 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] (1/8) 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 09:42:25,692 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=49778.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 09:42:47,423 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=49788.0, num_to_drop=1, layers_to_drop={1} 2023-04-28 09:43:09,763 INFO [train.py:904] (1/8) Epoch 5, batch 9200, loss[loss=0.1898, simple_loss=0.2631, pruned_loss=0.05827, over 12357.00 frames. ], tot_loss[loss=0.2044, simple_loss=0.2906, pruned_loss=0.05911, over 3058055.92 frames. ], batch size: 247, lr: 1.28e-02, grad_scale: 8.0 2023-04-28 09:43:34,800 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.9477, 1.8569, 2.2520, 3.2435, 1.9491, 2.5350, 2.2780, 1.9403], device='cuda:1'), covar=tensor([0.0565, 0.2247, 0.1001, 0.0347, 0.2976, 0.1207, 0.1912, 0.2506], device='cuda:1'), in_proj_covar=tensor([0.0304, 0.0302, 0.0254, 0.0297, 0.0360, 0.0302, 0.0283, 0.0359], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-28 09:43:58,031 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=49827.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 09:44:46,685 INFO [train.py:904] (1/8) Epoch 5, batch 9250, loss[loss=0.198, simple_loss=0.2954, pruned_loss=0.05036, over 16940.00 frames. ], tot_loss[loss=0.2046, simple_loss=0.2904, pruned_loss=0.05939, over 3053195.12 frames. ], batch size: 109, lr: 1.28e-02, grad_scale: 8.0 2023-04-28 09:44:49,646 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([1.9321, 2.6343, 2.6292, 1.9811, 2.4648, 2.6630, 2.5476, 1.7803], device='cuda:1'), covar=tensor([0.0258, 0.0024, 0.0043, 0.0209, 0.0072, 0.0046, 0.0043, 0.0306], device='cuda:1'), in_proj_covar=tensor([0.0113, 0.0052, 0.0057, 0.0112, 0.0057, 0.0063, 0.0061, 0.0106], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0001, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:1') 2023-04-28 09:45:18,929 INFO [optim.py:368] (1/8) 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 09:46:21,510 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=49892.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 09:46:40,050 INFO [train.py:904] (1/8) Epoch 5, batch 9300, loss[loss=0.1919, simple_loss=0.2711, pruned_loss=0.05637, over 16939.00 frames. ], tot_loss[loss=0.2022, simple_loss=0.2882, pruned_loss=0.05809, over 3055592.10 frames. ], batch size: 125, lr: 1.28e-02, grad_scale: 8.0 2023-04-28 09:47:33,859 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=3.99 vs. limit=5.0 2023-04-28 09:47:36,392 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.39 vs. limit=2.0 2023-04-28 09:48:05,856 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=49940.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 09:48:26,419 INFO [train.py:904] (1/8) Epoch 5, batch 9350, loss[loss=0.1991, simple_loss=0.2809, pruned_loss=0.05865, over 12324.00 frames. ], tot_loss[loss=0.2015, simple_loss=0.2877, pruned_loss=0.0576, over 3064582.09 frames. ], batch size: 248, lr: 1.28e-02, grad_scale: 8.0 2023-04-28 09:49:00,529 INFO [optim.py:368] (1/8) Clipping_scale=2.0, grad-norm quartiles 1.703e+02 3.133e+02 3.801e+02 4.449e+02 7.989e+02, threshold=7.602e+02, percent-clipped=0.0 2023-04-28 09:50:11,380 INFO [train.py:904] (1/8) Epoch 5, batch 9400, loss[loss=0.2104, simple_loss=0.3006, pruned_loss=0.06012, over 15320.00 frames. ], tot_loss[loss=0.2011, simple_loss=0.2875, pruned_loss=0.05732, over 3061228.22 frames. ], batch size: 190, lr: 1.28e-02, grad_scale: 8.0 2023-04-28 09:50:27,792 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=2.63 vs. limit=5.0 2023-04-28 09:50:53,744 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.1080, 3.1204, 3.1555, 1.6566, 3.3503, 3.4003, 2.9073, 2.7003], device='cuda:1'), covar=tensor([0.0739, 0.0137, 0.0125, 0.1078, 0.0065, 0.0074, 0.0308, 0.0355], device='cuda:1'), in_proj_covar=tensor([0.0136, 0.0087, 0.0074, 0.0138, 0.0066, 0.0074, 0.0111, 0.0122], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003, 0.0003], device='cuda:1') 2023-04-28 09:51:24,719 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.3787, 2.9912, 2.5958, 2.1842, 2.0327, 1.9700, 2.8837, 2.8785], device='cuda:1'), covar=tensor([0.1878, 0.0667, 0.1224, 0.1476, 0.2342, 0.1541, 0.0438, 0.0828], device='cuda:1'), in_proj_covar=tensor([0.0273, 0.0241, 0.0264, 0.0237, 0.0241, 0.0195, 0.0235, 0.0233], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-28 09:51:51,664 INFO [train.py:904] (1/8) Epoch 5, batch 9450, loss[loss=0.2139, simple_loss=0.2919, pruned_loss=0.06792, over 12468.00 frames. ], tot_loss[loss=0.2035, simple_loss=0.2902, pruned_loss=0.05841, over 3057843.21 frames. ], batch size: 248, lr: 1.28e-02, grad_scale: 8.0 2023-04-28 09:52:21,841 INFO [optim.py:368] (1/8) Clipping_scale=2.0, grad-norm quartiles 1.776e+02 2.955e+02 3.683e+02 5.023e+02 1.227e+03, threshold=7.366e+02, percent-clipped=5.0 2023-04-28 09:52:43,501 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.7160, 2.0755, 1.7001, 1.7945, 2.5730, 2.3189, 2.8006, 2.7473], device='cuda:1'), covar=tensor([0.0034, 0.0232, 0.0272, 0.0277, 0.0124, 0.0187, 0.0066, 0.0095], device='cuda:1'), in_proj_covar=tensor([0.0071, 0.0151, 0.0150, 0.0147, 0.0143, 0.0150, 0.0118, 0.0128], device='cuda:1'), out_proj_covar=tensor([8.4810e-05, 1.8207e-04, 1.7649e-04, 1.7395e-04, 1.7463e-04, 1.8120e-04, 1.3525e-04, 1.5476e-04], device='cuda:1') 2023-04-28 09:52:47,365 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=50078.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 09:53:08,989 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=50088.0, num_to_drop=1, layers_to_drop={0} 2023-04-28 09:53:13,397 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.1382, 3.3029, 3.2866, 2.3765, 3.1300, 3.2901, 3.1923, 1.9087], device='cuda:1'), covar=tensor([0.0262, 0.0016, 0.0027, 0.0187, 0.0034, 0.0037, 0.0032, 0.0277], device='cuda:1'), in_proj_covar=tensor([0.0111, 0.0051, 0.0056, 0.0111, 0.0056, 0.0063, 0.0060, 0.0104], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0001, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:1') 2023-04-28 09:53:33,284 INFO [train.py:904] (1/8) Epoch 5, batch 9500, loss[loss=0.2026, simple_loss=0.2923, pruned_loss=0.05652, over 16901.00 frames. ], tot_loss[loss=0.2016, simple_loss=0.2886, pruned_loss=0.05728, over 3064974.96 frames. ], batch size: 116, lr: 1.28e-02, grad_scale: 8.0 2023-04-28 09:53:57,288 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.31 vs. limit=2.0 2023-04-28 09:54:17,577 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=50122.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 09:54:25,577 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=50126.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 09:54:29,264 INFO [zipformer.py:625] (1/8) 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] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=50136.0, num_to_drop=1, layers_to_drop={1} 2023-04-28 09:55:14,406 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.3112, 2.9975, 2.5864, 2.2910, 2.1336, 2.1416, 2.9338, 3.0133], device='cuda:1'), covar=tensor([0.1808, 0.0780, 0.1160, 0.1472, 0.1706, 0.1360, 0.0404, 0.0731], device='cuda:1'), in_proj_covar=tensor([0.0274, 0.0240, 0.0264, 0.0240, 0.0241, 0.0196, 0.0236, 0.0232], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-28 09:55:18,505 INFO [train.py:904] (1/8) Epoch 5, batch 9550, loss[loss=0.1966, simple_loss=0.292, pruned_loss=0.05058, over 16847.00 frames. ], tot_loss[loss=0.2012, simple_loss=0.288, pruned_loss=0.05718, over 3055942.63 frames. ], batch size: 116, lr: 1.28e-02, grad_scale: 8.0 2023-04-28 09:55:53,349 INFO [optim.py:368] (1/8) Clipping_scale=2.0, grad-norm quartiles 1.669e+02 2.772e+02 3.459e+02 4.281e+02 6.687e+02, threshold=6.919e+02, percent-clipped=0.0 2023-04-28 09:56:38,160 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=50189.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 09:56:52,895 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=50197.0, num_to_drop=1, layers_to_drop={0} 2023-04-28 09:56:59,038 INFO [train.py:904] (1/8) Epoch 5, batch 9600, loss[loss=0.1719, simple_loss=0.2572, pruned_loss=0.04329, over 16839.00 frames. ], tot_loss[loss=0.2032, simple_loss=0.2894, pruned_loss=0.0585, over 3030420.55 frames. ], batch size: 42, lr: 1.28e-02, grad_scale: 8.0 2023-04-28 09:58:47,432 INFO [train.py:904] (1/8) Epoch 5, batch 9650, loss[loss=0.2177, simple_loss=0.2973, pruned_loss=0.0691, over 16227.00 frames. ], tot_loss[loss=0.2049, simple_loss=0.2915, pruned_loss=0.0591, over 3029076.09 frames. ], batch size: 165, lr: 1.28e-02, grad_scale: 8.0 2023-04-28 09:59:02,965 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.91 vs. limit=2.0 2023-04-28 09:59:05,750 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=50258.0, num_to_drop=1, layers_to_drop={3} 2023-04-28 09:59:27,461 INFO [optim.py:368] (1/8) 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 09:59:58,647 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.88 vs. limit=2.0 2023-04-28 10:00:33,606 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.3657, 3.2718, 3.4031, 3.5237, 3.5119, 3.2396, 3.5089, 3.5451], device='cuda:1'), covar=tensor([0.0717, 0.0623, 0.0906, 0.0481, 0.0524, 0.1582, 0.0677, 0.0503], device='cuda:1'), in_proj_covar=tensor([0.0344, 0.0431, 0.0540, 0.0446, 0.0331, 0.0329, 0.0354, 0.0371], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-28 10:00:35,786 INFO [train.py:904] (1/8) Epoch 5, batch 9700, loss[loss=0.1855, simple_loss=0.2729, pruned_loss=0.04901, over 17027.00 frames. ], tot_loss[loss=0.204, simple_loss=0.2907, pruned_loss=0.05867, over 3043593.06 frames. ], batch size: 55, lr: 1.28e-02, grad_scale: 8.0 2023-04-28 10:02:18,556 INFO [train.py:904] (1/8) Epoch 5, batch 9750, loss[loss=0.2089, simple_loss=0.2939, pruned_loss=0.06193, over 15353.00 frames. ], tot_loss[loss=0.2037, simple_loss=0.29, pruned_loss=0.05868, over 3055958.98 frames. ], batch size: 191, lr: 1.28e-02, grad_scale: 8.0 2023-04-28 10:02:50,575 INFO [optim.py:368] (1/8) 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 10:03:13,982 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.46 vs. limit=2.0 2023-04-28 10:03:56,153 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.68 vs. limit=2.0 2023-04-28 10:03:56,536 INFO [train.py:904] (1/8) Epoch 5, batch 9800, loss[loss=0.1907, simple_loss=0.2731, pruned_loss=0.05417, over 12724.00 frames. ], tot_loss[loss=0.2024, simple_loss=0.29, pruned_loss=0.05733, over 3070898.44 frames. ], batch size: 247, lr: 1.27e-02, grad_scale: 8.0 2023-04-28 10:04:36,880 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=50422.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 10:05:41,977 INFO [train.py:904] (1/8) Epoch 5, batch 9850, loss[loss=0.1913, simple_loss=0.2831, pruned_loss=0.04978, over 16839.00 frames. ], tot_loss[loss=0.2023, simple_loss=0.2912, pruned_loss=0.05671, over 3094800.82 frames. ], batch size: 124, lr: 1.27e-02, grad_scale: 8.0 2023-04-28 10:06:14,680 INFO [optim.py:368] (1/8) 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 10:06:21,303 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=50470.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 10:06:52,451 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=50484.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 10:07:31,481 INFO [train.py:904] (1/8) Epoch 5, batch 9900, loss[loss=0.2214, simple_loss=0.2929, pruned_loss=0.07492, over 12104.00 frames. ], tot_loss[loss=0.2032, simple_loss=0.2924, pruned_loss=0.05701, over 3104778.63 frames. ], batch size: 246, lr: 1.27e-02, grad_scale: 8.0 2023-04-28 10:09:17,011 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=50546.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 10:09:27,670 INFO [train.py:904] (1/8) Epoch 5, batch 9950, loss[loss=0.2031, simple_loss=0.2914, pruned_loss=0.05738, over 16702.00 frames. ], tot_loss[loss=0.2046, simple_loss=0.2943, pruned_loss=0.05745, over 3103265.36 frames. ], batch size: 57, lr: 1.27e-02, grad_scale: 8.0 2023-04-28 10:09:33,659 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=50553.0, num_to_drop=1, layers_to_drop={2} 2023-04-28 10:10:04,528 INFO [optim.py:368] (1/8) 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 10:10:09,171 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.7940, 1.9137, 2.1580, 3.2083, 1.9577, 2.4695, 2.2224, 1.9341], device='cuda:1'), covar=tensor([0.0561, 0.2150, 0.1050, 0.0372, 0.2725, 0.1154, 0.2034, 0.2287], device='cuda:1'), in_proj_covar=tensor([0.0303, 0.0307, 0.0261, 0.0302, 0.0359, 0.0306, 0.0287, 0.0360], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-28 10:11:08,347 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=3.86 vs. limit=5.0 2023-04-28 10:11:29,333 INFO [train.py:904] (1/8) Epoch 5, batch 10000, loss[loss=0.1983, simple_loss=0.2763, pruned_loss=0.0601, over 12624.00 frames. ], tot_loss[loss=0.2032, simple_loss=0.2925, pruned_loss=0.05698, over 3102129.19 frames. ], batch size: 250, lr: 1.27e-02, grad_scale: 8.0 2023-04-28 10:11:43,856 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=50607.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 10:13:08,522 INFO [train.py:904] (1/8) Epoch 5, batch 10050, loss[loss=0.2111, simple_loss=0.3056, pruned_loss=0.05827, over 16291.00 frames. ], tot_loss[loss=0.2027, simple_loss=0.2923, pruned_loss=0.05658, over 3111322.57 frames. ], batch size: 165, lr: 1.27e-02, grad_scale: 8.0 2023-04-28 10:13:38,879 INFO [optim.py:368] (1/8) 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 10:14:38,999 INFO [train.py:904] (1/8) Epoch 5, batch 10100, loss[loss=0.2059, simple_loss=0.288, pruned_loss=0.06187, over 16939.00 frames. ], tot_loss[loss=0.2033, simple_loss=0.2926, pruned_loss=0.05705, over 3104358.04 frames. ], batch size: 109, lr: 1.27e-02, grad_scale: 8.0 2023-04-28 10:15:24,297 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.3653, 3.2956, 3.4260, 3.5253, 3.5572, 3.2024, 3.5474, 3.5656], device='cuda:1'), covar=tensor([0.0744, 0.0673, 0.0950, 0.0561, 0.0450, 0.1695, 0.0615, 0.0475], device='cuda:1'), in_proj_covar=tensor([0.0348, 0.0442, 0.0552, 0.0461, 0.0338, 0.0336, 0.0358, 0.0374], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-28 10:16:20,269 INFO [train.py:904] (1/8) Epoch 6, batch 0, loss[loss=0.3649, simple_loss=0.3825, pruned_loss=0.1736, over 16816.00 frames. ], tot_loss[loss=0.3649, simple_loss=0.3825, pruned_loss=0.1736, over 16816.00 frames. ], batch size: 102, lr: 1.19e-02, grad_scale: 8.0 2023-04-28 10:16:20,269 INFO [train.py:929] (1/8) Computing validation loss 2023-04-28 10:16:27,645 INFO [train.py:938] (1/8) Epoch 6, validation: loss=0.1727, simple_loss=0.2755, pruned_loss=0.03501, over 944034.00 frames. 2023-04-28 10:16:27,645 INFO [train.py:939] (1/8) Maximum memory allocated so far is 17676MB 2023-04-28 10:16:52,397 INFO [optim.py:368] (1/8) 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 10:17:10,804 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=50782.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 10:17:13,080 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=50784.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 10:17:35,095 INFO [train.py:904] (1/8) Epoch 6, batch 50, loss[loss=0.2339, simple_loss=0.3177, pruned_loss=0.07509, over 17135.00 frames. ], tot_loss[loss=0.2447, simple_loss=0.3135, pruned_loss=0.08799, over 752682.63 frames. ], batch size: 48, lr: 1.19e-02, grad_scale: 2.0 2023-04-28 10:17:46,513 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=50808.0, num_to_drop=1, layers_to_drop={0} 2023-04-28 10:17:52,113 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=4.13 vs. limit=5.0 2023-04-28 10:18:20,193 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=50832.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 10:18:34,279 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=50843.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 10:18:47,054 INFO [train.py:904] (1/8) Epoch 6, batch 100, loss[loss=0.2497, simple_loss=0.303, pruned_loss=0.09817, over 16774.00 frames. ], tot_loss[loss=0.2357, simple_loss=0.3072, pruned_loss=0.08211, over 1319669.32 frames. ], batch size: 83, lr: 1.18e-02, grad_scale: 2.0 2023-04-28 10:18:49,498 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=50853.0, num_to_drop=1, layers_to_drop={0} 2023-04-28 10:19:03,578 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.9668, 4.6317, 4.9745, 5.2250, 5.3738, 4.6572, 5.3606, 5.3035], device='cuda:1'), covar=tensor([0.1059, 0.0951, 0.1307, 0.0545, 0.0448, 0.0725, 0.0408, 0.0388], device='cuda:1'), in_proj_covar=tensor([0.0380, 0.0482, 0.0605, 0.0493, 0.0367, 0.0361, 0.0389, 0.0408], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-28 10:19:11,789 INFO [optim.py:368] (1/8) 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 10:19:12,289 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=50869.0, num_to_drop=1, layers_to_drop={1} 2023-04-28 10:19:54,852 INFO [train.py:904] (1/8) Epoch 6, batch 150, loss[loss=0.1917, simple_loss=0.2756, pruned_loss=0.05388, over 17209.00 frames. ], tot_loss[loss=0.2286, simple_loss=0.302, pruned_loss=0.07759, over 1773375.43 frames. ], batch size: 45, lr: 1.18e-02, grad_scale: 2.0 2023-04-28 10:19:55,127 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=50901.0, num_to_drop=1, layers_to_drop={0} 2023-04-28 10:19:56,137 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=50902.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 10:20:13,394 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.5831, 4.3291, 4.0708, 1.9836, 3.2094, 2.5258, 3.8210, 3.9867], device='cuda:1'), covar=tensor([0.0262, 0.0472, 0.0424, 0.1555, 0.0630, 0.0917, 0.0609, 0.0948], device='cuda:1'), in_proj_covar=tensor([0.0136, 0.0118, 0.0149, 0.0139, 0.0127, 0.0125, 0.0135, 0.0131], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-04-28 10:20:42,158 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.26 vs. limit=2.0 2023-04-28 10:20:57,647 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=50946.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 10:21:05,275 INFO [train.py:904] (1/8) Epoch 6, batch 200, loss[loss=0.199, simple_loss=0.2844, pruned_loss=0.0568, over 17098.00 frames. ], tot_loss[loss=0.2292, simple_loss=0.303, pruned_loss=0.07772, over 2111508.56 frames. ], batch size: 49, lr: 1.18e-02, grad_scale: 2.0 2023-04-28 10:21:28,627 INFO [optim.py:368] (1/8) 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 10:22:12,666 INFO [train.py:904] (1/8) Epoch 6, batch 250, loss[loss=0.191, simple_loss=0.2719, pruned_loss=0.05505, over 17210.00 frames. ], tot_loss[loss=0.2263, simple_loss=0.2997, pruned_loss=0.07644, over 2381084.58 frames. ], batch size: 46, lr: 1.18e-02, grad_scale: 2.0 2023-04-28 10:22:20,792 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=51007.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 10:22:24,662 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.8495, 2.0705, 2.3546, 4.6390, 1.8655, 3.0042, 2.4088, 2.2083], device='cuda:1'), covar=tensor([0.0566, 0.2553, 0.1269, 0.0268, 0.3476, 0.1265, 0.2133, 0.3046], device='cuda:1'), in_proj_covar=tensor([0.0315, 0.0320, 0.0267, 0.0311, 0.0369, 0.0324, 0.0295, 0.0381], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-28 10:22:29,407 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.26 vs. limit=2.0 2023-04-28 10:23:20,591 INFO [train.py:904] (1/8) Epoch 6, batch 300, loss[loss=0.2681, simple_loss=0.3203, pruned_loss=0.108, over 16874.00 frames. ], tot_loss[loss=0.2219, simple_loss=0.296, pruned_loss=0.07387, over 2595841.11 frames. ], batch size: 109, lr: 1.18e-02, grad_scale: 2.0 2023-04-28 10:23:24,338 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=5.01 vs. limit=5.0 2023-04-28 10:23:36,720 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.56 vs. limit=2.0 2023-04-28 10:23:37,401 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=51062.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 10:23:45,700 INFO [optim.py:368] (1/8) Clipping_scale=2.0, grad-norm quartiles 1.679e+02 2.797e+02 3.691e+02 4.478e+02 8.058e+02, threshold=7.381e+02, percent-clipped=1.0 2023-04-28 10:24:30,409 INFO [train.py:904] (1/8) Epoch 6, batch 350, loss[loss=0.1868, simple_loss=0.2698, pruned_loss=0.05189, over 17041.00 frames. ], tot_loss[loss=0.219, simple_loss=0.2934, pruned_loss=0.07231, over 2758642.32 frames. ], batch size: 50, lr: 1.18e-02, grad_scale: 2.0 2023-04-28 10:25:01,812 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=51123.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 10:25:19,470 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=51138.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 10:25:37,100 INFO [train.py:904] (1/8) Epoch 6, batch 400, loss[loss=0.2113, simple_loss=0.2994, pruned_loss=0.06162, over 17220.00 frames. ], tot_loss[loss=0.2174, simple_loss=0.2909, pruned_loss=0.07197, over 2886355.73 frames. ], batch size: 52, lr: 1.18e-02, grad_scale: 4.0 2023-04-28 10:25:55,420 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=51164.0, num_to_drop=1, layers_to_drop={0} 2023-04-28 10:26:01,706 INFO [optim.py:368] (1/8) 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 10:26:02,445 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=2.02 vs. limit=2.0 2023-04-28 10:26:05,660 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=51172.0, num_to_drop=1, layers_to_drop={0} 2023-04-28 10:26:20,591 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.1758, 4.5731, 3.4116, 2.7264, 3.1495, 2.4827, 4.7226, 4.4478], device='cuda:1'), covar=tensor([0.1968, 0.0460, 0.1116, 0.1565, 0.2352, 0.1548, 0.0325, 0.0580], device='cuda:1'), in_proj_covar=tensor([0.0278, 0.0250, 0.0271, 0.0246, 0.0271, 0.0203, 0.0244, 0.0256], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-28 10:26:45,056 INFO [train.py:904] (1/8) Epoch 6, batch 450, loss[loss=0.1827, simple_loss=0.2545, pruned_loss=0.05542, over 15923.00 frames. ], tot_loss[loss=0.216, simple_loss=0.2898, pruned_loss=0.07108, over 2981094.14 frames. ], batch size: 35, lr: 1.18e-02, grad_scale: 4.0 2023-04-28 10:26:46,579 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=51202.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 10:27:29,518 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=51233.0, num_to_drop=1, layers_to_drop={0} 2023-04-28 10:27:52,965 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=51250.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 10:27:53,833 INFO [train.py:904] (1/8) Epoch 6, batch 500, loss[loss=0.2286, simple_loss=0.2885, pruned_loss=0.0843, over 16313.00 frames. ], tot_loss[loss=0.2129, simple_loss=0.2873, pruned_loss=0.06927, over 3052153.86 frames. ], batch size: 165, lr: 1.18e-02, grad_scale: 4.0 2023-04-28 10:28:17,360 INFO [optim.py:368] (1/8) 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 10:29:00,268 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.88 vs. limit=2.0 2023-04-28 10:29:01,771 INFO [train.py:904] (1/8) Epoch 6, batch 550, loss[loss=0.1947, simple_loss=0.2852, pruned_loss=0.05207, over 17076.00 frames. ], tot_loss[loss=0.2109, simple_loss=0.2858, pruned_loss=0.06796, over 3111534.48 frames. ], batch size: 49, lr: 1.18e-02, grad_scale: 4.0 2023-04-28 10:29:03,265 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=51302.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 10:29:05,333 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.3191, 4.0079, 3.2388, 1.9286, 2.8615, 2.4136, 3.7184, 3.7424], device='cuda:1'), covar=tensor([0.0211, 0.0423, 0.0587, 0.1534, 0.0698, 0.0874, 0.0453, 0.0808], device='cuda:1'), in_proj_covar=tensor([0.0137, 0.0126, 0.0150, 0.0139, 0.0129, 0.0124, 0.0136, 0.0136], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-04-28 10:29:49,608 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.7486, 4.5836, 4.7528, 5.0279, 5.1540, 4.5592, 5.0766, 5.0971], device='cuda:1'), covar=tensor([0.0996, 0.0717, 0.1380, 0.0478, 0.0427, 0.0667, 0.0488, 0.0423], device='cuda:1'), in_proj_covar=tensor([0.0423, 0.0524, 0.0667, 0.0532, 0.0396, 0.0395, 0.0421, 0.0451], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-28 10:30:13,391 INFO [train.py:904] (1/8) Epoch 6, batch 600, loss[loss=0.2307, simple_loss=0.2882, pruned_loss=0.08657, over 12276.00 frames. ], tot_loss[loss=0.2106, simple_loss=0.2858, pruned_loss=0.06768, over 3153781.21 frames. ], batch size: 246, lr: 1.18e-02, grad_scale: 4.0 2023-04-28 10:30:17,739 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.43 vs. limit=2.0 2023-04-28 10:30:38,608 INFO [optim.py:368] (1/8) 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 10:30:55,682 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=51382.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 10:31:10,948 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=51392.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 10:31:23,785 INFO [train.py:904] (1/8) Epoch 6, batch 650, loss[loss=0.2081, simple_loss=0.2955, pruned_loss=0.06035, over 16625.00 frames. ], tot_loss[loss=0.2093, simple_loss=0.2841, pruned_loss=0.06727, over 3183354.94 frames. ], batch size: 62, lr: 1.18e-02, grad_scale: 4.0 2023-04-28 10:31:47,261 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=51418.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 10:31:55,768 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.92 vs. limit=2.0 2023-04-28 10:32:14,562 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=51438.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 10:32:21,250 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=51443.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 10:32:28,208 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.0665, 2.1287, 2.2466, 4.5999, 1.8347, 3.1217, 2.3436, 2.3154], device='cuda:1'), covar=tensor([0.0545, 0.2644, 0.1314, 0.0282, 0.3556, 0.1275, 0.2096, 0.2983], device='cuda:1'), in_proj_covar=tensor([0.0324, 0.0327, 0.0272, 0.0318, 0.0374, 0.0338, 0.0300, 0.0396], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-28 10:32:31,562 INFO [train.py:904] (1/8) Epoch 6, batch 700, loss[loss=0.2368, simple_loss=0.3005, pruned_loss=0.08658, over 16747.00 frames. ], tot_loss[loss=0.2087, simple_loss=0.2838, pruned_loss=0.06683, over 3212054.07 frames. ], batch size: 124, lr: 1.18e-02, grad_scale: 2.0 2023-04-28 10:32:35,210 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=51453.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 10:32:49,157 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=51464.0, num_to_drop=1, layers_to_drop={1} 2023-04-28 10:32:57,146 INFO [optim.py:368] (1/8) 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] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=51486.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 10:33:27,633 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.7008, 4.7454, 5.2766, 5.3096, 5.2934, 4.8636, 4.9052, 4.5201], device='cuda:1'), covar=tensor([0.0286, 0.0314, 0.0327, 0.0376, 0.0346, 0.0253, 0.0749, 0.0391], device='cuda:1'), in_proj_covar=tensor([0.0262, 0.0263, 0.0259, 0.0258, 0.0309, 0.0274, 0.0391, 0.0230], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:1') 2023-04-28 10:33:32,354 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 6, batch 750, loss[loss=0.1999, simple_loss=0.291, pruned_loss=0.05441, over 17275.00 frames. ], tot_loss[loss=0.211, simple_loss=0.2856, pruned_loss=0.06822, over 3234523.14 frames. ], batch size: 52, lr: 1.18e-02, grad_scale: 2.0 2023-04-28 10:33:56,283 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=51512.0, num_to_drop=1, layers_to_drop={1} 2023-04-28 10:34:19,284 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=51528.0, num_to_drop=1, layers_to_drop={0} 2023-04-28 10:34:53,035 INFO [train.py:904] (1/8) Epoch 6, batch 800, loss[loss=0.2186, simple_loss=0.2867, pruned_loss=0.07524, over 16234.00 frames. ], tot_loss[loss=0.2096, simple_loss=0.2846, pruned_loss=0.06729, over 3251340.47 frames. ], batch size: 165, lr: 1.18e-02, grad_scale: 4.0 2023-04-28 10:34:59,935 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=51556.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 10:35:19,940 INFO [optim.py:368] (1/8) 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 10:35:50,448 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.42 vs. limit=2.0 2023-04-28 10:36:01,901 INFO [train.py:904] (1/8) Epoch 6, batch 850, loss[loss=0.2147, simple_loss=0.2836, pruned_loss=0.07292, over 16693.00 frames. ], tot_loss[loss=0.2092, simple_loss=0.2844, pruned_loss=0.06702, over 3259924.41 frames. ], batch size: 134, lr: 1.18e-02, grad_scale: 4.0 2023-04-28 10:36:04,193 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=51602.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 10:36:26,678 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.1903, 3.7097, 3.1557, 1.9748, 2.7880, 2.2456, 3.6181, 3.6160], device='cuda:1'), covar=tensor([0.0203, 0.0476, 0.0563, 0.1399, 0.0645, 0.0907, 0.0493, 0.0665], device='cuda:1'), in_proj_covar=tensor([0.0141, 0.0130, 0.0153, 0.0141, 0.0131, 0.0125, 0.0139, 0.0140], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-04-28 10:37:10,632 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=51650.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 10:37:11,527 INFO [train.py:904] (1/8) Epoch 6, batch 900, loss[loss=0.1958, simple_loss=0.2868, pruned_loss=0.05241, over 16721.00 frames. ], tot_loss[loss=0.2063, simple_loss=0.2824, pruned_loss=0.06508, over 3281107.71 frames. ], batch size: 57, lr: 1.18e-02, grad_scale: 4.0 2023-04-28 10:37:39,516 INFO [optim.py:368] (1/8) Clipping_scale=2.0, grad-norm quartiles 1.821e+02 2.905e+02 3.526e+02 4.406e+02 8.198e+02, threshold=7.052e+02, percent-clipped=7.0 2023-04-28 10:38:22,587 INFO [train.py:904] (1/8) Epoch 6, batch 950, loss[loss=0.18, simple_loss=0.2518, pruned_loss=0.05409, over 16982.00 frames. ], tot_loss[loss=0.2057, simple_loss=0.2818, pruned_loss=0.06479, over 3297234.85 frames. ], batch size: 41, lr: 1.17e-02, grad_scale: 4.0 2023-04-28 10:38:46,223 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=51718.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 10:39:12,962 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=51738.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 10:39:25,795 INFO [zipformer.py:625] (1/8) 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] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=51748.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 10:39:30,564 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.92 vs. limit=2.0 2023-04-28 10:39:30,782 INFO [train.py:904] (1/8) Epoch 6, batch 1000, loss[loss=0.1794, simple_loss=0.261, pruned_loss=0.04884, over 17219.00 frames. ], tot_loss[loss=0.2053, simple_loss=0.2807, pruned_loss=0.065, over 3311098.04 frames. ], batch size: 44, lr: 1.17e-02, grad_scale: 4.0 2023-04-28 10:39:51,033 INFO [zipformer.py:625] (1/8) 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] (1/8) 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] (1/8) Epoch 6, batch 1050, loss[loss=0.191, simple_loss=0.2783, pruned_loss=0.0518, over 17070.00 frames. ], tot_loss[loss=0.2051, simple_loss=0.2807, pruned_loss=0.06472, over 3324265.46 frames. ], batch size: 53, lr: 1.17e-02, grad_scale: 4.0 2023-04-28 10:40:49,385 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=51808.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 10:41:17,898 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=51828.0, num_to_drop=1, layers_to_drop={2} 2023-04-28 10:41:49,453 INFO [train.py:904] (1/8) Epoch 6, batch 1100, loss[loss=0.2004, simple_loss=0.265, pruned_loss=0.06796, over 16481.00 frames. ], tot_loss[loss=0.204, simple_loss=0.28, pruned_loss=0.06396, over 3322353.04 frames. ], batch size: 146, lr: 1.17e-02, grad_scale: 4.0 2023-04-28 10:41:49,769 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=51851.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 10:42:02,729 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.4332, 4.0789, 3.6637, 1.9176, 3.0056, 2.3021, 3.6681, 3.8121], device='cuda:1'), covar=tensor([0.0256, 0.0506, 0.0495, 0.1575, 0.0711, 0.0937, 0.0617, 0.0819], device='cuda:1'), in_proj_covar=tensor([0.0141, 0.0132, 0.0153, 0.0141, 0.0132, 0.0125, 0.0139, 0.0141], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-04-28 10:42:16,638 INFO [optim.py:368] (1/8) 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] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=51876.0, num_to_drop=1, layers_to_drop={0} 2023-04-28 10:42:59,414 INFO [train.py:904] (1/8) Epoch 6, batch 1150, loss[loss=0.203, simple_loss=0.2685, pruned_loss=0.06876, over 16457.00 frames. ], tot_loss[loss=0.2035, simple_loss=0.2795, pruned_loss=0.06374, over 3327836.94 frames. ], batch size: 68, lr: 1.17e-02, grad_scale: 4.0 2023-04-28 10:43:40,638 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.0686, 4.2950, 4.5233, 1.9573, 4.8435, 4.8263, 3.4240, 3.8629], device='cuda:1'), covar=tensor([0.0614, 0.0137, 0.0177, 0.1153, 0.0048, 0.0052, 0.0298, 0.0258], device='cuda:1'), in_proj_covar=tensor([0.0143, 0.0094, 0.0083, 0.0142, 0.0072, 0.0084, 0.0117, 0.0127], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0002, 0.0002, 0.0003, 0.0003], device='cuda:1') 2023-04-28 10:44:08,004 INFO [train.py:904] (1/8) Epoch 6, batch 1200, loss[loss=0.1543, simple_loss=0.2431, pruned_loss=0.03274, over 16882.00 frames. ], tot_loss[loss=0.2015, simple_loss=0.2777, pruned_loss=0.06268, over 3325609.77 frames. ], batch size: 42, lr: 1.17e-02, grad_scale: 8.0 2023-04-28 10:44:33,629 INFO [optim.py:368] (1/8) 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 10:44:55,350 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.6517, 5.0192, 4.6878, 4.7846, 4.4059, 4.4015, 4.5011, 5.0161], device='cuda:1'), covar=tensor([0.0886, 0.0802, 0.0982, 0.0569, 0.0773, 0.0936, 0.0903, 0.0863], device='cuda:1'), in_proj_covar=tensor([0.0407, 0.0550, 0.0453, 0.0353, 0.0337, 0.0347, 0.0448, 0.0389], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-28 10:45:20,380 INFO [train.py:904] (1/8) Epoch 6, batch 1250, loss[loss=0.1759, simple_loss=0.2643, pruned_loss=0.04374, over 17218.00 frames. ], tot_loss[loss=0.2025, simple_loss=0.2779, pruned_loss=0.06353, over 3315131.84 frames. ], batch size: 45, lr: 1.17e-02, grad_scale: 8.0 2023-04-28 10:45:46,874 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.8939, 5.2672, 4.9607, 5.0878, 4.7115, 4.5279, 4.8286, 5.3181], device='cuda:1'), covar=tensor([0.0832, 0.0742, 0.0950, 0.0457, 0.0674, 0.0849, 0.0756, 0.0763], device='cuda:1'), in_proj_covar=tensor([0.0401, 0.0546, 0.0448, 0.0349, 0.0334, 0.0342, 0.0444, 0.0385], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-28 10:46:13,383 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=52038.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 10:46:15,805 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=52040.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 10:46:26,680 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 6, batch 1300, loss[loss=0.1817, simple_loss=0.2578, pruned_loss=0.05284, over 17033.00 frames. ], tot_loss[loss=0.2032, simple_loss=0.2785, pruned_loss=0.0639, over 3319241.78 frames. ], batch size: 41, lr: 1.17e-02, grad_scale: 4.0 2023-04-28 10:46:38,513 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.49 vs. limit=2.0 2023-04-28 10:46:48,396 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.0215, 2.1316, 2.2926, 4.6328, 2.0526, 3.1716, 2.2809, 2.4344], device='cuda:1'), covar=tensor([0.0507, 0.2437, 0.1284, 0.0260, 0.3011, 0.1181, 0.2160, 0.2644], device='cuda:1'), in_proj_covar=tensor([0.0327, 0.0329, 0.0275, 0.0320, 0.0374, 0.0348, 0.0302, 0.0403], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-28 10:46:58,326 INFO [optim.py:368] (1/8) 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 10:47:19,561 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=52086.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 10:47:33,414 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 6, batch 1350, loss[loss=0.2192, simple_loss=0.2835, pruned_loss=0.07741, over 16717.00 frames. ], tot_loss[loss=0.2023, simple_loss=0.2785, pruned_loss=0.06308, over 3329353.70 frames. ], batch size: 134, lr: 1.17e-02, grad_scale: 4.0 2023-04-28 10:47:42,011 INFO [zipformer.py:625] (1/8) 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] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=52103.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 10:48:07,124 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([5.3837, 5.3446, 5.1802, 4.9356, 4.5783, 5.2079, 5.2268, 4.8229], device='cuda:1'), covar=tensor([0.0426, 0.0363, 0.0191, 0.0208, 0.1034, 0.0322, 0.0216, 0.0483], device='cuda:1'), in_proj_covar=tensor([0.0203, 0.0218, 0.0239, 0.0209, 0.0272, 0.0243, 0.0171, 0.0272], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-04-28 10:48:40,874 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.9648, 1.8387, 2.4224, 2.8383, 2.6497, 3.2487, 1.9536, 3.1648], device='cuda:1'), covar=tensor([0.0095, 0.0234, 0.0147, 0.0127, 0.0123, 0.0089, 0.0226, 0.0058], device='cuda:1'), in_proj_covar=tensor([0.0127, 0.0146, 0.0130, 0.0130, 0.0133, 0.0097, 0.0141, 0.0082], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001], device='cuda:1') 2023-04-28 10:48:51,203 INFO [train.py:904] (1/8) Epoch 6, batch 1400, loss[loss=0.2207, simple_loss=0.3066, pruned_loss=0.0674, over 17072.00 frames. ], tot_loss[loss=0.2019, simple_loss=0.2783, pruned_loss=0.0628, over 3314607.87 frames. ], batch size: 53, lr: 1.17e-02, grad_scale: 4.0 2023-04-28 10:48:51,498 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=52151.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 10:49:19,188 INFO [optim.py:368] (1/8) 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 10:49:19,618 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.0135, 2.7713, 2.6996, 1.7554, 2.9316, 2.8636, 2.4372, 2.3884], device='cuda:1'), covar=tensor([0.0681, 0.0158, 0.0185, 0.0929, 0.0094, 0.0147, 0.0422, 0.0395], device='cuda:1'), in_proj_covar=tensor([0.0142, 0.0093, 0.0084, 0.0142, 0.0072, 0.0084, 0.0117, 0.0128], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0002, 0.0002, 0.0003, 0.0003], device='cuda:1') 2023-04-28 10:49:19,981 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.41 vs. limit=2.0 2023-04-28 10:49:49,791 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=52193.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 10:49:58,142 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=52199.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 10:50:00,568 INFO [train.py:904] (1/8) Epoch 6, batch 1450, loss[loss=0.2289, simple_loss=0.2954, pruned_loss=0.08123, over 16375.00 frames. ], tot_loss[loss=0.2011, simple_loss=0.2777, pruned_loss=0.06223, over 3320676.91 frames. ], batch size: 68, lr: 1.17e-02, grad_scale: 4.0 2023-04-28 10:50:33,799 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.5649, 3.7557, 2.0545, 3.9387, 2.6651, 3.8252, 1.9760, 2.8511], device='cuda:1'), covar=tensor([0.0166, 0.0314, 0.1334, 0.0108, 0.0668, 0.0458, 0.1257, 0.0563], device='cuda:1'), in_proj_covar=tensor([0.0122, 0.0161, 0.0180, 0.0086, 0.0160, 0.0193, 0.0186, 0.0163], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-04-28 10:51:10,831 INFO [train.py:904] (1/8) Epoch 6, batch 1500, loss[loss=0.1759, simple_loss=0.2576, pruned_loss=0.04707, over 17016.00 frames. ], tot_loss[loss=0.2003, simple_loss=0.2772, pruned_loss=0.06166, over 3319857.26 frames. ], batch size: 41, lr: 1.17e-02, grad_scale: 4.0 2023-04-28 10:51:15,476 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=52254.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 10:51:30,616 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.31 vs. limit=2.0 2023-04-28 10:51:38,534 INFO [optim.py:368] (1/8) Clipping_scale=2.0, grad-norm quartiles 1.881e+02 3.021e+02 3.666e+02 4.541e+02 9.509e+02, threshold=7.332e+02, percent-clipped=3.0 2023-04-28 10:51:52,581 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.89 vs. limit=2.0 2023-04-28 10:52:18,640 INFO [train.py:904] (1/8) Epoch 6, batch 1550, loss[loss=0.2112, simple_loss=0.2999, pruned_loss=0.06124, over 17132.00 frames. ], tot_loss[loss=0.2022, simple_loss=0.2787, pruned_loss=0.06288, over 3315739.06 frames. ], batch size: 47, lr: 1.17e-02, grad_scale: 4.0 2023-04-28 10:52:30,286 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([5.0176, 5.4670, 5.5391, 5.3438, 5.3139, 5.9872, 5.5533, 5.2781], device='cuda:1'), covar=tensor([0.0729, 0.1466, 0.1357, 0.1780, 0.2558, 0.0831, 0.1192, 0.2216], device='cuda:1'), in_proj_covar=tensor([0.0292, 0.0423, 0.0420, 0.0364, 0.0489, 0.0453, 0.0343, 0.0486], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-04-28 10:52:33,892 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=52312.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 10:53:28,025 INFO [train.py:904] (1/8) Epoch 6, batch 1600, loss[loss=0.2032, simple_loss=0.2992, pruned_loss=0.05361, over 17090.00 frames. ], tot_loss[loss=0.2046, simple_loss=0.2813, pruned_loss=0.06395, over 3318093.69 frames. ], batch size: 49, lr: 1.17e-02, grad_scale: 8.0 2023-04-28 10:53:55,822 INFO [optim.py:368] (1/8) 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 10:53:59,241 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=52373.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 10:54:30,442 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=52396.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 10:54:37,119 INFO [train.py:904] (1/8) Epoch 6, batch 1650, loss[loss=0.1624, simple_loss=0.2434, pruned_loss=0.04064, over 16802.00 frames. ], tot_loss[loss=0.2067, simple_loss=0.2829, pruned_loss=0.06531, over 3315059.30 frames. ], batch size: 39, lr: 1.17e-02, grad_scale: 8.0 2023-04-28 10:54:40,229 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=52403.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 10:55:25,795 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.0974, 4.1043, 4.5150, 3.2325, 4.0910, 4.3657, 4.0138, 2.6767], device='cuda:1'), covar=tensor([0.0234, 0.0026, 0.0016, 0.0180, 0.0036, 0.0032, 0.0033, 0.0230], device='cuda:1'), in_proj_covar=tensor([0.0115, 0.0061, 0.0059, 0.0112, 0.0061, 0.0070, 0.0064, 0.0105], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:1') 2023-04-28 10:55:45,567 INFO [train.py:904] (1/8) Epoch 6, batch 1700, loss[loss=0.1754, simple_loss=0.2502, pruned_loss=0.05028, over 16791.00 frames. ], tot_loss[loss=0.2091, simple_loss=0.2853, pruned_loss=0.06643, over 3310528.93 frames. ], batch size: 39, lr: 1.17e-02, grad_scale: 8.0 2023-04-28 10:55:45,910 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=52451.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 10:55:50,012 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.42 vs. limit=2.0 2023-04-28 10:56:14,195 INFO [optim.py:368] (1/8) 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 10:56:40,323 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([5.1560, 4.8914, 5.0766, 5.3785, 5.5309, 4.8366, 5.4738, 5.4683], device='cuda:1'), covar=tensor([0.0854, 0.0868, 0.1293, 0.0458, 0.0361, 0.0550, 0.0375, 0.0408], device='cuda:1'), in_proj_covar=tensor([0.0436, 0.0544, 0.0690, 0.0552, 0.0414, 0.0407, 0.0433, 0.0464], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-28 10:56:57,691 INFO [train.py:904] (1/8) Epoch 6, batch 1750, loss[loss=0.2401, simple_loss=0.3062, pruned_loss=0.08703, over 16876.00 frames. ], tot_loss[loss=0.2093, simple_loss=0.286, pruned_loss=0.06626, over 3314359.88 frames. ], batch size: 116, lr: 1.17e-02, grad_scale: 8.0 2023-04-28 10:57:06,595 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.26 vs. limit=2.0 2023-04-28 10:58:05,484 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 6, batch 1800, loss[loss=0.2182, simple_loss=0.2857, pruned_loss=0.07532, over 16775.00 frames. ], tot_loss[loss=0.2103, simple_loss=0.2872, pruned_loss=0.06671, over 3308833.43 frames. ], batch size: 102, lr: 1.17e-02, grad_scale: 8.0 2023-04-28 10:58:14,103 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=52555.0, num_to_drop=1, layers_to_drop={0} 2023-04-28 10:58:27,259 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([5.0303, 4.7277, 5.0148, 5.2683, 5.4343, 4.7393, 5.3835, 5.3181], device='cuda:1'), covar=tensor([0.1042, 0.0909, 0.1379, 0.0492, 0.0369, 0.0613, 0.0357, 0.0435], device='cuda:1'), in_proj_covar=tensor([0.0443, 0.0548, 0.0695, 0.0560, 0.0416, 0.0406, 0.0435, 0.0467], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-28 10:58:36,301 INFO [optim.py:368] (1/8) 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 10:58:55,248 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.0900, 2.1583, 2.3484, 4.8558, 1.9705, 3.2979, 2.3468, 2.4941], device='cuda:1'), covar=tensor([0.0551, 0.2589, 0.1318, 0.0233, 0.3217, 0.1232, 0.2151, 0.2654], device='cuda:1'), in_proj_covar=tensor([0.0329, 0.0331, 0.0274, 0.0318, 0.0374, 0.0348, 0.0301, 0.0403], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-28 10:58:57,476 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([5.1874, 5.4156, 5.1610, 5.0805, 4.2323, 5.3274, 5.2903, 4.8560], device='cuda:1'), covar=tensor([0.0611, 0.0323, 0.0285, 0.0213, 0.1578, 0.0266, 0.0193, 0.0430], device='cuda:1'), in_proj_covar=tensor([0.0204, 0.0221, 0.0243, 0.0215, 0.0277, 0.0248, 0.0173, 0.0276], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-04-28 10:59:17,730 INFO [train.py:904] (1/8) Epoch 6, batch 1850, loss[loss=0.2321, simple_loss=0.3023, pruned_loss=0.08089, over 16421.00 frames. ], tot_loss[loss=0.2088, simple_loss=0.2865, pruned_loss=0.06559, over 3317855.33 frames. ], batch size: 146, lr: 1.16e-02, grad_scale: 8.0 2023-04-28 10:59:38,373 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=52616.0, num_to_drop=1, layers_to_drop={3} 2023-04-28 10:59:39,367 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.9501, 5.0658, 4.8651, 4.7050, 4.0754, 4.9468, 5.0546, 4.6302], device='cuda:1'), covar=tensor([0.0610, 0.0281, 0.0285, 0.0273, 0.1174, 0.0328, 0.0238, 0.0546], device='cuda:1'), in_proj_covar=tensor([0.0205, 0.0222, 0.0243, 0.0215, 0.0278, 0.0248, 0.0173, 0.0277], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-04-28 10:59:54,450 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.6182, 4.3625, 3.9518, 1.8966, 3.1163, 2.6040, 3.8512, 3.9788], device='cuda:1'), covar=tensor([0.0236, 0.0502, 0.0440, 0.1543, 0.0663, 0.0888, 0.0658, 0.0911], device='cuda:1'), in_proj_covar=tensor([0.0143, 0.0134, 0.0151, 0.0139, 0.0130, 0.0124, 0.0139, 0.0142], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-04-28 11:00:24,461 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([5.2966, 5.3062, 5.2125, 4.5917, 5.1648, 2.1264, 4.9591, 5.2573], device='cuda:1'), covar=tensor([0.0053, 0.0049, 0.0081, 0.0289, 0.0052, 0.1554, 0.0080, 0.0099], device='cuda:1'), in_proj_covar=tensor([0.0103, 0.0090, 0.0139, 0.0137, 0.0104, 0.0150, 0.0120, 0.0132], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-28 11:00:27,090 INFO [train.py:904] (1/8) Epoch 6, batch 1900, loss[loss=0.2, simple_loss=0.2903, pruned_loss=0.05482, over 17062.00 frames. ], tot_loss[loss=0.2071, simple_loss=0.2857, pruned_loss=0.06427, over 3322857.72 frames. ], batch size: 53, lr: 1.16e-02, grad_scale: 8.0 2023-04-28 11:00:51,241 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=52668.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 11:00:52,610 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.6732, 4.8182, 5.3007, 5.3616, 5.2982, 4.8867, 4.8761, 4.5294], device='cuda:1'), covar=tensor([0.0294, 0.0382, 0.0298, 0.0329, 0.0377, 0.0266, 0.0719, 0.0370], device='cuda:1'), in_proj_covar=tensor([0.0269, 0.0267, 0.0265, 0.0263, 0.0322, 0.0280, 0.0395, 0.0233], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:1') 2023-04-28 11:00:54,522 INFO [optim.py:368] (1/8) 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 11:01:10,413 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.1215, 4.2245, 4.6124, 4.6555, 4.6704, 4.2232, 4.2692, 4.0655], device='cuda:1'), covar=tensor([0.0314, 0.0430, 0.0324, 0.0399, 0.0339, 0.0293, 0.0737, 0.0471], device='cuda:1'), in_proj_covar=tensor([0.0269, 0.0267, 0.0266, 0.0263, 0.0322, 0.0280, 0.0396, 0.0233], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:1') 2023-04-28 11:01:26,536 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.0383, 3.8702, 4.0804, 4.2593, 4.3579, 3.9230, 3.9964, 4.3368], device='cuda:1'), covar=tensor([0.0943, 0.0822, 0.1199, 0.0562, 0.0445, 0.1127, 0.1484, 0.0487], device='cuda:1'), in_proj_covar=tensor([0.0446, 0.0549, 0.0699, 0.0564, 0.0420, 0.0415, 0.0439, 0.0472], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-28 11:01:30,561 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=52696.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 11:01:36,633 INFO [train.py:904] (1/8) Epoch 6, batch 1950, loss[loss=0.2381, simple_loss=0.309, pruned_loss=0.08361, over 15619.00 frames. ], tot_loss[loss=0.2082, simple_loss=0.2868, pruned_loss=0.06482, over 3320974.12 frames. ], batch size: 191, lr: 1.16e-02, grad_scale: 8.0 2023-04-28 11:01:57,089 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.5724, 3.7107, 1.8692, 3.8319, 2.5863, 3.8081, 1.9294, 2.7878], device='cuda:1'), covar=tensor([0.0122, 0.0256, 0.1347, 0.0100, 0.0699, 0.0487, 0.1241, 0.0539], device='cuda:1'), in_proj_covar=tensor([0.0122, 0.0161, 0.0176, 0.0085, 0.0160, 0.0193, 0.0185, 0.0161], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-04-28 11:02:36,311 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 6, batch 2000, loss[loss=0.1936, simple_loss=0.2736, pruned_loss=0.05678, over 16454.00 frames. ], tot_loss[loss=0.208, simple_loss=0.2862, pruned_loss=0.0649, over 3320491.85 frames. ], batch size: 75, lr: 1.16e-02, grad_scale: 8.0 2023-04-28 11:02:49,188 INFO [zipformer.py:625] (1/8) 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] (1/8) Clipping_scale=2.0, grad-norm quartiles 2.007e+02 2.923e+02 3.553e+02 4.132e+02 6.210e+02, threshold=7.105e+02, percent-clipped=0.0 2023-04-28 11:03:51,853 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.6286, 4.5776, 5.0453, 5.1081, 5.1099, 4.7190, 4.6497, 4.3951], device='cuda:1'), covar=tensor([0.0270, 0.0404, 0.0400, 0.0401, 0.0425, 0.0268, 0.0793, 0.0382], device='cuda:1'), in_proj_covar=tensor([0.0265, 0.0266, 0.0264, 0.0262, 0.0319, 0.0279, 0.0394, 0.0229], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:1') 2023-04-28 11:03:57,170 INFO [train.py:904] (1/8) Epoch 6, batch 2050, loss[loss=0.2095, simple_loss=0.278, pruned_loss=0.07048, over 16781.00 frames. ], tot_loss[loss=0.2079, simple_loss=0.286, pruned_loss=0.06487, over 3325014.36 frames. ], batch size: 83, lr: 1.16e-02, grad_scale: 8.0 2023-04-28 11:04:05,303 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([5.6624, 5.9990, 5.6992, 5.9221, 5.3897, 5.1749, 5.5435, 6.1543], device='cuda:1'), covar=tensor([0.0771, 0.0741, 0.0988, 0.0493, 0.0655, 0.0531, 0.0635, 0.0725], device='cuda:1'), in_proj_covar=tensor([0.0409, 0.0551, 0.0452, 0.0354, 0.0336, 0.0345, 0.0447, 0.0389], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-28 11:04:14,428 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=52813.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 11:05:06,240 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 6, batch 2100, loss[loss=0.2357, simple_loss=0.3053, pruned_loss=0.08302, over 11977.00 frames. ], tot_loss[loss=0.2082, simple_loss=0.2865, pruned_loss=0.06493, over 3319062.12 frames. ], batch size: 247, lr: 1.16e-02, grad_scale: 8.0 2023-04-28 11:05:36,281 INFO [optim.py:368] (1/8) Clipping_scale=2.0, grad-norm quartiles 1.586e+02 2.933e+02 3.490e+02 4.297e+02 7.985e+02, threshold=6.979e+02, percent-clipped=3.0 2023-04-28 11:06:12,919 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 6, batch 2150, loss[loss=0.2381, simple_loss=0.3001, pruned_loss=0.08806, over 16701.00 frames. ], tot_loss[loss=0.2117, simple_loss=0.2891, pruned_loss=0.06713, over 3311129.81 frames. ], batch size: 83, lr: 1.16e-02, grad_scale: 8.0 2023-04-28 11:06:33,073 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=52911.0, num_to_drop=1, layers_to_drop={3} 2023-04-28 11:07:14,966 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.2266, 3.7687, 3.1443, 1.9604, 2.7432, 2.1920, 3.6647, 3.4587], device='cuda:1'), covar=tensor([0.0212, 0.0435, 0.0553, 0.1415, 0.0688, 0.0916, 0.0417, 0.0709], device='cuda:1'), in_proj_covar=tensor([0.0142, 0.0135, 0.0152, 0.0140, 0.0130, 0.0125, 0.0139, 0.0143], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-04-28 11:07:29,309 INFO [train.py:904] (1/8) Epoch 6, batch 2200, loss[loss=0.2251, simple_loss=0.2879, pruned_loss=0.08113, over 16783.00 frames. ], tot_loss[loss=0.2131, simple_loss=0.2902, pruned_loss=0.06796, over 3306756.90 frames. ], batch size: 102, lr: 1.16e-02, grad_scale: 8.0 2023-04-28 11:07:52,256 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=52968.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 11:07:56,704 INFO [optim.py:368] (1/8) 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 11:08:38,583 INFO [train.py:904] (1/8) Epoch 6, batch 2250, loss[loss=0.2223, simple_loss=0.2916, pruned_loss=0.07656, over 16709.00 frames. ], tot_loss[loss=0.2133, simple_loss=0.2902, pruned_loss=0.06824, over 3315107.83 frames. ], batch size: 134, lr: 1.16e-02, grad_scale: 8.0 2023-04-28 11:08:59,896 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=53016.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 11:09:47,159 INFO [train.py:904] (1/8) Epoch 6, batch 2300, loss[loss=0.1807, simple_loss=0.2674, pruned_loss=0.04699, over 17251.00 frames. ], tot_loss[loss=0.2121, simple_loss=0.2895, pruned_loss=0.06739, over 3318663.06 frames. ], batch size: 44, lr: 1.16e-02, grad_scale: 8.0 2023-04-28 11:09:54,542 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=53056.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 11:10:15,630 INFO [optim.py:368] (1/8) 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 11:10:57,588 INFO [train.py:904] (1/8) Epoch 6, batch 2350, loss[loss=0.2176, simple_loss=0.3028, pruned_loss=0.06616, over 17030.00 frames. ], tot_loss[loss=0.2132, simple_loss=0.2903, pruned_loss=0.0681, over 3323393.26 frames. ], batch size: 55, lr: 1.16e-02, grad_scale: 8.0 2023-04-28 11:11:08,083 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=53108.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 11:11:20,196 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=53117.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 11:11:29,692 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.16 vs. limit=2.0 2023-04-28 11:12:08,734 INFO [train.py:904] (1/8) Epoch 6, batch 2400, loss[loss=0.2323, simple_loss=0.3019, pruned_loss=0.08138, over 16676.00 frames. ], tot_loss[loss=0.2135, simple_loss=0.2907, pruned_loss=0.06815, over 3326200.64 frames. ], batch size: 134, lr: 1.16e-02, grad_scale: 8.0 2023-04-28 11:12:36,945 INFO [optim.py:368] (1/8) Clipping_scale=2.0, grad-norm quartiles 1.873e+02 2.840e+02 3.422e+02 4.151e+02 8.672e+02, threshold=6.844e+02, percent-clipped=2.0 2023-04-28 11:13:12,697 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([1.9286, 2.7148, 2.6844, 2.0434, 2.5136, 2.6556, 2.5455, 1.7119], device='cuda:1'), covar=tensor([0.0267, 0.0065, 0.0041, 0.0192, 0.0074, 0.0062, 0.0053, 0.0275], device='cuda:1'), in_proj_covar=tensor([0.0118, 0.0063, 0.0062, 0.0115, 0.0063, 0.0072, 0.0065, 0.0109], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:1') 2023-04-28 11:13:19,222 INFO [train.py:904] (1/8) Epoch 6, batch 2450, loss[loss=0.1886, simple_loss=0.2774, pruned_loss=0.0499, over 17228.00 frames. ], tot_loss[loss=0.2131, simple_loss=0.2911, pruned_loss=0.06752, over 3319945.34 frames. ], batch size: 46, lr: 1.16e-02, grad_scale: 8.0 2023-04-28 11:13:32,921 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=53211.0, num_to_drop=1, layers_to_drop={2} 2023-04-28 11:13:37,142 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=4.82 vs. limit=5.0 2023-04-28 11:13:50,114 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.8857, 4.7854, 4.7579, 4.5508, 4.3683, 4.7762, 4.7444, 4.4545], device='cuda:1'), covar=tensor([0.0532, 0.0370, 0.0213, 0.0231, 0.0830, 0.0321, 0.0298, 0.0584], device='cuda:1'), in_proj_covar=tensor([0.0208, 0.0226, 0.0246, 0.0219, 0.0281, 0.0250, 0.0175, 0.0278], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-04-28 11:14:15,988 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 6, batch 2500, loss[loss=0.202, simple_loss=0.2708, pruned_loss=0.06657, over 16870.00 frames. ], tot_loss[loss=0.2126, simple_loss=0.2908, pruned_loss=0.06721, over 3321538.05 frames. ], batch size: 96, lr: 1.16e-02, grad_scale: 8.0 2023-04-28 11:14:39,970 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=53259.0, num_to_drop=1, layers_to_drop={1} 2023-04-28 11:14:57,074 INFO [optim.py:368] (1/8) 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] (1/8) Epoch 6, batch 2550, loss[loss=0.2356, simple_loss=0.3108, pruned_loss=0.08022, over 15544.00 frames. ], tot_loss[loss=0.2134, simple_loss=0.2915, pruned_loss=0.06766, over 3312135.49 frames. ], batch size: 190, lr: 1.16e-02, grad_scale: 8.0 2023-04-28 11:15:40,052 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=53302.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 11:15:43,242 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=53304.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 11:15:58,254 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=53315.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 11:16:42,062 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([5.0081, 4.7710, 4.9328, 5.2299, 5.3449, 4.6837, 5.2974, 5.3106], device='cuda:1'), covar=tensor([0.0835, 0.0658, 0.1343, 0.0426, 0.0390, 0.0504, 0.0384, 0.0321], device='cuda:1'), in_proj_covar=tensor([0.0438, 0.0536, 0.0695, 0.0555, 0.0420, 0.0412, 0.0435, 0.0465], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-28 11:16:48,720 INFO [train.py:904] (1/8) Epoch 6, batch 2600, loss[loss=0.1753, simple_loss=0.2625, pruned_loss=0.04406, over 17246.00 frames. ], tot_loss[loss=0.2121, simple_loss=0.2914, pruned_loss=0.06636, over 3320196.39 frames. ], batch size: 43, lr: 1.16e-02, grad_scale: 8.0 2023-04-28 11:17:09,359 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=53365.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 11:17:11,640 INFO [zipformer.py:625] (1/8) 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] (1/8) Clipping_scale=2.0, grad-norm quartiles 1.993e+02 2.796e+02 3.424e+02 4.100e+02 9.306e+02, threshold=6.847e+02, percent-clipped=3.0 2023-04-28 11:17:17,116 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=3.36 vs. limit=5.0 2023-04-28 11:17:24,067 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=53376.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 11:17:59,458 INFO [train.py:904] (1/8) Epoch 6, batch 2650, loss[loss=0.2147, simple_loss=0.2895, pruned_loss=0.06989, over 16050.00 frames. ], tot_loss[loss=0.212, simple_loss=0.2919, pruned_loss=0.06605, over 3324241.94 frames. ], batch size: 35, lr: 1.16e-02, grad_scale: 8.0 2023-04-28 11:18:10,399 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=53408.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 11:18:14,692 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=53412.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 11:18:37,167 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=53428.0, num_to_drop=1, layers_to_drop={2} 2023-04-28 11:18:56,700 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=4.75 vs. limit=5.0 2023-04-28 11:19:09,533 INFO [train.py:904] (1/8) Epoch 6, batch 2700, loss[loss=0.1841, simple_loss=0.2697, pruned_loss=0.04926, over 16988.00 frames. ], tot_loss[loss=0.2112, simple_loss=0.2916, pruned_loss=0.06539, over 3319887.86 frames. ], batch size: 41, lr: 1.16e-02, grad_scale: 4.0 2023-04-28 11:19:16,987 INFO [zipformer.py:625] (1/8) 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] (1/8) Clipping_scale=2.0, grad-norm quartiles 1.962e+02 2.946e+02 3.581e+02 4.267e+02 8.371e+02, threshold=7.163e+02, percent-clipped=3.0 2023-04-28 11:20:19,909 INFO [train.py:904] (1/8) Epoch 6, batch 2750, loss[loss=0.2002, simple_loss=0.2886, pruned_loss=0.05589, over 17184.00 frames. ], tot_loss[loss=0.21, simple_loss=0.2908, pruned_loss=0.06464, over 3329290.15 frames. ], batch size: 46, lr: 1.16e-02, grad_scale: 4.0 2023-04-28 11:20:21,679 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.4571, 4.2683, 3.7358, 1.9434, 3.2078, 2.5073, 3.8193, 3.9673], device='cuda:1'), covar=tensor([0.0308, 0.0521, 0.0490, 0.1565, 0.0648, 0.0850, 0.0670, 0.1042], device='cuda:1'), in_proj_covar=tensor([0.0144, 0.0137, 0.0155, 0.0140, 0.0130, 0.0124, 0.0140, 0.0145], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-04-28 11:20:24,560 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.6716, 2.4715, 1.7137, 2.4290, 2.9682, 2.8233, 3.6918, 3.1981], device='cuda:1'), covar=tensor([0.0027, 0.0190, 0.0294, 0.0206, 0.0123, 0.0161, 0.0093, 0.0119], device='cuda:1'), in_proj_covar=tensor([0.0095, 0.0167, 0.0166, 0.0164, 0.0163, 0.0170, 0.0156, 0.0152], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-28 11:21:31,892 INFO [train.py:904] (1/8) Epoch 6, batch 2800, loss[loss=0.2186, simple_loss=0.2917, pruned_loss=0.07274, over 16923.00 frames. ], tot_loss[loss=0.2106, simple_loss=0.2909, pruned_loss=0.06519, over 3328305.41 frames. ], batch size: 116, lr: 1.15e-02, grad_scale: 8.0 2023-04-28 11:21:53,265 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.0267, 4.2656, 2.0312, 4.5068, 2.8630, 4.5644, 2.3349, 3.0008], device='cuda:1'), covar=tensor([0.0128, 0.0197, 0.1462, 0.0053, 0.0687, 0.0260, 0.1213, 0.0554], device='cuda:1'), in_proj_covar=tensor([0.0124, 0.0162, 0.0177, 0.0088, 0.0161, 0.0194, 0.0184, 0.0159], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0004, 0.0004, 0.0002, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-04-28 11:22:01,718 INFO [optim.py:368] (1/8) 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 11:22:03,753 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.79 vs. limit=2.0 2023-04-28 11:22:37,078 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=53597.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 11:22:43,360 INFO [train.py:904] (1/8) Epoch 6, batch 2850, loss[loss=0.2406, simple_loss=0.3307, pruned_loss=0.07522, over 16637.00 frames. ], tot_loss[loss=0.2093, simple_loss=0.2896, pruned_loss=0.06447, over 3333244.07 frames. ], batch size: 57, lr: 1.15e-02, grad_scale: 8.0 2023-04-28 11:23:51,283 INFO [train.py:904] (1/8) Epoch 6, batch 2900, loss[loss=0.2191, simple_loss=0.2823, pruned_loss=0.07797, over 16938.00 frames. ], tot_loss[loss=0.2095, simple_loss=0.2892, pruned_loss=0.06494, over 3336227.33 frames. ], batch size: 96, lr: 1.15e-02, grad_scale: 8.0 2023-04-28 11:24:04,493 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=53660.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 11:24:19,312 INFO [zipformer.py:625] (1/8) 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] (1/8) 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 11:24:59,790 INFO [train.py:904] (1/8) Epoch 6, batch 2950, loss[loss=0.1855, simple_loss=0.2699, pruned_loss=0.05056, over 16982.00 frames. ], tot_loss[loss=0.2099, simple_loss=0.2882, pruned_loss=0.06576, over 3331140.55 frames. ], batch size: 41, lr: 1.15e-02, grad_scale: 8.0 2023-04-28 11:25:10,497 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.9442, 4.2699, 3.1364, 2.3769, 3.1375, 2.4452, 4.4640, 4.2447], device='cuda:1'), covar=tensor([0.2203, 0.0662, 0.1207, 0.1832, 0.2109, 0.1521, 0.0332, 0.0646], device='cuda:1'), in_proj_covar=tensor([0.0280, 0.0257, 0.0270, 0.0251, 0.0291, 0.0205, 0.0246, 0.0272], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-28 11:25:15,495 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=53712.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 11:25:30,128 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=53723.0, num_to_drop=1, layers_to_drop={3} 2023-04-28 11:26:08,567 INFO [train.py:904] (1/8) Epoch 6, batch 3000, loss[loss=0.1942, simple_loss=0.2764, pruned_loss=0.05603, over 17229.00 frames. ], tot_loss[loss=0.2102, simple_loss=0.2885, pruned_loss=0.06599, over 3334625.45 frames. ], batch size: 44, lr: 1.15e-02, grad_scale: 8.0 2023-04-28 11:26:08,568 INFO [train.py:929] (1/8) Computing validation loss 2023-04-28 11:26:17,401 INFO [train.py:938] (1/8) Epoch 6, validation: loss=0.1514, simple_loss=0.258, pruned_loss=0.02246, over 944034.00 frames. 2023-04-28 11:26:17,402 INFO [train.py:939] (1/8) Maximum memory allocated so far is 17676MB 2023-04-28 11:26:29,470 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=53760.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 11:26:38,620 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.4597, 4.4214, 4.2905, 4.1781, 4.0239, 4.3503, 4.1685, 4.0905], device='cuda:1'), covar=tensor([0.0431, 0.0307, 0.0204, 0.0219, 0.0709, 0.0267, 0.0477, 0.0474], device='cuda:1'), in_proj_covar=tensor([0.0209, 0.0228, 0.0248, 0.0220, 0.0282, 0.0249, 0.0174, 0.0277], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-04-28 11:26:45,969 INFO [optim.py:368] (1/8) 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 11:26:47,773 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.17 vs. limit=2.0 2023-04-28 11:27:25,831 INFO [train.py:904] (1/8) Epoch 6, batch 3050, loss[loss=0.1748, simple_loss=0.2577, pruned_loss=0.04598, over 16794.00 frames. ], tot_loss[loss=0.2084, simple_loss=0.2873, pruned_loss=0.0648, over 3339662.97 frames. ], batch size: 39, lr: 1.15e-02, grad_scale: 8.0 2023-04-28 11:28:35,659 INFO [train.py:904] (1/8) Epoch 6, batch 3100, loss[loss=0.2248, simple_loss=0.2962, pruned_loss=0.07667, over 16755.00 frames. ], tot_loss[loss=0.2083, simple_loss=0.2864, pruned_loss=0.0651, over 3335903.62 frames. ], batch size: 83, lr: 1.15e-02, grad_scale: 8.0 2023-04-28 11:29:05,910 INFO [optim.py:368] (1/8) Clipping_scale=2.0, grad-norm quartiles 1.546e+02 2.841e+02 3.355e+02 4.130e+02 6.611e+02, threshold=6.710e+02, percent-clipped=0.0 2023-04-28 11:29:40,276 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=53897.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 11:29:44,822 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.44 vs. limit=2.0 2023-04-28 11:29:45,137 INFO [train.py:904] (1/8) Epoch 6, batch 3150, loss[loss=0.2006, simple_loss=0.275, pruned_loss=0.06313, over 16603.00 frames. ], tot_loss[loss=0.2069, simple_loss=0.2851, pruned_loss=0.06434, over 3339192.24 frames. ], batch size: 89, lr: 1.15e-02, grad_scale: 8.0 2023-04-28 11:30:09,185 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([5.0788, 4.7569, 4.9757, 5.2521, 5.4343, 4.7208, 5.4295, 5.3740], device='cuda:1'), covar=tensor([0.1041, 0.0805, 0.1553, 0.0535, 0.0403, 0.0619, 0.0390, 0.0400], device='cuda:1'), in_proj_covar=tensor([0.0440, 0.0535, 0.0699, 0.0552, 0.0416, 0.0415, 0.0435, 0.0467], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-28 11:30:35,904 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=4.99 vs. limit=5.0 2023-04-28 11:30:49,057 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=53945.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 11:30:56,671 INFO [train.py:904] (1/8) Epoch 6, batch 3200, loss[loss=0.2024, simple_loss=0.275, pruned_loss=0.06493, over 16696.00 frames. ], tot_loss[loss=0.2059, simple_loss=0.2844, pruned_loss=0.06371, over 3333988.27 frames. ], batch size: 89, lr: 1.15e-02, grad_scale: 8.0 2023-04-28 11:31:09,485 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=53960.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 11:31:25,789 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=53971.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 11:31:26,625 INFO [optim.py:368] (1/8) 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 11:32:09,695 INFO [train.py:904] (1/8) Epoch 6, batch 3250, loss[loss=0.2053, simple_loss=0.2803, pruned_loss=0.06521, over 15826.00 frames. ], tot_loss[loss=0.2074, simple_loss=0.2854, pruned_loss=0.06471, over 3312503.13 frames. ], batch size: 35, lr: 1.15e-02, grad_scale: 8.0 2023-04-28 11:32:19,694 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=54008.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 11:32:35,403 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=54019.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 11:32:40,062 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=54023.0, num_to_drop=1, layers_to_drop={0} 2023-04-28 11:33:17,040 INFO [train.py:904] (1/8) Epoch 6, batch 3300, loss[loss=0.1986, simple_loss=0.2717, pruned_loss=0.06275, over 16809.00 frames. ], tot_loss[loss=0.208, simple_loss=0.286, pruned_loss=0.06498, over 3320122.68 frames. ], batch size: 83, lr: 1.15e-02, grad_scale: 8.0 2023-04-28 11:33:45,159 INFO [zipformer.py:625] (1/8) 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] (1/8) 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 11:33:53,617 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.0849, 4.2979, 4.5352, 3.2309, 3.8455, 4.4664, 4.0662, 2.8861], device='cuda:1'), covar=tensor([0.0258, 0.0021, 0.0020, 0.0202, 0.0044, 0.0030, 0.0033, 0.0243], device='cuda:1'), in_proj_covar=tensor([0.0115, 0.0060, 0.0061, 0.0111, 0.0062, 0.0070, 0.0064, 0.0105], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:1') 2023-04-28 11:34:03,687 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.62 vs. limit=2.0 2023-04-28 11:34:26,275 INFO [train.py:904] (1/8) Epoch 6, batch 3350, loss[loss=0.1853, simple_loss=0.2753, pruned_loss=0.04768, over 17124.00 frames. ], tot_loss[loss=0.208, simple_loss=0.2864, pruned_loss=0.06482, over 3321181.48 frames. ], batch size: 48, lr: 1.15e-02, grad_scale: 8.0 2023-04-28 11:35:25,486 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=4.94 vs. limit=5.0 2023-04-28 11:35:34,480 INFO [train.py:904] (1/8) Epoch 6, batch 3400, loss[loss=0.205, simple_loss=0.2931, pruned_loss=0.05844, over 17135.00 frames. ], tot_loss[loss=0.2071, simple_loss=0.2856, pruned_loss=0.06428, over 3316184.29 frames. ], batch size: 47, lr: 1.15e-02, grad_scale: 8.0 2023-04-28 11:36:05,329 INFO [optim.py:368] (1/8) Clipping_scale=2.0, grad-norm quartiles 1.856e+02 2.871e+02 3.468e+02 4.191e+02 6.688e+02, threshold=6.936e+02, percent-clipped=0.0 2023-04-28 11:36:46,918 INFO [train.py:904] (1/8) Epoch 6, batch 3450, loss[loss=0.2003, simple_loss=0.2714, pruned_loss=0.06455, over 16901.00 frames. ], tot_loss[loss=0.2052, simple_loss=0.2845, pruned_loss=0.063, over 3320986.21 frames. ], batch size: 96, lr: 1.15e-02, grad_scale: 8.0 2023-04-28 11:37:06,305 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.4523, 1.5534, 2.0553, 2.4676, 2.5947, 2.5433, 1.6178, 2.5082], device='cuda:1'), covar=tensor([0.0080, 0.0224, 0.0148, 0.0107, 0.0092, 0.0117, 0.0203, 0.0077], device='cuda:1'), in_proj_covar=tensor([0.0135, 0.0153, 0.0137, 0.0137, 0.0142, 0.0104, 0.0145, 0.0091], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001], device='cuda:1') 2023-04-28 11:37:48,857 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.81 vs. limit=2.0 2023-04-28 11:37:58,612 INFO [train.py:904] (1/8) Epoch 6, batch 3500, loss[loss=0.2167, simple_loss=0.302, pruned_loss=0.06572, over 16779.00 frames. ], tot_loss[loss=0.2038, simple_loss=0.2836, pruned_loss=0.06197, over 3334506.24 frames. ], batch size: 57, lr: 1.15e-02, grad_scale: 8.0 2023-04-28 11:38:28,902 INFO [optim.py:368] (1/8) Clipping_scale=2.0, grad-norm quartiles 2.000e+02 2.700e+02 3.164e+02 3.995e+02 8.205e+02, threshold=6.329e+02, percent-clipped=2.0 2023-04-28 11:39:10,636 INFO [train.py:904] (1/8) Epoch 6, batch 3550, loss[loss=0.2021, simple_loss=0.2805, pruned_loss=0.06186, over 17215.00 frames. ], tot_loss[loss=0.2039, simple_loss=0.2832, pruned_loss=0.06225, over 3331241.90 frames. ], batch size: 44, lr: 1.15e-02, grad_scale: 8.0 2023-04-28 11:39:57,262 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.23 vs. limit=2.0 2023-04-28 11:40:21,754 INFO [train.py:904] (1/8) Epoch 6, batch 3600, loss[loss=0.2268, simple_loss=0.2864, pruned_loss=0.08357, over 16836.00 frames. ], tot_loss[loss=0.2027, simple_loss=0.2818, pruned_loss=0.06185, over 3332809.24 frames. ], batch size: 116, lr: 1.15e-02, grad_scale: 8.0 2023-04-28 11:40:31,662 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.68 vs. limit=2.0 2023-04-28 11:40:51,130 INFO [optim.py:368] (1/8) 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] (1/8) Epoch 6, batch 3650, loss[loss=0.1755, simple_loss=0.2414, pruned_loss=0.05484, over 16650.00 frames. ], tot_loss[loss=0.2037, simple_loss=0.2808, pruned_loss=0.06324, over 3325644.44 frames. ], batch size: 89, lr: 1.15e-02, grad_scale: 8.0 2023-04-28 11:42:46,526 INFO [train.py:904] (1/8) Epoch 6, batch 3700, loss[loss=0.1887, simple_loss=0.2584, pruned_loss=0.05951, over 16674.00 frames. ], tot_loss[loss=0.2048, simple_loss=0.2799, pruned_loss=0.06486, over 3291466.66 frames. ], batch size: 89, lr: 1.14e-02, grad_scale: 8.0 2023-04-28 11:43:17,520 INFO [optim.py:368] (1/8) 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 11:43:59,568 INFO [train.py:904] (1/8) Epoch 6, batch 3750, loss[loss=0.2281, simple_loss=0.283, pruned_loss=0.08666, over 16687.00 frames. ], tot_loss[loss=0.2068, simple_loss=0.2804, pruned_loss=0.06653, over 3272334.57 frames. ], batch size: 134, lr: 1.14e-02, grad_scale: 8.0 2023-04-28 11:44:45,985 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.93 vs. limit=2.0 2023-04-28 11:44:49,706 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.3971, 3.8856, 3.2781, 2.0107, 2.9217, 2.4167, 3.7657, 3.6746], device='cuda:1'), covar=tensor([0.0195, 0.0482, 0.0600, 0.1483, 0.0651, 0.0875, 0.0504, 0.0699], device='cuda:1'), in_proj_covar=tensor([0.0142, 0.0136, 0.0154, 0.0140, 0.0131, 0.0124, 0.0138, 0.0145], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-04-28 11:44:56,282 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.41 vs. limit=2.0 2023-04-28 11:45:13,202 INFO [train.py:904] (1/8) Epoch 6, batch 3800, loss[loss=0.2227, simple_loss=0.2918, pruned_loss=0.07682, over 16821.00 frames. ], tot_loss[loss=0.2096, simple_loss=0.2827, pruned_loss=0.06825, over 3268283.35 frames. ], batch size: 102, lr: 1.14e-02, grad_scale: 8.0 2023-04-28 11:45:46,019 INFO [optim.py:368] (1/8) 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 11:46:03,897 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.3864, 4.0450, 3.9520, 4.4484, 4.6086, 4.2188, 4.3385, 4.5602], device='cuda:1'), covar=tensor([0.0829, 0.0892, 0.1954, 0.0856, 0.0605, 0.0935, 0.1474, 0.0739], device='cuda:1'), in_proj_covar=tensor([0.0427, 0.0524, 0.0671, 0.0543, 0.0403, 0.0400, 0.0421, 0.0456], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-28 11:46:28,643 INFO [train.py:904] (1/8) Epoch 6, batch 3850, loss[loss=0.201, simple_loss=0.2764, pruned_loss=0.0628, over 17024.00 frames. ], tot_loss[loss=0.2095, simple_loss=0.2823, pruned_loss=0.0684, over 3273825.92 frames. ], batch size: 53, lr: 1.14e-02, grad_scale: 8.0 2023-04-28 11:47:40,654 INFO [train.py:904] (1/8) Epoch 6, batch 3900, loss[loss=0.1903, simple_loss=0.2637, pruned_loss=0.05847, over 16748.00 frames. ], tot_loss[loss=0.2099, simple_loss=0.2817, pruned_loss=0.06904, over 3280980.95 frames. ], batch size: 102, lr: 1.14e-02, grad_scale: 8.0 2023-04-28 11:47:59,688 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=54664.0, num_to_drop=1, layers_to_drop={1} 2023-04-28 11:48:10,861 INFO [optim.py:368] (1/8) 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 11:48:53,476 INFO [train.py:904] (1/8) Epoch 6, batch 3950, loss[loss=0.2155, simple_loss=0.2844, pruned_loss=0.07332, over 15441.00 frames. ], tot_loss[loss=0.2102, simple_loss=0.2812, pruned_loss=0.06962, over 3279552.90 frames. ], batch size: 190, lr: 1.14e-02, grad_scale: 8.0 2023-04-28 11:49:27,755 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=54725.0, num_to_drop=1, layers_to_drop={2} 2023-04-28 11:50:04,727 INFO [train.py:904] (1/8) Epoch 6, batch 4000, loss[loss=0.1828, simple_loss=0.2587, pruned_loss=0.05345, over 16996.00 frames. ], tot_loss[loss=0.2104, simple_loss=0.281, pruned_loss=0.06985, over 3284120.29 frames. ], batch size: 41, lr: 1.14e-02, grad_scale: 8.0 2023-04-28 11:50:36,654 INFO [optim.py:368] (1/8) 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 11:51:17,347 INFO [train.py:904] (1/8) Epoch 6, batch 4050, loss[loss=0.1996, simple_loss=0.2763, pruned_loss=0.06141, over 16640.00 frames. ], tot_loss[loss=0.2085, simple_loss=0.2805, pruned_loss=0.06823, over 3287618.16 frames. ], batch size: 62, lr: 1.14e-02, grad_scale: 4.0 2023-04-28 11:51:42,570 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.26 vs. limit=2.0 2023-04-28 11:52:22,411 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.8747, 4.1585, 3.2511, 2.4929, 3.1211, 2.6199, 4.4218, 4.2374], device='cuda:1'), covar=tensor([0.2076, 0.0593, 0.1282, 0.1554, 0.2062, 0.1282, 0.0365, 0.0554], device='cuda:1'), in_proj_covar=tensor([0.0282, 0.0254, 0.0274, 0.0256, 0.0299, 0.0208, 0.0252, 0.0278], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003], device='cuda:1') 2023-04-28 11:52:28,288 INFO [train.py:904] (1/8) Epoch 6, batch 4100, loss[loss=0.2242, simple_loss=0.3045, pruned_loss=0.07193, over 16846.00 frames. ], tot_loss[loss=0.2076, simple_loss=0.2813, pruned_loss=0.06692, over 3274324.23 frames. ], batch size: 96, lr: 1.14e-02, grad_scale: 4.0 2023-04-28 11:52:45,818 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=54862.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 11:52:59,606 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.37 vs. limit=2.0 2023-04-28 11:53:01,852 INFO [optim.py:368] (1/8) Clipping_scale=2.0, grad-norm quartiles 1.687e+02 2.498e+02 3.136e+02 3.951e+02 8.635e+02, threshold=6.272e+02, percent-clipped=2.0 2023-04-28 11:53:45,439 INFO [train.py:904] (1/8) Epoch 6, batch 4150, loss[loss=0.2161, simple_loss=0.3025, pruned_loss=0.06484, over 16875.00 frames. ], tot_loss[loss=0.2155, simple_loss=0.2897, pruned_loss=0.07069, over 3237838.19 frames. ], batch size: 109, lr: 1.14e-02, grad_scale: 4.0 2023-04-28 11:54:19,552 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=54923.0, num_to_drop=1, layers_to_drop={1} 2023-04-28 11:54:56,495 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=4.69 vs. limit=5.0 2023-04-28 11:55:01,101 INFO [train.py:904] (1/8) Epoch 6, batch 4200, loss[loss=0.2312, simple_loss=0.3247, pruned_loss=0.06879, over 16707.00 frames. ], tot_loss[loss=0.2218, simple_loss=0.2974, pruned_loss=0.07305, over 3206807.35 frames. ], batch size: 89, lr: 1.14e-02, grad_scale: 4.0 2023-04-28 11:55:33,715 INFO [optim.py:368] (1/8) 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 11:56:00,080 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([1.8005, 2.7564, 2.6135, 1.8993, 2.5180, 2.6596, 2.5541, 1.7640], device='cuda:1'), covar=tensor([0.0280, 0.0036, 0.0040, 0.0219, 0.0048, 0.0046, 0.0045, 0.0240], device='cuda:1'), in_proj_covar=tensor([0.0117, 0.0057, 0.0060, 0.0114, 0.0062, 0.0070, 0.0064, 0.0106], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:1') 2023-04-28 11:56:15,284 INFO [train.py:904] (1/8) Epoch 6, batch 4250, loss[loss=0.2087, simple_loss=0.2908, pruned_loss=0.06329, over 16745.00 frames. ], tot_loss[loss=0.2243, simple_loss=0.3011, pruned_loss=0.07375, over 3176754.33 frames. ], batch size: 124, lr: 1.14e-02, grad_scale: 4.0 2023-04-28 11:56:44,000 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=55020.0, num_to_drop=1, layers_to_drop={2} 2023-04-28 11:57:29,394 INFO [train.py:904] (1/8) Epoch 6, batch 4300, loss[loss=0.2523, simple_loss=0.3245, pruned_loss=0.09003, over 11523.00 frames. ], tot_loss[loss=0.2233, simple_loss=0.3018, pruned_loss=0.07234, over 3173705.02 frames. ], batch size: 247, lr: 1.14e-02, grad_scale: 4.0 2023-04-28 11:57:49,430 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.64 vs. limit=2.0 2023-04-28 11:57:51,872 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.47 vs. limit=2.0 2023-04-28 11:58:01,800 INFO [optim.py:368] (1/8) 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 11:58:12,005 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.51 vs. limit=2.0 2023-04-28 11:58:43,132 INFO [train.py:904] (1/8) Epoch 6, batch 4350, loss[loss=0.2343, simple_loss=0.3184, pruned_loss=0.07511, over 16774.00 frames. ], tot_loss[loss=0.2259, simple_loss=0.3053, pruned_loss=0.07332, over 3179690.54 frames. ], batch size: 39, lr: 1.14e-02, grad_scale: 4.0 2023-04-28 11:59:28,383 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.5083, 4.1703, 3.9107, 1.9692, 3.0569, 2.6324, 3.9715, 3.8598], device='cuda:1'), covar=tensor([0.0208, 0.0445, 0.0433, 0.1594, 0.0689, 0.0805, 0.0561, 0.0827], device='cuda:1'), in_proj_covar=tensor([0.0141, 0.0133, 0.0153, 0.0139, 0.0132, 0.0125, 0.0139, 0.0144], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-04-28 11:59:39,484 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.7829, 3.4999, 3.1743, 1.7725, 2.6083, 2.1817, 3.3213, 3.3776], device='cuda:1'), covar=tensor([0.0255, 0.0516, 0.0505, 0.1658, 0.0787, 0.0887, 0.0660, 0.0862], device='cuda:1'), in_proj_covar=tensor([0.0141, 0.0133, 0.0154, 0.0140, 0.0132, 0.0125, 0.0139, 0.0144], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-04-28 11:59:53,027 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=55148.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 11:59:56,586 INFO [train.py:904] (1/8) Epoch 6, batch 4400, loss[loss=0.2455, simple_loss=0.3293, pruned_loss=0.0808, over 17018.00 frames. ], tot_loss[loss=0.2275, simple_loss=0.3072, pruned_loss=0.07392, over 3190415.23 frames. ], batch size: 55, lr: 1.14e-02, grad_scale: 8.0 2023-04-28 12:00:27,147 INFO [optim.py:368] (1/8) 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 12:01:06,617 INFO [train.py:904] (1/8) Epoch 6, batch 4450, loss[loss=0.2211, simple_loss=0.2989, pruned_loss=0.07169, over 16372.00 frames. ], tot_loss[loss=0.2297, simple_loss=0.3102, pruned_loss=0.07461, over 3187145.33 frames. ], batch size: 35, lr: 1.14e-02, grad_scale: 8.0 2023-04-28 12:01:14,499 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.6440, 2.7286, 1.6569, 2.8186, 2.1708, 2.8047, 1.7312, 2.2671], device='cuda:1'), covar=tensor([0.0189, 0.0390, 0.1418, 0.0077, 0.0643, 0.0460, 0.1340, 0.0621], device='cuda:1'), in_proj_covar=tensor([0.0125, 0.0158, 0.0179, 0.0083, 0.0162, 0.0189, 0.0183, 0.0162], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-04-28 12:01:19,115 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.8066, 5.4598, 5.5312, 5.3484, 5.4725, 6.0498, 5.5064, 5.3239], device='cuda:1'), covar=tensor([0.0736, 0.1322, 0.1116, 0.1725, 0.2240, 0.0846, 0.0909, 0.1903], device='cuda:1'), in_proj_covar=tensor([0.0294, 0.0406, 0.0402, 0.0353, 0.0467, 0.0428, 0.0328, 0.0472], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-04-28 12:01:19,265 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=55209.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 12:01:21,583 INFO [zipformer.py:625] (1/8) 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] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=55218.0, num_to_drop=1, layers_to_drop={3} 2023-04-28 12:02:16,151 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([5.4926, 5.7870, 5.5147, 5.6754, 5.1298, 4.8699, 5.3694, 6.0074], device='cuda:1'), covar=tensor([0.0648, 0.0660, 0.0859, 0.0450, 0.0606, 0.0593, 0.0553, 0.0543], device='cuda:1'), in_proj_covar=tensor([0.0391, 0.0523, 0.0439, 0.0335, 0.0320, 0.0335, 0.0426, 0.0368], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-28 12:02:16,913 INFO [train.py:904] (1/8) Epoch 6, batch 4500, loss[loss=0.2196, simple_loss=0.2953, pruned_loss=0.07198, over 17143.00 frames. ], tot_loss[loss=0.2292, simple_loss=0.3096, pruned_loss=0.07437, over 3188660.91 frames. ], batch size: 49, lr: 1.14e-02, grad_scale: 8.0 2023-04-28 12:02:47,093 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=55272.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 12:02:48,411 INFO [optim.py:368] (1/8) Clipping_scale=2.0, grad-norm quartiles 1.800e+02 2.322e+02 2.860e+02 3.458e+02 5.535e+02, threshold=5.719e+02, percent-clipped=0.0 2023-04-28 12:03:29,864 INFO [train.py:904] (1/8) Epoch 6, batch 4550, loss[loss=0.2488, simple_loss=0.3366, pruned_loss=0.08052, over 16875.00 frames. ], tot_loss[loss=0.2305, simple_loss=0.3105, pruned_loss=0.07521, over 3199459.17 frames. ], batch size: 96, lr: 1.14e-02, grad_scale: 8.0 2023-04-28 12:03:57,376 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 6, batch 4600, loss[loss=0.2531, simple_loss=0.3188, pruned_loss=0.09368, over 11671.00 frames. ], tot_loss[loss=0.2307, simple_loss=0.3113, pruned_loss=0.07504, over 3199781.98 frames. ], batch size: 247, lr: 1.14e-02, grad_scale: 8.0 2023-04-28 12:05:07,045 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=55368.0, num_to_drop=1, layers_to_drop={1} 2023-04-28 12:05:13,321 INFO [optim.py:368] (1/8) 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 12:05:51,981 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.5887, 2.3888, 1.5897, 2.1359, 2.9483, 2.5093, 3.3873, 3.2679], device='cuda:1'), covar=tensor([0.0023, 0.0202, 0.0351, 0.0262, 0.0116, 0.0215, 0.0085, 0.0095], device='cuda:1'), in_proj_covar=tensor([0.0089, 0.0160, 0.0162, 0.0161, 0.0157, 0.0164, 0.0146, 0.0147], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-28 12:05:52,743 INFO [train.py:904] (1/8) Epoch 6, batch 4650, loss[loss=0.2433, simple_loss=0.3272, pruned_loss=0.0797, over 17036.00 frames. ], tot_loss[loss=0.229, simple_loss=0.3096, pruned_loss=0.07422, over 3205639.32 frames. ], batch size: 50, lr: 1.14e-02, grad_scale: 8.0 2023-04-28 12:05:56,576 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.20 vs. limit=2.0 2023-04-28 12:07:03,016 INFO [train.py:904] (1/8) Epoch 6, batch 4700, loss[loss=0.2001, simple_loss=0.2834, pruned_loss=0.05845, over 16497.00 frames. ], tot_loss[loss=0.2267, simple_loss=0.3069, pruned_loss=0.07323, over 3204832.59 frames. ], batch size: 68, lr: 1.13e-02, grad_scale: 8.0 2023-04-28 12:07:34,099 INFO [optim.py:368] (1/8) 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] (1/8) Epoch 6, batch 4750, loss[loss=0.2154, simple_loss=0.2887, pruned_loss=0.07104, over 16367.00 frames. ], tot_loss[loss=0.2213, simple_loss=0.3019, pruned_loss=0.07036, over 3226503.23 frames. ], batch size: 146, lr: 1.13e-02, grad_scale: 8.0 2023-04-28 12:08:17,376 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=55504.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 12:08:19,183 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=2.01 vs. limit=2.0 2023-04-28 12:08:36,487 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=55518.0, num_to_drop=1, layers_to_drop={0} 2023-04-28 12:08:45,773 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.29 vs. limit=2.0 2023-04-28 12:09:22,945 INFO [train.py:904] (1/8) Epoch 6, batch 4800, loss[loss=0.2379, simple_loss=0.3026, pruned_loss=0.08657, over 12190.00 frames. ], tot_loss[loss=0.2194, simple_loss=0.2998, pruned_loss=0.06948, over 3195632.49 frames. ], batch size: 248, lr: 1.13e-02, grad_scale: 8.0 2023-04-28 12:09:45,464 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=55566.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 12:09:46,593 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=55567.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 12:09:53,989 INFO [optim.py:368] (1/8) Clipping_scale=2.0, grad-norm quartiles 1.426e+02 2.469e+02 2.878e+02 3.568e+02 5.612e+02, threshold=5.756e+02, percent-clipped=0.0 2023-04-28 12:10:35,249 INFO [train.py:904] (1/8) Epoch 6, batch 4850, loss[loss=0.2342, simple_loss=0.3213, pruned_loss=0.07358, over 15473.00 frames. ], tot_loss[loss=0.2194, simple_loss=0.3004, pruned_loss=0.06917, over 3175149.42 frames. ], batch size: 190, lr: 1.13e-02, grad_scale: 8.0 2023-04-28 12:11:13,998 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.76 vs. limit=2.0 2023-04-28 12:11:48,113 INFO [train.py:904] (1/8) Epoch 6, batch 4900, loss[loss=0.2075, simple_loss=0.2993, pruned_loss=0.05784, over 16248.00 frames. ], tot_loss[loss=0.2172, simple_loss=0.2993, pruned_loss=0.06757, over 3172737.84 frames. ], batch size: 165, lr: 1.13e-02, grad_scale: 8.0 2023-04-28 12:12:19,090 INFO [optim.py:368] (1/8) 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 12:12:42,218 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=3.13 vs. limit=5.0 2023-04-28 12:12:59,386 INFO [train.py:904] (1/8) Epoch 6, batch 4950, loss[loss=0.1969, simple_loss=0.2843, pruned_loss=0.05481, over 16414.00 frames. ], tot_loss[loss=0.2167, simple_loss=0.2994, pruned_loss=0.06703, over 3179665.74 frames. ], batch size: 68, lr: 1.13e-02, grad_scale: 8.0 2023-04-28 12:14:05,694 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 6, batch 5000, loss[loss=0.206, simple_loss=0.2913, pruned_loss=0.06034, over 17110.00 frames. ], tot_loss[loss=0.2179, simple_loss=0.3009, pruned_loss=0.06748, over 3188798.79 frames. ], batch size: 49, lr: 1.13e-02, grad_scale: 8.0 2023-04-28 12:14:26,492 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.3105, 3.5589, 2.4303, 2.0528, 2.6340, 1.9197, 3.4979, 3.5781], device='cuda:1'), covar=tensor([0.2399, 0.0694, 0.1571, 0.1704, 0.1810, 0.1751, 0.0544, 0.0670], device='cuda:1'), in_proj_covar=tensor([0.0281, 0.0251, 0.0270, 0.0250, 0.0283, 0.0201, 0.0247, 0.0263], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-28 12:14:38,593 INFO [optim.py:368] (1/8) 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 12:15:21,196 INFO [train.py:904] (1/8) Epoch 6, batch 5050, loss[loss=0.2367, simple_loss=0.3154, pruned_loss=0.07899, over 16693.00 frames. ], tot_loss[loss=0.2178, simple_loss=0.3013, pruned_loss=0.06715, over 3195661.70 frames. ], batch size: 62, lr: 1.13e-02, grad_scale: 8.0 2023-04-28 12:15:25,492 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=55804.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 12:15:31,728 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.21 vs. limit=2.0 2023-04-28 12:15:33,281 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=55809.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 12:16:00,078 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.7626, 2.1221, 2.3470, 4.3008, 1.8621, 2.7581, 2.2443, 2.2990], device='cuda:1'), covar=tensor([0.0615, 0.2499, 0.1297, 0.0325, 0.3298, 0.1529, 0.2244, 0.2588], device='cuda:1'), in_proj_covar=tensor([0.0329, 0.0338, 0.0278, 0.0313, 0.0379, 0.0350, 0.0304, 0.0400], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-28 12:16:28,623 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=55848.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 12:16:31,559 INFO [train.py:904] (1/8) Epoch 6, batch 5100, loss[loss=0.2036, simple_loss=0.2836, pruned_loss=0.06184, over 16590.00 frames. ], tot_loss[loss=0.2161, simple_loss=0.2995, pruned_loss=0.06631, over 3192883.39 frames. ], batch size: 57, lr: 1.13e-02, grad_scale: 8.0 2023-04-28 12:16:32,963 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=55852.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 12:16:54,835 INFO [zipformer.py:625] (1/8) 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] (1/8) 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] (1/8) Epoch 6, batch 5150, loss[loss=0.2168, simple_loss=0.3074, pruned_loss=0.06307, over 16872.00 frames. ], tot_loss[loss=0.2144, simple_loss=0.2986, pruned_loss=0.06514, over 3202260.85 frames. ], batch size: 96, lr: 1.13e-02, grad_scale: 8.0 2023-04-28 12:17:57,070 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=55909.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 12:18:05,650 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=55915.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 12:18:31,892 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.89 vs. limit=2.0 2023-04-28 12:18:58,946 INFO [train.py:904] (1/8) Epoch 6, batch 5200, loss[loss=0.1999, simple_loss=0.2847, pruned_loss=0.05757, over 16669.00 frames. ], tot_loss[loss=0.2133, simple_loss=0.2974, pruned_loss=0.0646, over 3193355.60 frames. ], batch size: 89, lr: 1.13e-02, grad_scale: 8.0 2023-04-28 12:19:30,989 INFO [optim.py:368] (1/8) 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] (1/8) Epoch 6, batch 5250, loss[loss=0.2304, simple_loss=0.306, pruned_loss=0.07741, over 16916.00 frames. ], tot_loss[loss=0.2108, simple_loss=0.2942, pruned_loss=0.06369, over 3209594.84 frames. ], batch size: 109, lr: 1.13e-02, grad_scale: 8.0 2023-04-28 12:20:17,624 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.9451, 1.6565, 2.3399, 3.0163, 2.7361, 3.3787, 1.9359, 3.2814], device='cuda:1'), covar=tensor([0.0075, 0.0243, 0.0142, 0.0104, 0.0107, 0.0045, 0.0215, 0.0030], device='cuda:1'), in_proj_covar=tensor([0.0127, 0.0148, 0.0128, 0.0131, 0.0135, 0.0096, 0.0141, 0.0085], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001], device='cuda:1') 2023-04-28 12:21:28,014 INFO [train.py:904] (1/8) Epoch 6, batch 5300, loss[loss=0.175, simple_loss=0.2538, pruned_loss=0.04807, over 17110.00 frames. ], tot_loss[loss=0.2082, simple_loss=0.2912, pruned_loss=0.06261, over 3212692.28 frames. ], batch size: 49, lr: 1.13e-02, grad_scale: 8.0 2023-04-28 12:21:59,804 INFO [optim.py:368] (1/8) 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 12:22:27,493 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.1938, 1.8492, 2.1260, 3.7963, 1.7204, 2.5480, 2.0654, 2.0068], device='cuda:1'), covar=tensor([0.0773, 0.2647, 0.1336, 0.0355, 0.3440, 0.1447, 0.2303, 0.2491], device='cuda:1'), in_proj_covar=tensor([0.0328, 0.0335, 0.0276, 0.0311, 0.0378, 0.0348, 0.0303, 0.0395], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-28 12:22:40,508 INFO [train.py:904] (1/8) Epoch 6, batch 5350, loss[loss=0.2388, simple_loss=0.3226, pruned_loss=0.07745, over 15333.00 frames. ], tot_loss[loss=0.2075, simple_loss=0.29, pruned_loss=0.06251, over 3201173.75 frames. ], batch size: 190, lr: 1.13e-02, grad_scale: 8.0 2023-04-28 12:22:45,575 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=56104.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 12:23:44,977 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=56145.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 12:23:52,598 INFO [train.py:904] (1/8) Epoch 6, batch 5400, loss[loss=0.2226, simple_loss=0.3106, pruned_loss=0.06726, over 16762.00 frames. ], tot_loss[loss=0.2106, simple_loss=0.2936, pruned_loss=0.06382, over 3213896.91 frames. ], batch size: 124, lr: 1.13e-02, grad_scale: 4.0 2023-04-28 12:24:26,635 INFO [optim.py:368] (1/8) 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 12:25:08,991 INFO [train.py:904] (1/8) Epoch 6, batch 5450, loss[loss=0.2207, simple_loss=0.2946, pruned_loss=0.07341, over 12055.00 frames. ], tot_loss[loss=0.2135, simple_loss=0.2963, pruned_loss=0.06534, over 3192133.33 frames. ], batch size: 247, lr: 1.13e-02, grad_scale: 4.0 2023-04-28 12:25:14,383 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=56204.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 12:25:17,606 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=56206.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 12:26:08,022 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.95 vs. limit=2.0 2023-04-28 12:26:18,061 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.3353, 3.1838, 2.6363, 2.1690, 2.3092, 2.1315, 3.2095, 3.3505], device='cuda:1'), covar=tensor([0.2216, 0.0737, 0.1287, 0.1681, 0.2217, 0.1493, 0.0486, 0.0741], device='cuda:1'), in_proj_covar=tensor([0.0278, 0.0247, 0.0266, 0.0249, 0.0280, 0.0201, 0.0246, 0.0258], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-28 12:26:27,336 INFO [train.py:904] (1/8) Epoch 6, batch 5500, loss[loss=0.2419, simple_loss=0.3211, pruned_loss=0.08132, over 16647.00 frames. ], tot_loss[loss=0.2245, simple_loss=0.3053, pruned_loss=0.07185, over 3169564.57 frames. ], batch size: 57, lr: 1.13e-02, grad_scale: 4.0 2023-04-28 12:26:50,253 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.1429, 1.6476, 2.4724, 3.1346, 2.9130, 3.3541, 1.7752, 3.1459], device='cuda:1'), covar=tensor([0.0081, 0.0237, 0.0147, 0.0087, 0.0099, 0.0058, 0.0218, 0.0049], device='cuda:1'), in_proj_covar=tensor([0.0126, 0.0148, 0.0129, 0.0131, 0.0136, 0.0097, 0.0142, 0.0088], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001], device='cuda:1') 2023-04-28 12:27:01,683 INFO [optim.py:368] (1/8) 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] (1/8) Epoch 6, batch 5550, loss[loss=0.2546, simple_loss=0.3264, pruned_loss=0.09137, over 17006.00 frames. ], tot_loss[loss=0.2346, simple_loss=0.3131, pruned_loss=0.07804, over 3152348.74 frames. ], batch size: 41, lr: 1.13e-02, grad_scale: 4.0 2023-04-28 12:29:07,880 INFO [train.py:904] (1/8) Epoch 6, batch 5600, loss[loss=0.2556, simple_loss=0.3276, pruned_loss=0.09181, over 16843.00 frames. ], tot_loss[loss=0.2456, simple_loss=0.3209, pruned_loss=0.08515, over 3117613.47 frames. ], batch size: 116, lr: 1.13e-02, grad_scale: 8.0 2023-04-28 12:29:45,253 INFO [optim.py:368] (1/8) 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 12:30:25,650 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.2168, 3.6176, 3.7596, 1.7496, 3.9865, 3.9425, 2.7618, 2.8220], device='cuda:1'), covar=tensor([0.0913, 0.0147, 0.0140, 0.1153, 0.0048, 0.0075, 0.0383, 0.0446], device='cuda:1'), in_proj_covar=tensor([0.0142, 0.0093, 0.0079, 0.0139, 0.0071, 0.0080, 0.0115, 0.0127], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003, 0.0003], device='cuda:1') 2023-04-28 12:30:29,969 INFO [train.py:904] (1/8) Epoch 6, batch 5650, loss[loss=0.2341, simple_loss=0.3168, pruned_loss=0.07569, over 16482.00 frames. ], tot_loss[loss=0.2533, simple_loss=0.3271, pruned_loss=0.0898, over 3117540.49 frames. ], batch size: 75, lr: 1.13e-02, grad_scale: 8.0 2023-04-28 12:30:34,823 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=56404.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 12:31:37,824 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.22 vs. limit=2.0 2023-04-28 12:31:43,010 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.54 vs. limit=2.0 2023-04-28 12:31:48,511 INFO [train.py:904] (1/8) Epoch 6, batch 5700, loss[loss=0.2815, simple_loss=0.3541, pruned_loss=0.1045, over 16326.00 frames. ], tot_loss[loss=0.2547, simple_loss=0.3277, pruned_loss=0.09088, over 3099788.69 frames. ], batch size: 165, lr: 1.12e-02, grad_scale: 8.0 2023-04-28 12:31:51,183 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=56452.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 12:32:25,511 INFO [optim.py:368] (1/8) Clipping_scale=2.0, grad-norm quartiles 2.692e+02 4.150e+02 5.062e+02 6.862e+02 1.724e+03, threshold=1.012e+03, percent-clipped=2.0 2023-04-28 12:33:08,899 INFO [train.py:904] (1/8) Epoch 6, batch 5750, loss[loss=0.2513, simple_loss=0.3253, pruned_loss=0.08865, over 15224.00 frames. ], tot_loss[loss=0.2588, simple_loss=0.3313, pruned_loss=0.09309, over 3064826.20 frames. ], batch size: 190, lr: 1.12e-02, grad_scale: 8.0 2023-04-28 12:33:09,307 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=56501.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 12:33:13,908 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=56504.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 12:33:44,533 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=56523.0, num_to_drop=1, layers_to_drop={0} 2023-04-28 12:34:19,135 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.6418, 2.0727, 1.6738, 1.9410, 2.5192, 2.2548, 2.6571, 2.6911], device='cuda:1'), covar=tensor([0.0050, 0.0187, 0.0243, 0.0236, 0.0127, 0.0178, 0.0108, 0.0112], device='cuda:1'), in_proj_covar=tensor([0.0086, 0.0159, 0.0163, 0.0160, 0.0155, 0.0165, 0.0145, 0.0144], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-28 12:34:30,329 INFO [train.py:904] (1/8) Epoch 6, batch 5800, loss[loss=0.2425, simple_loss=0.3247, pruned_loss=0.08012, over 16813.00 frames. ], tot_loss[loss=0.256, simple_loss=0.33, pruned_loss=0.09104, over 3064694.50 frames. ], batch size: 42, lr: 1.12e-02, grad_scale: 8.0 2023-04-28 12:34:32,752 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=56552.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 12:34:39,024 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.9383, 4.9198, 4.8424, 4.7383, 4.3547, 4.8850, 4.8921, 4.5458], device='cuda:1'), covar=tensor([0.0677, 0.0422, 0.0256, 0.0209, 0.0941, 0.0327, 0.0227, 0.0517], device='cuda:1'), in_proj_covar=tensor([0.0191, 0.0209, 0.0224, 0.0197, 0.0252, 0.0226, 0.0157, 0.0252], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-28 12:35:05,968 INFO [optim.py:368] (1/8) 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 12:35:22,937 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=56584.0, num_to_drop=1, layers_to_drop={1} 2023-04-28 12:35:49,149 INFO [train.py:904] (1/8) Epoch 6, batch 5850, loss[loss=0.236, simple_loss=0.3164, pruned_loss=0.0778, over 16687.00 frames. ], tot_loss[loss=0.2528, simple_loss=0.3276, pruned_loss=0.08901, over 3072556.33 frames. ], batch size: 57, lr: 1.12e-02, grad_scale: 8.0 2023-04-28 12:36:30,293 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.8521, 4.8847, 4.7162, 4.5290, 4.2351, 4.7745, 4.6647, 4.4254], device='cuda:1'), covar=tensor([0.0511, 0.0330, 0.0213, 0.0211, 0.0920, 0.0278, 0.0305, 0.0539], device='cuda:1'), in_proj_covar=tensor([0.0192, 0.0212, 0.0226, 0.0200, 0.0255, 0.0229, 0.0160, 0.0256], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002], device='cuda:1') 2023-04-28 12:37:11,644 INFO [train.py:904] (1/8) Epoch 6, batch 5900, loss[loss=0.2296, simple_loss=0.3137, pruned_loss=0.07277, over 16864.00 frames. ], tot_loss[loss=0.2521, simple_loss=0.3267, pruned_loss=0.0887, over 3057869.20 frames. ], batch size: 116, lr: 1.12e-02, grad_scale: 8.0 2023-04-28 12:37:52,136 INFO [optim.py:368] (1/8) 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 12:38:33,941 INFO [train.py:904] (1/8) Epoch 6, batch 5950, loss[loss=0.2274, simple_loss=0.3072, pruned_loss=0.07379, over 17208.00 frames. ], tot_loss[loss=0.2512, simple_loss=0.3276, pruned_loss=0.0874, over 3049931.97 frames. ], batch size: 44, lr: 1.12e-02, grad_scale: 8.0 2023-04-28 12:39:11,074 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.7579, 3.8891, 1.8926, 4.2484, 2.6221, 4.1316, 2.2897, 2.8197], device='cuda:1'), covar=tensor([0.0136, 0.0230, 0.1604, 0.0042, 0.0749, 0.0379, 0.1378, 0.0663], device='cuda:1'), in_proj_covar=tensor([0.0122, 0.0152, 0.0174, 0.0081, 0.0160, 0.0185, 0.0184, 0.0160], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-04-28 12:39:52,035 INFO [train.py:904] (1/8) Epoch 6, batch 6000, loss[loss=0.2168, simple_loss=0.2959, pruned_loss=0.06886, over 16731.00 frames. ], tot_loss[loss=0.2507, simple_loss=0.3268, pruned_loss=0.08732, over 3050352.23 frames. ], batch size: 124, lr: 1.12e-02, grad_scale: 8.0 2023-04-28 12:39:52,035 INFO [train.py:929] (1/8) Computing validation loss 2023-04-28 12:40:01,521 INFO [train.py:938] (1/8) Epoch 6, validation: loss=0.18, simple_loss=0.2922, pruned_loss=0.03386, over 944034.00 frames. 2023-04-28 12:40:01,521 INFO [train.py:939] (1/8) Maximum memory allocated so far is 17676MB 2023-04-28 12:40:36,511 INFO [optim.py:368] (1/8) 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 12:41:19,690 INFO [train.py:904] (1/8) Epoch 6, batch 6050, loss[loss=0.2405, simple_loss=0.3217, pruned_loss=0.07967, over 16645.00 frames. ], tot_loss[loss=0.2488, simple_loss=0.325, pruned_loss=0.08631, over 3058814.25 frames. ], batch size: 68, lr: 1.12e-02, grad_scale: 8.0 2023-04-28 12:41:20,687 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=56801.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 12:41:23,973 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=56803.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 12:42:00,380 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.37 vs. limit=2.0 2023-04-28 12:42:33,596 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 6, batch 6100, loss[loss=0.2325, simple_loss=0.321, pruned_loss=0.07198, over 16780.00 frames. ], tot_loss[loss=0.2464, simple_loss=0.3233, pruned_loss=0.08477, over 3061485.84 frames. ], batch size: 83, lr: 1.12e-02, grad_scale: 8.0 2023-04-28 12:42:56,841 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([1.8319, 3.4795, 2.7936, 5.2209, 4.3358, 4.6482, 1.9417, 2.9814], device='cuda:1'), covar=tensor([0.1243, 0.0476, 0.1067, 0.0076, 0.0371, 0.0285, 0.1168, 0.0837], device='cuda:1'), in_proj_covar=tensor([0.0147, 0.0146, 0.0170, 0.0096, 0.0200, 0.0191, 0.0165, 0.0172], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0003, 0.0004], device='cuda:1') 2023-04-28 12:42:58,724 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=56864.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 12:43:12,453 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.0266, 3.1438, 3.0110, 5.1071, 4.2586, 4.6451, 1.6864, 3.3285], device='cuda:1'), covar=tensor([0.1178, 0.0574, 0.0973, 0.0084, 0.0329, 0.0258, 0.1306, 0.0681], device='cuda:1'), in_proj_covar=tensor([0.0146, 0.0145, 0.0170, 0.0096, 0.0199, 0.0190, 0.0165, 0.0171], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0003, 0.0004], device='cuda:1') 2023-04-28 12:43:13,026 INFO [optim.py:368] (1/8) 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 12:43:20,783 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=56879.0, num_to_drop=1, layers_to_drop={3} 2023-04-28 12:43:46,857 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.3214, 1.5882, 2.6253, 3.2552, 2.9036, 3.4889, 1.6606, 3.3680], device='cuda:1'), covar=tensor([0.0065, 0.0282, 0.0137, 0.0083, 0.0103, 0.0067, 0.0293, 0.0042], device='cuda:1'), in_proj_covar=tensor([0.0124, 0.0148, 0.0130, 0.0129, 0.0133, 0.0099, 0.0144, 0.0086], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001], device='cuda:1') 2023-04-28 12:43:50,670 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.83 vs. limit=2.0 2023-04-28 12:43:56,260 INFO [train.py:904] (1/8) Epoch 6, batch 6150, loss[loss=0.2684, simple_loss=0.3325, pruned_loss=0.1022, over 11562.00 frames. ], tot_loss[loss=0.2444, simple_loss=0.3213, pruned_loss=0.08375, over 3084768.22 frames. ], batch size: 246, lr: 1.12e-02, grad_scale: 8.0 2023-04-28 12:44:08,079 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=56908.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 12:45:12,463 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.2922, 3.9924, 3.6192, 1.8548, 3.0716, 2.3860, 3.7145, 3.8179], device='cuda:1'), covar=tensor([0.0254, 0.0523, 0.0507, 0.1659, 0.0689, 0.0881, 0.0544, 0.0757], device='cuda:1'), in_proj_covar=tensor([0.0139, 0.0126, 0.0152, 0.0139, 0.0132, 0.0124, 0.0138, 0.0138], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-04-28 12:45:17,263 INFO [train.py:904] (1/8) Epoch 6, batch 6200, loss[loss=0.2463, simple_loss=0.3223, pruned_loss=0.08514, over 16715.00 frames. ], tot_loss[loss=0.2428, simple_loss=0.3195, pruned_loss=0.08301, over 3104020.27 frames. ], batch size: 134, lr: 1.12e-02, grad_scale: 8.0 2023-04-28 12:45:45,652 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=56969.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 12:45:53,728 INFO [optim.py:368] (1/8) 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 12:45:59,623 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.3859, 1.9342, 2.1741, 3.9236, 1.9046, 2.6785, 2.1706, 2.0938], device='cuda:1'), covar=tensor([0.0720, 0.2528, 0.1380, 0.0360, 0.3125, 0.1394, 0.2267, 0.2474], device='cuda:1'), in_proj_covar=tensor([0.0327, 0.0339, 0.0278, 0.0315, 0.0385, 0.0349, 0.0304, 0.0399], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-28 12:46:22,659 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=56993.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 12:46:26,556 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=56995.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 12:46:34,327 INFO [train.py:904] (1/8) Epoch 6, batch 6250, loss[loss=0.2654, simple_loss=0.3368, pruned_loss=0.09701, over 16696.00 frames. ], tot_loss[loss=0.2423, simple_loss=0.3191, pruned_loss=0.08276, over 3111265.28 frames. ], batch size: 134, lr: 1.12e-02, grad_scale: 4.0 2023-04-28 12:47:50,596 INFO [train.py:904] (1/8) Epoch 6, batch 6300, loss[loss=0.2669, simple_loss=0.3215, pruned_loss=0.1062, over 11404.00 frames. ], tot_loss[loss=0.2425, simple_loss=0.3197, pruned_loss=0.0827, over 3115274.17 frames. ], batch size: 248, lr: 1.12e-02, grad_scale: 4.0 2023-04-28 12:47:56,058 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=57054.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 12:47:59,211 INFO [zipformer.py:625] (1/8) 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] (1/8) Clipping_scale=2.0, grad-norm quartiles 2.167e+02 3.660e+02 4.420e+02 5.766e+02 1.393e+03, threshold=8.841e+02, percent-clipped=5.0 2023-04-28 12:48:30,425 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([1.8291, 1.6803, 1.4305, 1.5400, 1.8648, 1.6028, 1.7226, 1.8764], device='cuda:1'), covar=tensor([0.0052, 0.0134, 0.0198, 0.0160, 0.0094, 0.0120, 0.0095, 0.0091], device='cuda:1'), in_proj_covar=tensor([0.0086, 0.0160, 0.0163, 0.0161, 0.0156, 0.0165, 0.0148, 0.0144], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-28 12:49:09,382 INFO [train.py:904] (1/8) Epoch 6, batch 6350, loss[loss=0.2416, simple_loss=0.3228, pruned_loss=0.0802, over 16401.00 frames. ], tot_loss[loss=0.2462, simple_loss=0.3221, pruned_loss=0.08513, over 3091417.58 frames. ], batch size: 146, lr: 1.12e-02, grad_scale: 4.0 2023-04-28 12:49:29,354 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.45 vs. limit=2.0 2023-04-28 12:49:51,319 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.49 vs. limit=2.0 2023-04-28 12:50:26,695 INFO [train.py:904] (1/8) Epoch 6, batch 6400, loss[loss=0.254, simple_loss=0.3272, pruned_loss=0.09041, over 16706.00 frames. ], tot_loss[loss=0.2484, simple_loss=0.3231, pruned_loss=0.08687, over 3071483.59 frames. ], batch size: 134, lr: 1.12e-02, grad_scale: 8.0 2023-04-28 12:50:38,655 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=57159.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 12:50:49,404 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.1470, 2.9450, 2.7343, 2.0883, 2.5208, 2.0853, 2.7531, 2.8400], device='cuda:1'), covar=tensor([0.0302, 0.0470, 0.0482, 0.1267, 0.0653, 0.0865, 0.0519, 0.0648], device='cuda:1'), in_proj_covar=tensor([0.0138, 0.0126, 0.0151, 0.0137, 0.0131, 0.0123, 0.0137, 0.0137], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:1') 2023-04-28 12:51:01,322 INFO [optim.py:368] (1/8) 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 12:51:07,811 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=57179.0, num_to_drop=1, layers_to_drop={2} 2023-04-28 12:51:19,355 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=57186.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 12:51:35,097 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([1.8760, 1.6420, 1.4845, 1.5442, 1.8941, 1.6411, 1.7691, 1.9449], device='cuda:1'), covar=tensor([0.0054, 0.0131, 0.0199, 0.0172, 0.0113, 0.0124, 0.0114, 0.0107], device='cuda:1'), in_proj_covar=tensor([0.0086, 0.0158, 0.0162, 0.0160, 0.0154, 0.0164, 0.0147, 0.0143], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-28 12:51:36,916 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=57197.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 12:51:39,296 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.1641, 4.0940, 4.0028, 2.3079, 3.6879, 4.0117, 3.8501, 2.2052], device='cuda:1'), covar=tensor([0.0363, 0.0015, 0.0023, 0.0276, 0.0039, 0.0050, 0.0031, 0.0297], device='cuda:1'), in_proj_covar=tensor([0.0115, 0.0053, 0.0058, 0.0112, 0.0060, 0.0070, 0.0063, 0.0105], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0001, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:1') 2023-04-28 12:51:42,172 INFO [train.py:904] (1/8) Epoch 6, batch 6450, loss[loss=0.2246, simple_loss=0.309, pruned_loss=0.07015, over 16380.00 frames. ], tot_loss[loss=0.2485, simple_loss=0.3236, pruned_loss=0.08673, over 3042352.75 frames. ], batch size: 146, lr: 1.12e-02, grad_scale: 8.0 2023-04-28 12:52:26,739 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=57227.0, num_to_drop=1, layers_to_drop={0} 2023-04-28 12:52:50,330 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.5654, 3.2655, 2.8485, 1.8323, 2.6276, 2.1389, 3.0239, 3.1989], device='cuda:1'), covar=tensor([0.0314, 0.0485, 0.0596, 0.1611, 0.0733, 0.0942, 0.0668, 0.0655], device='cuda:1'), in_proj_covar=tensor([0.0139, 0.0127, 0.0154, 0.0139, 0.0132, 0.0124, 0.0138, 0.0138], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-04-28 12:52:52,973 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=57244.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 12:52:57,226 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 6, batch 6500, loss[loss=0.2166, simple_loss=0.2996, pruned_loss=0.06676, over 16800.00 frames. ], tot_loss[loss=0.2445, simple_loss=0.3198, pruned_loss=0.08456, over 3057633.07 frames. ], batch size: 124, lr: 1.12e-02, grad_scale: 8.0 2023-04-28 12:53:15,514 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=57258.0, num_to_drop=1, layers_to_drop={0} 2023-04-28 12:53:25,363 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=57264.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 12:53:26,748 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=2.63 vs. limit=5.0 2023-04-28 12:53:36,565 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.41 vs. limit=2.0 2023-04-28 12:53:41,777 INFO [optim.py:368] (1/8) 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] (1/8) Epoch 6, batch 6550, loss[loss=0.2251, simple_loss=0.3213, pruned_loss=0.06448, over 16479.00 frames. ], tot_loss[loss=0.2476, simple_loss=0.3234, pruned_loss=0.08588, over 3064494.89 frames. ], batch size: 146, lr: 1.12e-02, grad_scale: 8.0 2023-04-28 12:54:31,695 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=57305.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 12:55:04,890 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.8551, 3.8461, 3.0424, 2.4607, 3.0626, 2.3981, 4.2743, 3.9346], device='cuda:1'), covar=tensor([0.2134, 0.0675, 0.1174, 0.1568, 0.1821, 0.1431, 0.0368, 0.0699], device='cuda:1'), in_proj_covar=tensor([0.0282, 0.0250, 0.0268, 0.0254, 0.0285, 0.0204, 0.0250, 0.0263], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-28 12:55:40,098 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 6, batch 6600, loss[loss=0.265, simple_loss=0.3467, pruned_loss=0.09167, over 16795.00 frames. ], tot_loss[loss=0.2494, simple_loss=0.3259, pruned_loss=0.0864, over 3092040.00 frames. ], batch size: 124, lr: 1.12e-02, grad_scale: 8.0 2023-04-28 12:55:42,555 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=57351.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 12:56:18,766 INFO [optim.py:368] (1/8) 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 12:56:40,097 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=4.72 vs. limit=5.0 2023-04-28 12:56:56,540 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=57398.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 12:57:00,521 INFO [train.py:904] (1/8) Epoch 6, batch 6650, loss[loss=0.236, simple_loss=0.3105, pruned_loss=0.08081, over 15334.00 frames. ], tot_loss[loss=0.2491, simple_loss=0.3254, pruned_loss=0.08638, over 3098767.02 frames. ], batch size: 190, lr: 1.12e-02, grad_scale: 8.0 2023-04-28 12:57:13,278 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([1.8111, 1.6356, 1.4814, 1.4599, 1.8078, 1.5704, 1.6894, 1.8403], device='cuda:1'), covar=tensor([0.0039, 0.0135, 0.0171, 0.0164, 0.0090, 0.0126, 0.0078, 0.0092], device='cuda:1'), in_proj_covar=tensor([0.0086, 0.0160, 0.0163, 0.0162, 0.0156, 0.0166, 0.0148, 0.0144], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-28 12:58:15,912 INFO [train.py:904] (1/8) Epoch 6, batch 6700, loss[loss=0.2507, simple_loss=0.3245, pruned_loss=0.08842, over 16694.00 frames. ], tot_loss[loss=0.2482, simple_loss=0.3237, pruned_loss=0.08636, over 3093491.56 frames. ], batch size: 124, lr: 1.11e-02, grad_scale: 8.0 2023-04-28 12:58:29,452 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=57459.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 12:58:29,559 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=57459.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 12:58:34,595 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=57462.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 12:58:44,666 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.55 vs. limit=2.0 2023-04-28 12:58:54,092 INFO [optim.py:368] (1/8) Clipping_scale=2.0, grad-norm quartiles 2.152e+02 3.718e+02 4.462e+02 5.305e+02 7.989e+02, threshold=8.923e+02, percent-clipped=0.0 2023-04-28 12:59:24,802 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=57495.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 12:59:33,607 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.24 vs. limit=2.0 2023-04-28 12:59:34,110 INFO [train.py:904] (1/8) Epoch 6, batch 6750, loss[loss=0.2985, simple_loss=0.3557, pruned_loss=0.1206, over 12130.00 frames. ], tot_loss[loss=0.2491, simple_loss=0.3236, pruned_loss=0.08726, over 3072708.71 frames. ], batch size: 248, lr: 1.11e-02, grad_scale: 8.0 2023-04-28 12:59:43,067 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=57507.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 13:00:08,002 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=57523.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 13:00:36,584 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=57542.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 13:00:49,718 INFO [train.py:904] (1/8) Epoch 6, batch 6800, loss[loss=0.2592, simple_loss=0.3437, pruned_loss=0.08735, over 16326.00 frames. ], tot_loss[loss=0.2484, simple_loss=0.3233, pruned_loss=0.08669, over 3070459.10 frames. ], batch size: 166, lr: 1.11e-02, grad_scale: 8.0 2023-04-28 13:00:54,406 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=57553.0, num_to_drop=1, layers_to_drop={0} 2023-04-28 13:00:58,233 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=57556.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 13:01:03,123 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.5081, 4.5026, 4.2972, 3.6897, 4.3086, 1.6806, 4.0831, 4.2084], device='cuda:1'), covar=tensor([0.0060, 0.0058, 0.0099, 0.0321, 0.0070, 0.1873, 0.0096, 0.0142], device='cuda:1'), in_proj_covar=tensor([0.0097, 0.0085, 0.0130, 0.0129, 0.0098, 0.0149, 0.0114, 0.0126], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-28 13:01:09,873 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=57564.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 13:01:27,503 INFO [optim.py:368] (1/8) 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 13:02:06,741 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=57600.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 13:02:07,769 INFO [train.py:904] (1/8) Epoch 6, batch 6850, loss[loss=0.2117, simple_loss=0.3176, pruned_loss=0.05286, over 16913.00 frames. ], tot_loss[loss=0.2494, simple_loss=0.3243, pruned_loss=0.08722, over 3062323.33 frames. ], batch size: 96, lr: 1.11e-02, grad_scale: 8.0 2023-04-28 13:02:24,690 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=57612.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 13:03:11,874 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.96 vs. limit=2.0 2023-04-28 13:03:22,307 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=57649.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 13:03:24,933 INFO [train.py:904] (1/8) Epoch 6, batch 6900, loss[loss=0.2583, simple_loss=0.3369, pruned_loss=0.08988, over 16638.00 frames. ], tot_loss[loss=0.2483, simple_loss=0.3255, pruned_loss=0.08553, over 3090904.17 frames. ], batch size: 89, lr: 1.11e-02, grad_scale: 8.0 2023-04-28 13:03:25,350 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=57651.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 13:03:34,764 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.6655, 3.6560, 4.0820, 4.0383, 4.0302, 3.7185, 3.8056, 3.7866], device='cuda:1'), covar=tensor([0.0265, 0.0470, 0.0289, 0.0378, 0.0367, 0.0350, 0.0722, 0.0374], device='cuda:1'), in_proj_covar=tensor([0.0255, 0.0247, 0.0254, 0.0252, 0.0299, 0.0270, 0.0372, 0.0220], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0002], device='cuda:1') 2023-04-28 13:04:02,709 INFO [optim.py:368] (1/8) 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 13:04:37,887 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=57697.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 13:04:40,787 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 6, batch 6950, loss[loss=0.2277, simple_loss=0.3079, pruned_loss=0.07373, over 16572.00 frames. ], tot_loss[loss=0.2504, simple_loss=0.3268, pruned_loss=0.08695, over 3099329.98 frames. ], batch size: 62, lr: 1.11e-02, grad_scale: 8.0 2023-04-28 13:04:48,545 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.7962, 3.9748, 3.1672, 2.4982, 3.0207, 2.4820, 4.2556, 3.9817], device='cuda:1'), covar=tensor([0.2402, 0.0743, 0.1441, 0.1744, 0.2074, 0.1470, 0.0417, 0.0734], device='cuda:1'), in_proj_covar=tensor([0.0279, 0.0248, 0.0266, 0.0250, 0.0279, 0.0200, 0.0245, 0.0257], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-28 13:05:15,042 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([1.6304, 1.2334, 1.5309, 1.5207, 1.6578, 1.8452, 1.3556, 1.6852], device='cuda:1'), covar=tensor([0.0127, 0.0203, 0.0109, 0.0142, 0.0113, 0.0085, 0.0219, 0.0058], device='cuda:1'), in_proj_covar=tensor([0.0124, 0.0148, 0.0130, 0.0129, 0.0134, 0.0099, 0.0145, 0.0086], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001], device='cuda:1') 2023-04-28 13:05:34,187 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=57733.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 13:05:41,140 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.48 vs. limit=2.0 2023-04-28 13:05:42,216 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.8178, 2.2868, 2.4593, 4.6024, 2.0955, 3.0214, 2.3673, 2.4942], device='cuda:1'), covar=tensor([0.0640, 0.2747, 0.1373, 0.0232, 0.3375, 0.1374, 0.2451, 0.2519], device='cuda:1'), in_proj_covar=tensor([0.0327, 0.0338, 0.0280, 0.0314, 0.0387, 0.0352, 0.0305, 0.0400], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-28 13:06:01,329 INFO [train.py:904] (1/8) Epoch 6, batch 7000, loss[loss=0.2747, simple_loss=0.3561, pruned_loss=0.09668, over 16481.00 frames. ], tot_loss[loss=0.2501, simple_loss=0.3272, pruned_loss=0.0865, over 3095182.52 frames. ], batch size: 146, lr: 1.11e-02, grad_scale: 8.0 2023-04-28 13:06:05,921 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=57754.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 13:06:38,242 INFO [optim.py:368] (1/8) 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 13:07:07,541 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 6, batch 7050, loss[loss=0.2751, simple_loss=0.3293, pruned_loss=0.1104, over 11354.00 frames. ], tot_loss[loss=0.2504, simple_loss=0.3281, pruned_loss=0.08633, over 3090525.94 frames. ], batch size: 246, lr: 1.11e-02, grad_scale: 8.0 2023-04-28 13:07:44,521 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=57818.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 13:08:20,759 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 6, batch 7100, loss[loss=0.2485, simple_loss=0.3354, pruned_loss=0.0808, over 16935.00 frames. ], tot_loss[loss=0.2496, simple_loss=0.3268, pruned_loss=0.08621, over 3092533.83 frames. ], batch size: 109, lr: 1.11e-02, grad_scale: 8.0 2023-04-28 13:08:34,651 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=57851.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 13:08:37,897 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=57853.0, num_to_drop=1, layers_to_drop={0} 2023-04-28 13:09:10,524 INFO [optim.py:368] (1/8) 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 13:09:32,900 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=57890.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 13:09:48,235 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 6, batch 7150, loss[loss=0.218, simple_loss=0.3105, pruned_loss=0.06272, over 16920.00 frames. ], tot_loss[loss=0.2463, simple_loss=0.3235, pruned_loss=0.08456, over 3115357.16 frames. ], batch size: 96, lr: 1.11e-02, grad_scale: 8.0 2023-04-28 13:09:49,972 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=57901.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 13:10:23,460 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.5584, 4.0999, 4.1112, 2.7673, 3.6749, 4.0294, 3.8539, 2.4036], device='cuda:1'), covar=tensor([0.0302, 0.0017, 0.0020, 0.0220, 0.0049, 0.0051, 0.0035, 0.0259], device='cuda:1'), in_proj_covar=tensor([0.0119, 0.0056, 0.0059, 0.0116, 0.0063, 0.0074, 0.0065, 0.0108], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0001, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:1') 2023-04-28 13:11:00,229 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 6, batch 7200, loss[loss=0.2418, simple_loss=0.3216, pruned_loss=0.08101, over 16786.00 frames. ], tot_loss[loss=0.2441, simple_loss=0.3213, pruned_loss=0.08342, over 3087093.32 frames. ], batch size: 39, lr: 1.11e-02, grad_scale: 8.0 2023-04-28 13:11:41,799 INFO [optim.py:368] (1/8) 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] (1/8) Epoch 6, batch 7250, loss[loss=0.215, simple_loss=0.2889, pruned_loss=0.07054, over 16712.00 frames. ], tot_loss[loss=0.2402, simple_loss=0.3179, pruned_loss=0.08127, over 3087056.08 frames. ], batch size: 62, lr: 1.11e-02, grad_scale: 8.0 2023-04-28 13:12:42,941 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.5376, 3.2666, 2.7576, 1.6848, 2.5007, 1.9900, 3.0161, 3.2839], device='cuda:1'), covar=tensor([0.0278, 0.0482, 0.0766, 0.1979, 0.1026, 0.1157, 0.0714, 0.0618], device='cuda:1'), in_proj_covar=tensor([0.0137, 0.0127, 0.0153, 0.0139, 0.0132, 0.0125, 0.0137, 0.0137], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-04-28 13:12:58,398 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.7669, 3.7116, 3.8681, 3.7383, 3.8421, 4.1691, 3.9214, 3.7064], device='cuda:1'), covar=tensor([0.1726, 0.1844, 0.1459, 0.1990, 0.2171, 0.1394, 0.1301, 0.2427], device='cuda:1'), in_proj_covar=tensor([0.0293, 0.0397, 0.0402, 0.0350, 0.0457, 0.0433, 0.0331, 0.0472], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-04-28 13:13:04,448 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([1.8786, 1.6881, 1.5242, 1.4608, 1.7792, 1.6571, 1.7339, 1.8819], device='cuda:1'), covar=tensor([0.0049, 0.0145, 0.0197, 0.0208, 0.0105, 0.0140, 0.0095, 0.0108], device='cuda:1'), in_proj_covar=tensor([0.0086, 0.0159, 0.0164, 0.0162, 0.0155, 0.0165, 0.0145, 0.0146], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-28 13:13:44,368 INFO [train.py:904] (1/8) Epoch 6, batch 7300, loss[loss=0.2422, simple_loss=0.3224, pruned_loss=0.08098, over 16773.00 frames. ], tot_loss[loss=0.2393, simple_loss=0.317, pruned_loss=0.08085, over 3084816.48 frames. ], batch size: 124, lr: 1.11e-02, grad_scale: 8.0 2023-04-28 13:13:49,486 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=58054.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 13:13:49,989 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.79 vs. limit=2.0 2023-04-28 13:14:21,941 INFO [optim.py:368] (1/8) 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 13:14:35,106 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.16 vs. limit=2.0 2023-04-28 13:14:43,296 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 6, batch 7350, loss[loss=0.2708, simple_loss=0.3308, pruned_loss=0.1054, over 11042.00 frames. ], tot_loss[loss=0.2395, simple_loss=0.3167, pruned_loss=0.08112, over 3062325.16 frames. ], batch size: 247, lr: 1.11e-02, grad_scale: 8.0 2023-04-28 13:15:03,735 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=58102.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 13:15:28,118 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=58118.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 13:15:40,951 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.83 vs. limit=2.0 2023-04-28 13:16:04,400 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=58141.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 13:16:08,481 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 6, batch 7400, loss[loss=0.257, simple_loss=0.3398, pruned_loss=0.08706, over 16654.00 frames. ], tot_loss[loss=0.2424, simple_loss=0.3192, pruned_loss=0.08279, over 3056245.25 frames. ], batch size: 76, lr: 1.11e-02, grad_scale: 8.0 2023-04-28 13:16:19,778 INFO [zipformer.py:625] (1/8) 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] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=58166.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 13:16:57,162 INFO [optim.py:368] (1/8) 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 13:17:19,634 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.6008, 2.6123, 1.7650, 2.7607, 2.1778, 2.7237, 1.9934, 2.2843], device='cuda:1'), covar=tensor([0.0190, 0.0329, 0.1237, 0.0096, 0.0608, 0.0452, 0.1115, 0.0574], device='cuda:1'), in_proj_covar=tensor([0.0125, 0.0152, 0.0178, 0.0084, 0.0161, 0.0187, 0.0187, 0.0165], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-04-28 13:17:34,282 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=58199.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 13:17:37,740 INFO [train.py:904] (1/8) Epoch 6, batch 7450, loss[loss=0.2706, simple_loss=0.3247, pruned_loss=0.1082, over 11592.00 frames. ], tot_loss[loss=0.2447, simple_loss=0.3209, pruned_loss=0.08432, over 3060143.88 frames. ], batch size: 247, lr: 1.11e-02, grad_scale: 8.0 2023-04-28 13:17:39,909 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=58202.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 13:17:45,735 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=58205.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 13:18:59,285 INFO [train.py:904] (1/8) Epoch 6, batch 7500, loss[loss=0.2218, simple_loss=0.3022, pruned_loss=0.07072, over 16442.00 frames. ], tot_loss[loss=0.2446, simple_loss=0.3216, pruned_loss=0.08379, over 3050528.66 frames. ], batch size: 68, lr: 1.11e-02, grad_scale: 4.0 2023-04-28 13:19:39,439 INFO [optim.py:368] (1/8) 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] (1/8) Epoch 6, batch 7550, loss[loss=0.2822, simple_loss=0.3349, pruned_loss=0.1148, over 11413.00 frames. ], tot_loss[loss=0.2452, simple_loss=0.3213, pruned_loss=0.08453, over 3043499.84 frames. ], batch size: 247, lr: 1.11e-02, grad_scale: 4.0 2023-04-28 13:20:33,903 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.6856, 4.8174, 4.9134, 4.7988, 4.8302, 5.3658, 4.9566, 4.6904], device='cuda:1'), covar=tensor([0.0947, 0.1504, 0.1517, 0.1588, 0.2145, 0.0954, 0.1280, 0.2275], device='cuda:1'), in_proj_covar=tensor([0.0300, 0.0407, 0.0416, 0.0360, 0.0469, 0.0446, 0.0342, 0.0481], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-04-28 13:21:33,625 INFO [train.py:904] (1/8) Epoch 6, batch 7600, loss[loss=0.2346, simple_loss=0.3163, pruned_loss=0.07644, over 16947.00 frames. ], tot_loss[loss=0.2442, simple_loss=0.3202, pruned_loss=0.08412, over 3061991.02 frames. ], batch size: 109, lr: 1.11e-02, grad_scale: 8.0 2023-04-28 13:22:14,309 INFO [optim.py:368] (1/8) 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 13:22:33,777 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=58389.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 13:22:52,233 INFO [train.py:904] (1/8) Epoch 6, batch 7650, loss[loss=0.2925, simple_loss=0.3397, pruned_loss=0.1227, over 10918.00 frames. ], tot_loss[loss=0.2449, simple_loss=0.3208, pruned_loss=0.08446, over 3070954.77 frames. ], batch size: 246, lr: 1.11e-02, grad_scale: 4.0 2023-04-28 13:23:49,834 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=58437.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 13:23:57,327 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.3486, 3.8364, 4.0171, 1.9300, 4.2847, 4.2856, 2.8740, 3.0384], device='cuda:1'), covar=tensor([0.0849, 0.0135, 0.0154, 0.1144, 0.0045, 0.0064, 0.0408, 0.0422], device='cuda:1'), in_proj_covar=tensor([0.0142, 0.0092, 0.0082, 0.0141, 0.0073, 0.0082, 0.0118, 0.0128], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0002, 0.0002, 0.0003, 0.0003], device='cuda:1') 2023-04-28 13:24:02,106 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.1608, 2.3390, 1.8568, 2.1035, 2.6761, 2.4339, 3.0473, 2.9472], device='cuda:1'), covar=tensor([0.0039, 0.0216, 0.0297, 0.0254, 0.0139, 0.0210, 0.0115, 0.0135], device='cuda:1'), in_proj_covar=tensor([0.0085, 0.0157, 0.0163, 0.0161, 0.0154, 0.0164, 0.0145, 0.0144], device='cuda:1'), out_proj_covar=tensor([9.9864e-05, 1.8141e-04, 1.8599e-04, 1.8371e-04, 1.7962e-04, 1.9082e-04, 1.6392e-04, 1.6718e-04], device='cuda:1') 2023-04-28 13:24:11,082 INFO [train.py:904] (1/8) Epoch 6, batch 7700, loss[loss=0.2522, simple_loss=0.334, pruned_loss=0.08521, over 16717.00 frames. ], tot_loss[loss=0.2456, simple_loss=0.321, pruned_loss=0.08511, over 3071237.53 frames. ], batch size: 124, lr: 1.11e-02, grad_scale: 4.0 2023-04-28 13:24:20,100 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.54 vs. limit=2.0 2023-04-28 13:24:51,993 INFO [optim.py:368] (1/8) 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 13:25:23,658 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=58497.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 13:25:28,185 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 6, batch 7750, loss[loss=0.239, simple_loss=0.3195, pruned_loss=0.07923, over 16804.00 frames. ], tot_loss[loss=0.2438, simple_loss=0.3202, pruned_loss=0.08367, over 3103966.48 frames. ], batch size: 83, lr: 1.10e-02, grad_scale: 4.0 2023-04-28 13:25:37,746 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.3532, 3.3940, 1.7221, 3.7184, 2.3085, 3.6486, 1.9598, 2.5614], device='cuda:1'), covar=tensor([0.0175, 0.0342, 0.1654, 0.0076, 0.0868, 0.0537, 0.1459, 0.0710], device='cuda:1'), in_proj_covar=tensor([0.0127, 0.0154, 0.0181, 0.0087, 0.0164, 0.0191, 0.0190, 0.0168], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-04-28 13:26:46,071 INFO [train.py:904] (1/8) Epoch 6, batch 7800, loss[loss=0.244, simple_loss=0.3221, pruned_loss=0.08298, over 16432.00 frames. ], tot_loss[loss=0.2449, simple_loss=0.321, pruned_loss=0.08444, over 3113544.44 frames. ], batch size: 146, lr: 1.10e-02, grad_scale: 4.0 2023-04-28 13:27:26,213 INFO [optim.py:368] (1/8) 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 13:28:01,711 INFO [train.py:904] (1/8) Epoch 6, batch 7850, loss[loss=0.2426, simple_loss=0.3227, pruned_loss=0.08123, over 16712.00 frames. ], tot_loss[loss=0.2442, simple_loss=0.3216, pruned_loss=0.08346, over 3121843.28 frames. ], batch size: 124, lr: 1.10e-02, grad_scale: 4.0 2023-04-28 13:29:14,704 INFO [train.py:904] (1/8) Epoch 6, batch 7900, loss[loss=0.2778, simple_loss=0.3342, pruned_loss=0.1107, over 11198.00 frames. ], tot_loss[loss=0.2441, simple_loss=0.3211, pruned_loss=0.08353, over 3111461.48 frames. ], batch size: 246, lr: 1.10e-02, grad_scale: 4.0 2023-04-28 13:29:53,308 INFO [optim.py:368] (1/8) 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] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=58681.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 13:30:09,625 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.91 vs. limit=2.0 2023-04-28 13:30:31,227 INFO [train.py:904] (1/8) Epoch 6, batch 7950, loss[loss=0.2309, simple_loss=0.3027, pruned_loss=0.07958, over 16426.00 frames. ], tot_loss[loss=0.245, simple_loss=0.3214, pruned_loss=0.08428, over 3090437.62 frames. ], batch size: 68, lr: 1.10e-02, grad_scale: 4.0 2023-04-28 13:31:33,496 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=58742.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 13:31:37,337 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.75 vs. limit=2.0 2023-04-28 13:31:46,455 INFO [train.py:904] (1/8) Epoch 6, batch 8000, loss[loss=0.3068, simple_loss=0.3496, pruned_loss=0.132, over 11223.00 frames. ], tot_loss[loss=0.2464, simple_loss=0.3224, pruned_loss=0.08524, over 3089293.25 frames. ], batch size: 248, lr: 1.10e-02, grad_scale: 8.0 2023-04-28 13:32:25,107 INFO [zipformer.py:625] (1/8) 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] (1/8) 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 13:32:57,308 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=58797.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 13:33:02,691 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 6, batch 8050, loss[loss=0.2555, simple_loss=0.3167, pruned_loss=0.09717, over 11699.00 frames. ], tot_loss[loss=0.2457, simple_loss=0.3217, pruned_loss=0.08487, over 3075861.63 frames. ], batch size: 246, lr: 1.10e-02, grad_scale: 8.0 2023-04-28 13:33:31,592 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.6649, 4.8512, 5.0146, 4.9556, 4.9116, 5.4613, 4.9810, 4.7476], device='cuda:1'), covar=tensor([0.0883, 0.1566, 0.1457, 0.1612, 0.2279, 0.0849, 0.1103, 0.2055], device='cuda:1'), in_proj_covar=tensor([0.0297, 0.0407, 0.0418, 0.0359, 0.0466, 0.0444, 0.0338, 0.0479], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-04-28 13:33:58,011 INFO [zipformer.py:625] (1/8) 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] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=58845.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 13:34:16,877 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=58848.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 13:34:20,745 INFO [train.py:904] (1/8) Epoch 6, batch 8100, loss[loss=0.2373, simple_loss=0.3165, pruned_loss=0.07906, over 16423.00 frames. ], tot_loss[loss=0.2446, simple_loss=0.321, pruned_loss=0.08407, over 3066567.00 frames. ], batch size: 75, lr: 1.10e-02, grad_scale: 2.0 2023-04-28 13:35:03,715 INFO [optim.py:368] (1/8) 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 13:35:38,303 INFO [train.py:904] (1/8) Epoch 6, batch 8150, loss[loss=0.2305, simple_loss=0.304, pruned_loss=0.07855, over 16876.00 frames. ], tot_loss[loss=0.2428, simple_loss=0.3191, pruned_loss=0.08328, over 3070084.82 frames. ], batch size: 96, lr: 1.10e-02, grad_scale: 2.0 2023-04-28 13:35:46,330 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=58906.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 13:36:11,256 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.76 vs. limit=2.0 2023-04-28 13:36:28,109 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.6086, 3.8751, 4.1209, 2.0529, 4.3563, 4.2520, 3.0878, 3.0658], device='cuda:1'), covar=tensor([0.0725, 0.0113, 0.0118, 0.1090, 0.0035, 0.0066, 0.0353, 0.0407], device='cuda:1'), in_proj_covar=tensor([0.0142, 0.0092, 0.0084, 0.0142, 0.0072, 0.0083, 0.0119, 0.0127], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0002, 0.0002, 0.0003, 0.0003], device='cuda:1') 2023-04-28 13:36:54,140 INFO [train.py:904] (1/8) Epoch 6, batch 8200, loss[loss=0.2126, simple_loss=0.295, pruned_loss=0.06507, over 16471.00 frames. ], tot_loss[loss=0.2398, simple_loss=0.3158, pruned_loss=0.08191, over 3095021.78 frames. ], batch size: 68, lr: 1.10e-02, grad_scale: 2.0 2023-04-28 13:37:21,756 INFO [zipformer.py:625] (1/8) 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] (1/8) 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] (1/8) Epoch 6, batch 8250, loss[loss=0.1884, simple_loss=0.2692, pruned_loss=0.05377, over 11937.00 frames. ], tot_loss[loss=0.2376, simple_loss=0.3155, pruned_loss=0.07982, over 3078910.66 frames. ], batch size: 247, lr: 1.10e-02, grad_scale: 2.0 2023-04-28 13:39:16,122 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=59037.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 13:39:27,893 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.4868, 3.5241, 3.2335, 3.0422, 3.0828, 3.3288, 3.2333, 3.1725], device='cuda:1'), covar=tensor([0.0490, 0.0390, 0.0199, 0.0189, 0.0532, 0.0287, 0.1019, 0.0406], device='cuda:1'), in_proj_covar=tensor([0.0189, 0.0212, 0.0222, 0.0192, 0.0249, 0.0226, 0.0160, 0.0254], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-28 13:39:37,602 INFO [train.py:904] (1/8) Epoch 6, batch 8300, loss[loss=0.2127, simple_loss=0.287, pruned_loss=0.06923, over 12000.00 frames. ], tot_loss[loss=0.2321, simple_loss=0.3119, pruned_loss=0.0762, over 3073591.86 frames. ], batch size: 247, lr: 1.10e-02, grad_scale: 2.0 2023-04-28 13:40:22,347 INFO [optim.py:368] (1/8) 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 13:40:59,217 INFO [train.py:904] (1/8) Epoch 6, batch 8350, loss[loss=0.1948, simple_loss=0.2865, pruned_loss=0.05156, over 16514.00 frames. ], tot_loss[loss=0.2287, simple_loss=0.3099, pruned_loss=0.07374, over 3054599.11 frames. ], batch size: 68, lr: 1.10e-02, grad_scale: 2.0 2023-04-28 13:41:26,354 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.5054, 3.4892, 3.4504, 2.8962, 3.3906, 2.0261, 3.2200, 2.9782], device='cuda:1'), covar=tensor([0.0097, 0.0079, 0.0107, 0.0194, 0.0074, 0.1802, 0.0102, 0.0145], device='cuda:1'), in_proj_covar=tensor([0.0096, 0.0082, 0.0129, 0.0125, 0.0098, 0.0151, 0.0112, 0.0123], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-28 13:41:49,267 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=59131.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 13:41:54,498 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.48 vs. limit=2.0 2023-04-28 13:42:21,175 INFO [train.py:904] (1/8) Epoch 6, batch 8400, loss[loss=0.2, simple_loss=0.2878, pruned_loss=0.05613, over 16315.00 frames. ], tot_loss[loss=0.2251, simple_loss=0.3069, pruned_loss=0.07171, over 3028576.50 frames. ], batch size: 68, lr: 1.10e-02, grad_scale: 4.0 2023-04-28 13:42:42,293 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=59164.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 13:43:05,301 INFO [optim.py:368] (1/8) 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] (1/8) Epoch 6, batch 8450, loss[loss=0.2098, simple_loss=0.2883, pruned_loss=0.0656, over 12372.00 frames. ], tot_loss[loss=0.2215, simple_loss=0.3045, pruned_loss=0.06922, over 3035095.89 frames. ], batch size: 247, lr: 1.10e-02, grad_scale: 4.0 2023-04-28 13:44:19,005 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 6, batch 8500, loss[loss=0.1836, simple_loss=0.2603, pruned_loss=0.05347, over 11839.00 frames. ], tot_loss[loss=0.2156, simple_loss=0.2996, pruned_loss=0.06581, over 3032838.10 frames. ], batch size: 247, lr: 1.10e-02, grad_scale: 4.0 2023-04-28 13:45:19,656 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=59262.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 13:45:46,052 INFO [optim.py:368] (1/8) 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 13:45:49,786 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.0352, 3.9881, 3.9194, 3.4088, 3.8671, 1.6893, 3.7379, 3.6742], device='cuda:1'), covar=tensor([0.0069, 0.0066, 0.0096, 0.0225, 0.0070, 0.1933, 0.0098, 0.0142], device='cuda:1'), in_proj_covar=tensor([0.0094, 0.0081, 0.0127, 0.0122, 0.0096, 0.0149, 0.0110, 0.0120], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-28 13:46:25,537 INFO [train.py:904] (1/8) Epoch 6, batch 8550, loss[loss=0.1825, simple_loss=0.2775, pruned_loss=0.04374, over 16910.00 frames. ], tot_loss[loss=0.2127, simple_loss=0.2967, pruned_loss=0.0644, over 3022643.77 frames. ], batch size: 96, lr: 1.10e-02, grad_scale: 4.0 2023-04-28 13:47:25,468 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([1.8943, 2.5215, 2.3119, 3.1544, 2.4770, 3.3981, 1.5754, 2.7121], device='cuda:1'), covar=tensor([0.1195, 0.0448, 0.0879, 0.0096, 0.0140, 0.0336, 0.1312, 0.0660], device='cuda:1'), in_proj_covar=tensor([0.0145, 0.0142, 0.0167, 0.0093, 0.0182, 0.0186, 0.0164, 0.0164], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0003, 0.0004], device='cuda:1') 2023-04-28 13:47:38,193 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=59337.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 13:48:07,962 INFO [train.py:904] (1/8) Epoch 6, batch 8600, loss[loss=0.1927, simple_loss=0.2858, pruned_loss=0.04984, over 16732.00 frames. ], tot_loss[loss=0.2118, simple_loss=0.297, pruned_loss=0.06326, over 3036122.29 frames. ], batch size: 89, lr: 1.10e-02, grad_scale: 4.0 2023-04-28 13:49:03,381 INFO [optim.py:368] (1/8) 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] (1/8) 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] (1/8) Epoch 6, batch 8650, loss[loss=0.2075, simple_loss=0.2957, pruned_loss=0.05965, over 16903.00 frames. ], tot_loss[loss=0.2087, simple_loss=0.2945, pruned_loss=0.06141, over 3026613.80 frames. ], batch size: 116, lr: 1.10e-02, grad_scale: 4.0 2023-04-28 13:50:55,035 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 6, batch 8700, loss[loss=0.2043, simple_loss=0.2942, pruned_loss=0.05727, over 16742.00 frames. ], tot_loss[loss=0.2058, simple_loss=0.2918, pruned_loss=0.05994, over 3028812.89 frames. ], batch size: 134, lr: 1.10e-02, grad_scale: 4.0 2023-04-28 13:51:52,765 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=2.69 vs. limit=5.0 2023-04-28 13:52:04,641 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.88 vs. limit=2.0 2023-04-28 13:52:21,220 INFO [optim.py:368] (1/8) 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] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=59479.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 13:53:05,570 INFO [train.py:904] (1/8) Epoch 6, batch 8750, loss[loss=0.2348, simple_loss=0.3268, pruned_loss=0.07142, over 15343.00 frames. ], tot_loss[loss=0.2052, simple_loss=0.2918, pruned_loss=0.05928, over 3047166.79 frames. ], batch size: 191, lr: 1.10e-02, grad_scale: 4.0 2023-04-28 13:53:08,805 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=59502.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 13:53:16,755 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.20 vs. limit=2.0 2023-04-28 13:53:54,428 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=59520.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 13:53:54,626 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.6747, 2.1069, 1.7678, 1.7592, 2.3626, 2.1944, 2.4738, 2.5447], device='cuda:1'), covar=tensor([0.0040, 0.0225, 0.0278, 0.0294, 0.0126, 0.0214, 0.0079, 0.0136], device='cuda:1'), in_proj_covar=tensor([0.0084, 0.0160, 0.0161, 0.0161, 0.0156, 0.0161, 0.0139, 0.0141], device='cuda:1'), 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:1') 2023-04-28 13:54:55,917 INFO [train.py:904] (1/8) Epoch 6, batch 8800, loss[loss=0.187, simple_loss=0.2767, pruned_loss=0.04864, over 15369.00 frames. ], tot_loss[loss=0.2022, simple_loss=0.2887, pruned_loss=0.05782, over 3016928.05 frames. ], batch size: 191, lr: 1.10e-02, grad_scale: 8.0 2023-04-28 13:55:17,746 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=59562.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 13:55:19,834 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=59563.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 13:55:30,627 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.5073, 4.4490, 4.2691, 3.8203, 4.3508, 1.5434, 4.1474, 4.0956], device='cuda:1'), covar=tensor([0.0050, 0.0049, 0.0103, 0.0204, 0.0057, 0.1997, 0.0078, 0.0125], device='cuda:1'), in_proj_covar=tensor([0.0092, 0.0078, 0.0124, 0.0117, 0.0093, 0.0146, 0.0108, 0.0116], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-28 13:55:52,197 INFO [optim.py:368] (1/8) 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 13:55:55,407 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.9446, 5.2531, 5.0596, 5.0158, 4.6988, 4.5442, 4.7198, 5.3148], device='cuda:1'), covar=tensor([0.0720, 0.0616, 0.0704, 0.0390, 0.0534, 0.0785, 0.0673, 0.0685], device='cuda:1'), in_proj_covar=tensor([0.0383, 0.0500, 0.0416, 0.0325, 0.0307, 0.0331, 0.0408, 0.0362], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:1') 2023-04-28 13:56:37,975 INFO [train.py:904] (1/8) Epoch 6, batch 8850, loss[loss=0.2016, simple_loss=0.3001, pruned_loss=0.05153, over 16736.00 frames. ], tot_loss[loss=0.2025, simple_loss=0.2911, pruned_loss=0.05696, over 3024901.95 frames. ], batch size: 134, lr: 1.09e-02, grad_scale: 8.0 2023-04-28 13:56:56,205 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=59610.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 13:58:21,080 INFO [train.py:904] (1/8) Epoch 6, batch 8900, loss[loss=0.1826, simple_loss=0.2731, pruned_loss=0.04602, over 16519.00 frames. ], tot_loss[loss=0.2014, simple_loss=0.2906, pruned_loss=0.05613, over 3009648.82 frames. ], batch size: 75, lr: 1.09e-02, grad_scale: 8.0 2023-04-28 13:59:22,389 INFO [optim.py:368] (1/8) 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 14:00:23,136 INFO [train.py:904] (1/8) Epoch 6, batch 8950, loss[loss=0.1945, simple_loss=0.2917, pruned_loss=0.04859, over 16384.00 frames. ], tot_loss[loss=0.2026, simple_loss=0.2912, pruned_loss=0.05696, over 3017742.26 frames. ], batch size: 146, lr: 1.09e-02, grad_scale: 8.0 2023-04-28 14:01:29,348 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.8486, 4.1176, 3.9407, 3.9596, 3.5968, 3.7048, 3.8306, 4.0951], device='cuda:1'), covar=tensor([0.0893, 0.0925, 0.0962, 0.0587, 0.0800, 0.1582, 0.0779, 0.0982], device='cuda:1'), in_proj_covar=tensor([0.0387, 0.0505, 0.0422, 0.0327, 0.0309, 0.0336, 0.0413, 0.0367], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:1') 2023-04-28 14:02:12,299 INFO [train.py:904] (1/8) Epoch 6, batch 9000, loss[loss=0.1647, simple_loss=0.2575, pruned_loss=0.03596, over 16856.00 frames. ], tot_loss[loss=0.1985, simple_loss=0.2871, pruned_loss=0.05495, over 3040404.26 frames. ], batch size: 96, lr: 1.09e-02, grad_scale: 4.0 2023-04-28 14:02:12,300 INFO [train.py:929] (1/8) Computing validation loss 2023-04-28 14:02:22,190 INFO [train.py:938] (1/8) Epoch 6, validation: loss=0.1682, simple_loss=0.2716, pruned_loss=0.03235, over 944034.00 frames. 2023-04-28 14:02:22,190 INFO [train.py:939] (1/8) Maximum memory allocated so far is 17676MB 2023-04-28 14:03:21,581 INFO [optim.py:368] (1/8) Clipping_scale=2.0, grad-norm quartiles 1.989e+02 2.787e+02 3.475e+02 4.351e+02 1.064e+03, threshold=6.950e+02, percent-clipped=4.0 2023-04-28 14:04:06,172 INFO [train.py:904] (1/8) Epoch 6, batch 9050, loss[loss=0.2038, simple_loss=0.2853, pruned_loss=0.06114, over 16455.00 frames. ], tot_loss[loss=0.2006, simple_loss=0.2891, pruned_loss=0.05601, over 3043557.25 frames. ], batch size: 146, lr: 1.09e-02, grad_scale: 4.0 2023-04-28 14:04:44,591 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=59820.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 14:05:45,962 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 6, batch 9100, loss[loss=0.198, simple_loss=0.2958, pruned_loss=0.05016, over 15344.00 frames. ], tot_loss[loss=0.2013, simple_loss=0.2889, pruned_loss=0.05685, over 3036777.21 frames. ], batch size: 191, lr: 1.09e-02, grad_scale: 4.0 2023-04-28 14:06:03,678 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=59858.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 14:06:24,264 INFO [zipformer.py:625] (1/8) 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] (1/8) 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 14:07:19,888 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=59889.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 14:07:47,000 INFO [train.py:904] (1/8) Epoch 6, batch 9150, loss[loss=0.1942, simple_loss=0.2758, pruned_loss=0.0563, over 12150.00 frames. ], tot_loss[loss=0.2014, simple_loss=0.2895, pruned_loss=0.05663, over 3026190.65 frames. ], batch size: 250, lr: 1.09e-02, grad_scale: 4.0 2023-04-28 14:08:06,735 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=59909.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 14:08:37,761 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.59 vs. limit=2.0 2023-04-28 14:09:29,924 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 6, batch 9200, loss[loss=0.1612, simple_loss=0.2425, pruned_loss=0.03993, over 12051.00 frames. ], tot_loss[loss=0.1974, simple_loss=0.2847, pruned_loss=0.05508, over 3023332.90 frames. ], batch size: 248, lr: 1.09e-02, grad_scale: 8.0 2023-04-28 14:10:22,279 INFO [optim.py:368] (1/8) 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] (1/8) Epoch 6, batch 9250, loss[loss=0.1832, simple_loss=0.2576, pruned_loss=0.0544, over 11917.00 frames. ], tot_loss[loss=0.1976, simple_loss=0.2845, pruned_loss=0.0554, over 3025649.54 frames. ], batch size: 246, lr: 1.09e-02, grad_scale: 8.0 2023-04-28 14:12:33,893 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=60040.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 14:12:58,205 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.48 vs. limit=2.0 2023-04-28 14:12:58,477 INFO [train.py:904] (1/8) Epoch 6, batch 9300, loss[loss=0.1875, simple_loss=0.2678, pruned_loss=0.05362, over 16807.00 frames. ], tot_loss[loss=0.1962, simple_loss=0.2832, pruned_loss=0.05458, over 3046362.87 frames. ], batch size: 124, lr: 1.09e-02, grad_scale: 8.0 2023-04-28 14:13:01,057 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=60052.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 14:13:05,246 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.5890, 3.3880, 3.4316, 3.7854, 3.8263, 3.5094, 3.9213, 3.8663], device='cuda:1'), covar=tensor([0.1077, 0.1013, 0.1571, 0.0706, 0.0820, 0.1406, 0.0666, 0.0797], device='cuda:1'), in_proj_covar=tensor([0.0380, 0.0464, 0.0584, 0.0473, 0.0361, 0.0351, 0.0371, 0.0401], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002], device='cuda:1') 2023-04-28 14:13:33,144 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=60065.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 14:13:49,661 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.5082, 3.5409, 3.3622, 3.1860, 3.1824, 3.4415, 3.2664, 3.2672], device='cuda:1'), covar=tensor([0.0466, 0.0443, 0.0248, 0.0203, 0.0578, 0.0347, 0.0987, 0.0428], device='cuda:1'), in_proj_covar=tensor([0.0180, 0.0202, 0.0217, 0.0187, 0.0236, 0.0218, 0.0153, 0.0243], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-28 14:13:53,805 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.6494, 3.6595, 4.0762, 4.0362, 4.0173, 3.7672, 3.8053, 3.7099], device='cuda:1'), covar=tensor([0.0279, 0.0419, 0.0326, 0.0425, 0.0445, 0.0338, 0.0707, 0.0417], device='cuda:1'), in_proj_covar=tensor([0.0238, 0.0232, 0.0240, 0.0233, 0.0280, 0.0254, 0.0344, 0.0208], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0002], device='cuda:1') 2023-04-28 14:13:59,602 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.75 vs. limit=2.0 2023-04-28 14:14:01,689 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=60078.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 14:14:05,091 INFO [optim.py:368] (1/8) Clipping_scale=2.0, grad-norm quartiles 1.815e+02 2.894e+02 3.605e+02 4.509e+02 1.051e+03, threshold=7.210e+02, percent-clipped=5.0 2023-04-28 14:14:23,268 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.0659, 2.7476, 2.6374, 1.8065, 2.8845, 2.8638, 2.4749, 2.4663], device='cuda:1'), covar=tensor([0.0649, 0.0147, 0.0187, 0.1001, 0.0073, 0.0100, 0.0354, 0.0362], device='cuda:1'), in_proj_covar=tensor([0.0137, 0.0088, 0.0078, 0.0138, 0.0066, 0.0079, 0.0113, 0.0121], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003, 0.0003], device='cuda:1') 2023-04-28 14:14:44,861 INFO [train.py:904] (1/8) Epoch 6, batch 9350, loss[loss=0.2052, simple_loss=0.2929, pruned_loss=0.05879, over 16925.00 frames. ], tot_loss[loss=0.1965, simple_loss=0.2839, pruned_loss=0.05452, over 3064060.38 frames. ], batch size: 116, lr: 1.09e-02, grad_scale: 8.0 2023-04-28 14:14:45,728 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=60101.0, num_to_drop=1, layers_to_drop={2} 2023-04-28 14:15:12,279 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=60113.0, num_to_drop=1, layers_to_drop={3} 2023-04-28 14:15:36,674 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=60126.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 14:15:46,319 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=3.94 vs. limit=5.0 2023-04-28 14:16:03,908 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=60139.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 14:16:25,945 INFO [train.py:904] (1/8) Epoch 6, batch 9400, loss[loss=0.1984, simple_loss=0.2913, pruned_loss=0.05271, over 16878.00 frames. ], tot_loss[loss=0.1961, simple_loss=0.2839, pruned_loss=0.05417, over 3066590.04 frames. ], batch size: 116, lr: 1.09e-02, grad_scale: 8.0 2023-04-28 14:16:34,343 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.0471, 2.7638, 2.6688, 1.8058, 2.9043, 2.8887, 2.4664, 2.4814], device='cuda:1'), covar=tensor([0.0660, 0.0134, 0.0157, 0.0944, 0.0069, 0.0105, 0.0349, 0.0374], device='cuda:1'), in_proj_covar=tensor([0.0140, 0.0090, 0.0080, 0.0141, 0.0068, 0.0081, 0.0115, 0.0124], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0004, 0.0002, 0.0002, 0.0003, 0.0003], device='cuda:1') 2023-04-28 14:16:40,784 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=60158.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 14:17:07,260 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.92 vs. limit=2.0 2023-04-28 14:17:25,097 INFO [optim.py:368] (1/8) 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 14:18:05,528 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=60198.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 14:18:09,611 INFO [train.py:904] (1/8) Epoch 6, batch 9450, loss[loss=0.196, simple_loss=0.2824, pruned_loss=0.05482, over 15488.00 frames. ], tot_loss[loss=0.1977, simple_loss=0.2859, pruned_loss=0.0547, over 3059470.15 frames. ], batch size: 192, lr: 1.09e-02, grad_scale: 8.0 2023-04-28 14:18:14,623 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=60204.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 14:18:19,279 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=60206.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 14:19:39,940 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=60245.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 14:19:50,583 INFO [train.py:904] (1/8) Epoch 6, batch 9500, loss[loss=0.2005, simple_loss=0.2899, pruned_loss=0.05558, over 17146.00 frames. ], tot_loss[loss=0.1958, simple_loss=0.2843, pruned_loss=0.0536, over 3065062.86 frames. ], batch size: 46, lr: 1.09e-02, grad_scale: 8.0 2023-04-28 14:20:11,543 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=60259.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 14:20:51,246 INFO [optim.py:368] (1/8) 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 14:21:39,645 INFO [train.py:904] (1/8) Epoch 6, batch 9550, loss[loss=0.2062, simple_loss=0.2984, pruned_loss=0.05698, over 15322.00 frames. ], tot_loss[loss=0.1951, simple_loss=0.2836, pruned_loss=0.05336, over 3057194.37 frames. ], batch size: 191, lr: 1.09e-02, grad_scale: 8.0 2023-04-28 14:23:09,336 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.1469, 4.0943, 4.0150, 3.4581, 3.8588, 1.6571, 3.7348, 3.8038], device='cuda:1'), covar=tensor([0.0068, 0.0060, 0.0100, 0.0249, 0.0084, 0.1960, 0.0106, 0.0155], device='cuda:1'), in_proj_covar=tensor([0.0093, 0.0079, 0.0123, 0.0115, 0.0094, 0.0148, 0.0109, 0.0115], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-28 14:23:21,496 INFO [train.py:904] (1/8) Epoch 6, batch 9600, loss[loss=0.2143, simple_loss=0.2847, pruned_loss=0.07194, over 12364.00 frames. ], tot_loss[loss=0.1977, simple_loss=0.2855, pruned_loss=0.05499, over 3028424.19 frames. ], batch size: 250, lr: 1.09e-02, grad_scale: 8.0 2023-04-28 14:23:26,359 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.84 vs. limit=2.0 2023-04-28 14:24:17,215 INFO [optim.py:368] (1/8) 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 14:24:31,074 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.58 vs. limit=2.0 2023-04-28 14:24:43,645 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([1.5026, 2.6155, 2.4312, 4.0565, 2.9106, 4.1437, 1.3452, 2.8833], device='cuda:1'), covar=tensor([0.1565, 0.0730, 0.1133, 0.0101, 0.0161, 0.0315, 0.1600, 0.0740], device='cuda:1'), in_proj_covar=tensor([0.0145, 0.0142, 0.0168, 0.0092, 0.0163, 0.0187, 0.0163, 0.0165], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0004, 0.0003, 0.0004], device='cuda:1') 2023-04-28 14:24:56,963 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=60396.0, num_to_drop=1, layers_to_drop={3} 2023-04-28 14:25:09,709 INFO [train.py:904] (1/8) Epoch 6, batch 9650, loss[loss=0.2016, simple_loss=0.2904, pruned_loss=0.05641, over 16195.00 frames. ], tot_loss[loss=0.1985, simple_loss=0.287, pruned_loss=0.05498, over 3042664.68 frames. ], batch size: 165, lr: 1.09e-02, grad_scale: 8.0 2023-04-28 14:25:28,079 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=60408.0, num_to_drop=1, layers_to_drop={1} 2023-04-28 14:25:57,580 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=60421.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 14:26:22,096 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=60434.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 14:26:51,245 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.26 vs. limit=2.0 2023-04-28 14:26:58,006 INFO [train.py:904] (1/8) Epoch 6, batch 9700, loss[loss=0.197, simple_loss=0.2824, pruned_loss=0.05585, over 16708.00 frames. ], tot_loss[loss=0.1978, simple_loss=0.2858, pruned_loss=0.05494, over 3038884.70 frames. ], batch size: 134, lr: 1.09e-02, grad_scale: 8.0 2023-04-28 14:27:09,769 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.6160, 4.7675, 4.8976, 4.8469, 4.8001, 5.3143, 4.9393, 4.6290], device='cuda:1'), covar=tensor([0.0731, 0.1549, 0.1575, 0.1619, 0.2134, 0.0944, 0.1144, 0.2123], device='cuda:1'), in_proj_covar=tensor([0.0265, 0.0377, 0.0382, 0.0324, 0.0430, 0.0413, 0.0312, 0.0435], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-28 14:27:52,344 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.33 vs. limit=2.0 2023-04-28 14:27:59,429 INFO [optim.py:368] (1/8) Clipping_scale=2.0, grad-norm quartiles 1.972e+02 2.947e+02 3.794e+02 4.727e+02 9.838e+02, threshold=7.588e+02, percent-clipped=5.0 2023-04-28 14:28:41,229 INFO [train.py:904] (1/8) Epoch 6, batch 9750, loss[loss=0.2001, simple_loss=0.2939, pruned_loss=0.05312, over 16631.00 frames. ], tot_loss[loss=0.1972, simple_loss=0.2848, pruned_loss=0.05476, over 3065716.29 frames. ], batch size: 134, lr: 1.09e-02, grad_scale: 8.0 2023-04-28 14:28:47,886 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=60504.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 14:30:11,471 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=60545.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 14:30:19,564 INFO [train.py:904] (1/8) Epoch 6, batch 9800, loss[loss=0.1993, simple_loss=0.2791, pruned_loss=0.05973, over 12503.00 frames. ], tot_loss[loss=0.1968, simple_loss=0.2854, pruned_loss=0.05415, over 3074636.50 frames. ], batch size: 248, lr: 1.09e-02, grad_scale: 8.0 2023-04-28 14:30:21,862 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=60552.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 14:30:27,282 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=60554.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 14:30:40,770 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=4.70 vs. limit=5.0 2023-04-28 14:31:00,128 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.0169, 3.3909, 3.4614, 2.4700, 3.2033, 3.3361, 3.2950, 1.9406], device='cuda:1'), covar=tensor([0.0340, 0.0022, 0.0034, 0.0211, 0.0045, 0.0058, 0.0044, 0.0293], device='cuda:1'), in_proj_covar=tensor([0.0117, 0.0053, 0.0058, 0.0115, 0.0062, 0.0072, 0.0063, 0.0111], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0001, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:1') 2023-04-28 14:31:14,368 INFO [optim.py:368] (1/8) 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] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=60593.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 14:31:53,528 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.1749, 3.7586, 3.6220, 1.8452, 3.0457, 2.3373, 3.4922, 3.5000], device='cuda:1'), covar=tensor([0.0236, 0.0415, 0.0474, 0.1535, 0.0636, 0.0937, 0.0678, 0.0827], device='cuda:1'), in_proj_covar=tensor([0.0135, 0.0119, 0.0152, 0.0138, 0.0131, 0.0124, 0.0135, 0.0131], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-04-28 14:32:03,897 INFO [train.py:904] (1/8) Epoch 6, batch 9850, loss[loss=0.2154, simple_loss=0.2984, pruned_loss=0.06619, over 15373.00 frames. ], tot_loss[loss=0.1972, simple_loss=0.287, pruned_loss=0.05369, over 3100253.60 frames. ], batch size: 191, lr: 1.09e-02, grad_scale: 8.0 2023-04-28 14:33:02,961 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([1.9976, 1.2714, 1.6929, 1.9570, 2.0761, 2.1514, 1.4978, 2.0916], device='cuda:1'), covar=tensor([0.0134, 0.0282, 0.0169, 0.0168, 0.0146, 0.0113, 0.0262, 0.0061], device='cuda:1'), in_proj_covar=tensor([0.0126, 0.0147, 0.0128, 0.0127, 0.0133, 0.0094, 0.0141, 0.0080], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001], device='cuda:1') 2023-04-28 14:33:56,848 INFO [train.py:904] (1/8) Epoch 6, batch 9900, loss[loss=0.1987, simple_loss=0.2933, pruned_loss=0.05198, over 15451.00 frames. ], tot_loss[loss=0.1974, simple_loss=0.2876, pruned_loss=0.05356, over 3098300.53 frames. ], batch size: 191, lr: 1.09e-02, grad_scale: 8.0 2023-04-28 14:35:04,552 INFO [optim.py:368] (1/8) 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 14:35:42,389 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=60696.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 14:35:42,836 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=4.95 vs. limit=5.0 2023-04-28 14:35:52,849 INFO [train.py:904] (1/8) Epoch 6, batch 9950, loss[loss=0.1983, simple_loss=0.2945, pruned_loss=0.05101, over 16190.00 frames. ], tot_loss[loss=0.1984, simple_loss=0.2892, pruned_loss=0.05377, over 3091399.53 frames. ], batch size: 165, lr: 1.08e-02, grad_scale: 8.0 2023-04-28 14:36:09,332 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=60708.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 14:36:34,772 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=60718.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 14:36:42,401 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=60721.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 14:37:13,118 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=60734.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 14:37:38,501 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 6, batch 10000, loss[loss=0.2071, simple_loss=0.2966, pruned_loss=0.05876, over 15323.00 frames. ], tot_loss[loss=0.1975, simple_loss=0.288, pruned_loss=0.0535, over 3079420.41 frames. ], batch size: 191, lr: 1.08e-02, grad_scale: 8.0 2023-04-28 14:38:06,643 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=60756.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 14:38:15,884 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.4445, 4.4035, 4.8487, 4.7893, 4.8420, 4.4793, 4.5252, 4.3100], device='cuda:1'), covar=tensor([0.0213, 0.0313, 0.0380, 0.0463, 0.0291, 0.0251, 0.0587, 0.0312], device='cuda:1'), in_proj_covar=tensor([0.0246, 0.0241, 0.0247, 0.0238, 0.0281, 0.0262, 0.0349, 0.0215], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0002], device='cuda:1') 2023-04-28 14:38:29,853 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=60769.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 14:38:50,600 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=60779.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 14:38:51,220 INFO [optim.py:368] (1/8) 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 14:38:56,775 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=60782.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 14:39:33,418 INFO [train.py:904] (1/8) Epoch 6, batch 10050, loss[loss=0.1953, simple_loss=0.2812, pruned_loss=0.05474, over 12296.00 frames. ], tot_loss[loss=0.1966, simple_loss=0.2875, pruned_loss=0.05291, over 3103502.90 frames. ], batch size: 250, lr: 1.08e-02, grad_scale: 8.0 2023-04-28 14:40:19,623 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.51 vs. limit=2.0 2023-04-28 14:41:05,281 INFO [train.py:904] (1/8) Epoch 6, batch 10100, loss[loss=0.2011, simple_loss=0.2794, pruned_loss=0.06135, over 12667.00 frames. ], tot_loss[loss=0.1966, simple_loss=0.2874, pruned_loss=0.05295, over 3096725.87 frames. ], batch size: 250, lr: 1.08e-02, grad_scale: 8.0 2023-04-28 14:41:10,986 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=60854.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 14:41:17,772 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.75 vs. limit=2.0 2023-04-28 14:42:01,260 INFO [optim.py:368] (1/8) 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 14:42:22,730 INFO [train.py:904] (1/8) Epoch 6, batch 10150, loss[loss=0.1974, simple_loss=0.2747, pruned_loss=0.06007, over 12324.00 frames. ], tot_loss[loss=0.1955, simple_loss=0.2852, pruned_loss=0.0529, over 3076734.44 frames. ], batch size: 246, lr: 1.08e-02, grad_scale: 8.0 2023-04-28 14:42:48,314 INFO [train.py:904] (1/8) Epoch 7, batch 0, loss[loss=0.21, simple_loss=0.2947, pruned_loss=0.06263, over 17253.00 frames. ], tot_loss[loss=0.21, simple_loss=0.2947, pruned_loss=0.06263, over 17253.00 frames. ], batch size: 52, lr: 1.02e-02, grad_scale: 8.0 2023-04-28 14:42:48,314 INFO [train.py:929] (1/8) Computing validation loss 2023-04-28 14:42:55,783 INFO [train.py:938] (1/8) Epoch 7, validation: loss=0.1665, simple_loss=0.2702, pruned_loss=0.03141, over 944034.00 frames. 2023-04-28 14:42:55,783 INFO [train.py:939] (1/8) Maximum memory allocated so far is 17676MB 2023-04-28 14:42:55,992 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=60902.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 14:43:20,647 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.9027, 5.4191, 5.5208, 5.4085, 5.4342, 5.9778, 5.5626, 5.2848], device='cuda:1'), covar=tensor([0.0812, 0.1859, 0.1513, 0.1995, 0.2555, 0.1098, 0.1261, 0.2503], device='cuda:1'), in_proj_covar=tensor([0.0279, 0.0393, 0.0398, 0.0337, 0.0447, 0.0432, 0.0330, 0.0456], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002], device='cuda:1') 2023-04-28 14:44:05,495 INFO [train.py:904] (1/8) Epoch 7, batch 50, loss[loss=0.233, simple_loss=0.295, pruned_loss=0.08544, over 16851.00 frames. ], tot_loss[loss=0.2376, simple_loss=0.309, pruned_loss=0.08307, over 757307.86 frames. ], batch size: 96, lr: 1.01e-02, grad_scale: 1.0 2023-04-28 14:44:49,441 INFO [optim.py:368] (1/8) 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 14:44:56,118 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.97 vs. limit=2.0 2023-04-28 14:44:56,140 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.36 vs. limit=2.0 2023-04-28 14:45:15,337 INFO [train.py:904] (1/8) Epoch 7, batch 100, loss[loss=0.2203, simple_loss=0.2923, pruned_loss=0.0741, over 16812.00 frames. ], tot_loss[loss=0.2256, simple_loss=0.3002, pruned_loss=0.07548, over 1322641.92 frames. ], batch size: 102, lr: 1.01e-02, grad_scale: 1.0 2023-04-28 14:46:24,664 INFO [train.py:904] (1/8) Epoch 7, batch 150, loss[loss=0.2834, simple_loss=0.3502, pruned_loss=0.1083, over 11939.00 frames. ], tot_loss[loss=0.2208, simple_loss=0.2972, pruned_loss=0.07222, over 1759820.00 frames. ], batch size: 246, lr: 1.01e-02, grad_scale: 1.0 2023-04-28 14:46:37,312 INFO [zipformer.py:625] (1/8) 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] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=61074.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 14:47:07,163 INFO [optim.py:368] (1/8) 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] (1/8) Epoch 7, batch 200, loss[loss=0.1873, simple_loss=0.2718, pruned_loss=0.05142, over 17236.00 frames. ], tot_loss[loss=0.2205, simple_loss=0.2963, pruned_loss=0.07231, over 2109007.06 frames. ], batch size: 45, lr: 1.01e-02, grad_scale: 1.0 2023-04-28 14:48:01,690 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=61122.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 14:48:44,080 INFO [train.py:904] (1/8) Epoch 7, batch 250, loss[loss=0.2144, simple_loss=0.3057, pruned_loss=0.06159, over 17054.00 frames. ], tot_loss[loss=0.2167, simple_loss=0.2928, pruned_loss=0.07028, over 2380603.56 frames. ], batch size: 53, lr: 1.01e-02, grad_scale: 1.0 2023-04-28 14:48:50,819 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.1288, 3.8338, 4.1096, 4.2884, 4.4339, 3.9627, 4.2032, 4.3924], device='cuda:1'), covar=tensor([0.1141, 0.0922, 0.1250, 0.0560, 0.0470, 0.1138, 0.1212, 0.0452], device='cuda:1'), in_proj_covar=tensor([0.0428, 0.0525, 0.0657, 0.0523, 0.0398, 0.0396, 0.0417, 0.0449], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-28 14:49:24,840 INFO [optim.py:368] (1/8) 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 14:49:52,021 INFO [train.py:904] (1/8) Epoch 7, batch 300, loss[loss=0.1558, simple_loss=0.236, pruned_loss=0.03783, over 16812.00 frames. ], tot_loss[loss=0.2132, simple_loss=0.2896, pruned_loss=0.06847, over 2590569.13 frames. ], batch size: 39, lr: 1.01e-02, grad_scale: 1.0 2023-04-28 14:49:54,853 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([5.4082, 5.3416, 5.1439, 4.8502, 4.6483, 5.1389, 5.1865, 4.7862], device='cuda:1'), covar=tensor([0.0479, 0.0259, 0.0254, 0.0222, 0.1071, 0.0410, 0.0197, 0.0599], device='cuda:1'), in_proj_covar=tensor([0.0201, 0.0225, 0.0236, 0.0209, 0.0269, 0.0241, 0.0167, 0.0271], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002], device='cuda:1') 2023-04-28 14:49:56,271 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=3.28 vs. limit=5.0 2023-04-28 14:50:59,331 INFO [train.py:904] (1/8) Epoch 7, batch 350, loss[loss=0.2124, simple_loss=0.2737, pruned_loss=0.07555, over 16444.00 frames. ], tot_loss[loss=0.211, simple_loss=0.2871, pruned_loss=0.06742, over 2757938.13 frames. ], batch size: 68, lr: 1.01e-02, grad_scale: 1.0 2023-04-28 14:51:41,729 INFO [optim.py:368] (1/8) Clipping_scale=2.0, grad-norm quartiles 1.762e+02 2.825e+02 3.324e+02 4.016e+02 8.512e+02, threshold=6.649e+02, percent-clipped=2.0 2023-04-28 14:51:58,814 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.1163, 3.1213, 3.1922, 1.6839, 3.3749, 3.3327, 2.7197, 2.5254], device='cuda:1'), covar=tensor([0.0790, 0.0169, 0.0162, 0.1135, 0.0082, 0.0129, 0.0391, 0.0487], device='cuda:1'), in_proj_covar=tensor([0.0143, 0.0094, 0.0083, 0.0143, 0.0071, 0.0087, 0.0120, 0.0130], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0002, 0.0002, 0.0003, 0.0003], device='cuda:1') 2023-04-28 14:52:08,572 INFO [train.py:904] (1/8) Epoch 7, batch 400, loss[loss=0.2339, simple_loss=0.2984, pruned_loss=0.08474, over 16438.00 frames. ], tot_loss[loss=0.2088, simple_loss=0.2853, pruned_loss=0.06614, over 2884935.30 frames. ], batch size: 146, lr: 1.01e-02, grad_scale: 2.0 2023-04-28 14:52:32,423 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.6184, 1.5903, 2.0895, 2.5341, 2.6050, 2.5614, 1.7652, 2.6220], device='cuda:1'), covar=tensor([0.0104, 0.0260, 0.0194, 0.0153, 0.0115, 0.0141, 0.0234, 0.0076], device='cuda:1'), in_proj_covar=tensor([0.0133, 0.0152, 0.0135, 0.0135, 0.0138, 0.0101, 0.0147, 0.0085], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001], device='cuda:1') 2023-04-28 14:53:09,942 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.8987, 2.0179, 2.2014, 3.1756, 2.0728, 2.3908, 2.2870, 2.0443], device='cuda:1'), covar=tensor([0.0707, 0.2438, 0.1336, 0.0508, 0.2927, 0.1598, 0.2143, 0.2632], device='cuda:1'), in_proj_covar=tensor([0.0330, 0.0346, 0.0288, 0.0318, 0.0387, 0.0369, 0.0312, 0.0410], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-28 14:53:17,203 INFO [train.py:904] (1/8) Epoch 7, batch 450, loss[loss=0.1678, simple_loss=0.2536, pruned_loss=0.041, over 16995.00 frames. ], tot_loss[loss=0.2061, simple_loss=0.283, pruned_loss=0.06461, over 2976461.24 frames. ], batch size: 41, lr: 1.01e-02, grad_scale: 2.0 2023-04-28 14:53:48,652 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=61374.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 14:53:52,520 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.4373, 4.4060, 4.9271, 4.8725, 4.9148, 4.5503, 4.5521, 4.3768], device='cuda:1'), covar=tensor([0.0243, 0.0455, 0.0298, 0.0375, 0.0339, 0.0289, 0.0696, 0.0396], device='cuda:1'), in_proj_covar=tensor([0.0274, 0.0271, 0.0276, 0.0262, 0.0319, 0.0289, 0.0389, 0.0238], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003], device='cuda:1') 2023-04-28 14:54:00,435 INFO [optim.py:368] (1/8) 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 14:54:27,749 INFO [train.py:904] (1/8) Epoch 7, batch 500, loss[loss=0.223, simple_loss=0.294, pruned_loss=0.07599, over 16889.00 frames. ], tot_loss[loss=0.2046, simple_loss=0.2811, pruned_loss=0.06406, over 3053873.91 frames. ], batch size: 96, lr: 1.01e-02, grad_scale: 2.0 2023-04-28 14:54:47,693 INFO [zipformer.py:625] (1/8) 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] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=61422.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 14:55:35,822 INFO [train.py:904] (1/8) Epoch 7, batch 550, loss[loss=0.1827, simple_loss=0.2704, pruned_loss=0.04749, over 17119.00 frames. ], tot_loss[loss=0.2047, simple_loss=0.2809, pruned_loss=0.0643, over 3108153.82 frames. ], batch size: 48, lr: 1.01e-02, grad_scale: 2.0 2023-04-28 14:56:05,094 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.0585, 1.7649, 2.2509, 2.8072, 2.6579, 3.3218, 1.9492, 3.2515], device='cuda:1'), covar=tensor([0.0108, 0.0267, 0.0189, 0.0160, 0.0139, 0.0098, 0.0269, 0.0062], device='cuda:1'), in_proj_covar=tensor([0.0134, 0.0151, 0.0136, 0.0135, 0.0139, 0.0100, 0.0147, 0.0085], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001], device='cuda:1') 2023-04-28 14:56:17,396 INFO [optim.py:368] (1/8) 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 14:56:39,062 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.4330, 3.5627, 1.9217, 3.7136, 2.6200, 3.7367, 1.8173, 2.7726], device='cuda:1'), covar=tensor([0.0182, 0.0353, 0.1359, 0.0155, 0.0710, 0.0410, 0.1401, 0.0564], device='cuda:1'), in_proj_covar=tensor([0.0130, 0.0160, 0.0181, 0.0094, 0.0160, 0.0197, 0.0190, 0.0169], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-04-28 14:56:43,946 INFO [train.py:904] (1/8) Epoch 7, batch 600, loss[loss=0.19, simple_loss=0.2604, pruned_loss=0.0598, over 16442.00 frames. ], tot_loss[loss=0.204, simple_loss=0.2801, pruned_loss=0.06395, over 3152692.67 frames. ], batch size: 68, lr: 1.01e-02, grad_scale: 2.0 2023-04-28 14:57:53,638 INFO [train.py:904] (1/8) Epoch 7, batch 650, loss[loss=0.2015, simple_loss=0.2747, pruned_loss=0.0641, over 15328.00 frames. ], tot_loss[loss=0.2018, simple_loss=0.2781, pruned_loss=0.06279, over 3184310.96 frames. ], batch size: 191, lr: 1.01e-02, grad_scale: 2.0 2023-04-28 14:58:21,993 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.4538, 2.0192, 2.2833, 4.0954, 1.9745, 2.6742, 2.1171, 2.1966], device='cuda:1'), covar=tensor([0.0682, 0.2610, 0.1485, 0.0323, 0.3044, 0.1602, 0.2540, 0.2616], device='cuda:1'), in_proj_covar=tensor([0.0332, 0.0345, 0.0288, 0.0319, 0.0387, 0.0371, 0.0312, 0.0411], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-28 14:58:35,762 INFO [optim.py:368] (1/8) 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] (1/8) Epoch 7, batch 700, loss[loss=0.184, simple_loss=0.2643, pruned_loss=0.05187, over 16841.00 frames. ], tot_loss[loss=0.2014, simple_loss=0.2775, pruned_loss=0.06269, over 3216738.87 frames. ], batch size: 42, lr: 1.01e-02, grad_scale: 2.0 2023-04-28 14:59:03,829 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.27 vs. limit=2.0 2023-04-28 14:59:09,941 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([5.2151, 5.5971, 5.3003, 5.4767, 4.8930, 4.7453, 5.1524, 5.6897], device='cuda:1'), covar=tensor([0.0970, 0.1088, 0.1217, 0.0538, 0.0858, 0.0763, 0.0806, 0.0825], device='cuda:1'), in_proj_covar=tensor([0.0439, 0.0570, 0.0477, 0.0378, 0.0355, 0.0370, 0.0474, 0.0418], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-28 15:00:08,125 INFO [train.py:904] (1/8) Epoch 7, batch 750, loss[loss=0.1903, simple_loss=0.2757, pruned_loss=0.0525, over 17101.00 frames. ], tot_loss[loss=0.2016, simple_loss=0.278, pruned_loss=0.06263, over 3243798.06 frames. ], batch size: 47, lr: 1.01e-02, grad_scale: 2.0 2023-04-28 15:00:51,366 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.86 vs. limit=2.0 2023-04-28 15:00:51,555 INFO [optim.py:368] (1/8) 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 15:01:16,766 INFO [train.py:904] (1/8) Epoch 7, batch 800, loss[loss=0.2184, simple_loss=0.2873, pruned_loss=0.07475, over 16524.00 frames. ], tot_loss[loss=0.2012, simple_loss=0.2777, pruned_loss=0.06235, over 3263153.71 frames. ], batch size: 75, lr: 1.01e-02, grad_scale: 4.0 2023-04-28 15:01:37,369 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=61717.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 15:01:49,483 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.3872, 2.4300, 2.0980, 2.2781, 2.9129, 2.7293, 3.2923, 3.1191], device='cuda:1'), covar=tensor([0.0037, 0.0210, 0.0254, 0.0228, 0.0126, 0.0183, 0.0147, 0.0117], device='cuda:1'), in_proj_covar=tensor([0.0099, 0.0174, 0.0173, 0.0173, 0.0169, 0.0175, 0.0164, 0.0159], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-28 15:01:57,041 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=61732.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 15:02:12,599 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=61743.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 15:02:24,039 INFO [train.py:904] (1/8) Epoch 7, batch 850, loss[loss=0.2031, simple_loss=0.2832, pruned_loss=0.06151, over 16427.00 frames. ], tot_loss[loss=0.2003, simple_loss=0.2773, pruned_loss=0.06166, over 3267026.90 frames. ], batch size: 68, lr: 1.01e-02, grad_scale: 4.0 2023-04-28 15:02:42,574 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=61765.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 15:03:06,806 INFO [optim.py:368] (1/8) Clipping_scale=2.0, grad-norm quartiles 1.892e+02 2.876e+02 3.503e+02 4.271e+02 9.516e+02, threshold=7.005e+02, percent-clipped=8.0 2023-04-28 15:03:19,352 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=61793.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 15:03:19,459 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.6682, 3.7369, 2.9648, 2.2988, 2.5852, 2.3150, 3.7197, 3.5806], device='cuda:1'), covar=tensor([0.2132, 0.0528, 0.1185, 0.1946, 0.1986, 0.1572, 0.0427, 0.0921], device='cuda:1'), in_proj_covar=tensor([0.0288, 0.0258, 0.0275, 0.0257, 0.0279, 0.0213, 0.0253, 0.0275], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-28 15:03:19,484 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.4975, 2.0717, 2.3273, 4.0754, 1.9645, 2.7835, 2.2419, 2.2223], device='cuda:1'), covar=tensor([0.0707, 0.2639, 0.1508, 0.0365, 0.3104, 0.1531, 0.2456, 0.2373], device='cuda:1'), in_proj_covar=tensor([0.0336, 0.0345, 0.0290, 0.0321, 0.0389, 0.0375, 0.0314, 0.0414], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-28 15:03:33,686 INFO [train.py:904] (1/8) Epoch 7, batch 900, loss[loss=0.1858, simple_loss=0.2746, pruned_loss=0.0485, over 17128.00 frames. ], tot_loss[loss=0.1986, simple_loss=0.2765, pruned_loss=0.06038, over 3286297.61 frames. ], batch size: 47, lr: 1.01e-02, grad_scale: 4.0 2023-04-28 15:03:36,363 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=61804.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 15:03:51,113 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([5.1792, 4.8731, 5.0860, 5.3697, 5.5763, 4.8763, 5.4336, 5.4903], device='cuda:1'), covar=tensor([0.1045, 0.0990, 0.1479, 0.0572, 0.0385, 0.0588, 0.0423, 0.0471], device='cuda:1'), in_proj_covar=tensor([0.0461, 0.0565, 0.0709, 0.0565, 0.0431, 0.0423, 0.0451, 0.0483], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-28 15:04:12,837 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.83 vs. limit=2.0 2023-04-28 15:04:40,613 INFO [train.py:904] (1/8) Epoch 7, batch 950, loss[loss=0.2105, simple_loss=0.2835, pruned_loss=0.06874, over 12424.00 frames. ], tot_loss[loss=0.1984, simple_loss=0.2758, pruned_loss=0.06047, over 3283062.65 frames. ], batch size: 248, lr: 1.01e-02, grad_scale: 4.0 2023-04-28 15:05:20,758 INFO [optim.py:368] (1/8) Clipping_scale=2.0, grad-norm quartiles 1.798e+02 2.637e+02 3.238e+02 3.897e+02 7.916e+02, threshold=6.477e+02, percent-clipped=2.0 2023-04-28 15:05:46,847 INFO [train.py:904] (1/8) Epoch 7, batch 1000, loss[loss=0.2077, simple_loss=0.2735, pruned_loss=0.07097, over 16852.00 frames. ], tot_loss[loss=0.1987, simple_loss=0.2751, pruned_loss=0.06117, over 3299417.68 frames. ], batch size: 116, lr: 1.01e-02, grad_scale: 4.0 2023-04-28 15:06:05,370 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=61914.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 15:06:40,663 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.1110, 2.2146, 2.3862, 4.7940, 2.0552, 2.8782, 2.3626, 2.4292], device='cuda:1'), covar=tensor([0.0559, 0.2691, 0.1588, 0.0260, 0.3302, 0.1746, 0.2370, 0.2781], device='cuda:1'), in_proj_covar=tensor([0.0334, 0.0344, 0.0289, 0.0319, 0.0385, 0.0375, 0.0313, 0.0412], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-28 15:06:56,257 INFO [train.py:904] (1/8) Epoch 7, batch 1050, loss[loss=0.1638, simple_loss=0.2384, pruned_loss=0.0446, over 16727.00 frames. ], tot_loss[loss=0.1982, simple_loss=0.2743, pruned_loss=0.06108, over 3302169.08 frames. ], batch size: 39, lr: 1.01e-02, grad_scale: 4.0 2023-04-28 15:07:08,761 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.47 vs. limit=2.0 2023-04-28 15:07:28,155 INFO [zipformer.py:625] (1/8) 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] (1/8) 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 15:07:57,826 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.24 vs. limit=2.0 2023-04-28 15:08:07,086 INFO [train.py:904] (1/8) Epoch 7, batch 1100, loss[loss=0.1647, simple_loss=0.2452, pruned_loss=0.04203, over 17228.00 frames. ], tot_loss[loss=0.1978, simple_loss=0.2741, pruned_loss=0.06079, over 3311822.78 frames. ], batch size: 45, lr: 1.01e-02, grad_scale: 4.0 2023-04-28 15:09:15,540 INFO [train.py:904] (1/8) Epoch 7, batch 1150, loss[loss=0.1958, simple_loss=0.27, pruned_loss=0.06076, over 12630.00 frames. ], tot_loss[loss=0.1961, simple_loss=0.273, pruned_loss=0.05967, over 3309102.76 frames. ], batch size: 248, lr: 1.01e-02, grad_scale: 4.0 2023-04-28 15:09:57,384 INFO [optim.py:368] (1/8) 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 15:10:00,833 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.1872, 3.9058, 2.5715, 5.3099, 4.6814, 4.8320, 2.1323, 2.8814], device='cuda:1'), covar=tensor([0.1035, 0.0382, 0.1056, 0.0112, 0.0251, 0.0293, 0.1084, 0.0777], device='cuda:1'), in_proj_covar=tensor([0.0144, 0.0146, 0.0169, 0.0102, 0.0190, 0.0196, 0.0165, 0.0168], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0003, 0.0004], device='cuda:1') 2023-04-28 15:10:04,104 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=62088.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 15:10:07,526 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([5.6790, 6.0668, 5.7526, 5.8412, 5.2805, 5.1805, 5.5024, 6.1398], device='cuda:1'), covar=tensor([0.0829, 0.0650, 0.1023, 0.0552, 0.0679, 0.0540, 0.0771, 0.0773], device='cuda:1'), in_proj_covar=tensor([0.0448, 0.0581, 0.0487, 0.0388, 0.0362, 0.0377, 0.0480, 0.0427], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-28 15:10:18,761 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 7, batch 1200, loss[loss=0.1696, simple_loss=0.2547, pruned_loss=0.04221, over 17202.00 frames. ], tot_loss[loss=0.1952, simple_loss=0.2719, pruned_loss=0.0592, over 3312738.52 frames. ], batch size: 44, lr: 1.01e-02, grad_scale: 8.0 2023-04-28 15:10:51,369 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=62122.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 15:11:22,887 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.89 vs. limit=2.0 2023-04-28 15:11:30,498 INFO [train.py:904] (1/8) Epoch 7, batch 1250, loss[loss=0.2159, simple_loss=0.2868, pruned_loss=0.07251, over 16452.00 frames. ], tot_loss[loss=0.1961, simple_loss=0.2727, pruned_loss=0.05975, over 3316973.80 frames. ], batch size: 68, lr: 1.01e-02, grad_scale: 8.0 2023-04-28 15:11:33,388 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.9226, 2.1231, 2.3145, 4.8106, 2.0159, 2.8466, 2.3575, 2.4467], device='cuda:1'), covar=tensor([0.0679, 0.3077, 0.1687, 0.0261, 0.3542, 0.1919, 0.2538, 0.3042], device='cuda:1'), in_proj_covar=tensor([0.0338, 0.0349, 0.0291, 0.0321, 0.0387, 0.0380, 0.0317, 0.0417], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-28 15:12:13,502 INFO [optim.py:368] (1/8) 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 15:12:13,907 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=62183.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 15:12:39,181 INFO [train.py:904] (1/8) Epoch 7, batch 1300, loss[loss=0.2087, simple_loss=0.2955, pruned_loss=0.06093, over 17297.00 frames. ], tot_loss[loss=0.1954, simple_loss=0.2722, pruned_loss=0.05923, over 3314511.87 frames. ], batch size: 52, lr: 1.00e-02, grad_scale: 8.0 2023-04-28 15:13:30,580 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([1.6917, 2.9314, 2.3531, 4.3217, 3.5546, 4.1785, 1.3374, 2.8387], device='cuda:1'), covar=tensor([0.1305, 0.0547, 0.1179, 0.0100, 0.0243, 0.0308, 0.1391, 0.0759], device='cuda:1'), in_proj_covar=tensor([0.0142, 0.0145, 0.0169, 0.0103, 0.0191, 0.0197, 0.0165, 0.0167], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0003, 0.0004], device='cuda:1') 2023-04-28 15:13:35,760 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.9388, 1.7470, 2.2533, 2.7901, 2.7373, 3.0603, 1.9552, 3.0245], device='cuda:1'), covar=tensor([0.0108, 0.0283, 0.0191, 0.0174, 0.0135, 0.0105, 0.0262, 0.0075], device='cuda:1'), in_proj_covar=tensor([0.0137, 0.0154, 0.0138, 0.0139, 0.0143, 0.0101, 0.0147, 0.0089], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001], device='cuda:1') 2023-04-28 15:13:49,052 INFO [train.py:904] (1/8) Epoch 7, batch 1350, loss[loss=0.194, simple_loss=0.2698, pruned_loss=0.05909, over 16786.00 frames. ], tot_loss[loss=0.1954, simple_loss=0.2728, pruned_loss=0.05905, over 3317853.81 frames. ], batch size: 124, lr: 1.00e-02, grad_scale: 8.0 2023-04-28 15:14:15,362 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=62270.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 15:14:17,815 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([5.3926, 5.3359, 5.2059, 4.9093, 4.6721, 5.1850, 5.2411, 4.8567], device='cuda:1'), covar=tensor([0.0449, 0.0245, 0.0211, 0.0195, 0.1024, 0.0278, 0.0199, 0.0518], device='cuda:1'), in_proj_covar=tensor([0.0216, 0.0247, 0.0256, 0.0229, 0.0293, 0.0257, 0.0179, 0.0293], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-04-28 15:14:19,019 INFO [zipformer.py:625] (1/8) 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] (1/8) 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] (1/8) Epoch 7, batch 1400, loss[loss=0.1731, simple_loss=0.261, pruned_loss=0.04264, over 16749.00 frames. ], tot_loss[loss=0.1959, simple_loss=0.2734, pruned_loss=0.05919, over 3323601.81 frames. ], batch size: 62, lr: 1.00e-02, grad_scale: 8.0 2023-04-28 15:15:21,452 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.3878, 4.3395, 4.2515, 4.0629, 3.9893, 4.3078, 4.0922, 4.0403], device='cuda:1'), covar=tensor([0.0443, 0.0362, 0.0218, 0.0211, 0.0654, 0.0317, 0.0500, 0.0507], device='cuda:1'), in_proj_covar=tensor([0.0218, 0.0249, 0.0258, 0.0231, 0.0295, 0.0259, 0.0181, 0.0295], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-04-28 15:15:42,031 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=62334.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 15:15:52,484 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.7775, 4.1067, 2.1259, 4.4025, 2.7763, 4.3942, 2.2245, 2.9467], device='cuda:1'), covar=tensor([0.0205, 0.0225, 0.1527, 0.0074, 0.0748, 0.0296, 0.1389, 0.0686], device='cuda:1'), in_proj_covar=tensor([0.0133, 0.0162, 0.0179, 0.0096, 0.0159, 0.0201, 0.0188, 0.0170], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-04-28 15:16:06,307 INFO [train.py:904] (1/8) Epoch 7, batch 1450, loss[loss=0.1842, simple_loss=0.2544, pruned_loss=0.057, over 16554.00 frames. ], tot_loss[loss=0.1952, simple_loss=0.2727, pruned_loss=0.05889, over 3319712.91 frames. ], batch size: 68, lr: 1.00e-02, grad_scale: 8.0 2023-04-28 15:16:47,642 INFO [optim.py:368] (1/8) 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 15:16:56,003 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=62388.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 15:16:59,985 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=62391.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 15:17:10,952 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 7, batch 1500, loss[loss=0.2078, simple_loss=0.2953, pruned_loss=0.06011, over 17042.00 frames. ], tot_loss[loss=0.1963, simple_loss=0.2737, pruned_loss=0.0594, over 3326757.93 frames. ], batch size: 50, lr: 1.00e-02, grad_scale: 8.0 2023-04-28 15:18:01,110 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=62436.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 15:18:15,914 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 7, batch 1550, loss[loss=0.1733, simple_loss=0.2589, pruned_loss=0.04386, over 16852.00 frames. ], tot_loss[loss=0.1982, simple_loss=0.2752, pruned_loss=0.06056, over 3329522.09 frames. ], batch size: 42, lr: 1.00e-02, grad_scale: 8.0 2023-04-28 15:18:24,127 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=62452.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 15:18:54,719 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([5.1901, 5.5974, 4.9851, 5.5812, 5.0569, 4.8158, 5.2531, 5.5627], device='cuda:1'), covar=tensor([0.1748, 0.1529, 0.2321, 0.0951, 0.1258, 0.1146, 0.1399, 0.1891], device='cuda:1'), in_proj_covar=tensor([0.0455, 0.0590, 0.0491, 0.0390, 0.0367, 0.0378, 0.0479, 0.0426], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-28 15:18:59,836 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=62478.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 15:19:06,020 INFO [optim.py:368] (1/8) 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 15:19:32,092 INFO [train.py:904] (1/8) Epoch 7, batch 1600, loss[loss=0.1966, simple_loss=0.2698, pruned_loss=0.06173, over 16729.00 frames. ], tot_loss[loss=0.1995, simple_loss=0.2771, pruned_loss=0.06097, over 3336823.84 frames. ], batch size: 134, lr: 1.00e-02, grad_scale: 8.0 2023-04-28 15:20:00,300 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.3267, 4.0654, 4.3587, 4.5385, 4.6148, 4.1325, 4.3283, 4.6265], device='cuda:1'), covar=tensor([0.1192, 0.0918, 0.1210, 0.0639, 0.0576, 0.1072, 0.1403, 0.0648], device='cuda:1'), in_proj_covar=tensor([0.0473, 0.0575, 0.0727, 0.0581, 0.0442, 0.0435, 0.0464, 0.0498], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-28 15:20:04,835 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=3.85 vs. limit=5.0 2023-04-28 15:20:26,563 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.76 vs. limit=2.0 2023-04-28 15:20:29,721 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.6397, 2.6590, 2.0942, 2.4576, 2.9598, 2.8090, 3.6543, 3.2147], device='cuda:1'), covar=tensor([0.0053, 0.0212, 0.0275, 0.0244, 0.0147, 0.0215, 0.0110, 0.0134], device='cuda:1'), in_proj_covar=tensor([0.0105, 0.0174, 0.0172, 0.0173, 0.0170, 0.0176, 0.0167, 0.0160], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-28 15:20:31,557 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=62545.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 15:20:35,190 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=2.69 vs. limit=5.0 2023-04-28 15:20:40,967 INFO [train.py:904] (1/8) Epoch 7, batch 1650, loss[loss=0.242, simple_loss=0.3035, pruned_loss=0.0902, over 16486.00 frames. ], tot_loss[loss=0.2017, simple_loss=0.279, pruned_loss=0.0622, over 3341267.18 frames. ], batch size: 75, lr: 1.00e-02, grad_scale: 8.0 2023-04-28 15:21:04,824 INFO [zipformer.py:625] (1/8) 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] (1/8) 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 15:21:25,350 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=62585.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 15:21:36,673 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.5677, 4.5209, 4.6034, 4.6540, 4.5463, 5.1557, 4.8150, 4.5085], device='cuda:1'), covar=tensor([0.1260, 0.1777, 0.1602, 0.1786, 0.2863, 0.1076, 0.1089, 0.2315], device='cuda:1'), in_proj_covar=tensor([0.0311, 0.0448, 0.0449, 0.0376, 0.0507, 0.0478, 0.0358, 0.0507], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-04-28 15:21:49,314 INFO [train.py:904] (1/8) Epoch 7, batch 1700, loss[loss=0.208, simple_loss=0.2783, pruned_loss=0.06885, over 16905.00 frames. ], tot_loss[loss=0.203, simple_loss=0.2807, pruned_loss=0.06271, over 3330898.38 frames. ], batch size: 116, lr: 1.00e-02, grad_scale: 8.0 2023-04-28 15:21:49,571 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([5.0954, 5.5704, 5.6735, 5.5459, 5.4168, 6.0403, 5.6704, 5.3855], device='cuda:1'), covar=tensor([0.0746, 0.1660, 0.1388, 0.1605, 0.2677, 0.0895, 0.1098, 0.2446], device='cuda:1'), in_proj_covar=tensor([0.0310, 0.0448, 0.0448, 0.0375, 0.0506, 0.0478, 0.0357, 0.0507], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-04-28 15:21:55,179 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=62606.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 15:22:12,979 INFO [zipformer.py:625] (1/8) 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] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=62629.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 15:22:51,332 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=62646.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 15:22:58,546 INFO [train.py:904] (1/8) Epoch 7, batch 1750, loss[loss=0.2086, simple_loss=0.2785, pruned_loss=0.06934, over 16789.00 frames. ], tot_loss[loss=0.2036, simple_loss=0.2816, pruned_loss=0.06282, over 3324684.20 frames. ], batch size: 42, lr: 1.00e-02, grad_scale: 8.0 2023-04-28 15:23:18,058 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=62666.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 15:23:41,547 INFO [optim.py:368] (1/8) 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 15:23:58,578 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.6238, 3.9500, 4.2340, 3.0864, 3.7657, 4.0667, 3.9345, 2.2483], device='cuda:1'), covar=tensor([0.0326, 0.0045, 0.0024, 0.0222, 0.0050, 0.0054, 0.0039, 0.0344], device='cuda:1'), in_proj_covar=tensor([0.0117, 0.0062, 0.0061, 0.0114, 0.0064, 0.0076, 0.0066, 0.0111], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:1') 2023-04-28 15:24:02,838 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=4.41 vs. limit=5.0 2023-04-28 15:24:08,264 INFO [train.py:904] (1/8) Epoch 7, batch 1800, loss[loss=0.2143, simple_loss=0.285, pruned_loss=0.0718, over 16746.00 frames. ], tot_loss[loss=0.2043, simple_loss=0.2828, pruned_loss=0.06292, over 3330973.87 frames. ], batch size: 124, lr: 1.00e-02, grad_scale: 8.0 2023-04-28 15:24:42,387 INFO [zipformer.py:625] (1/8) 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] (1/8) 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] (1/8) Epoch 7, batch 1850, loss[loss=0.1954, simple_loss=0.2734, pruned_loss=0.05868, over 16744.00 frames. ], tot_loss[loss=0.2035, simple_loss=0.2825, pruned_loss=0.06228, over 3334684.91 frames. ], batch size: 83, lr: 1.00e-02, grad_scale: 8.0 2023-04-28 15:25:25,580 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.6379, 3.7322, 2.8359, 2.3249, 2.5848, 2.1549, 3.6796, 3.5229], device='cuda:1'), covar=tensor([0.1976, 0.0597, 0.1277, 0.1892, 0.1994, 0.1714, 0.0441, 0.0926], device='cuda:1'), in_proj_covar=tensor([0.0282, 0.0257, 0.0273, 0.0256, 0.0282, 0.0209, 0.0251, 0.0276], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-28 15:25:33,123 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.4499, 4.0900, 4.0367, 2.0172, 3.1896, 2.5626, 4.0034, 3.6960], device='cuda:1'), covar=tensor([0.0258, 0.0565, 0.0435, 0.1604, 0.0692, 0.0900, 0.0535, 0.1054], device='cuda:1'), in_proj_covar=tensor([0.0140, 0.0137, 0.0156, 0.0141, 0.0134, 0.0125, 0.0139, 0.0147], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-04-28 15:25:45,793 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=62773.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 15:25:52,509 INFO [zipformer.py:625] (1/8) 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] (1/8) 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 15:26:13,632 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.0499, 3.0777, 3.4116, 2.2949, 3.0812, 3.4291, 3.3038, 1.9848], device='cuda:1'), covar=tensor([0.0308, 0.0079, 0.0027, 0.0229, 0.0061, 0.0056, 0.0040, 0.0270], device='cuda:1'), in_proj_covar=tensor([0.0118, 0.0062, 0.0062, 0.0115, 0.0065, 0.0076, 0.0066, 0.0111], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:1') 2023-04-28 15:26:26,217 INFO [train.py:904] (1/8) Epoch 7, batch 1900, loss[loss=0.1938, simple_loss=0.2779, pruned_loss=0.05483, over 16281.00 frames. ], tot_loss[loss=0.2026, simple_loss=0.282, pruned_loss=0.0616, over 3333630.32 frames. ], batch size: 165, lr: 1.00e-02, grad_scale: 8.0 2023-04-28 15:27:00,818 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=62826.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 15:27:11,589 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 7, batch 1950, loss[loss=0.1913, simple_loss=0.2772, pruned_loss=0.05266, over 17110.00 frames. ], tot_loss[loss=0.2011, simple_loss=0.2813, pruned_loss=0.06046, over 3319833.79 frames. ], batch size: 49, lr: 9.99e-03, grad_scale: 8.0 2023-04-28 15:28:19,258 INFO [optim.py:368] (1/8) 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 15:28:44,328 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 7, batch 2000, loss[loss=0.2085, simple_loss=0.301, pruned_loss=0.05799, over 17110.00 frames. ], tot_loss[loss=0.2022, simple_loss=0.2819, pruned_loss=0.06122, over 3315466.31 frames. ], batch size: 48, lr: 9.99e-03, grad_scale: 8.0 2023-04-28 15:29:24,392 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=62929.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 15:29:39,598 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=62941.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 15:29:55,688 INFO [train.py:904] (1/8) Epoch 7, batch 2050, loss[loss=0.1894, simple_loss=0.2878, pruned_loss=0.04543, over 17128.00 frames. ], tot_loss[loss=0.2029, simple_loss=0.2822, pruned_loss=0.06185, over 3312847.68 frames. ], batch size: 49, lr: 9.99e-03, grad_scale: 16.0 2023-04-28 15:30:31,319 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=62977.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 15:30:39,209 INFO [optim.py:368] (1/8) 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 15:31:03,650 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.89 vs. limit=2.0 2023-04-28 15:31:05,139 INFO [train.py:904] (1/8) Epoch 7, batch 2100, loss[loss=0.1796, simple_loss=0.2672, pruned_loss=0.04601, over 17180.00 frames. ], tot_loss[loss=0.2048, simple_loss=0.2836, pruned_loss=0.06303, over 3316882.74 frames. ], batch size: 46, lr: 9.98e-03, grad_scale: 8.0 2023-04-28 15:31:33,158 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=63022.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 15:32:09,400 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 7, batch 2150, loss[loss=0.1877, simple_loss=0.2827, pruned_loss=0.04633, over 17252.00 frames. ], tot_loss[loss=0.2046, simple_loss=0.2836, pruned_loss=0.06281, over 3317545.49 frames. ], batch size: 52, lr: 9.98e-03, grad_scale: 8.0 2023-04-28 15:33:00,561 INFO [optim.py:368] (1/8) Clipping_scale=2.0, grad-norm quartiles 1.945e+02 2.751e+02 3.323e+02 4.118e+02 9.925e+02, threshold=6.647e+02, percent-clipped=2.0 2023-04-28 15:33:17,098 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 7, batch 2200, loss[loss=0.1855, simple_loss=0.266, pruned_loss=0.05251, over 16783.00 frames. ], tot_loss[loss=0.2052, simple_loss=0.2843, pruned_loss=0.06308, over 3315631.00 frames. ], batch size: 39, lr: 9.98e-03, grad_scale: 8.0 2023-04-28 15:34:03,305 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 7, batch 2250, loss[loss=0.2139, simple_loss=0.2974, pruned_loss=0.06519, over 17075.00 frames. ], tot_loss[loss=0.2055, simple_loss=0.2846, pruned_loss=0.06318, over 3321078.95 frames. ], batch size: 55, lr: 9.97e-03, grad_scale: 8.0 2023-04-28 15:35:18,521 INFO [optim.py:368] (1/8) 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 15:35:42,238 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 7, batch 2300, loss[loss=0.223, simple_loss=0.3068, pruned_loss=0.06958, over 16561.00 frames. ], tot_loss[loss=0.2053, simple_loss=0.2846, pruned_loss=0.06304, over 3327020.05 frames. ], batch size: 68, lr: 9.97e-03, grad_scale: 8.0 2023-04-28 15:36:04,379 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.9326, 1.7409, 2.3084, 2.7912, 2.7418, 2.7681, 1.8018, 3.0184], device='cuda:1'), covar=tensor([0.0121, 0.0267, 0.0206, 0.0149, 0.0132, 0.0168, 0.0275, 0.0070], device='cuda:1'), in_proj_covar=tensor([0.0144, 0.0159, 0.0142, 0.0143, 0.0146, 0.0103, 0.0151, 0.0094], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001], device='cuda:1') 2023-04-28 15:36:10,206 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=63221.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 15:36:20,586 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([1.7252, 3.6273, 2.7427, 5.1567, 4.5463, 4.7868, 1.7806, 3.5591], device='cuda:1'), covar=tensor([0.1287, 0.0458, 0.1062, 0.0095, 0.0233, 0.0303, 0.1298, 0.0542], device='cuda:1'), in_proj_covar=tensor([0.0144, 0.0147, 0.0169, 0.0104, 0.0199, 0.0201, 0.0166, 0.0166], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0003, 0.0004], device='cuda:1') 2023-04-28 15:36:36,867 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=63241.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 15:36:45,591 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.99 vs. limit=2.0 2023-04-28 15:36:47,591 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=63249.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 15:36:51,525 INFO [train.py:904] (1/8) Epoch 7, batch 2350, loss[loss=0.2019, simple_loss=0.2815, pruned_loss=0.0612, over 16024.00 frames. ], tot_loss[loss=0.206, simple_loss=0.2848, pruned_loss=0.06358, over 3315164.94 frames. ], batch size: 35, lr: 9.96e-03, grad_scale: 4.0 2023-04-28 15:36:51,868 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([5.2934, 5.6336, 5.3703, 5.4473, 4.9226, 4.7548, 5.1553, 5.7402], device='cuda:1'), covar=tensor([0.0844, 0.0734, 0.1043, 0.0488, 0.0727, 0.0719, 0.0743, 0.0697], device='cuda:1'), in_proj_covar=tensor([0.0454, 0.0589, 0.0493, 0.0389, 0.0365, 0.0379, 0.0484, 0.0430], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-28 15:37:19,084 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.8504, 5.1211, 5.2795, 5.1323, 5.0044, 5.7082, 5.2905, 5.0067], device='cuda:1'), covar=tensor([0.0936, 0.1729, 0.1683, 0.1546, 0.2685, 0.0989, 0.1263, 0.2302], device='cuda:1'), in_proj_covar=tensor([0.0323, 0.0463, 0.0459, 0.0386, 0.0527, 0.0493, 0.0371, 0.0525], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-04-28 15:37:34,035 INFO [zipformer.py:625] (1/8) 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] (1/8) Clipping_scale=2.0, grad-norm quartiles 2.053e+02 2.699e+02 3.274e+02 4.399e+02 8.377e+02, threshold=6.549e+02, percent-clipped=2.0 2023-04-28 15:37:43,610 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=63289.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 15:38:00,171 INFO [train.py:904] (1/8) Epoch 7, batch 2400, loss[loss=0.2048, simple_loss=0.2907, pruned_loss=0.05947, over 17023.00 frames. ], tot_loss[loss=0.2081, simple_loss=0.2871, pruned_loss=0.06455, over 3308405.79 frames. ], batch size: 50, lr: 9.96e-03, grad_scale: 8.0 2023-04-28 15:38:27,598 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=63322.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 15:38:58,879 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([5.3831, 5.3768, 5.2070, 4.6413, 5.2726, 2.4413, 4.9554, 5.3071], device='cuda:1'), covar=tensor([0.0063, 0.0041, 0.0097, 0.0275, 0.0052, 0.1405, 0.0075, 0.0090], device='cuda:1'), in_proj_covar=tensor([0.0111, 0.0095, 0.0147, 0.0141, 0.0113, 0.0156, 0.0129, 0.0139], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-28 15:39:10,358 INFO [train.py:904] (1/8) Epoch 7, batch 2450, loss[loss=0.1883, simple_loss=0.277, pruned_loss=0.04985, over 17143.00 frames. ], tot_loss[loss=0.2058, simple_loss=0.2855, pruned_loss=0.06311, over 3314716.45 frames. ], batch size: 47, lr: 9.96e-03, grad_scale: 8.0 2023-04-28 15:39:34,203 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=63370.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 15:39:55,096 INFO [optim.py:368] (1/8) 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 15:40:19,245 INFO [train.py:904] (1/8) Epoch 7, batch 2500, loss[loss=0.204, simple_loss=0.2875, pruned_loss=0.06029, over 15965.00 frames. ], tot_loss[loss=0.2061, simple_loss=0.2856, pruned_loss=0.06332, over 3324272.50 frames. ], batch size: 35, lr: 9.95e-03, grad_scale: 4.0 2023-04-28 15:40:43,667 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=63420.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 15:40:46,504 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=63422.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 15:40:55,007 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 7, batch 2550, loss[loss=0.1855, simple_loss=0.2739, pruned_loss=0.04859, over 17159.00 frames. ], tot_loss[loss=0.2069, simple_loss=0.2861, pruned_loss=0.06381, over 3322832.71 frames. ], batch size: 46, lr: 9.95e-03, grad_scale: 4.0 2023-04-28 15:42:01,822 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=63477.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 15:42:08,244 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=63481.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 15:42:11,016 INFO [zipformer.py:625] (1/8) 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] (1/8) Clipping_scale=2.0, grad-norm quartiles 1.872e+02 3.238e+02 3.870e+02 4.521e+02 8.304e+02, threshold=7.740e+02, percent-clipped=2.0 2023-04-28 15:42:35,898 INFO [train.py:904] (1/8) Epoch 7, batch 2600, loss[loss=0.1783, simple_loss=0.2637, pruned_loss=0.04651, over 16986.00 frames. ], tot_loss[loss=0.2067, simple_loss=0.2863, pruned_loss=0.06357, over 3309982.81 frames. ], batch size: 41, lr: 9.94e-03, grad_scale: 4.0 2023-04-28 15:42:51,559 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.0454, 4.0232, 4.4726, 4.4706, 4.4929, 4.1188, 4.1809, 4.0614], device='cuda:1'), covar=tensor([0.0284, 0.0463, 0.0303, 0.0360, 0.0364, 0.0314, 0.0770, 0.0464], device='cuda:1'), in_proj_covar=tensor([0.0282, 0.0287, 0.0289, 0.0273, 0.0329, 0.0305, 0.0410, 0.0246], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:1') 2023-04-28 15:43:42,047 INFO [train.py:904] (1/8) Epoch 7, batch 2650, loss[loss=0.2206, simple_loss=0.2951, pruned_loss=0.07302, over 12571.00 frames. ], tot_loss[loss=0.2069, simple_loss=0.2868, pruned_loss=0.0635, over 3300078.49 frames. ], batch size: 246, lr: 9.94e-03, grad_scale: 4.0 2023-04-28 15:44:05,505 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=63569.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 15:44:15,583 INFO [zipformer.py:625] (1/8) 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] (1/8) 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 15:44:27,924 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=63586.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 15:44:48,755 INFO [train.py:904] (1/8) Epoch 7, batch 2700, loss[loss=0.2219, simple_loss=0.3076, pruned_loss=0.06806, over 17086.00 frames. ], tot_loss[loss=0.2059, simple_loss=0.2864, pruned_loss=0.06268, over 3297677.43 frames. ], batch size: 55, lr: 9.94e-03, grad_scale: 4.0 2023-04-28 15:45:28,077 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=63630.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 15:45:45,082 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.5936, 4.4188, 4.4382, 4.3161, 4.1126, 4.5109, 4.3768, 4.2102], device='cuda:1'), covar=tensor([0.0494, 0.0451, 0.0232, 0.0226, 0.0871, 0.0332, 0.0400, 0.0559], device='cuda:1'), in_proj_covar=tensor([0.0222, 0.0258, 0.0264, 0.0237, 0.0299, 0.0266, 0.0187, 0.0302], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-04-28 15:45:47,248 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([5.4381, 5.8150, 5.5421, 5.6343, 5.0838, 4.8668, 5.2014, 5.8911], device='cuda:1'), covar=tensor([0.0966, 0.0793, 0.0909, 0.0541, 0.0779, 0.0716, 0.0911, 0.0867], device='cuda:1'), in_proj_covar=tensor([0.0460, 0.0595, 0.0499, 0.0393, 0.0374, 0.0385, 0.0493, 0.0439], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-28 15:45:51,335 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=63647.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 15:45:57,996 INFO [train.py:904] (1/8) Epoch 7, batch 2750, loss[loss=0.1876, simple_loss=0.2763, pruned_loss=0.04947, over 16083.00 frames. ], tot_loss[loss=0.2049, simple_loss=0.2859, pruned_loss=0.06192, over 3304233.01 frames. ], batch size: 35, lr: 9.93e-03, grad_scale: 4.0 2023-04-28 15:46:09,318 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([5.1299, 5.6443, 5.8713, 5.6533, 5.7110, 6.1690, 5.8575, 5.6024], device='cuda:1'), covar=tensor([0.0702, 0.1481, 0.1375, 0.1705, 0.2353, 0.0912, 0.1019, 0.2113], device='cuda:1'), in_proj_covar=tensor([0.0321, 0.0463, 0.0459, 0.0388, 0.0526, 0.0492, 0.0370, 0.0528], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-04-28 15:46:45,849 INFO [optim.py:368] (1/8) 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 15:47:08,542 INFO [train.py:904] (1/8) Epoch 7, batch 2800, loss[loss=0.1771, simple_loss=0.2705, pruned_loss=0.04186, over 17071.00 frames. ], tot_loss[loss=0.2049, simple_loss=0.2856, pruned_loss=0.06211, over 3305459.63 frames. ], batch size: 50, lr: 9.93e-03, grad_scale: 8.0 2023-04-28 15:47:10,212 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.9119, 1.9987, 2.2290, 3.2099, 2.0472, 2.4015, 2.1735, 2.0363], device='cuda:1'), covar=tensor([0.0725, 0.2382, 0.1299, 0.0397, 0.2715, 0.1524, 0.2254, 0.2407], device='cuda:1'), in_proj_covar=tensor([0.0341, 0.0352, 0.0297, 0.0324, 0.0388, 0.0389, 0.0320, 0.0420], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-28 15:47:12,372 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.9380, 3.4440, 3.0824, 1.9093, 2.6953, 2.1649, 3.3191, 3.4271], device='cuda:1'), covar=tensor([0.0234, 0.0550, 0.0510, 0.1514, 0.0705, 0.0884, 0.0550, 0.0710], device='cuda:1'), in_proj_covar=tensor([0.0141, 0.0140, 0.0157, 0.0141, 0.0134, 0.0124, 0.0139, 0.0148], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-04-28 15:48:12,842 INFO [train.py:904] (1/8) Epoch 7, batch 2850, loss[loss=0.1919, simple_loss=0.2798, pruned_loss=0.05197, over 17016.00 frames. ], tot_loss[loss=0.2054, simple_loss=0.2854, pruned_loss=0.06271, over 3298705.51 frames. ], batch size: 50, lr: 9.92e-03, grad_scale: 8.0 2023-04-28 15:48:47,329 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=63776.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 15:48:50,124 INFO [zipformer.py:625] (1/8) 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] (1/8) 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 15:49:14,294 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.88 vs. limit=2.0 2023-04-28 15:49:23,210 INFO [train.py:904] (1/8) Epoch 7, batch 2900, loss[loss=0.2012, simple_loss=0.2647, pruned_loss=0.06883, over 16910.00 frames. ], tot_loss[loss=0.2048, simple_loss=0.2844, pruned_loss=0.06262, over 3308677.65 frames. ], batch size: 109, lr: 9.92e-03, grad_scale: 8.0 2023-04-28 15:49:35,836 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=63810.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 15:50:22,351 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=63845.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 15:50:32,270 INFO [train.py:904] (1/8) Epoch 7, batch 2950, loss[loss=0.2059, simple_loss=0.272, pruned_loss=0.06995, over 16918.00 frames. ], tot_loss[loss=0.2062, simple_loss=0.2845, pruned_loss=0.06395, over 3308268.64 frames. ], batch size: 109, lr: 9.92e-03, grad_scale: 8.0 2023-04-28 15:50:58,312 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=63871.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 15:51:05,314 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=63877.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 15:51:17,103 INFO [optim.py:368] (1/8) 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 15:51:38,106 INFO [train.py:904] (1/8) Epoch 7, batch 3000, loss[loss=0.2026, simple_loss=0.2796, pruned_loss=0.06281, over 16673.00 frames. ], tot_loss[loss=0.2052, simple_loss=0.2837, pruned_loss=0.06339, over 3319909.13 frames. ], batch size: 89, lr: 9.91e-03, grad_scale: 8.0 2023-04-28 15:51:38,107 INFO [train.py:929] (1/8) Computing validation loss 2023-04-28 15:51:46,841 INFO [train.py:938] (1/8) Epoch 7, validation: loss=0.1489, simple_loss=0.2553, pruned_loss=0.02124, over 944034.00 frames. 2023-04-28 15:51:46,842 INFO [train.py:939] (1/8) Maximum memory allocated so far is 17676MB 2023-04-28 15:51:49,143 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.2182, 3.9550, 3.6955, 2.0662, 2.9104, 2.3554, 3.5869, 3.8212], device='cuda:1'), covar=tensor([0.0332, 0.0569, 0.0465, 0.1586, 0.0773, 0.0916, 0.0647, 0.0815], device='cuda:1'), in_proj_covar=tensor([0.0141, 0.0140, 0.0155, 0.0140, 0.0132, 0.0124, 0.0138, 0.0148], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-04-28 15:51:54,296 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=63906.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 15:52:06,763 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=63916.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 15:52:19,423 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=63925.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 15:52:19,436 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=63925.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 15:52:42,091 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=63942.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 15:52:55,695 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([5.0930, 5.4868, 5.6886, 5.4295, 5.4559, 6.0886, 5.6663, 5.3900], device='cuda:1'), covar=tensor([0.0673, 0.1611, 0.1509, 0.1693, 0.2275, 0.0815, 0.1092, 0.2148], device='cuda:1'), in_proj_covar=tensor([0.0320, 0.0461, 0.0458, 0.0384, 0.0522, 0.0491, 0.0370, 0.0523], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-04-28 15:52:56,534 INFO [train.py:904] (1/8) Epoch 7, batch 3050, loss[loss=0.1897, simple_loss=0.2786, pruned_loss=0.05039, over 17077.00 frames. ], tot_loss[loss=0.2048, simple_loss=0.2834, pruned_loss=0.06308, over 3318379.49 frames. ], batch size: 47, lr: 9.91e-03, grad_scale: 4.0 2023-04-28 15:53:08,779 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([5.2648, 5.1583, 5.0626, 4.8184, 4.5781, 5.1244, 5.0427, 4.7231], device='cuda:1'), covar=tensor([0.0386, 0.0294, 0.0208, 0.0207, 0.0977, 0.0276, 0.0239, 0.0501], device='cuda:1'), in_proj_covar=tensor([0.0215, 0.0253, 0.0262, 0.0233, 0.0296, 0.0262, 0.0184, 0.0297], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-04-28 15:53:30,245 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=63977.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 15:53:35,122 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.17 vs. limit=2.0 2023-04-28 15:53:40,112 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.1839, 2.2359, 2.4144, 4.7903, 2.2131, 3.1226, 2.3396, 2.5542], device='cuda:1'), covar=tensor([0.0581, 0.2892, 0.1535, 0.0278, 0.3248, 0.1509, 0.2385, 0.2814], device='cuda:1'), in_proj_covar=tensor([0.0343, 0.0356, 0.0296, 0.0327, 0.0389, 0.0389, 0.0321, 0.0421], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-28 15:53:43,706 INFO [optim.py:368] (1/8) 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 15:53:55,498 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.64 vs. limit=2.0 2023-04-28 15:54:06,475 INFO [train.py:904] (1/8) Epoch 7, batch 3100, loss[loss=0.1842, simple_loss=0.275, pruned_loss=0.04667, over 17220.00 frames. ], tot_loss[loss=0.2033, simple_loss=0.2823, pruned_loss=0.06213, over 3327903.47 frames. ], batch size: 45, lr: 9.91e-03, grad_scale: 4.0 2023-04-28 15:54:23,282 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.44 vs. limit=2.0 2023-04-28 15:54:36,532 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 7, batch 3150, loss[loss=0.2124, simple_loss=0.2769, pruned_loss=0.07397, over 16751.00 frames. ], tot_loss[loss=0.2028, simple_loss=0.2815, pruned_loss=0.062, over 3330122.47 frames. ], batch size: 124, lr: 9.90e-03, grad_scale: 4.0 2023-04-28 15:55:49,771 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=64076.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 15:55:52,164 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=64078.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 15:56:00,116 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=64084.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 15:56:00,596 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.56 vs. limit=2.0 2023-04-28 15:56:03,681 INFO [optim.py:368] (1/8) Clipping_scale=2.0, grad-norm quartiles 1.615e+02 2.613e+02 3.331e+02 4.706e+02 9.304e+02, threshold=6.662e+02, percent-clipped=4.0 2023-04-28 15:56:24,282 INFO [train.py:904] (1/8) Epoch 7, batch 3200, loss[loss=0.1759, simple_loss=0.253, pruned_loss=0.04937, over 16657.00 frames. ], tot_loss[loss=0.2015, simple_loss=0.2805, pruned_loss=0.0612, over 3324014.33 frames. ], batch size: 89, lr: 9.90e-03, grad_scale: 8.0 2023-04-28 15:56:54,785 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=64124.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 15:56:57,816 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=64126.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 15:57:17,638 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.97 vs. limit=2.0 2023-04-28 15:57:33,480 INFO [train.py:904] (1/8) Epoch 7, batch 3250, loss[loss=0.1767, simple_loss=0.2577, pruned_loss=0.0479, over 16779.00 frames. ], tot_loss[loss=0.2017, simple_loss=0.2804, pruned_loss=0.06153, over 3327321.91 frames. ], batch size: 39, lr: 9.89e-03, grad_scale: 8.0 2023-04-28 15:57:53,162 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=64166.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 15:58:21,282 INFO [optim.py:368] (1/8) 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 15:58:42,121 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 7, batch 3300, loss[loss=0.2263, simple_loss=0.2944, pruned_loss=0.07912, over 16919.00 frames. ], tot_loss[loss=0.2028, simple_loss=0.2817, pruned_loss=0.06192, over 3326511.26 frames. ], batch size: 116, lr: 9.89e-03, grad_scale: 8.0 2023-04-28 15:59:04,148 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([5.6456, 6.0197, 5.7366, 5.8302, 5.3882, 5.0227, 5.5149, 6.1104], device='cuda:1'), covar=tensor([0.0813, 0.0729, 0.0939, 0.0514, 0.0682, 0.0686, 0.0694, 0.0756], device='cuda:1'), in_proj_covar=tensor([0.0479, 0.0623, 0.0516, 0.0405, 0.0388, 0.0401, 0.0507, 0.0454], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-28 15:59:09,268 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=4.78 vs. limit=5.0 2023-04-28 15:59:10,099 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.8476, 3.7255, 3.8143, 3.6993, 3.8110, 4.2118, 3.9307, 3.6730], device='cuda:1'), covar=tensor([0.1921, 0.1952, 0.1615, 0.2340, 0.2676, 0.1793, 0.1266, 0.2445], device='cuda:1'), in_proj_covar=tensor([0.0319, 0.0457, 0.0458, 0.0382, 0.0518, 0.0489, 0.0366, 0.0521], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-04-28 15:59:16,267 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=64225.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 15:59:19,176 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=5.15 vs. limit=5.0 2023-04-28 15:59:20,173 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([5.0651, 4.7423, 4.9760, 5.2345, 5.4131, 4.7497, 5.3634, 5.3390], device='cuda:1'), covar=tensor([0.1037, 0.0924, 0.1459, 0.0580, 0.0425, 0.0753, 0.0392, 0.0439], device='cuda:1'), in_proj_covar=tensor([0.0494, 0.0604, 0.0776, 0.0616, 0.0466, 0.0474, 0.0478, 0.0527], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-28 15:59:24,783 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.33 vs. limit=2.0 2023-04-28 15:59:25,428 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=64232.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 15:59:40,486 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=64242.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 15:59:51,969 INFO [train.py:904] (1/8) Epoch 7, batch 3350, loss[loss=0.2025, simple_loss=0.2791, pruned_loss=0.06299, over 16823.00 frames. ], tot_loss[loss=0.2044, simple_loss=0.2839, pruned_loss=0.06251, over 3313554.42 frames. ], batch size: 102, lr: 9.89e-03, grad_scale: 8.0 2023-04-28 16:00:20,486 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=64272.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 16:00:22,692 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=64273.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 16:00:42,048 INFO [optim.py:368] (1/8) 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 16:00:45,939 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=64290.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 16:00:49,478 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=64293.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 16:00:49,752 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.96 vs. limit=2.0 2023-04-28 16:01:02,177 INFO [train.py:904] (1/8) Epoch 7, batch 3400, loss[loss=0.2146, simple_loss=0.2866, pruned_loss=0.07134, over 15414.00 frames. ], tot_loss[loss=0.2035, simple_loss=0.2828, pruned_loss=0.06208, over 3310302.30 frames. ], batch size: 190, lr: 9.88e-03, grad_scale: 8.0 2023-04-28 16:01:45,936 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.0898, 4.2472, 4.4878, 3.3089, 3.8451, 4.4282, 4.1085, 2.4039], device='cuda:1'), covar=tensor([0.0268, 0.0028, 0.0018, 0.0195, 0.0045, 0.0038, 0.0034, 0.0326], device='cuda:1'), in_proj_covar=tensor([0.0119, 0.0062, 0.0062, 0.0116, 0.0066, 0.0078, 0.0068, 0.0113], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:1') 2023-04-28 16:02:02,948 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=64345.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 16:02:10,852 INFO [train.py:904] (1/8) Epoch 7, batch 3450, loss[loss=0.1993, simple_loss=0.2804, pruned_loss=0.0591, over 16715.00 frames. ], tot_loss[loss=0.2017, simple_loss=0.2806, pruned_loss=0.06134, over 3318761.02 frames. ], batch size: 57, lr: 9.88e-03, grad_scale: 8.0 2023-04-28 16:02:43,544 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.39 vs. limit=2.0 2023-04-28 16:02:49,993 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=64379.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 16:03:01,781 INFO [optim.py:368] (1/8) Clipping_scale=2.0, grad-norm quartiles 1.744e+02 2.661e+02 3.216e+02 3.896e+02 8.084e+02, threshold=6.431e+02, percent-clipped=1.0 2023-04-28 16:03:13,344 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([5.3815, 5.2743, 5.1299, 4.5949, 5.1630, 2.1612, 4.9313, 5.2097], device='cuda:1'), covar=tensor([0.0053, 0.0050, 0.0104, 0.0281, 0.0061, 0.1637, 0.0081, 0.0111], device='cuda:1'), in_proj_covar=tensor([0.0114, 0.0099, 0.0149, 0.0146, 0.0116, 0.0158, 0.0133, 0.0144], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-28 16:03:21,642 INFO [train.py:904] (1/8) Epoch 7, batch 3500, loss[loss=0.1763, simple_loss=0.2592, pruned_loss=0.04674, over 16835.00 frames. ], tot_loss[loss=0.2009, simple_loss=0.2799, pruned_loss=0.061, over 3316907.19 frames. ], batch size: 39, lr: 9.87e-03, grad_scale: 8.0 2023-04-28 16:03:28,739 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=64406.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 16:04:30,074 INFO [train.py:904] (1/8) Epoch 7, batch 3550, loss[loss=0.2083, simple_loss=0.2983, pruned_loss=0.0592, over 17284.00 frames. ], tot_loss[loss=0.1998, simple_loss=0.2789, pruned_loss=0.06038, over 3323562.60 frames. ], batch size: 52, lr: 9.87e-03, grad_scale: 8.0 2023-04-28 16:04:48,188 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=64465.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 16:04:49,884 INFO [zipformer.py:625] (1/8) 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] (1/8) 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 16:05:38,456 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 7, batch 3600, loss[loss=0.2191, simple_loss=0.2777, pruned_loss=0.08027, over 16895.00 frames. ], tot_loss[loss=0.1984, simple_loss=0.2776, pruned_loss=0.0596, over 3318609.90 frames. ], batch size: 109, lr: 9.87e-03, grad_scale: 8.0 2023-04-28 16:05:56,510 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=64514.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 16:06:14,291 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=64526.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 16:06:45,516 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=64549.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 16:06:48,841 INFO [train.py:904] (1/8) Epoch 7, batch 3650, loss[loss=0.1916, simple_loss=0.2608, pruned_loss=0.06118, over 16729.00 frames. ], tot_loss[loss=0.198, simple_loss=0.2763, pruned_loss=0.05986, over 3313175.93 frames. ], batch size: 89, lr: 9.86e-03, grad_scale: 8.0 2023-04-28 16:07:18,642 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=64572.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 16:07:30,925 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.6823, 5.0034, 4.7404, 4.7426, 4.4929, 4.4448, 4.5098, 5.0545], device='cuda:1'), covar=tensor([0.0942, 0.0742, 0.1010, 0.0552, 0.0625, 0.0780, 0.0770, 0.0704], device='cuda:1'), in_proj_covar=tensor([0.0460, 0.0597, 0.0495, 0.0388, 0.0370, 0.0386, 0.0485, 0.0434], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-28 16:07:40,347 INFO [optim.py:368] (1/8) 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 16:07:42,461 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=64588.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 16:07:44,430 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.6347, 4.9938, 4.7004, 4.7562, 4.4826, 4.3517, 4.4911, 5.0159], device='cuda:1'), covar=tensor([0.1059, 0.0786, 0.1156, 0.0580, 0.0664, 0.1078, 0.0820, 0.0867], device='cuda:1'), in_proj_covar=tensor([0.0460, 0.0597, 0.0495, 0.0388, 0.0370, 0.0387, 0.0485, 0.0435], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-28 16:07:44,514 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 7, batch 3700, loss[loss=0.1882, simple_loss=0.2628, pruned_loss=0.05679, over 16516.00 frames. ], tot_loss[loss=0.2003, simple_loss=0.2764, pruned_loss=0.06212, over 3290856.90 frames. ], batch size: 146, lr: 9.86e-03, grad_scale: 8.0 2023-04-28 16:08:27,963 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=64620.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 16:08:42,979 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([1.9620, 2.4934, 2.3223, 2.9348, 2.6335, 3.2321, 1.7575, 2.6525], device='cuda:1'), covar=tensor([0.1048, 0.0488, 0.0936, 0.0122, 0.0192, 0.0351, 0.1129, 0.0656], device='cuda:1'), in_proj_covar=tensor([0.0144, 0.0145, 0.0168, 0.0107, 0.0198, 0.0199, 0.0165, 0.0166], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0003, 0.0004], device='cuda:1') 2023-04-28 16:08:53,100 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.8885, 4.1222, 3.0749, 2.4398, 2.8547, 2.2074, 3.9480, 3.7998], device='cuda:1'), covar=tensor([0.2052, 0.0500, 0.1221, 0.1918, 0.2175, 0.1722, 0.0457, 0.0867], device='cuda:1'), in_proj_covar=tensor([0.0288, 0.0261, 0.0277, 0.0262, 0.0293, 0.0215, 0.0256, 0.0286], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003], device='cuda:1') 2023-04-28 16:08:54,749 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=4.98 vs. limit=5.0 2023-04-28 16:09:09,946 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=64648.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 16:09:12,275 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 7, batch 3750, loss[loss=0.2094, simple_loss=0.2712, pruned_loss=0.07387, over 16715.00 frames. ], tot_loss[loss=0.2023, simple_loss=0.2773, pruned_loss=0.06365, over 3272521.15 frames. ], batch size: 89, lr: 9.86e-03, grad_scale: 8.0 2023-04-28 16:09:55,076 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=64679.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 16:10:04,722 INFO [optim.py:368] (1/8) 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 16:10:23,849 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.9837, 4.9604, 4.7897, 4.6278, 4.4101, 4.8884, 4.7629, 4.5922], device='cuda:1'), covar=tensor([0.0436, 0.0309, 0.0199, 0.0214, 0.0844, 0.0280, 0.0306, 0.0493], device='cuda:1'), in_proj_covar=tensor([0.0213, 0.0251, 0.0257, 0.0231, 0.0294, 0.0258, 0.0179, 0.0293], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-04-28 16:10:24,915 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 7, batch 3800, loss[loss=0.2184, simple_loss=0.284, pruned_loss=0.07634, over 16898.00 frames. ], tot_loss[loss=0.2044, simple_loss=0.2783, pruned_loss=0.06528, over 3258596.65 frames. ], batch size: 116, lr: 9.85e-03, grad_scale: 4.0 2023-04-28 16:10:28,001 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=64703.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 16:10:38,577 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=64709.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 16:11:00,638 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.76 vs. limit=2.0 2023-04-28 16:11:05,638 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=64727.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 16:11:09,546 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.7113, 3.9462, 2.9369, 2.2918, 2.7743, 2.2769, 3.8456, 3.6881], device='cuda:1'), covar=tensor([0.2125, 0.0469, 0.1193, 0.2004, 0.1976, 0.1574, 0.0466, 0.0802], device='cuda:1'), in_proj_covar=tensor([0.0288, 0.0259, 0.0277, 0.0263, 0.0294, 0.0215, 0.0256, 0.0285], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003], device='cuda:1') 2023-04-28 16:11:31,486 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=64745.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 16:11:40,000 INFO [train.py:904] (1/8) Epoch 7, batch 3850, loss[loss=0.1977, simple_loss=0.2634, pruned_loss=0.06598, over 16412.00 frames. ], tot_loss[loss=0.2048, simple_loss=0.2782, pruned_loss=0.06568, over 3268780.65 frames. ], batch size: 146, lr: 9.85e-03, grad_scale: 4.0 2023-04-28 16:11:40,483 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=64752.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 16:11:54,042 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.8893, 5.1409, 4.5721, 5.0560, 4.7122, 4.4819, 4.7241, 5.1320], device='cuda:1'), covar=tensor([0.1340, 0.1117, 0.2089, 0.0767, 0.1184, 0.1438, 0.1453, 0.1466], device='cuda:1'), in_proj_covar=tensor([0.0451, 0.0583, 0.0487, 0.0380, 0.0363, 0.0380, 0.0477, 0.0428], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-28 16:11:58,604 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=64764.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 16:12:33,098 INFO [optim.py:368] (1/8) Clipping_scale=2.0, grad-norm quartiles 1.798e+02 2.797e+02 3.034e+02 3.668e+02 1.110e+03, threshold=6.069e+02, percent-clipped=1.0 2023-04-28 16:12:52,610 INFO [train.py:904] (1/8) Epoch 7, batch 3900, loss[loss=0.1867, simple_loss=0.2618, pruned_loss=0.05577, over 16735.00 frames. ], tot_loss[loss=0.2041, simple_loss=0.2769, pruned_loss=0.06562, over 3270448.69 frames. ], batch size: 83, lr: 9.84e-03, grad_scale: 4.0 2023-04-28 16:13:00,369 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=64806.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 16:13:05,071 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.26 vs. limit=2.0 2023-04-28 16:13:08,264 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=64813.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 16:13:20,517 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=64821.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 16:14:02,899 INFO [train.py:904] (1/8) Epoch 7, batch 3950, loss[loss=0.1753, simple_loss=0.2597, pruned_loss=0.04549, over 16751.00 frames. ], tot_loss[loss=0.2039, simple_loss=0.2763, pruned_loss=0.06571, over 3269612.83 frames. ], batch size: 39, lr: 9.84e-03, grad_scale: 4.0 2023-04-28 16:14:05,037 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([5.0816, 5.0125, 4.8318, 4.1417, 4.9772, 1.9599, 4.7130, 4.7074], device='cuda:1'), covar=tensor([0.0061, 0.0046, 0.0108, 0.0311, 0.0060, 0.1824, 0.0087, 0.0137], device='cuda:1'), in_proj_covar=tensor([0.0112, 0.0099, 0.0150, 0.0146, 0.0115, 0.0158, 0.0133, 0.0143], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-28 16:14:53,035 INFO [optim.py:368] (1/8) 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 16:14:53,501 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=64888.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 16:14:57,046 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.7793, 4.7234, 5.2398, 5.1667, 5.1974, 4.7961, 4.7739, 4.3683], device='cuda:1'), covar=tensor([0.0236, 0.0441, 0.0267, 0.0395, 0.0436, 0.0232, 0.0714, 0.0428], device='cuda:1'), in_proj_covar=tensor([0.0289, 0.0293, 0.0291, 0.0282, 0.0337, 0.0306, 0.0413, 0.0250], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:1') 2023-04-28 16:15:12,236 INFO [train.py:904] (1/8) Epoch 7, batch 4000, loss[loss=0.1876, simple_loss=0.2681, pruned_loss=0.0536, over 17007.00 frames. ], tot_loss[loss=0.2053, simple_loss=0.2769, pruned_loss=0.06684, over 3265857.40 frames. ], batch size: 50, lr: 9.84e-03, grad_scale: 8.0 2023-04-28 16:15:53,634 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.75 vs. limit=2.0 2023-04-28 16:16:02,046 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=64936.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 16:16:16,964 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 7, batch 4050, loss[loss=0.1872, simple_loss=0.2682, pruned_loss=0.0531, over 16903.00 frames. ], tot_loss[loss=0.2028, simple_loss=0.276, pruned_loss=0.06482, over 3279902.67 frames. ], batch size: 96, lr: 9.83e-03, grad_scale: 8.0 2023-04-28 16:16:59,707 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=64976.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 16:17:14,198 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.2792, 4.2786, 4.6846, 4.6099, 4.6215, 4.2879, 4.2820, 4.0406], device='cuda:1'), covar=tensor([0.0250, 0.0447, 0.0290, 0.0398, 0.0385, 0.0284, 0.0711, 0.0421], device='cuda:1'), in_proj_covar=tensor([0.0285, 0.0288, 0.0287, 0.0278, 0.0335, 0.0302, 0.0407, 0.0244], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:1') 2023-04-28 16:17:16,149 INFO [optim.py:368] (1/8) 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 16:17:35,698 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 7, batch 4100, loss[loss=0.2127, simple_loss=0.297, pruned_loss=0.0642, over 16836.00 frames. ], tot_loss[loss=0.2026, simple_loss=0.2775, pruned_loss=0.06381, over 3284194.78 frames. ], batch size: 96, lr: 9.83e-03, grad_scale: 8.0 2023-04-28 16:17:39,308 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=65004.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 16:18:29,126 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=65037.0, num_to_drop=1, layers_to_drop={2} 2023-04-28 16:18:30,900 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.0721, 2.8451, 2.6418, 1.9385, 2.5272, 2.0327, 2.7485, 2.8968], device='cuda:1'), covar=tensor([0.0261, 0.0572, 0.0496, 0.1534, 0.0691, 0.0877, 0.0510, 0.0576], device='cuda:1'), in_proj_covar=tensor([0.0136, 0.0138, 0.0155, 0.0140, 0.0133, 0.0124, 0.0137, 0.0147], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-04-28 16:18:47,166 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=65049.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 16:18:51,077 INFO [train.py:904] (1/8) Epoch 7, batch 4150, loss[loss=0.2445, simple_loss=0.3243, pruned_loss=0.08236, over 15328.00 frames. ], tot_loss[loss=0.21, simple_loss=0.2856, pruned_loss=0.06725, over 3249819.41 frames. ], batch size: 190, lr: 9.83e-03, grad_scale: 8.0 2023-04-28 16:18:56,787 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=65055.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 16:19:01,897 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=65059.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 16:19:44,054 INFO [optim.py:368] (1/8) 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 16:20:03,712 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=65101.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 16:20:04,500 INFO [train.py:904] (1/8) Epoch 7, batch 4200, loss[loss=0.2443, simple_loss=0.3283, pruned_loss=0.08015, over 16607.00 frames. ], tot_loss[loss=0.2158, simple_loss=0.2931, pruned_loss=0.06918, over 3222263.51 frames. ], batch size: 62, lr: 9.82e-03, grad_scale: 8.0 2023-04-28 16:20:13,269 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=65108.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 16:20:24,675 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=65116.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 16:20:32,053 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=65121.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 16:21:03,955 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.8751, 2.0035, 2.2364, 3.1971, 2.0633, 2.3700, 2.1937, 2.0188], device='cuda:1'), covar=tensor([0.0743, 0.2564, 0.1342, 0.0413, 0.2835, 0.1566, 0.2161, 0.2684], device='cuda:1'), in_proj_covar=tensor([0.0337, 0.0355, 0.0294, 0.0318, 0.0382, 0.0386, 0.0318, 0.0420], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-28 16:21:16,726 INFO [train.py:904] (1/8) Epoch 7, batch 4250, loss[loss=0.2116, simple_loss=0.2904, pruned_loss=0.06645, over 15305.00 frames. ], tot_loss[loss=0.2166, simple_loss=0.2956, pruned_loss=0.06878, over 3201950.45 frames. ], batch size: 191, lr: 9.82e-03, grad_scale: 8.0 2023-04-28 16:21:42,348 INFO [zipformer.py:625] (1/8) 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] (1/8) 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 16:22:26,328 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.30 vs. limit=2.0 2023-04-28 16:22:29,057 INFO [train.py:904] (1/8) Epoch 7, batch 4300, loss[loss=0.2239, simple_loss=0.3102, pruned_loss=0.06874, over 17112.00 frames. ], tot_loss[loss=0.2163, simple_loss=0.2967, pruned_loss=0.06796, over 3200443.06 frames. ], batch size: 49, lr: 9.81e-03, grad_scale: 8.0 2023-04-28 16:22:37,947 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.3228, 4.1720, 4.3534, 4.5645, 4.7045, 4.2366, 4.6371, 4.6880], device='cuda:1'), covar=tensor([0.1065, 0.0838, 0.1155, 0.0468, 0.0372, 0.0785, 0.0453, 0.0449], device='cuda:1'), in_proj_covar=tensor([0.0437, 0.0529, 0.0667, 0.0538, 0.0406, 0.0413, 0.0418, 0.0464], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-28 16:23:31,791 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=65245.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 16:23:41,835 INFO [train.py:904] (1/8) Epoch 7, batch 4350, loss[loss=0.2189, simple_loss=0.3014, pruned_loss=0.06817, over 16786.00 frames. ], tot_loss[loss=0.2196, simple_loss=0.3004, pruned_loss=0.06937, over 3182094.62 frames. ], batch size: 83, lr: 9.81e-03, grad_scale: 8.0 2023-04-28 16:24:30,786 INFO [zipformer.py:625] (1/8) 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] (1/8) 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 16:24:42,172 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=65293.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 16:24:54,959 INFO [train.py:904] (1/8) Epoch 7, batch 4400, loss[loss=0.222, simple_loss=0.3002, pruned_loss=0.07189, over 16766.00 frames. ], tot_loss[loss=0.2214, simple_loss=0.3023, pruned_loss=0.07029, over 3180004.68 frames. ], batch size: 124, lr: 9.81e-03, grad_scale: 8.0 2023-04-28 16:24:57,827 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=65304.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 16:25:17,337 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([5.4852, 5.4695, 5.2834, 5.1263, 4.8987, 5.2663, 5.2668, 4.9823], device='cuda:1'), covar=tensor([0.0381, 0.0133, 0.0164, 0.0124, 0.0760, 0.0199, 0.0149, 0.0398], device='cuda:1'), in_proj_covar=tensor([0.0201, 0.0233, 0.0241, 0.0215, 0.0272, 0.0240, 0.0170, 0.0272], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-28 16:25:39,689 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=65332.0, num_to_drop=1, layers_to_drop={1} 2023-04-28 16:25:59,703 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 7, batch 4450, loss[loss=0.2123, simple_loss=0.2984, pruned_loss=0.06305, over 17028.00 frames. ], tot_loss[loss=0.2245, simple_loss=0.3059, pruned_loss=0.07156, over 3171789.34 frames. ], batch size: 50, lr: 9.80e-03, grad_scale: 8.0 2023-04-28 16:26:08,113 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=65352.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 16:26:17,537 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=65359.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 16:26:58,766 INFO [optim.py:368] (1/8) 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 16:27:19,361 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 7, batch 4500, loss[loss=0.209, simple_loss=0.2966, pruned_loss=0.06072, over 16676.00 frames. ], tot_loss[loss=0.2241, simple_loss=0.3057, pruned_loss=0.0713, over 3184723.74 frames. ], batch size: 89, lr: 9.80e-03, grad_scale: 8.0 2023-04-28 16:27:26,478 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=65407.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 16:27:28,729 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=65408.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 16:27:32,342 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=65411.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 16:28:27,212 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=65449.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 16:28:31,285 INFO [train.py:904] (1/8) Epoch 7, batch 4550, loss[loss=0.2084, simple_loss=0.2962, pruned_loss=0.06028, over 16443.00 frames. ], tot_loss[loss=0.225, simple_loss=0.3066, pruned_loss=0.07167, over 3202059.05 frames. ], batch size: 75, lr: 9.80e-03, grad_scale: 8.0 2023-04-28 16:28:37,661 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=65456.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 16:28:43,660 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.7846, 4.1078, 3.1456, 2.4761, 3.0320, 2.3444, 4.4217, 4.0124], device='cuda:1'), covar=tensor([0.2272, 0.0613, 0.1297, 0.1625, 0.2143, 0.1584, 0.0367, 0.0677], device='cuda:1'), in_proj_covar=tensor([0.0289, 0.0253, 0.0272, 0.0260, 0.0290, 0.0211, 0.0253, 0.0274], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-28 16:29:21,410 INFO [optim.py:368] (1/8) 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 16:29:41,303 INFO [train.py:904] (1/8) Epoch 7, batch 4600, loss[loss=0.2593, simple_loss=0.3255, pruned_loss=0.09662, over 11633.00 frames. ], tot_loss[loss=0.2255, simple_loss=0.3072, pruned_loss=0.07191, over 3192266.35 frames. ], batch size: 247, lr: 9.79e-03, grad_scale: 8.0 2023-04-28 16:30:14,360 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.1878, 3.8207, 3.4702, 1.9047, 2.9496, 2.3825, 3.4349, 3.7539], device='cuda:1'), covar=tensor([0.0272, 0.0557, 0.0536, 0.1808, 0.0755, 0.0911, 0.0717, 0.0876], device='cuda:1'), in_proj_covar=tensor([0.0134, 0.0134, 0.0153, 0.0139, 0.0131, 0.0124, 0.0135, 0.0142], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-04-28 16:30:24,198 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=65532.0, num_to_drop=1, layers_to_drop={0} 2023-04-28 16:30:50,589 INFO [train.py:904] (1/8) Epoch 7, batch 4650, loss[loss=0.2068, simple_loss=0.2847, pruned_loss=0.06446, over 16785.00 frames. ], tot_loss[loss=0.2239, simple_loss=0.3053, pruned_loss=0.07127, over 3197204.21 frames. ], batch size: 39, lr: 9.79e-03, grad_scale: 8.0 2023-04-28 16:30:52,294 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([1.6590, 2.4823, 2.2705, 3.4116, 2.5843, 3.5422, 1.2988, 2.7052], device='cuda:1'), covar=tensor([0.1364, 0.0643, 0.1128, 0.0109, 0.0206, 0.0332, 0.1600, 0.0737], device='cuda:1'), in_proj_covar=tensor([0.0145, 0.0146, 0.0169, 0.0103, 0.0200, 0.0194, 0.0167, 0.0168], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0003, 0.0004], device='cuda:1') 2023-04-28 16:31:02,111 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.8591, 5.2997, 5.4287, 5.3276, 5.2739, 5.9483, 5.4962, 5.2679], device='cuda:1'), covar=tensor([0.0726, 0.1414, 0.1219, 0.1474, 0.2365, 0.0807, 0.1014, 0.1936], device='cuda:1'), in_proj_covar=tensor([0.0308, 0.0436, 0.0433, 0.0364, 0.0498, 0.0468, 0.0350, 0.0500], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-04-28 16:31:41,970 INFO [optim.py:368] (1/8) 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 16:31:49,838 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=65593.0, num_to_drop=1, layers_to_drop={0} 2023-04-28 16:32:03,152 INFO [train.py:904] (1/8) Epoch 7, batch 4700, loss[loss=0.2155, simple_loss=0.2875, pruned_loss=0.07177, over 12153.00 frames. ], tot_loss[loss=0.2203, simple_loss=0.3019, pruned_loss=0.06936, over 3211887.41 frames. ], batch size: 246, lr: 9.78e-03, grad_scale: 8.0 2023-04-28 16:32:24,897 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.8413, 4.1597, 3.9430, 4.0243, 3.5445, 3.7252, 3.7135, 4.1215], device='cuda:1'), covar=tensor([0.0782, 0.0698, 0.0798, 0.0460, 0.0670, 0.1436, 0.0770, 0.0855], device='cuda:1'), in_proj_covar=tensor([0.0424, 0.0548, 0.0465, 0.0358, 0.0346, 0.0367, 0.0455, 0.0404], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-28 16:32:30,793 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.9031, 3.9230, 3.9179, 3.2340, 3.8795, 1.7592, 3.6807, 3.6800], device='cuda:1'), covar=tensor([0.0109, 0.0078, 0.0103, 0.0396, 0.0086, 0.1840, 0.0117, 0.0182], device='cuda:1'), in_proj_covar=tensor([0.0104, 0.0091, 0.0139, 0.0138, 0.0107, 0.0151, 0.0123, 0.0131], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-28 16:32:37,906 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.0717, 4.1445, 1.9380, 4.8439, 2.8168, 4.6776, 2.0697, 3.0669], device='cuda:1'), covar=tensor([0.0130, 0.0222, 0.1645, 0.0028, 0.0685, 0.0227, 0.1560, 0.0600], device='cuda:1'), in_proj_covar=tensor([0.0131, 0.0156, 0.0178, 0.0092, 0.0160, 0.0195, 0.0187, 0.0166], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-04-28 16:32:45,692 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=65632.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 16:32:56,219 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=65641.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 16:33:12,631 INFO [train.py:904] (1/8) Epoch 7, batch 4750, loss[loss=0.1927, simple_loss=0.2755, pruned_loss=0.05494, over 16871.00 frames. ], tot_loss[loss=0.2169, simple_loss=0.2984, pruned_loss=0.06767, over 3207621.53 frames. ], batch size: 116, lr: 9.78e-03, grad_scale: 8.0 2023-04-28 16:33:49,946 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.6957, 2.1579, 2.3867, 4.4184, 2.0640, 2.7477, 2.3431, 2.4188], device='cuda:1'), covar=tensor([0.0653, 0.2702, 0.1547, 0.0278, 0.3133, 0.1643, 0.2410, 0.2492], device='cuda:1'), in_proj_covar=tensor([0.0338, 0.0354, 0.0297, 0.0319, 0.0391, 0.0387, 0.0318, 0.0421], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-28 16:33:51,048 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.2982, 4.1414, 4.1470, 2.6695, 3.5630, 4.0373, 3.7435, 2.5361], device='cuda:1'), covar=tensor([0.0364, 0.0016, 0.0015, 0.0242, 0.0044, 0.0045, 0.0041, 0.0256], device='cuda:1'), in_proj_covar=tensor([0.0119, 0.0057, 0.0059, 0.0116, 0.0065, 0.0075, 0.0067, 0.0111], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:1') 2023-04-28 16:33:53,098 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=65680.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 16:34:03,335 INFO [optim.py:368] (1/8) 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 16:34:22,789 INFO [train.py:904] (1/8) Epoch 7, batch 4800, loss[loss=0.2425, simple_loss=0.3157, pruned_loss=0.08467, over 12050.00 frames. ], tot_loss[loss=0.2127, simple_loss=0.2947, pruned_loss=0.06535, over 3211576.02 frames. ], batch size: 246, lr: 9.78e-03, grad_scale: 8.0 2023-04-28 16:34:36,305 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=65711.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 16:35:09,031 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.34 vs. limit=2.0 2023-04-28 16:35:36,916 INFO [train.py:904] (1/8) Epoch 7, batch 4850, loss[loss=0.1924, simple_loss=0.2742, pruned_loss=0.05527, over 16847.00 frames. ], tot_loss[loss=0.214, simple_loss=0.2961, pruned_loss=0.06591, over 3179457.91 frames. ], batch size: 42, lr: 9.77e-03, grad_scale: 8.0 2023-04-28 16:35:45,498 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=65759.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 16:35:49,727 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([1.7700, 2.8286, 2.4972, 4.3801, 3.1024, 4.2497, 1.4978, 3.1147], device='cuda:1'), covar=tensor([0.1241, 0.0598, 0.1034, 0.0075, 0.0149, 0.0251, 0.1441, 0.0632], device='cuda:1'), in_proj_covar=tensor([0.0144, 0.0146, 0.0170, 0.0102, 0.0198, 0.0194, 0.0167, 0.0168], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0003, 0.0004], device='cuda:1') 2023-04-28 16:36:27,839 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.4445, 2.0336, 1.6118, 1.7529, 2.3113, 2.1142, 2.3138, 2.4927], device='cuda:1'), covar=tensor([0.0051, 0.0228, 0.0283, 0.0267, 0.0112, 0.0215, 0.0090, 0.0135], device='cuda:1'), in_proj_covar=tensor([0.0102, 0.0171, 0.0171, 0.0170, 0.0165, 0.0173, 0.0162, 0.0159], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-28 16:36:28,553 INFO [optim.py:368] (1/8) Clipping_scale=2.0, grad-norm quartiles 1.537e+02 2.411e+02 2.781e+02 3.259e+02 6.496e+02, threshold=5.561e+02, percent-clipped=1.0 2023-04-28 16:36:48,885 INFO [train.py:904] (1/8) Epoch 7, batch 4900, loss[loss=0.2316, simple_loss=0.3088, pruned_loss=0.07723, over 12295.00 frames. ], tot_loss[loss=0.2136, simple_loss=0.2964, pruned_loss=0.06534, over 3158249.54 frames. ], batch size: 246, lr: 9.77e-03, grad_scale: 8.0 2023-04-28 16:37:02,544 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.2265, 4.2311, 4.1880, 3.5053, 4.1515, 1.6318, 3.9274, 4.0134], device='cuda:1'), covar=tensor([0.0071, 0.0060, 0.0082, 0.0346, 0.0072, 0.1931, 0.0093, 0.0139], device='cuda:1'), in_proj_covar=tensor([0.0100, 0.0089, 0.0135, 0.0133, 0.0104, 0.0147, 0.0119, 0.0128], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-28 16:37:43,413 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=65838.0, num_to_drop=1, layers_to_drop={0} 2023-04-28 16:38:02,371 INFO [train.py:904] (1/8) Epoch 7, batch 4950, loss[loss=0.2203, simple_loss=0.3004, pruned_loss=0.07007, over 17111.00 frames. ], tot_loss[loss=0.2123, simple_loss=0.2956, pruned_loss=0.06455, over 3175542.36 frames. ], batch size: 49, lr: 9.77e-03, grad_scale: 8.0 2023-04-28 16:38:28,506 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.44 vs. limit=2.0 2023-04-28 16:38:37,911 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=4.90 vs. limit=5.0 2023-04-28 16:38:53,490 INFO [optim.py:368] (1/8) 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] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=65888.0, num_to_drop=1, layers_to_drop={0} 2023-04-28 16:39:08,323 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=65899.0, num_to_drop=1, layers_to_drop={2} 2023-04-28 16:39:12,681 INFO [train.py:904] (1/8) Epoch 7, batch 5000, loss[loss=0.2134, simple_loss=0.3132, pruned_loss=0.05684, over 16811.00 frames. ], tot_loss[loss=0.2135, simple_loss=0.2975, pruned_loss=0.06471, over 3181794.44 frames. ], batch size: 96, lr: 9.76e-03, grad_scale: 8.0 2023-04-28 16:39:18,128 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.58 vs. limit=2.0 2023-04-28 16:39:46,739 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=2.31 vs. limit=5.0 2023-04-28 16:40:09,098 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=65941.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 16:40:18,513 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=2.02 vs. limit=2.0 2023-04-28 16:40:23,703 INFO [train.py:904] (1/8) Epoch 7, batch 5050, loss[loss=0.2075, simple_loss=0.3011, pruned_loss=0.05697, over 16869.00 frames. ], tot_loss[loss=0.2126, simple_loss=0.2973, pruned_loss=0.06392, over 3201207.05 frames. ], batch size: 96, lr: 9.76e-03, grad_scale: 8.0 2023-04-28 16:40:24,183 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.2326, 1.9173, 1.5077, 1.7210, 2.2751, 2.0516, 2.2240, 2.4465], device='cuda:1'), covar=tensor([0.0059, 0.0246, 0.0308, 0.0285, 0.0118, 0.0219, 0.0090, 0.0123], device='cuda:1'), in_proj_covar=tensor([0.0102, 0.0174, 0.0174, 0.0173, 0.0169, 0.0176, 0.0163, 0.0161], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-28 16:41:14,576 INFO [optim.py:368] (1/8) 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 16:41:16,134 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 7, batch 5100, loss[loss=0.2582, simple_loss=0.3221, pruned_loss=0.09718, over 11992.00 frames. ], tot_loss[loss=0.2101, simple_loss=0.2948, pruned_loss=0.06269, over 3205641.71 frames. ], batch size: 247, lr: 9.75e-03, grad_scale: 8.0 2023-04-28 16:41:50,904 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([5.4096, 5.5863, 5.0002, 5.7004, 5.1127, 4.6690, 5.3492, 5.7634], device='cuda:1'), covar=tensor([0.1509, 0.1348, 0.2807, 0.0728, 0.1350, 0.1231, 0.1381, 0.1395], device='cuda:1'), in_proj_covar=tensor([0.0436, 0.0566, 0.0478, 0.0369, 0.0352, 0.0376, 0.0464, 0.0411], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-28 16:42:26,397 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.9263, 4.8232, 4.7806, 3.9501, 4.7447, 1.8875, 4.5074, 4.6475], device='cuda:1'), covar=tensor([0.0075, 0.0060, 0.0080, 0.0445, 0.0064, 0.1851, 0.0090, 0.0159], device='cuda:1'), in_proj_covar=tensor([0.0104, 0.0092, 0.0138, 0.0137, 0.0107, 0.0152, 0.0122, 0.0132], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-28 16:42:46,394 INFO [train.py:904] (1/8) Epoch 7, batch 5150, loss[loss=0.208, simple_loss=0.2854, pruned_loss=0.06532, over 16260.00 frames. ], tot_loss[loss=0.2096, simple_loss=0.2947, pruned_loss=0.06223, over 3184385.86 frames. ], batch size: 35, lr: 9.75e-03, grad_scale: 8.0 2023-04-28 16:42:58,927 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.0468, 2.3175, 2.4576, 4.6433, 2.1102, 3.0304, 2.4358, 2.7403], device='cuda:1'), covar=tensor([0.0556, 0.2533, 0.1454, 0.0243, 0.3032, 0.1435, 0.2305, 0.1990], device='cuda:1'), in_proj_covar=tensor([0.0331, 0.0347, 0.0292, 0.0316, 0.0384, 0.0378, 0.0312, 0.0412], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-28 16:43:37,179 INFO [optim.py:368] (1/8) 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 16:43:55,946 INFO [train.py:904] (1/8) Epoch 7, batch 5200, loss[loss=0.1802, simple_loss=0.2622, pruned_loss=0.04907, over 17023.00 frames. ], tot_loss[loss=0.2091, simple_loss=0.2938, pruned_loss=0.06222, over 3188224.25 frames. ], batch size: 55, lr: 9.75e-03, grad_scale: 8.0 2023-04-28 16:44:56,199 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 7, batch 5250, loss[loss=0.2052, simple_loss=0.2873, pruned_loss=0.06159, over 16820.00 frames. ], tot_loss[loss=0.2073, simple_loss=0.291, pruned_loss=0.06178, over 3189430.29 frames. ], batch size: 83, lr: 9.74e-03, grad_scale: 8.0 2023-04-28 16:45:18,978 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=66160.0, num_to_drop=1, layers_to_drop={0} 2023-04-28 16:45:59,716 INFO [optim.py:368] (1/8) 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 16:46:00,094 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=66188.0, num_to_drop=1, layers_to_drop={0} 2023-04-28 16:46:07,004 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=66194.0, num_to_drop=1, layers_to_drop={0} 2023-04-28 16:46:18,394 INFO [train.py:904] (1/8) Epoch 7, batch 5300, loss[loss=0.1879, simple_loss=0.2596, pruned_loss=0.05805, over 16623.00 frames. ], tot_loss[loss=0.2042, simple_loss=0.2874, pruned_loss=0.06053, over 3191508.04 frames. ], batch size: 35, lr: 9.74e-03, grad_scale: 8.0 2023-04-28 16:46:23,347 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=66205.0, num_to_drop=1, layers_to_drop={1} 2023-04-28 16:46:45,712 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=66221.0, num_to_drop=1, layers_to_drop={2} 2023-04-28 16:47:00,364 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([5.7012, 5.9993, 5.6308, 5.8101, 5.4534, 5.1363, 5.5176, 6.0905], device='cuda:1'), covar=tensor([0.0737, 0.0600, 0.0907, 0.0491, 0.0567, 0.0533, 0.0657, 0.0608], device='cuda:1'), in_proj_covar=tensor([0.0442, 0.0568, 0.0482, 0.0373, 0.0355, 0.0378, 0.0468, 0.0415], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-28 16:47:06,138 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=66236.0, num_to_drop=1, layers_to_drop={0} 2023-04-28 16:47:12,776 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.4463, 3.6643, 3.8748, 1.6895, 4.0473, 4.0770, 3.0755, 2.9315], device='cuda:1'), covar=tensor([0.0746, 0.0154, 0.0097, 0.1181, 0.0040, 0.0056, 0.0305, 0.0420], device='cuda:1'), in_proj_covar=tensor([0.0142, 0.0097, 0.0082, 0.0140, 0.0071, 0.0085, 0.0120, 0.0128], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0003, 0.0002, 0.0004, 0.0002, 0.0002, 0.0003, 0.0003], device='cuda:1') 2023-04-28 16:47:28,949 INFO [train.py:904] (1/8) Epoch 7, batch 5350, loss[loss=0.193, simple_loss=0.2869, pruned_loss=0.04949, over 16855.00 frames. ], tot_loss[loss=0.2033, simple_loss=0.2866, pruned_loss=0.06003, over 3199140.60 frames. ], batch size: 90, lr: 9.74e-03, grad_scale: 8.0 2023-04-28 16:47:36,955 INFO [zipformer.py:625] (1/8) 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] (1/8) 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 16:48:37,974 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.43 vs. limit=2.0 2023-04-28 16:48:40,524 INFO [train.py:904] (1/8) Epoch 7, batch 5400, loss[loss=0.2454, simple_loss=0.3238, pruned_loss=0.08351, over 15254.00 frames. ], tot_loss[loss=0.2057, simple_loss=0.2897, pruned_loss=0.06083, over 3204899.80 frames. ], batch size: 190, lr: 9.73e-03, grad_scale: 8.0 2023-04-28 16:49:05,974 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=66319.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 16:49:41,980 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.36 vs. limit=2.0 2023-04-28 16:49:55,998 INFO [train.py:904] (1/8) Epoch 7, batch 5450, loss[loss=0.2674, simple_loss=0.344, pruned_loss=0.0954, over 16453.00 frames. ], tot_loss[loss=0.2094, simple_loss=0.2934, pruned_loss=0.06268, over 3200735.05 frames. ], batch size: 146, lr: 9.73e-03, grad_scale: 8.0 2023-04-28 16:50:50,980 INFO [optim.py:368] (1/8) 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 16:51:12,547 INFO [train.py:904] (1/8) Epoch 7, batch 5500, loss[loss=0.2392, simple_loss=0.3271, pruned_loss=0.07569, over 16747.00 frames. ], tot_loss[loss=0.2197, simple_loss=0.302, pruned_loss=0.06871, over 3175648.51 frames. ], batch size: 124, lr: 9.73e-03, grad_scale: 8.0 2023-04-28 16:52:29,858 INFO [train.py:904] (1/8) Epoch 7, batch 5550, loss[loss=0.2301, simple_loss=0.3109, pruned_loss=0.07462, over 16435.00 frames. ], tot_loss[loss=0.2304, simple_loss=0.31, pruned_loss=0.07541, over 3138842.89 frames. ], batch size: 146, lr: 9.72e-03, grad_scale: 8.0 2023-04-28 16:53:27,052 INFO [optim.py:368] (1/8) 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 16:53:35,854 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=66494.0, num_to_drop=1, layers_to_drop={0} 2023-04-28 16:53:47,080 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=66500.0, num_to_drop=1, layers_to_drop={3} 2023-04-28 16:53:49,152 INFO [train.py:904] (1/8) Epoch 7, batch 5600, loss[loss=0.2496, simple_loss=0.3224, pruned_loss=0.0884, over 16825.00 frames. ], tot_loss[loss=0.2391, simple_loss=0.3164, pruned_loss=0.08091, over 3105536.13 frames. ], batch size: 116, lr: 9.72e-03, grad_scale: 8.0 2023-04-28 16:54:11,868 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=66516.0, num_to_drop=1, layers_to_drop={2} 2023-04-28 16:54:20,943 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=66521.0, num_to_drop=1, layers_to_drop={0} 2023-04-28 16:54:54,202 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=66542.0, num_to_drop=1, layers_to_drop={1} 2023-04-28 16:55:08,328 INFO [train.py:904] (1/8) Epoch 7, batch 5650, loss[loss=0.231, simple_loss=0.3123, pruned_loss=0.07478, over 16600.00 frames. ], tot_loss[loss=0.2476, simple_loss=0.3228, pruned_loss=0.08616, over 3083588.07 frames. ], batch size: 62, lr: 9.71e-03, grad_scale: 8.0 2023-04-28 16:55:53,677 INFO [zipformer.py:625] (1/8) 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] (1/8) 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 16:56:23,903 INFO [train.py:904] (1/8) Epoch 7, batch 5700, loss[loss=0.3332, simple_loss=0.3734, pruned_loss=0.1465, over 11253.00 frames. ], tot_loss[loss=0.2507, simple_loss=0.3249, pruned_loss=0.08824, over 3060554.75 frames. ], batch size: 248, lr: 9.71e-03, grad_scale: 8.0 2023-04-28 16:56:42,483 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=66614.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 16:57:41,036 INFO [train.py:904] (1/8) Epoch 7, batch 5750, loss[loss=0.2933, simple_loss=0.3485, pruned_loss=0.119, over 10967.00 frames. ], tot_loss[loss=0.2535, simple_loss=0.3274, pruned_loss=0.08976, over 3037123.29 frames. ], batch size: 249, lr: 9.71e-03, grad_scale: 8.0 2023-04-28 16:58:39,454 INFO [optim.py:368] (1/8) 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] (1/8) Epoch 7, batch 5800, loss[loss=0.2339, simple_loss=0.3215, pruned_loss=0.07309, over 16825.00 frames. ], tot_loss[loss=0.2537, simple_loss=0.3281, pruned_loss=0.08965, over 3013870.31 frames. ], batch size: 96, lr: 9.70e-03, grad_scale: 16.0 2023-04-28 16:59:21,196 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.5634, 4.2945, 4.0579, 1.9609, 3.2466, 2.6678, 3.9527, 4.2872], device='cuda:1'), covar=tensor([0.0237, 0.0446, 0.0463, 0.1746, 0.0705, 0.0856, 0.0532, 0.0588], device='cuda:1'), in_proj_covar=tensor([0.0135, 0.0131, 0.0154, 0.0141, 0.0132, 0.0124, 0.0136, 0.0139], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-04-28 16:59:23,231 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.4680, 3.4103, 2.7084, 2.1507, 2.5166, 2.1565, 3.4549, 3.3560], device='cuda:1'), covar=tensor([0.2315, 0.0657, 0.1312, 0.1811, 0.1912, 0.1594, 0.0447, 0.0838], device='cuda:1'), in_proj_covar=tensor([0.0286, 0.0251, 0.0273, 0.0257, 0.0279, 0.0207, 0.0249, 0.0269], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-28 16:59:48,303 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=66731.0, num_to_drop=1, layers_to_drop={1} 2023-04-28 17:00:19,932 INFO [train.py:904] (1/8) Epoch 7, batch 5850, loss[loss=0.2184, simple_loss=0.3085, pruned_loss=0.06415, over 16862.00 frames. ], tot_loss[loss=0.2493, simple_loss=0.3247, pruned_loss=0.08696, over 3021255.73 frames. ], batch size: 102, lr: 9.70e-03, grad_scale: 8.0 2023-04-28 17:00:57,211 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([1.9965, 3.2232, 3.2834, 1.5049, 3.4526, 3.4934, 2.7854, 2.5172], device='cuda:1'), covar=tensor([0.0853, 0.0146, 0.0159, 0.1279, 0.0066, 0.0098, 0.0342, 0.0489], device='cuda:1'), in_proj_covar=tensor([0.0140, 0.0095, 0.0081, 0.0138, 0.0071, 0.0086, 0.0117, 0.0126], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0002, 0.0002, 0.0003, 0.0003], device='cuda:1') 2023-04-28 17:01:20,003 INFO [optim.py:368] (1/8) 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 17:01:25,718 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=66792.0, num_to_drop=1, layers_to_drop={2} 2023-04-28 17:01:38,959 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=66800.0, num_to_drop=1, layers_to_drop={1} 2023-04-28 17:01:41,104 INFO [train.py:904] (1/8) Epoch 7, batch 5900, loss[loss=0.2975, simple_loss=0.3446, pruned_loss=0.1252, over 11699.00 frames. ], tot_loss[loss=0.2477, simple_loss=0.3235, pruned_loss=0.08592, over 3044126.68 frames. ], batch size: 246, lr: 9.70e-03, grad_scale: 8.0 2023-04-28 17:02:07,225 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=66816.0, num_to_drop=1, layers_to_drop={1} 2023-04-28 17:02:56,754 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=66848.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 17:03:02,861 INFO [train.py:904] (1/8) Epoch 7, batch 5950, loss[loss=0.2391, simple_loss=0.3339, pruned_loss=0.07212, over 15392.00 frames. ], tot_loss[loss=0.2466, simple_loss=0.3243, pruned_loss=0.08448, over 3047991.51 frames. ], batch size: 191, lr: 9.69e-03, grad_scale: 8.0 2023-04-28 17:03:04,884 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.55 vs. limit=2.0 2023-04-28 17:03:22,514 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=66864.0, num_to_drop=1, layers_to_drop={1} 2023-04-28 17:03:41,704 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=66877.0, num_to_drop=1, layers_to_drop={2} 2023-04-28 17:03:44,302 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.27 vs. limit=2.0 2023-04-28 17:03:53,690 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.9878, 3.5770, 3.2212, 1.8277, 2.8232, 2.2188, 3.3784, 3.4356], device='cuda:1'), covar=tensor([0.0275, 0.0523, 0.0545, 0.1660, 0.0722, 0.0882, 0.0602, 0.0799], device='cuda:1'), in_proj_covar=tensor([0.0135, 0.0130, 0.0153, 0.0140, 0.0132, 0.0124, 0.0136, 0.0140], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-04-28 17:04:00,336 INFO [optim.py:368] (1/8) 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 17:04:22,260 INFO [train.py:904] (1/8) Epoch 7, batch 6000, loss[loss=0.2311, simple_loss=0.3067, pruned_loss=0.0777, over 16800.00 frames. ], tot_loss[loss=0.2469, simple_loss=0.3241, pruned_loss=0.08479, over 3044821.61 frames. ], batch size: 39, lr: 9.69e-03, grad_scale: 8.0 2023-04-28 17:04:22,260 INFO [train.py:929] (1/8) Computing validation loss 2023-04-28 17:04:32,878 INFO [train.py:938] (1/8) Epoch 7, validation: loss=0.1758, simple_loss=0.2891, pruned_loss=0.03127, over 944034.00 frames. 2023-04-28 17:04:32,878 INFO [train.py:939] (1/8) Maximum memory allocated so far is 17676MB 2023-04-28 17:04:51,820 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=66914.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 17:05:53,513 INFO [train.py:904] (1/8) Epoch 7, batch 6050, loss[loss=0.2204, simple_loss=0.3175, pruned_loss=0.06168, over 16794.00 frames. ], tot_loss[loss=0.2443, simple_loss=0.3214, pruned_loss=0.08362, over 3048500.09 frames. ], batch size: 83, lr: 9.69e-03, grad_scale: 8.0 2023-04-28 17:06:02,567 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=66957.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 17:06:09,561 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=66962.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 17:06:50,891 INFO [optim.py:368] (1/8) 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 17:07:12,273 INFO [train.py:904] (1/8) Epoch 7, batch 6100, loss[loss=0.2793, simple_loss=0.3384, pruned_loss=0.1101, over 11656.00 frames. ], tot_loss[loss=0.2422, simple_loss=0.3203, pruned_loss=0.08202, over 3055977.19 frames. ], batch size: 246, lr: 9.68e-03, grad_scale: 8.0 2023-04-28 17:07:39,961 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=67018.0, num_to_drop=1, layers_to_drop={1} 2023-04-28 17:08:32,436 INFO [train.py:904] (1/8) Epoch 7, batch 6150, loss[loss=0.2082, simple_loss=0.2881, pruned_loss=0.0642, over 16644.00 frames. ], tot_loss[loss=0.2398, simple_loss=0.318, pruned_loss=0.08082, over 3077367.79 frames. ], batch size: 62, lr: 9.68e-03, grad_scale: 8.0 2023-04-28 17:09:23,253 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=67083.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 17:09:29,146 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=67087.0, num_to_drop=1, layers_to_drop={0} 2023-04-28 17:09:31,359 INFO [optim.py:368] (1/8) 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 17:09:51,770 INFO [train.py:904] (1/8) Epoch 7, batch 6200, loss[loss=0.2274, simple_loss=0.3096, pruned_loss=0.07258, over 16440.00 frames. ], tot_loss[loss=0.2392, simple_loss=0.3168, pruned_loss=0.08082, over 3073673.86 frames. ], batch size: 75, lr: 9.67e-03, grad_scale: 4.0 2023-04-28 17:11:02,472 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=67144.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 17:11:14,122 INFO [train.py:904] (1/8) Epoch 7, batch 6250, loss[loss=0.2196, simple_loss=0.3027, pruned_loss=0.0683, over 17050.00 frames. ], tot_loss[loss=0.2373, simple_loss=0.3158, pruned_loss=0.07937, over 3108785.95 frames. ], batch size: 53, lr: 9.67e-03, grad_scale: 4.0 2023-04-28 17:11:37,519 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.1558, 3.2829, 3.5832, 3.5539, 3.5271, 3.2790, 3.3355, 3.3838], device='cuda:1'), covar=tensor([0.0412, 0.0596, 0.0408, 0.0446, 0.0504, 0.0432, 0.0815, 0.0482], device='cuda:1'), in_proj_covar=tensor([0.0279, 0.0280, 0.0284, 0.0272, 0.0331, 0.0301, 0.0404, 0.0242], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:1') 2023-04-28 17:11:37,958 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.30 vs. limit=2.0 2023-04-28 17:11:52,517 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=67177.0, num_to_drop=1, layers_to_drop={0} 2023-04-28 17:12:11,782 INFO [optim.py:368] (1/8) 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 17:12:30,899 INFO [train.py:904] (1/8) Epoch 7, batch 6300, loss[loss=0.2109, simple_loss=0.2947, pruned_loss=0.0636, over 16717.00 frames. ], tot_loss[loss=0.237, simple_loss=0.3158, pruned_loss=0.07907, over 3110040.27 frames. ], batch size: 62, lr: 9.67e-03, grad_scale: 4.0 2023-04-28 17:13:08,168 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=67225.0, num_to_drop=1, layers_to_drop={0} 2023-04-28 17:13:10,642 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.42 vs. limit=2.0 2023-04-28 17:13:49,867 INFO [train.py:904] (1/8) Epoch 7, batch 6350, loss[loss=0.2204, simple_loss=0.3071, pruned_loss=0.06688, over 16875.00 frames. ], tot_loss[loss=0.2388, simple_loss=0.3166, pruned_loss=0.08049, over 3108946.27 frames. ], batch size: 102, lr: 9.66e-03, grad_scale: 4.0 2023-04-28 17:14:47,872 INFO [optim.py:368] (1/8) 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] (1/8) Epoch 7, batch 6400, loss[loss=0.2439, simple_loss=0.3255, pruned_loss=0.0811, over 16230.00 frames. ], tot_loss[loss=0.2393, simple_loss=0.3163, pruned_loss=0.08112, over 3111883.09 frames. ], batch size: 165, lr: 9.66e-03, grad_scale: 8.0 2023-04-28 17:15:22,343 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=67313.0, num_to_drop=1, layers_to_drop={3} 2023-04-28 17:15:46,864 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.24 vs. limit=2.0 2023-04-28 17:16:01,773 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=4.71 vs. limit=5.0 2023-04-28 17:16:20,298 INFO [train.py:904] (1/8) Epoch 7, batch 6450, loss[loss=0.2136, simple_loss=0.2995, pruned_loss=0.06385, over 17262.00 frames. ], tot_loss[loss=0.2383, simple_loss=0.3159, pruned_loss=0.08035, over 3103841.99 frames. ], batch size: 52, lr: 9.66e-03, grad_scale: 4.0 2023-04-28 17:17:16,545 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=67387.0, num_to_drop=1, layers_to_drop={1} 2023-04-28 17:17:18,149 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.95 vs. limit=2.0 2023-04-28 17:17:22,054 INFO [optim.py:368] (1/8) 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 17:17:29,549 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([1.7010, 2.5871, 2.3077, 3.3927, 2.5915, 3.6764, 1.4622, 2.6957], device='cuda:1'), covar=tensor([0.1403, 0.0587, 0.1118, 0.0145, 0.0227, 0.0389, 0.1602, 0.0755], device='cuda:1'), in_proj_covar=tensor([0.0146, 0.0148, 0.0168, 0.0104, 0.0197, 0.0196, 0.0167, 0.0169], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0003, 0.0004], device='cuda:1') 2023-04-28 17:17:38,435 INFO [train.py:904] (1/8) Epoch 7, batch 6500, loss[loss=0.2562, simple_loss=0.3338, pruned_loss=0.08937, over 16540.00 frames. ], tot_loss[loss=0.2366, simple_loss=0.3136, pruned_loss=0.07977, over 3086658.11 frames. ], batch size: 75, lr: 9.65e-03, grad_scale: 4.0 2023-04-28 17:18:29,447 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=67435.0, num_to_drop=1, layers_to_drop={1} 2023-04-28 17:18:36,081 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 7, batch 6550, loss[loss=0.23, simple_loss=0.3302, pruned_loss=0.06491, over 16624.00 frames. ], tot_loss[loss=0.2397, simple_loss=0.3173, pruned_loss=0.08108, over 3094319.20 frames. ], batch size: 76, lr: 9.65e-03, grad_scale: 4.0 2023-04-28 17:19:56,001 INFO [optim.py:368] (1/8) 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] (1/8) Epoch 7, batch 6600, loss[loss=0.27, simple_loss=0.3417, pruned_loss=0.09916, over 15375.00 frames. ], tot_loss[loss=0.2409, simple_loss=0.3196, pruned_loss=0.08114, over 3102558.66 frames. ], batch size: 191, lr: 9.65e-03, grad_scale: 4.0 2023-04-28 17:20:30,798 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=67513.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 17:21:02,281 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.7439, 4.2567, 4.0872, 2.1051, 3.2309, 3.0091, 3.9906, 4.2002], device='cuda:1'), covar=tensor([0.0195, 0.0470, 0.0414, 0.1645, 0.0756, 0.0705, 0.0577, 0.0684], device='cuda:1'), in_proj_covar=tensor([0.0137, 0.0130, 0.0155, 0.0141, 0.0134, 0.0125, 0.0136, 0.0140], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-04-28 17:21:29,300 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.7768, 2.1937, 1.7363, 1.9572, 2.6199, 2.3041, 2.8307, 2.8167], device='cuda:1'), covar=tensor([0.0065, 0.0249, 0.0323, 0.0318, 0.0138, 0.0226, 0.0114, 0.0143], device='cuda:1'), in_proj_covar=tensor([0.0099, 0.0173, 0.0172, 0.0171, 0.0167, 0.0173, 0.0165, 0.0158], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-28 17:21:29,933 INFO [train.py:904] (1/8) Epoch 7, batch 6650, loss[loss=0.2184, simple_loss=0.3019, pruned_loss=0.06744, over 16891.00 frames. ], tot_loss[loss=0.2431, simple_loss=0.3206, pruned_loss=0.08275, over 3080877.58 frames. ], batch size: 109, lr: 9.64e-03, grad_scale: 4.0 2023-04-28 17:21:43,842 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.20 vs. limit=2.0 2023-04-28 17:21:48,520 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([1.5647, 2.6765, 2.4005, 4.0519, 2.9245, 3.9097, 1.3853, 2.8739], device='cuda:1'), covar=tensor([0.1458, 0.0663, 0.1191, 0.0104, 0.0237, 0.0376, 0.1595, 0.0790], device='cuda:1'), in_proj_covar=tensor([0.0148, 0.0147, 0.0169, 0.0104, 0.0197, 0.0197, 0.0166, 0.0169], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0003, 0.0004], device='cuda:1') 2023-04-28 17:22:02,861 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=67574.0, num_to_drop=1, layers_to_drop={3} 2023-04-28 17:22:27,318 INFO [optim.py:368] (1/8) 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 17:22:27,884 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.6377, 4.6071, 4.4707, 4.2827, 4.0798, 4.5078, 4.4061, 4.1706], device='cuda:1'), covar=tensor([0.0470, 0.0296, 0.0227, 0.0209, 0.0879, 0.0341, 0.0325, 0.0588], device='cuda:1'), in_proj_covar=tensor([0.0208, 0.0241, 0.0243, 0.0216, 0.0274, 0.0246, 0.0171, 0.0283], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-28 17:22:43,252 INFO [train.py:904] (1/8) Epoch 7, batch 6700, loss[loss=0.2318, simple_loss=0.3131, pruned_loss=0.07523, over 16473.00 frames. ], tot_loss[loss=0.2418, simple_loss=0.3192, pruned_loss=0.08224, over 3091423.66 frames. ], batch size: 68, lr: 9.64e-03, grad_scale: 4.0 2023-04-28 17:22:59,725 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=67613.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 17:23:47,249 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.1866, 1.6454, 2.3986, 3.0855, 2.8288, 3.4622, 1.8397, 3.3562], device='cuda:1'), covar=tensor([0.0098, 0.0353, 0.0204, 0.0131, 0.0165, 0.0079, 0.0340, 0.0068], device='cuda:1'), in_proj_covar=tensor([0.0138, 0.0154, 0.0137, 0.0137, 0.0146, 0.0102, 0.0155, 0.0094], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001], device='cuda:1') 2023-04-28 17:23:56,587 INFO [train.py:904] (1/8) Epoch 7, batch 6750, loss[loss=0.2884, simple_loss=0.3378, pruned_loss=0.1195, over 11798.00 frames. ], tot_loss[loss=0.2413, simple_loss=0.318, pruned_loss=0.0823, over 3080645.29 frames. ], batch size: 248, lr: 9.64e-03, grad_scale: 4.0 2023-04-28 17:24:10,415 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=67661.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 17:24:37,888 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=67680.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 17:24:54,246 INFO [optim.py:368] (1/8) 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 17:24:55,990 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=67692.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 17:25:10,438 INFO [train.py:904] (1/8) Epoch 7, batch 6800, loss[loss=0.2343, simple_loss=0.3179, pruned_loss=0.07538, over 16755.00 frames. ], tot_loss[loss=0.2403, simple_loss=0.3175, pruned_loss=0.08156, over 3085090.11 frames. ], batch size: 124, lr: 9.63e-03, grad_scale: 8.0 2023-04-28 17:25:40,688 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=3.10 vs. limit=5.0 2023-04-28 17:26:06,255 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=67739.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 17:26:10,104 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 7, batch 6850, loss[loss=0.2245, simple_loss=0.3279, pruned_loss=0.06058, over 16568.00 frames. ], tot_loss[loss=0.242, simple_loss=0.3194, pruned_loss=0.0823, over 3088122.04 frames. ], batch size: 62, lr: 9.63e-03, grad_scale: 8.0 2023-04-28 17:26:25,734 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=67753.0, num_to_drop=1, layers_to_drop={1} 2023-04-28 17:27:10,889 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.8669, 4.9207, 5.4486, 5.4838, 5.3714, 5.0662, 4.9923, 4.8212], device='cuda:1'), covar=tensor([0.0313, 0.0393, 0.0407, 0.0313, 0.0386, 0.0271, 0.0887, 0.0347], device='cuda:1'), in_proj_covar=tensor([0.0275, 0.0271, 0.0275, 0.0271, 0.0325, 0.0293, 0.0393, 0.0237], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0002], device='cuda:1') 2023-04-28 17:27:14,418 INFO [zipformer.py:625] (1/8) 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] (1/8) 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] (1/8) Epoch 7, batch 6900, loss[loss=0.2885, simple_loss=0.354, pruned_loss=0.1115, over 16452.00 frames. ], tot_loss[loss=0.2411, simple_loss=0.3208, pruned_loss=0.08065, over 3118293.79 frames. ], batch size: 146, lr: 9.63e-03, grad_scale: 2.0 2023-04-28 17:28:35,675 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.82 vs. limit=2.0 2023-04-28 17:28:46,544 INFO [train.py:904] (1/8) Epoch 7, batch 6950, loss[loss=0.3075, simple_loss=0.3543, pruned_loss=0.1303, over 11417.00 frames. ], tot_loss[loss=0.2444, simple_loss=0.3228, pruned_loss=0.08301, over 3099605.91 frames. ], batch size: 247, lr: 9.62e-03, grad_scale: 2.0 2023-04-28 17:29:07,891 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.44 vs. limit=2.0 2023-04-28 17:29:13,147 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=67869.0, num_to_drop=1, layers_to_drop={0} 2023-04-28 17:29:38,099 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.93 vs. limit=2.0 2023-04-28 17:29:46,384 INFO [optim.py:368] (1/8) 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] (1/8) Epoch 7, batch 7000, loss[loss=0.2057, simple_loss=0.303, pruned_loss=0.05421, over 16670.00 frames. ], tot_loss[loss=0.2429, simple_loss=0.3226, pruned_loss=0.08163, over 3116718.63 frames. ], batch size: 57, lr: 9.62e-03, grad_scale: 2.0 2023-04-28 17:31:13,270 INFO [train.py:904] (1/8) Epoch 7, batch 7050, loss[loss=0.2969, simple_loss=0.3488, pruned_loss=0.1225, over 11648.00 frames. ], tot_loss[loss=0.2428, simple_loss=0.323, pruned_loss=0.08128, over 3114979.60 frames. ], batch size: 246, lr: 9.61e-03, grad_scale: 2.0 2023-04-28 17:31:57,375 INFO [zipformer.py:625] (1/8) 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] (1/8) 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] (1/8) Epoch 7, batch 7100, loss[loss=0.2215, simple_loss=0.3071, pruned_loss=0.06799, over 16724.00 frames. ], tot_loss[loss=0.2411, simple_loss=0.3209, pruned_loss=0.08065, over 3115045.53 frames. ], batch size: 89, lr: 9.61e-03, grad_scale: 2.0 2023-04-28 17:33:16,915 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.4779, 2.1029, 1.5338, 1.8060, 2.4000, 2.1386, 2.5602, 2.6510], device='cuda:1'), covar=tensor([0.0071, 0.0215, 0.0329, 0.0299, 0.0152, 0.0236, 0.0122, 0.0128], device='cuda:1'), in_proj_covar=tensor([0.0101, 0.0173, 0.0175, 0.0172, 0.0169, 0.0176, 0.0166, 0.0160], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-28 17:33:20,753 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=68036.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 17:33:29,664 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=68042.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 17:33:39,220 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=68048.0, num_to_drop=1, layers_to_drop={1} 2023-04-28 17:33:44,554 INFO [train.py:904] (1/8) Epoch 7, batch 7150, loss[loss=0.2368, simple_loss=0.3179, pruned_loss=0.07786, over 16409.00 frames. ], tot_loss[loss=0.2407, simple_loss=0.3194, pruned_loss=0.08097, over 3118335.21 frames. ], batch size: 146, lr: 9.61e-03, grad_scale: 2.0 2023-04-28 17:34:25,469 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.21 vs. limit=2.0 2023-04-28 17:34:36,417 INFO [zipformer.py:625] (1/8) 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] (1/8) Clipping_scale=2.0, grad-norm quartiles 2.291e+02 3.518e+02 4.375e+02 5.869e+02 1.278e+03, threshold=8.749e+02, percent-clipped=4.0 2023-04-28 17:34:58,213 INFO [train.py:904] (1/8) Epoch 7, batch 7200, loss[loss=0.2349, simple_loss=0.3233, pruned_loss=0.07327, over 16711.00 frames. ], tot_loss[loss=0.237, simple_loss=0.3165, pruned_loss=0.07877, over 3120427.53 frames. ], batch size: 134, lr: 9.60e-03, grad_scale: 4.0 2023-04-28 17:35:28,050 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.5190, 2.6931, 1.7338, 2.7380, 2.0798, 2.7707, 2.0500, 2.3376], device='cuda:1'), covar=tensor([0.0186, 0.0311, 0.1261, 0.0100, 0.0663, 0.0420, 0.1098, 0.0559], device='cuda:1'), in_proj_covar=tensor([0.0134, 0.0155, 0.0179, 0.0094, 0.0162, 0.0192, 0.0189, 0.0168], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-04-28 17:35:29,413 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.7394, 4.3674, 4.4862, 2.3043, 3.3979, 2.8571, 3.9569, 4.3536], device='cuda:1'), covar=tensor([0.0215, 0.0460, 0.0322, 0.1562, 0.0674, 0.0779, 0.0587, 0.0800], device='cuda:1'), in_proj_covar=tensor([0.0136, 0.0129, 0.0153, 0.0139, 0.0133, 0.0124, 0.0135, 0.0140], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-04-28 17:36:11,153 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=68148.0, num_to_drop=1, layers_to_drop={1} 2023-04-28 17:36:16,028 INFO [train.py:904] (1/8) Epoch 7, batch 7250, loss[loss=0.1895, simple_loss=0.2705, pruned_loss=0.05425, over 17244.00 frames. ], tot_loss[loss=0.2345, simple_loss=0.3138, pruned_loss=0.07761, over 3098804.71 frames. ], batch size: 52, lr: 9.60e-03, grad_scale: 4.0 2023-04-28 17:36:41,062 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=68169.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 17:37:16,279 INFO [optim.py:368] (1/8) Clipping_scale=2.0, grad-norm quartiles 2.105e+02 3.214e+02 3.991e+02 5.016e+02 1.051e+03, threshold=7.982e+02, percent-clipped=2.0 2023-04-28 17:37:29,948 INFO [train.py:904] (1/8) Epoch 7, batch 7300, loss[loss=0.224, simple_loss=0.3097, pruned_loss=0.0691, over 16587.00 frames. ], tot_loss[loss=0.2337, simple_loss=0.3129, pruned_loss=0.0773, over 3105354.97 frames. ], batch size: 57, lr: 9.60e-03, grad_scale: 4.0 2023-04-28 17:37:52,952 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=68217.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 17:38:22,432 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.8552, 3.9620, 2.1422, 4.5367, 2.7144, 4.4498, 2.3028, 3.0555], device='cuda:1'), covar=tensor([0.0154, 0.0265, 0.1464, 0.0041, 0.0803, 0.0264, 0.1381, 0.0586], device='cuda:1'), in_proj_covar=tensor([0.0133, 0.0153, 0.0179, 0.0094, 0.0161, 0.0190, 0.0187, 0.0167], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-04-28 17:38:44,625 INFO [train.py:904] (1/8) Epoch 7, batch 7350, loss[loss=0.1954, simple_loss=0.2847, pruned_loss=0.05304, over 16857.00 frames. ], tot_loss[loss=0.235, simple_loss=0.3135, pruned_loss=0.07827, over 3062380.24 frames. ], batch size: 96, lr: 9.59e-03, grad_scale: 4.0 2023-04-28 17:39:10,004 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.2067, 3.6357, 3.7937, 2.3196, 3.4033, 3.6859, 3.4237, 1.8618], device='cuda:1'), covar=tensor([0.0350, 0.0022, 0.0024, 0.0261, 0.0050, 0.0070, 0.0047, 0.0333], device='cuda:1'), in_proj_covar=tensor([0.0118, 0.0055, 0.0059, 0.0115, 0.0064, 0.0076, 0.0066, 0.0110], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0001, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:1') 2023-04-28 17:39:30,486 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.4410, 3.4524, 2.8388, 2.1181, 2.4781, 2.2019, 3.5322, 3.3704], device='cuda:1'), covar=tensor([0.2604, 0.0723, 0.1418, 0.2054, 0.2026, 0.1615, 0.0516, 0.0891], device='cuda:1'), in_proj_covar=tensor([0.0294, 0.0254, 0.0278, 0.0263, 0.0286, 0.0212, 0.0257, 0.0274], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-28 17:39:48,567 INFO [optim.py:368] (1/8) Clipping_scale=2.0, grad-norm quartiles 1.908e+02 3.085e+02 3.772e+02 4.630e+02 1.239e+03, threshold=7.544e+02, percent-clipped=2.0 2023-04-28 17:40:02,137 INFO [train.py:904] (1/8) Epoch 7, batch 7400, loss[loss=0.269, simple_loss=0.3538, pruned_loss=0.09212, over 16760.00 frames. ], tot_loss[loss=0.2371, simple_loss=0.3153, pruned_loss=0.07938, over 3064477.11 frames. ], batch size: 62, lr: 9.59e-03, grad_scale: 4.0 2023-04-28 17:40:40,786 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.5946, 3.7005, 2.9744, 2.2430, 2.6586, 2.1723, 3.9288, 3.6157], device='cuda:1'), covar=tensor([0.2486, 0.0746, 0.1395, 0.1936, 0.2173, 0.1694, 0.0450, 0.0850], device='cuda:1'), in_proj_covar=tensor([0.0294, 0.0253, 0.0278, 0.0262, 0.0285, 0.0211, 0.0256, 0.0273], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-28 17:40:53,652 INFO [zipformer.py:625] (1/8) 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] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=68337.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 17:41:14,099 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=68348.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 17:41:19,093 INFO [train.py:904] (1/8) Epoch 7, batch 7450, loss[loss=0.2556, simple_loss=0.3385, pruned_loss=0.08634, over 16300.00 frames. ], tot_loss[loss=0.2399, simple_loss=0.3174, pruned_loss=0.08119, over 3061812.26 frames. ], batch size: 165, lr: 9.59e-03, grad_scale: 4.0 2023-04-28 17:41:28,006 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([5.0617, 5.0639, 4.8588, 4.2294, 4.8585, 1.7319, 4.6521, 4.7428], device='cuda:1'), covar=tensor([0.0057, 0.0052, 0.0100, 0.0279, 0.0058, 0.2031, 0.0106, 0.0126], device='cuda:1'), in_proj_covar=tensor([0.0105, 0.0092, 0.0140, 0.0134, 0.0106, 0.0157, 0.0124, 0.0130], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-28 17:42:11,320 INFO [zipformer.py:625] (1/8) 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] (1/8) 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] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=68396.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 17:42:35,723 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=68399.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 17:42:39,845 INFO [train.py:904] (1/8) Epoch 7, batch 7500, loss[loss=0.202, simple_loss=0.3, pruned_loss=0.05205, over 16924.00 frames. ], tot_loss[loss=0.2394, simple_loss=0.3178, pruned_loss=0.08047, over 3073570.73 frames. ], batch size: 96, lr: 9.58e-03, grad_scale: 4.0 2023-04-28 17:42:42,314 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.6638, 3.6862, 4.1132, 4.0345, 4.0232, 3.7397, 3.8072, 3.7547], device='cuda:1'), covar=tensor([0.0306, 0.0495, 0.0298, 0.0425, 0.0495, 0.0355, 0.0776, 0.0410], device='cuda:1'), in_proj_covar=tensor([0.0271, 0.0271, 0.0273, 0.0267, 0.0321, 0.0294, 0.0389, 0.0241], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003], device='cuda:1') 2023-04-28 17:43:43,716 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=68443.0, num_to_drop=1, layers_to_drop={1} 2023-04-28 17:43:57,919 INFO [train.py:904] (1/8) Epoch 7, batch 7550, loss[loss=0.2207, simple_loss=0.2926, pruned_loss=0.07444, over 17041.00 frames. ], tot_loss[loss=0.2386, simple_loss=0.3164, pruned_loss=0.08041, over 3066353.09 frames. ], batch size: 55, lr: 9.58e-03, grad_scale: 4.0 2023-04-28 17:44:10,814 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=68460.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 17:44:59,736 INFO [optim.py:368] (1/8) 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 17:45:12,776 INFO [train.py:904] (1/8) Epoch 7, batch 7600, loss[loss=0.243, simple_loss=0.3162, pruned_loss=0.08485, over 15370.00 frames. ], tot_loss[loss=0.2388, simple_loss=0.3161, pruned_loss=0.08075, over 3058242.17 frames. ], batch size: 190, lr: 9.58e-03, grad_scale: 8.0 2023-04-28 17:45:29,701 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.7147, 4.7256, 4.5573, 4.4001, 4.1774, 4.6119, 4.5758, 4.3254], device='cuda:1'), covar=tensor([0.0592, 0.0325, 0.0250, 0.0234, 0.0952, 0.0391, 0.0327, 0.0568], device='cuda:1'), in_proj_covar=tensor([0.0202, 0.0235, 0.0236, 0.0209, 0.0263, 0.0238, 0.0165, 0.0271], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-28 17:46:28,266 INFO [train.py:904] (1/8) Epoch 7, batch 7650, loss[loss=0.2402, simple_loss=0.3226, pruned_loss=0.07889, over 16739.00 frames. ], tot_loss[loss=0.241, simple_loss=0.3174, pruned_loss=0.08227, over 3040806.52 frames. ], batch size: 124, lr: 9.57e-03, grad_scale: 8.0 2023-04-28 17:47:08,534 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=68578.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 17:47:29,675 INFO [optim.py:368] (1/8) Clipping_scale=2.0, grad-norm quartiles 2.494e+02 3.559e+02 4.201e+02 5.687e+02 1.184e+03, threshold=8.403e+02, percent-clipped=4.0 2023-04-28 17:47:42,699 INFO [train.py:904] (1/8) Epoch 7, batch 7700, loss[loss=0.2382, simple_loss=0.3311, pruned_loss=0.07268, over 16817.00 frames. ], tot_loss[loss=0.2406, simple_loss=0.317, pruned_loss=0.08213, over 3054546.98 frames. ], batch size: 102, lr: 9.57e-03, grad_scale: 8.0 2023-04-28 17:48:34,696 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=68637.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 17:48:38,389 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=68639.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 17:48:43,246 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([5.0641, 4.9214, 4.8921, 4.6868, 4.4622, 4.8389, 4.8664, 4.5970], device='cuda:1'), covar=tensor([0.0486, 0.0435, 0.0215, 0.0206, 0.0872, 0.0374, 0.0238, 0.0515], device='cuda:1'), in_proj_covar=tensor([0.0203, 0.0236, 0.0235, 0.0209, 0.0265, 0.0238, 0.0165, 0.0272], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-28 17:48:57,123 INFO [train.py:904] (1/8) Epoch 7, batch 7750, loss[loss=0.2364, simple_loss=0.3188, pruned_loss=0.07699, over 17195.00 frames. ], tot_loss[loss=0.2405, simple_loss=0.317, pruned_loss=0.08198, over 3056156.50 frames. ], batch size: 46, lr: 9.57e-03, grad_scale: 8.0 2023-04-28 17:49:44,494 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=68685.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 17:49:44,674 INFO [zipformer.py:625] (1/8) 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] (1/8) Clipping_scale=2.0, grad-norm quartiles 2.246e+02 3.570e+02 4.272e+02 5.150e+02 9.085e+02, threshold=8.545e+02, percent-clipped=2.0 2023-04-28 17:50:11,059 INFO [train.py:904] (1/8) Epoch 7, batch 7800, loss[loss=0.3107, simple_loss=0.3582, pruned_loss=0.1316, over 11469.00 frames. ], tot_loss[loss=0.2396, simple_loss=0.3167, pruned_loss=0.08127, over 3076437.77 frames. ], batch size: 246, lr: 9.56e-03, grad_scale: 8.0 2023-04-28 17:51:12,790 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=68743.0, num_to_drop=1, layers_to_drop={1} 2023-04-28 17:51:17,596 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=68746.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 17:51:25,160 INFO [train.py:904] (1/8) Epoch 7, batch 7850, loss[loss=0.2338, simple_loss=0.3187, pruned_loss=0.07446, over 16367.00 frames. ], tot_loss[loss=0.2405, simple_loss=0.318, pruned_loss=0.08146, over 3071316.82 frames. ], batch size: 146, lr: 9.56e-03, grad_scale: 8.0 2023-04-28 17:51:30,899 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=68755.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 17:52:24,299 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=68791.0, num_to_drop=1, layers_to_drop={1} 2023-04-28 17:52:26,860 INFO [optim.py:368] (1/8) 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] (1/8) Epoch 7, batch 7900, loss[loss=0.2334, simple_loss=0.3165, pruned_loss=0.07513, over 16482.00 frames. ], tot_loss[loss=0.2391, simple_loss=0.3168, pruned_loss=0.0807, over 3078464.24 frames. ], batch size: 146, lr: 9.56e-03, grad_scale: 8.0 2023-04-28 17:53:58,005 INFO [train.py:904] (1/8) Epoch 7, batch 7950, loss[loss=0.3065, simple_loss=0.3542, pruned_loss=0.1294, over 11717.00 frames. ], tot_loss[loss=0.2395, simple_loss=0.317, pruned_loss=0.08099, over 3079494.96 frames. ], batch size: 248, lr: 9.55e-03, grad_scale: 2.0 2023-04-28 17:54:10,622 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.9824, 2.9497, 2.7374, 1.9557, 2.6200, 2.0687, 2.7243, 3.0255], device='cuda:1'), covar=tensor([0.0269, 0.0497, 0.0499, 0.1590, 0.0696, 0.0978, 0.0572, 0.0543], device='cuda:1'), in_proj_covar=tensor([0.0136, 0.0131, 0.0153, 0.0140, 0.0134, 0.0123, 0.0135, 0.0138], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-04-28 17:55:03,363 INFO [optim.py:368] (1/8) 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 17:55:12,769 INFO [train.py:904] (1/8) Epoch 7, batch 8000, loss[loss=0.2409, simple_loss=0.3164, pruned_loss=0.08269, over 16691.00 frames. ], tot_loss[loss=0.2414, simple_loss=0.3181, pruned_loss=0.08235, over 3062035.09 frames. ], batch size: 134, lr: 9.55e-03, grad_scale: 4.0 2023-04-28 17:56:00,096 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=68934.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 17:56:09,608 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.0155, 3.1079, 1.7291, 3.2468, 2.2914, 3.2871, 1.9066, 2.5376], device='cuda:1'), covar=tensor([0.0215, 0.0347, 0.1550, 0.0107, 0.0716, 0.0538, 0.1408, 0.0618], device='cuda:1'), in_proj_covar=tensor([0.0136, 0.0155, 0.0179, 0.0095, 0.0162, 0.0192, 0.0188, 0.0168], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-04-28 17:56:25,018 INFO [train.py:904] (1/8) Epoch 7, batch 8050, loss[loss=0.2472, simple_loss=0.334, pruned_loss=0.08023, over 16609.00 frames. ], tot_loss[loss=0.2415, simple_loss=0.3182, pruned_loss=0.08239, over 3059956.73 frames. ], batch size: 68, lr: 9.54e-03, grad_scale: 4.0 2023-04-28 17:57:30,025 INFO [optim.py:368] (1/8) 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 17:57:41,497 INFO [train.py:904] (1/8) Epoch 7, batch 8100, loss[loss=0.2376, simple_loss=0.3177, pruned_loss=0.07873, over 16257.00 frames. ], tot_loss[loss=0.2402, simple_loss=0.3175, pruned_loss=0.08148, over 3059688.10 frames. ], batch size: 35, lr: 9.54e-03, grad_scale: 4.0 2023-04-28 17:58:40,375 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=69041.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 17:58:43,435 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 7, batch 8150, loss[loss=0.2654, simple_loss=0.3168, pruned_loss=0.107, over 11354.00 frames. ], tot_loss[loss=0.2395, simple_loss=0.3164, pruned_loss=0.08134, over 3049563.16 frames. ], batch size: 247, lr: 9.54e-03, grad_scale: 4.0 2023-04-28 17:58:57,621 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.5695, 2.6815, 1.7104, 2.7651, 2.1338, 2.7858, 1.9618, 2.3855], device='cuda:1'), covar=tensor([0.0235, 0.0369, 0.1272, 0.0121, 0.0620, 0.0495, 0.1153, 0.0539], device='cuda:1'), in_proj_covar=tensor([0.0136, 0.0156, 0.0180, 0.0096, 0.0163, 0.0193, 0.0189, 0.0168], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-04-28 17:59:00,737 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=69055.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 17:59:03,748 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.4975, 1.5893, 1.9656, 2.4567, 2.3618, 2.7981, 1.7756, 2.6736], device='cuda:1'), covar=tensor([0.0098, 0.0257, 0.0172, 0.0141, 0.0149, 0.0079, 0.0250, 0.0061], device='cuda:1'), in_proj_covar=tensor([0.0135, 0.0151, 0.0132, 0.0133, 0.0145, 0.0100, 0.0150, 0.0091], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001], device='cuda:1') 2023-04-28 17:59:18,624 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.0826, 1.8749, 2.1787, 3.4815, 1.8577, 2.3812, 2.1158, 2.0184], device='cuda:1'), covar=tensor([0.0808, 0.2724, 0.1617, 0.0399, 0.3275, 0.1615, 0.2379, 0.2676], device='cuda:1'), in_proj_covar=tensor([0.0339, 0.0359, 0.0298, 0.0323, 0.0395, 0.0386, 0.0318, 0.0420], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-28 17:59:35,261 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.8526, 4.1269, 3.9265, 3.9545, 3.6072, 3.7403, 3.8207, 4.0726], device='cuda:1'), covar=tensor([0.0919, 0.0739, 0.0836, 0.0573, 0.0778, 0.1288, 0.0816, 0.0973], device='cuda:1'), in_proj_covar=tensor([0.0452, 0.0567, 0.0486, 0.0377, 0.0357, 0.0387, 0.0473, 0.0421], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-28 17:59:36,819 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.26 vs. limit=2.0 2023-04-28 17:59:59,698 INFO [optim.py:368] (1/8) 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 18:00:10,162 INFO [train.py:904] (1/8) Epoch 7, batch 8200, loss[loss=0.258, simple_loss=0.3151, pruned_loss=0.1005, over 11412.00 frames. ], tot_loss[loss=0.2366, simple_loss=0.3133, pruned_loss=0.08001, over 3061595.02 frames. ], batch size: 246, lr: 9.53e-03, grad_scale: 4.0 2023-04-28 18:00:12,049 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=69103.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 18:00:13,587 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=69104.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 18:01:21,788 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.2063, 3.9207, 4.2072, 4.3870, 4.5100, 4.0915, 4.4913, 4.5214], device='cuda:1'), covar=tensor([0.1228, 0.0969, 0.1390, 0.0605, 0.0500, 0.0897, 0.0519, 0.0518], device='cuda:1'), in_proj_covar=tensor([0.0445, 0.0546, 0.0675, 0.0548, 0.0413, 0.0419, 0.0439, 0.0473], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-28 18:01:30,738 INFO [train.py:904] (1/8) Epoch 7, batch 8250, loss[loss=0.193, simple_loss=0.2848, pruned_loss=0.05058, over 16714.00 frames. ], tot_loss[loss=0.2336, simple_loss=0.3121, pruned_loss=0.07754, over 3057752.37 frames. ], batch size: 83, lr: 9.53e-03, grad_scale: 4.0 2023-04-28 18:01:47,250 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=69162.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 18:01:58,182 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=69168.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 18:02:40,829 INFO [optim.py:368] (1/8) 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 18:02:52,451 INFO [train.py:904] (1/8) Epoch 7, batch 8300, loss[loss=0.1978, simple_loss=0.2858, pruned_loss=0.05491, over 16730.00 frames. ], tot_loss[loss=0.2288, simple_loss=0.3085, pruned_loss=0.0745, over 3032460.37 frames. ], batch size: 57, lr: 9.53e-03, grad_scale: 4.0 2023-04-28 18:03:25,494 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=69223.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 18:03:29,148 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=69225.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 18:03:34,980 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=69229.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 18:03:42,649 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=69234.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 18:04:00,668 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 7, batch 8350, loss[loss=0.2037, simple_loss=0.2968, pruned_loss=0.0553, over 16201.00 frames. ], tot_loss[loss=0.2245, simple_loss=0.3066, pruned_loss=0.07117, over 3042927.20 frames. ], batch size: 165, lr: 9.52e-03, grad_scale: 4.0 2023-04-28 18:04:46,447 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=69273.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 18:04:59,334 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=69282.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 18:05:06,578 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=69286.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 18:05:20,092 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.1832, 1.2759, 1.7936, 2.0928, 2.1508, 2.2159, 1.4319, 2.2004], device='cuda:1'), covar=tensor([0.0121, 0.0337, 0.0168, 0.0169, 0.0157, 0.0152, 0.0344, 0.0068], device='cuda:1'), in_proj_covar=tensor([0.0137, 0.0154, 0.0134, 0.0135, 0.0144, 0.0100, 0.0152, 0.0092], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001], device='cuda:1') 2023-04-28 18:05:20,697 INFO [optim.py:368] (1/8) 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] (1/8) Epoch 7, batch 8400, loss[loss=0.1862, simple_loss=0.2655, pruned_loss=0.05341, over 12245.00 frames. ], tot_loss[loss=0.2196, simple_loss=0.3025, pruned_loss=0.06836, over 3021776.16 frames. ], batch size: 248, lr: 9.52e-03, grad_scale: 8.0 2023-04-28 18:05:39,352 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=69306.0, num_to_drop=1, layers_to_drop={3} 2023-04-28 18:06:02,086 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=69320.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 18:06:23,830 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=69334.0, num_to_drop=1, layers_to_drop={3} 2023-04-28 18:06:35,209 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=69341.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 18:06:42,957 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=5.01 vs. limit=5.0 2023-04-28 18:06:52,098 INFO [train.py:904] (1/8) Epoch 7, batch 8450, loss[loss=0.1998, simple_loss=0.289, pruned_loss=0.05529, over 16696.00 frames. ], tot_loss[loss=0.2169, simple_loss=0.301, pruned_loss=0.06634, over 3052762.11 frames. ], batch size: 124, lr: 9.52e-03, grad_scale: 8.0 2023-04-28 18:07:18,139 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=69368.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 18:07:38,252 INFO [zipformer.py:625] (1/8) 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] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=69389.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 18:07:58,563 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=69394.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 18:07:59,249 INFO [optim.py:368] (1/8) 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 18:08:06,556 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=69399.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 18:08:10,419 INFO [train.py:904] (1/8) Epoch 7, batch 8500, loss[loss=0.1852, simple_loss=0.2662, pruned_loss=0.05211, over 12100.00 frames. ], tot_loss[loss=0.2113, simple_loss=0.2964, pruned_loss=0.06313, over 3067042.81 frames. ], batch size: 247, lr: 9.51e-03, grad_scale: 8.0 2023-04-28 18:08:55,069 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=69429.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 18:09:33,176 INFO [train.py:904] (1/8) Epoch 7, batch 8550, loss[loss=0.2133, simple_loss=0.301, pruned_loss=0.06284, over 15228.00 frames. ], tot_loss[loss=0.2084, simple_loss=0.2935, pruned_loss=0.06165, over 3043679.24 frames. ], batch size: 190, lr: 9.51e-03, grad_scale: 8.0 2023-04-28 18:09:41,190 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=69455.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 18:10:58,769 INFO [optim.py:368] (1/8) 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 18:11:12,035 INFO [train.py:904] (1/8) Epoch 7, batch 8600, loss[loss=0.1898, simple_loss=0.2791, pruned_loss=0.05027, over 16524.00 frames. ], tot_loss[loss=0.2078, simple_loss=0.2938, pruned_loss=0.06091, over 3030375.05 frames. ], batch size: 62, lr: 9.51e-03, grad_scale: 8.0 2023-04-28 18:11:46,305 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=69518.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 18:11:55,636 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 7, batch 8650, loss[loss=0.2004, simple_loss=0.2897, pruned_loss=0.05551, over 16871.00 frames. ], tot_loss[loss=0.2049, simple_loss=0.2918, pruned_loss=0.05897, over 3049553.37 frames. ], batch size: 116, lr: 9.50e-03, grad_scale: 8.0 2023-04-28 18:13:15,896 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.0867, 2.6853, 2.6851, 1.8107, 2.8791, 2.9151, 2.5186, 2.4927], device='cuda:1'), covar=tensor([0.0564, 0.0158, 0.0157, 0.0906, 0.0067, 0.0107, 0.0347, 0.0351], device='cuda:1'), in_proj_covar=tensor([0.0136, 0.0090, 0.0077, 0.0134, 0.0063, 0.0080, 0.0112, 0.0120], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003, 0.0003], device='cuda:1') 2023-04-28 18:13:53,654 INFO [zipformer.py:625] (1/8) 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] (1/8) 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] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=69601.0, num_to_drop=1, layers_to_drop={0} 2023-04-28 18:14:32,611 INFO [train.py:904] (1/8) Epoch 7, batch 8700, loss[loss=0.1853, simple_loss=0.2759, pruned_loss=0.04731, over 16394.00 frames. ], tot_loss[loss=0.202, simple_loss=0.2891, pruned_loss=0.0574, over 3062030.24 frames. ], batch size: 146, lr: 9.50e-03, grad_scale: 8.0 2023-04-28 18:15:22,789 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=69629.0, num_to_drop=1, layers_to_drop={3} 2023-04-28 18:15:38,542 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.8221, 4.8032, 4.6576, 4.3959, 4.2989, 4.6840, 4.6207, 4.3880], device='cuda:1'), covar=tensor([0.0421, 0.0281, 0.0189, 0.0214, 0.0746, 0.0286, 0.0231, 0.0484], device='cuda:1'), in_proj_covar=tensor([0.0198, 0.0228, 0.0228, 0.0204, 0.0251, 0.0229, 0.0161, 0.0266], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-28 18:16:08,632 INFO [train.py:904] (1/8) Epoch 7, batch 8750, loss[loss=0.2103, simple_loss=0.3017, pruned_loss=0.0595, over 16888.00 frames. ], tot_loss[loss=0.2007, simple_loss=0.2883, pruned_loss=0.05658, over 3051672.04 frames. ], batch size: 116, lr: 9.50e-03, grad_scale: 8.0 2023-04-28 18:17:04,337 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=69674.0, num_to_drop=1, layers_to_drop={0} 2023-04-28 18:17:07,493 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=69676.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 18:17:19,739 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.2210, 3.3424, 3.6169, 3.5690, 3.5886, 3.3726, 3.4352, 3.4603], device='cuda:1'), covar=tensor([0.0303, 0.0541, 0.0393, 0.0518, 0.0439, 0.0410, 0.0773, 0.0385], device='cuda:1'), in_proj_covar=tensor([0.0258, 0.0263, 0.0262, 0.0256, 0.0304, 0.0281, 0.0371, 0.0231], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0002], device='cuda:1') 2023-04-28 18:17:45,985 INFO [optim.py:368] (1/8) 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 18:17:55,725 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=69699.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 18:18:01,174 INFO [train.py:904] (1/8) Epoch 7, batch 8800, loss[loss=0.1751, simple_loss=0.2744, pruned_loss=0.03795, over 16634.00 frames. ], tot_loss[loss=0.1988, simple_loss=0.2866, pruned_loss=0.05552, over 3035462.73 frames. ], batch size: 89, lr: 9.49e-03, grad_scale: 8.0 2023-04-28 18:18:29,858 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([5.0593, 5.3805, 5.1628, 5.1493, 4.8213, 4.7328, 4.8484, 5.3778], device='cuda:1'), covar=tensor([0.0832, 0.0674, 0.0704, 0.0485, 0.0640, 0.0683, 0.0803, 0.0812], device='cuda:1'), in_proj_covar=tensor([0.0428, 0.0546, 0.0461, 0.0365, 0.0343, 0.0370, 0.0456, 0.0402], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-28 18:18:47,926 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=69724.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 18:19:09,259 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.9766, 3.9907, 3.7957, 3.6785, 3.5727, 3.9040, 3.6979, 3.7528], device='cuda:1'), covar=tensor([0.0431, 0.0244, 0.0221, 0.0186, 0.0616, 0.0289, 0.0659, 0.0405], device='cuda:1'), in_proj_covar=tensor([0.0197, 0.0227, 0.0229, 0.0203, 0.0251, 0.0228, 0.0160, 0.0264], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-28 18:19:11,462 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=69735.0, num_to_drop=1, layers_to_drop={2} 2023-04-28 18:19:26,254 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.6098, 3.5470, 2.6645, 2.1050, 2.3486, 2.0651, 3.7679, 3.3427], device='cuda:1'), covar=tensor([0.2394, 0.0700, 0.1460, 0.2014, 0.2165, 0.1744, 0.0364, 0.0898], device='cuda:1'), in_proj_covar=tensor([0.0278, 0.0239, 0.0263, 0.0248, 0.0252, 0.0202, 0.0239, 0.0253], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-28 18:19:26,441 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.92 vs. limit=2.0 2023-04-28 18:19:36,495 INFO [zipformer.py:625] (1/8) 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] (1/8) 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] (1/8) Epoch 7, batch 8850, loss[loss=0.1834, simple_loss=0.2861, pruned_loss=0.04034, over 15161.00 frames. ], tot_loss[loss=0.1995, simple_loss=0.2893, pruned_loss=0.05483, over 3037707.85 frames. ], batch size: 190, lr: 9.49e-03, grad_scale: 4.0 2023-04-28 18:19:51,464 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.3973, 3.3651, 3.4136, 3.4969, 3.5127, 3.1998, 3.5485, 3.5916], device='cuda:1'), covar=tensor([0.0824, 0.0673, 0.0975, 0.0578, 0.0576, 0.2107, 0.0626, 0.0489], device='cuda:1'), in_proj_covar=tensor([0.0417, 0.0516, 0.0633, 0.0523, 0.0390, 0.0398, 0.0409, 0.0451], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-28 18:20:01,411 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.84 vs. limit=2.0 2023-04-28 18:21:21,321 INFO [optim.py:368] (1/8) Clipping_scale=2.0, grad-norm quartiles 1.819e+02 2.840e+02 3.473e+02 4.436e+02 7.105e+02, threshold=6.946e+02, percent-clipped=4.0 2023-04-28 18:21:34,922 INFO [train.py:904] (1/8) Epoch 7, batch 8900, loss[loss=0.1792, simple_loss=0.2674, pruned_loss=0.04553, over 12419.00 frames. ], tot_loss[loss=0.1988, simple_loss=0.2893, pruned_loss=0.05416, over 3039609.85 frames. ], batch size: 247, lr: 9.49e-03, grad_scale: 4.0 2023-04-28 18:22:06,695 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=69818.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 18:22:08,435 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.7847, 3.6417, 3.8320, 3.6873, 3.8968, 4.2711, 3.9587, 3.6896], device='cuda:1'), covar=tensor([0.1699, 0.2319, 0.1774, 0.2248, 0.2706, 0.1525, 0.1304, 0.2783], device='cuda:1'), in_proj_covar=tensor([0.0288, 0.0402, 0.0418, 0.0355, 0.0464, 0.0451, 0.0336, 0.0471], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002], device='cuda:1') 2023-04-28 18:22:21,384 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 7, batch 8950, loss[loss=0.1878, simple_loss=0.2759, pruned_loss=0.04987, over 16867.00 frames. ], tot_loss[loss=0.1988, simple_loss=0.2888, pruned_loss=0.05441, over 3053644.14 frames. ], batch size: 124, lr: 9.48e-03, grad_scale: 4.0 2023-04-28 18:24:09,508 INFO [zipformer.py:625] (1/8) 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] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=69872.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 18:24:42,905 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=69881.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 18:25:14,045 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.5354, 4.8436, 4.4845, 4.3899, 3.6778, 4.6480, 4.7039, 4.1812], device='cuda:1'), covar=tensor([0.0813, 0.0481, 0.0370, 0.0256, 0.1483, 0.0489, 0.0270, 0.0691], device='cuda:1'), in_proj_covar=tensor([0.0195, 0.0228, 0.0227, 0.0202, 0.0249, 0.0230, 0.0159, 0.0264], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-28 18:25:14,733 INFO [optim.py:368] (1/8) 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 18:25:27,282 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=69901.0, num_to_drop=1, layers_to_drop={0} 2023-04-28 18:25:27,970 INFO [train.py:904] (1/8) Epoch 7, batch 9000, loss[loss=0.1669, simple_loss=0.26, pruned_loss=0.03695, over 16887.00 frames. ], tot_loss[loss=0.1958, simple_loss=0.2857, pruned_loss=0.05294, over 3075624.83 frames. ], batch size: 96, lr: 9.48e-03, grad_scale: 4.0 2023-04-28 18:25:27,970 INFO [train.py:929] (1/8) Computing validation loss 2023-04-28 18:25:37,199 INFO [train.py:938] (1/8) Epoch 7, validation: loss=0.1647, simple_loss=0.2682, pruned_loss=0.03062, over 944034.00 frames. 2023-04-28 18:25:37,199 INFO [train.py:939] (1/8) Maximum memory allocated so far is 17676MB 2023-04-28 18:26:26,882 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([1.9738, 1.6827, 1.4249, 1.4160, 1.8763, 1.6158, 1.8061, 1.9552], device='cuda:1'), covar=tensor([0.0049, 0.0178, 0.0243, 0.0233, 0.0115, 0.0170, 0.0113, 0.0114], device='cuda:1'), in_proj_covar=tensor([0.0095, 0.0172, 0.0168, 0.0167, 0.0165, 0.0170, 0.0155, 0.0153], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-28 18:26:33,159 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=69929.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 18:26:33,278 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=69929.0, num_to_drop=1, layers_to_drop={2} 2023-04-28 18:27:14,335 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=69949.0, num_to_drop=1, layers_to_drop={1} 2023-04-28 18:27:19,718 INFO [train.py:904] (1/8) Epoch 7, batch 9050, loss[loss=0.1732, simple_loss=0.2696, pruned_loss=0.03839, over 16847.00 frames. ], tot_loss[loss=0.197, simple_loss=0.2868, pruned_loss=0.05361, over 3067069.20 frames. ], batch size: 83, lr: 9.48e-03, grad_scale: 4.0 2023-04-28 18:28:00,350 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.6333, 1.8948, 2.1556, 4.1511, 1.7523, 2.4130, 2.0917, 2.0256], device='cuda:1'), covar=tensor([0.0734, 0.3389, 0.1850, 0.0334, 0.4727, 0.2085, 0.2975, 0.3443], device='cuda:1'), in_proj_covar=tensor([0.0322, 0.0339, 0.0290, 0.0304, 0.0380, 0.0363, 0.0307, 0.0396], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-28 18:28:08,400 INFO [zipformer.py:625] (1/8) 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] (1/8) 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] (1/8) 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 18:29:06,535 INFO [train.py:904] (1/8) Epoch 7, batch 9100, loss[loss=0.2104, simple_loss=0.2879, pruned_loss=0.06643, over 12653.00 frames. ], tot_loss[loss=0.1965, simple_loss=0.2856, pruned_loss=0.05367, over 3068535.22 frames. ], batch size: 248, lr: 9.47e-03, grad_scale: 4.0 2023-04-28 18:29:45,735 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.21 vs. limit=2.0 2023-04-28 18:29:58,210 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=70024.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 18:29:58,350 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=70024.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 18:30:13,189 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=70030.0, num_to_drop=1, layers_to_drop={2} 2023-04-28 18:31:00,576 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=70050.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 18:31:03,190 INFO [train.py:904] (1/8) Epoch 7, batch 9150, loss[loss=0.1902, simple_loss=0.2788, pruned_loss=0.05078, over 16221.00 frames. ], tot_loss[loss=0.1966, simple_loss=0.2863, pruned_loss=0.0535, over 3060556.66 frames. ], batch size: 165, lr: 9.47e-03, grad_scale: 4.0 2023-04-28 18:31:06,487 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.4387, 3.6711, 1.9462, 3.9183, 2.4152, 3.8385, 1.9235, 2.7385], device='cuda:1'), covar=tensor([0.0170, 0.0246, 0.1546, 0.0066, 0.0848, 0.0338, 0.1563, 0.0630], device='cuda:1'), in_proj_covar=tensor([0.0130, 0.0149, 0.0177, 0.0090, 0.0157, 0.0180, 0.0185, 0.0162], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0004, 0.0004, 0.0003], device='cuda:1') 2023-04-28 18:31:39,071 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.3406, 4.6716, 4.4629, 4.4517, 4.1305, 4.1105, 4.1851, 4.6972], device='cuda:1'), covar=tensor([0.0929, 0.0741, 0.0784, 0.0537, 0.0665, 0.1169, 0.0732, 0.0744], device='cuda:1'), in_proj_covar=tensor([0.0423, 0.0537, 0.0452, 0.0360, 0.0338, 0.0362, 0.0446, 0.0396], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:1') 2023-04-28 18:31:47,574 INFO [zipformer.py:625] (1/8) 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] (1/8) 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 18:32:39,004 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=70098.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 18:32:41,960 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.1718, 3.3167, 3.5855, 3.5961, 3.5773, 3.3631, 3.3985, 3.4451], device='cuda:1'), covar=tensor([0.0324, 0.0544, 0.0418, 0.0409, 0.0428, 0.0413, 0.0745, 0.0446], device='cuda:1'), in_proj_covar=tensor([0.0257, 0.0262, 0.0260, 0.0251, 0.0299, 0.0281, 0.0366, 0.0229], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0002], device='cuda:1') 2023-04-28 18:32:45,149 INFO [train.py:904] (1/8) Epoch 7, batch 9200, loss[loss=0.1973, simple_loss=0.2846, pruned_loss=0.05497, over 16209.00 frames. ], tot_loss[loss=0.1927, simple_loss=0.2813, pruned_loss=0.05204, over 3072207.78 frames. ], batch size: 35, lr: 9.47e-03, grad_scale: 8.0 2023-04-28 18:33:01,827 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=70110.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 18:33:41,195 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.97 vs. limit=2.0 2023-04-28 18:34:22,192 INFO [train.py:904] (1/8) Epoch 7, batch 9250, loss[loss=0.1845, simple_loss=0.2722, pruned_loss=0.04839, over 15368.00 frames. ], tot_loss[loss=0.1934, simple_loss=0.2816, pruned_loss=0.05254, over 3065398.71 frames. ], batch size: 191, lr: 9.46e-03, grad_scale: 8.0 2023-04-28 18:35:01,701 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=70171.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 18:35:28,080 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.87 vs. limit=2.0 2023-04-28 18:36:01,191 INFO [optim.py:368] (1/8) 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 18:36:12,888 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.1908, 4.1649, 4.0083, 3.5692, 4.0550, 1.8035, 3.9212, 3.8724], device='cuda:1'), covar=tensor([0.0072, 0.0057, 0.0112, 0.0225, 0.0082, 0.1852, 0.0092, 0.0149], device='cuda:1'), in_proj_covar=tensor([0.0103, 0.0090, 0.0137, 0.0126, 0.0104, 0.0157, 0.0122, 0.0126], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-28 18:36:13,550 INFO [train.py:904] (1/8) Epoch 7, batch 9300, loss[loss=0.1875, simple_loss=0.2757, pruned_loss=0.04967, over 16221.00 frames. ], tot_loss[loss=0.1918, simple_loss=0.28, pruned_loss=0.05183, over 3056707.46 frames. ], batch size: 165, lr: 9.46e-03, grad_scale: 8.0 2023-04-28 18:36:14,574 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.79 vs. limit=2.0 2023-04-28 18:36:51,439 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.0820, 1.3937, 1.7082, 2.1593, 2.0800, 2.0956, 1.5750, 2.1874], device='cuda:1'), covar=tensor([0.0104, 0.0286, 0.0164, 0.0171, 0.0169, 0.0139, 0.0272, 0.0069], device='cuda:1'), in_proj_covar=tensor([0.0132, 0.0150, 0.0132, 0.0133, 0.0142, 0.0096, 0.0150, 0.0089], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001], device='cuda:1') 2023-04-28 18:37:59,027 INFO [train.py:904] (1/8) Epoch 7, batch 9350, loss[loss=0.1982, simple_loss=0.2843, pruned_loss=0.05608, over 15404.00 frames. ], tot_loss[loss=0.1918, simple_loss=0.2797, pruned_loss=0.05199, over 3052623.79 frames. ], batch size: 191, lr: 9.46e-03, grad_scale: 8.0 2023-04-28 18:38:18,306 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=70260.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 18:39:03,352 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.5187, 3.7634, 3.9032, 1.7770, 4.1346, 4.1346, 2.8989, 3.1070], device='cuda:1'), covar=tensor([0.0787, 0.0141, 0.0150, 0.1176, 0.0037, 0.0067, 0.0387, 0.0412], device='cuda:1'), in_proj_covar=tensor([0.0139, 0.0092, 0.0080, 0.0136, 0.0064, 0.0082, 0.0114, 0.0122], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003, 0.0003], device='cuda:1') 2023-04-28 18:39:31,294 INFO [optim.py:368] (1/8) 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] (1/8) Epoch 7, batch 9400, loss[loss=0.154, simple_loss=0.2409, pruned_loss=0.03356, over 12643.00 frames. ], tot_loss[loss=0.1911, simple_loss=0.2795, pruned_loss=0.05135, over 3056974.14 frames. ], batch size: 248, lr: 9.45e-03, grad_scale: 4.0 2023-04-28 18:40:18,280 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=70321.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 18:40:22,032 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.4717, 3.7608, 3.9080, 1.7830, 4.1301, 4.1293, 2.8738, 3.1377], device='cuda:1'), covar=tensor([0.0689, 0.0122, 0.0113, 0.1058, 0.0036, 0.0063, 0.0349, 0.0343], device='cuda:1'), in_proj_covar=tensor([0.0139, 0.0092, 0.0079, 0.0136, 0.0064, 0.0081, 0.0114, 0.0122], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003, 0.0003], device='cuda:1') 2023-04-28 18:40:36,567 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=70330.0, num_to_drop=1, layers_to_drop={0} 2023-04-28 18:40:54,514 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.7539, 3.3656, 3.2703, 1.7882, 2.7473, 2.0595, 3.2445, 3.2970], device='cuda:1'), covar=tensor([0.0273, 0.0545, 0.0451, 0.1655, 0.0714, 0.0917, 0.0643, 0.0762], device='cuda:1'), in_proj_covar=tensor([0.0133, 0.0123, 0.0152, 0.0138, 0.0132, 0.0123, 0.0132, 0.0133], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-04-28 18:41:20,563 INFO [train.py:904] (1/8) Epoch 7, batch 9450, loss[loss=0.1561, simple_loss=0.2532, pruned_loss=0.0295, over 16765.00 frames. ], tot_loss[loss=0.192, simple_loss=0.2807, pruned_loss=0.05163, over 3037221.43 frames. ], batch size: 83, lr: 9.45e-03, grad_scale: 4.0 2023-04-28 18:42:14,166 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=70378.0, num_to_drop=1, layers_to_drop={0} 2023-04-28 18:42:50,911 INFO [optim.py:368] (1/8) 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 18:43:02,096 INFO [train.py:904] (1/8) Epoch 7, batch 9500, loss[loss=0.1962, simple_loss=0.2865, pruned_loss=0.05298, over 16333.00 frames. ], tot_loss[loss=0.1913, simple_loss=0.2804, pruned_loss=0.0511, over 3058111.63 frames. ], batch size: 146, lr: 9.45e-03, grad_scale: 4.0 2023-04-28 18:43:23,692 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.38 vs. limit=2.0 2023-04-28 18:43:47,730 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.64 vs. limit=2.0 2023-04-28 18:44:02,108 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.9273, 5.3250, 5.0738, 5.0311, 4.7391, 4.6842, 4.8115, 5.3877], device='cuda:1'), covar=tensor([0.0927, 0.0854, 0.1060, 0.0608, 0.0724, 0.0789, 0.0772, 0.0822], device='cuda:1'), in_proj_covar=tensor([0.0423, 0.0538, 0.0454, 0.0363, 0.0344, 0.0364, 0.0444, 0.0399], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:1') 2023-04-28 18:44:16,075 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.46 vs. limit=2.0 2023-04-28 18:44:47,823 INFO [train.py:904] (1/8) Epoch 7, batch 9550, loss[loss=0.1936, simple_loss=0.2762, pruned_loss=0.05552, over 12491.00 frames. ], tot_loss[loss=0.1914, simple_loss=0.2799, pruned_loss=0.05143, over 3037462.56 frames. ], batch size: 250, lr: 9.44e-03, grad_scale: 4.0 2023-04-28 18:45:04,995 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=3.15 vs. limit=5.0 2023-04-28 18:45:18,541 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=70466.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 18:45:41,343 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.5432, 3.3535, 2.8294, 2.1040, 2.2645, 2.1230, 3.4641, 3.3058], device='cuda:1'), covar=tensor([0.2251, 0.0706, 0.1298, 0.2079, 0.1898, 0.1656, 0.0462, 0.0803], device='cuda:1'), in_proj_covar=tensor([0.0277, 0.0238, 0.0264, 0.0251, 0.0242, 0.0204, 0.0240, 0.0253], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-28 18:46:18,719 INFO [optim.py:368] (1/8) 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 18:46:21,776 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=5.01 vs. limit=5.0 2023-04-28 18:46:26,212 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([1.9042, 2.7061, 2.5707, 1.8702, 2.4894, 2.7025, 2.6367, 1.7752], device='cuda:1'), covar=tensor([0.0338, 0.0035, 0.0046, 0.0256, 0.0083, 0.0061, 0.0050, 0.0345], device='cuda:1'), in_proj_covar=tensor([0.0121, 0.0056, 0.0059, 0.0117, 0.0065, 0.0073, 0.0065, 0.0113], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0001, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:1') 2023-04-28 18:46:26,616 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=4.59 vs. limit=5.0 2023-04-28 18:46:26,838 INFO [train.py:904] (1/8) Epoch 7, batch 9600, loss[loss=0.1941, simple_loss=0.2742, pruned_loss=0.05695, over 12280.00 frames. ], tot_loss[loss=0.1925, simple_loss=0.2812, pruned_loss=0.0519, over 3050808.94 frames. ], batch size: 248, lr: 9.44e-03, grad_scale: 8.0 2023-04-28 18:46:35,599 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.32 vs. limit=2.0 2023-04-28 18:48:15,878 INFO [train.py:904] (1/8) Epoch 7, batch 9650, loss[loss=0.1946, simple_loss=0.2791, pruned_loss=0.05504, over 16590.00 frames. ], tot_loss[loss=0.194, simple_loss=0.2834, pruned_loss=0.05234, over 3063813.16 frames. ], batch size: 57, lr: 9.44e-03, grad_scale: 8.0 2023-04-28 18:49:18,987 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.7791, 3.6462, 3.8196, 3.7244, 3.8760, 4.2284, 3.9782, 3.6875], device='cuda:1'), covar=tensor([0.1791, 0.1974, 0.1530, 0.2185, 0.2612, 0.1338, 0.1165, 0.2262], device='cuda:1'), in_proj_covar=tensor([0.0274, 0.0394, 0.0410, 0.0344, 0.0449, 0.0433, 0.0328, 0.0455], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-28 18:49:55,419 INFO [optim.py:368] (1/8) Clipping_scale=2.0, grad-norm quartiles 1.918e+02 2.673e+02 3.296e+02 3.931e+02 1.187e+03, threshold=6.593e+02, percent-clipped=2.0 2023-04-28 18:50:06,099 INFO [train.py:904] (1/8) Epoch 7, batch 9700, loss[loss=0.1831, simple_loss=0.2654, pruned_loss=0.05037, over 12624.00 frames. ], tot_loss[loss=0.193, simple_loss=0.282, pruned_loss=0.05197, over 3049580.04 frames. ], batch size: 248, lr: 9.43e-03, grad_scale: 8.0 2023-04-28 18:50:28,974 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.0123, 1.6758, 2.3158, 2.8362, 2.5317, 2.8919, 1.9029, 3.0260], device='cuda:1'), covar=tensor([0.0096, 0.0304, 0.0191, 0.0143, 0.0167, 0.0126, 0.0289, 0.0085], device='cuda:1'), in_proj_covar=tensor([0.0133, 0.0150, 0.0133, 0.0133, 0.0141, 0.0095, 0.0149, 0.0088], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001], device='cuda:1') 2023-04-28 18:50:33,469 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=70616.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 18:51:45,334 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([1.8348, 1.7130, 1.5430, 1.4941, 1.8982, 1.6882, 1.7539, 1.9182], device='cuda:1'), covar=tensor([0.0059, 0.0185, 0.0274, 0.0233, 0.0122, 0.0157, 0.0112, 0.0125], device='cuda:1'), in_proj_covar=tensor([0.0097, 0.0174, 0.0170, 0.0170, 0.0166, 0.0171, 0.0155, 0.0153], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-28 18:51:47,801 INFO [train.py:904] (1/8) Epoch 7, batch 9750, loss[loss=0.1849, simple_loss=0.279, pruned_loss=0.04537, over 16421.00 frames. ], tot_loss[loss=0.1916, simple_loss=0.2802, pruned_loss=0.05154, over 3051816.47 frames. ], batch size: 146, lr: 9.43e-03, grad_scale: 8.0 2023-04-28 18:52:19,410 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.8138, 4.8142, 5.3073, 5.2269, 5.2836, 4.9474, 4.9144, 4.6683], device='cuda:1'), covar=tensor([0.0215, 0.0312, 0.0311, 0.0419, 0.0298, 0.0238, 0.0613, 0.0314], device='cuda:1'), in_proj_covar=tensor([0.0255, 0.0260, 0.0260, 0.0249, 0.0294, 0.0277, 0.0358, 0.0225], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0002], device='cuda:1') 2023-04-28 18:53:18,480 INFO [optim.py:368] (1/8) 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] (1/8) Epoch 7, batch 9800, loss[loss=0.1913, simple_loss=0.2802, pruned_loss=0.05116, over 16540.00 frames. ], tot_loss[loss=0.1902, simple_loss=0.2799, pruned_loss=0.05022, over 3065502.68 frames. ], batch size: 62, lr: 9.43e-03, grad_scale: 8.0 2023-04-28 18:55:03,219 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.1348, 3.4657, 3.4851, 2.5179, 3.2607, 3.4219, 3.4453, 1.8291], device='cuda:1'), covar=tensor([0.0348, 0.0024, 0.0027, 0.0225, 0.0059, 0.0057, 0.0037, 0.0376], device='cuda:1'), in_proj_covar=tensor([0.0120, 0.0057, 0.0059, 0.0117, 0.0065, 0.0073, 0.0066, 0.0114], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0001, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:1') 2023-04-28 18:55:12,739 INFO [train.py:904] (1/8) Epoch 7, batch 9850, loss[loss=0.1951, simple_loss=0.274, pruned_loss=0.05809, over 12568.00 frames. ], tot_loss[loss=0.1908, simple_loss=0.2811, pruned_loss=0.05021, over 3065990.27 frames. ], batch size: 248, lr: 9.42e-03, grad_scale: 8.0 2023-04-28 18:55:43,255 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=70766.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 18:56:31,732 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.3294, 3.6431, 3.7213, 1.6380, 3.9556, 3.9802, 2.7852, 3.0046], device='cuda:1'), covar=tensor([0.0779, 0.0147, 0.0140, 0.1187, 0.0039, 0.0064, 0.0399, 0.0360], device='cuda:1'), in_proj_covar=tensor([0.0138, 0.0090, 0.0077, 0.0135, 0.0064, 0.0080, 0.0114, 0.0121], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003, 0.0003], device='cuda:1') 2023-04-28 18:56:54,659 INFO [optim.py:368] (1/8) Clipping_scale=2.0, grad-norm quartiles 1.862e+02 2.569e+02 3.072e+02 3.785e+02 9.202e+02, threshold=6.144e+02, percent-clipped=3.0 2023-04-28 18:57:04,339 INFO [train.py:904] (1/8) Epoch 7, batch 9900, loss[loss=0.1733, simple_loss=0.2588, pruned_loss=0.04394, over 12410.00 frames. ], tot_loss[loss=0.1903, simple_loss=0.2812, pruned_loss=0.04971, over 3058746.74 frames. ], batch size: 246, lr: 9.42e-03, grad_scale: 8.0 2023-04-28 18:57:33,006 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=70814.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 18:59:01,223 INFO [train.py:904] (1/8) Epoch 7, batch 9950, loss[loss=0.171, simple_loss=0.2705, pruned_loss=0.03579, over 16947.00 frames. ], tot_loss[loss=0.1918, simple_loss=0.2834, pruned_loss=0.05004, over 3077912.48 frames. ], batch size: 102, lr: 9.42e-03, grad_scale: 8.0 2023-04-28 18:59:11,636 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.5239, 3.8972, 4.0280, 1.8781, 4.2335, 4.2411, 3.1299, 3.1859], device='cuda:1'), covar=tensor([0.0772, 0.0105, 0.0082, 0.1173, 0.0029, 0.0060, 0.0292, 0.0348], device='cuda:1'), in_proj_covar=tensor([0.0138, 0.0090, 0.0076, 0.0135, 0.0064, 0.0081, 0.0113, 0.0120], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003, 0.0003], device='cuda:1') 2023-04-28 18:59:14,422 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.86 vs. limit=2.0 2023-04-28 18:59:48,692 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([1.8434, 2.6678, 2.6655, 1.8760, 2.4674, 2.7310, 2.5700, 1.6732], device='cuda:1'), covar=tensor([0.0320, 0.0031, 0.0034, 0.0248, 0.0076, 0.0051, 0.0044, 0.0341], device='cuda:1'), in_proj_covar=tensor([0.0119, 0.0056, 0.0059, 0.0116, 0.0065, 0.0072, 0.0065, 0.0112], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0001, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:1') 2023-04-28 18:59:55,478 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.4132, 3.7910, 3.9207, 1.6690, 4.1340, 4.1498, 3.0279, 2.9979], device='cuda:1'), covar=tensor([0.0780, 0.0117, 0.0103, 0.1291, 0.0033, 0.0067, 0.0280, 0.0397], device='cuda:1'), in_proj_covar=tensor([0.0138, 0.0090, 0.0076, 0.0135, 0.0064, 0.0081, 0.0113, 0.0121], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003, 0.0003], device='cuda:1') 2023-04-28 19:00:39,434 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.7194, 2.0372, 2.2296, 4.4600, 1.9447, 2.5992, 2.1756, 2.2716], device='cuda:1'), covar=tensor([0.0625, 0.2792, 0.1756, 0.0256, 0.3361, 0.1768, 0.2704, 0.2784], device='cuda:1'), in_proj_covar=tensor([0.0324, 0.0338, 0.0290, 0.0305, 0.0380, 0.0362, 0.0307, 0.0396], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-28 19:00:47,805 INFO [optim.py:368] (1/8) 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 19:01:00,572 INFO [train.py:904] (1/8) Epoch 7, batch 10000, loss[loss=0.1931, simple_loss=0.2851, pruned_loss=0.05059, over 16747.00 frames. ], tot_loss[loss=0.1907, simple_loss=0.2821, pruned_loss=0.04963, over 3093622.20 frames. ], batch size: 134, lr: 9.41e-03, grad_scale: 8.0 2023-04-28 19:01:01,259 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.9349, 4.2215, 4.0204, 4.0426, 3.7043, 3.7974, 3.8784, 4.2024], device='cuda:1'), covar=tensor([0.0929, 0.0834, 0.0860, 0.0577, 0.0764, 0.1403, 0.0739, 0.0968], device='cuda:1'), in_proj_covar=tensor([0.0417, 0.0528, 0.0442, 0.0359, 0.0341, 0.0356, 0.0439, 0.0393], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:1') 2023-04-28 19:01:16,252 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.3675, 1.9363, 1.6291, 1.6416, 2.2133, 1.9759, 2.2184, 2.3356], device='cuda:1'), covar=tensor([0.0061, 0.0215, 0.0285, 0.0281, 0.0129, 0.0208, 0.0107, 0.0144], device='cuda:1'), in_proj_covar=tensor([0.0096, 0.0174, 0.0170, 0.0167, 0.0166, 0.0172, 0.0154, 0.0154], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-28 19:01:16,310 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.4268, 3.3416, 2.8392, 2.1171, 2.1465, 2.2001, 3.3974, 3.1272], device='cuda:1'), covar=tensor([0.2303, 0.0572, 0.1191, 0.1963, 0.2077, 0.1573, 0.0386, 0.0843], device='cuda:1'), in_proj_covar=tensor([0.0277, 0.0236, 0.0264, 0.0249, 0.0239, 0.0201, 0.0239, 0.0251], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-28 19:01:29,744 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=70916.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 19:02:38,791 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.0150, 3.5216, 3.3977, 1.7723, 2.9610, 2.2703, 3.4273, 3.4267], device='cuda:1'), covar=tensor([0.0278, 0.0581, 0.0520, 0.1697, 0.0676, 0.0915, 0.0737, 0.0859], device='cuda:1'), in_proj_covar=tensor([0.0131, 0.0121, 0.0151, 0.0137, 0.0128, 0.0121, 0.0131, 0.0132], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:1') 2023-04-28 19:02:40,827 INFO [train.py:904] (1/8) Epoch 7, batch 10050, loss[loss=0.2069, simple_loss=0.2955, pruned_loss=0.05914, over 16725.00 frames. ], tot_loss[loss=0.1908, simple_loss=0.2822, pruned_loss=0.04966, over 3083482.49 frames. ], batch size: 134, lr: 9.41e-03, grad_scale: 8.0 2023-04-28 19:03:04,143 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=70964.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 19:03:36,913 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.63 vs. limit=2.0 2023-04-28 19:03:56,355 INFO [zipformer.py:625] (1/8) 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] (1/8) 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] (1/8) Epoch 7, batch 10100, loss[loss=0.2183, simple_loss=0.2908, pruned_loss=0.0729, over 12342.00 frames. ], tot_loss[loss=0.1915, simple_loss=0.2829, pruned_loss=0.05007, over 3072315.98 frames. ], batch size: 248, lr: 9.41e-03, grad_scale: 8.0 2023-04-28 19:05:06,993 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=71030.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 19:05:15,853 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([5.1820, 5.4953, 5.2864, 5.2949, 4.9174, 4.7405, 5.0032, 5.5480], device='cuda:1'), covar=tensor([0.0776, 0.0643, 0.0760, 0.0434, 0.0603, 0.0725, 0.0658, 0.0732], device='cuda:1'), in_proj_covar=tensor([0.0415, 0.0528, 0.0441, 0.0359, 0.0341, 0.0357, 0.0440, 0.0396], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:1') 2023-04-28 19:05:23,187 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.4157, 3.0766, 3.0621, 1.8907, 2.6794, 2.1849, 2.9889, 2.9993], device='cuda:1'), covar=tensor([0.0284, 0.0648, 0.0480, 0.1655, 0.0711, 0.0890, 0.0651, 0.0845], device='cuda:1'), in_proj_covar=tensor([0.0132, 0.0121, 0.0151, 0.0137, 0.0129, 0.0122, 0.0132, 0.0132], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:1') 2023-04-28 19:05:54,945 INFO [train.py:904] (1/8) Epoch 8, batch 0, loss[loss=0.3366, simple_loss=0.3669, pruned_loss=0.1532, over 16679.00 frames. ], tot_loss[loss=0.3366, simple_loss=0.3669, pruned_loss=0.1532, over 16679.00 frames. ], batch size: 134, lr: 8.86e-03, grad_scale: 8.0 2023-04-28 19:05:54,945 INFO [train.py:929] (1/8) Computing validation loss 2023-04-28 19:06:02,577 INFO [train.py:938] (1/8) Epoch 8, validation: loss=0.1627, simple_loss=0.2663, pruned_loss=0.02958, over 944034.00 frames. 2023-04-28 19:06:02,578 INFO [train.py:939] (1/8) Maximum memory allocated so far is 17676MB 2023-04-28 19:06:04,008 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=71053.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 19:06:13,868 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.9105, 5.5928, 5.6713, 5.5026, 5.5189, 6.0693, 5.6338, 5.4468], device='cuda:1'), covar=tensor([0.0708, 0.1517, 0.1569, 0.1668, 0.2149, 0.0851, 0.1100, 0.2187], device='cuda:1'), in_proj_covar=tensor([0.0284, 0.0403, 0.0419, 0.0355, 0.0464, 0.0444, 0.0334, 0.0471], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-28 19:06:30,240 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.4592, 4.3859, 4.2878, 4.0765, 3.9167, 4.3518, 4.2993, 4.0819], device='cuda:1'), covar=tensor([0.0600, 0.0617, 0.0379, 0.0302, 0.0979, 0.0480, 0.0444, 0.0686], device='cuda:1'), in_proj_covar=tensor([0.0191, 0.0224, 0.0225, 0.0200, 0.0249, 0.0225, 0.0154, 0.0260], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-28 19:06:55,636 INFO [zipformer.py:625] (1/8) 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] (1/8) 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 19:07:09,988 INFO [train.py:904] (1/8) Epoch 8, batch 50, loss[loss=0.1836, simple_loss=0.2713, pruned_loss=0.04799, over 16856.00 frames. ], tot_loss[loss=0.2218, simple_loss=0.2964, pruned_loss=0.0736, over 759009.24 frames. ], batch size: 42, lr: 8.86e-03, grad_scale: 1.0 2023-04-28 19:07:14,587 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.3076, 3.1946, 3.6618, 2.5047, 3.4609, 3.5641, 3.5581, 2.1255], device='cuda:1'), covar=tensor([0.0345, 0.0112, 0.0031, 0.0243, 0.0066, 0.0065, 0.0052, 0.0304], device='cuda:1'), in_proj_covar=tensor([0.0122, 0.0060, 0.0061, 0.0118, 0.0066, 0.0074, 0.0066, 0.0114], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:1') 2023-04-28 19:08:17,892 INFO [train.py:904] (1/8) Epoch 8, batch 100, loss[loss=0.2334, simple_loss=0.2967, pruned_loss=0.08509, over 16461.00 frames. ], tot_loss[loss=0.2177, simple_loss=0.2933, pruned_loss=0.07099, over 1325995.45 frames. ], batch size: 146, lr: 8.85e-03, grad_scale: 1.0 2023-04-28 19:09:14,389 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.4315, 3.4281, 3.3793, 2.9050, 3.4151, 2.0956, 3.1331, 2.6821], device='cuda:1'), covar=tensor([0.0086, 0.0068, 0.0119, 0.0175, 0.0066, 0.1587, 0.0094, 0.0154], device='cuda:1'), in_proj_covar=tensor([0.0104, 0.0091, 0.0141, 0.0128, 0.0106, 0.0160, 0.0124, 0.0129], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-28 19:09:23,286 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.88 vs. limit=2.0 2023-04-28 19:09:23,827 INFO [optim.py:368] (1/8) 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 19:09:26,722 INFO [train.py:904] (1/8) Epoch 8, batch 150, loss[loss=0.1862, simple_loss=0.2804, pruned_loss=0.04606, over 17092.00 frames. ], tot_loss[loss=0.2135, simple_loss=0.291, pruned_loss=0.06796, over 1765044.04 frames. ], batch size: 49, lr: 8.85e-03, grad_scale: 1.0 2023-04-28 19:10:33,447 INFO [train.py:904] (1/8) Epoch 8, batch 200, loss[loss=0.2189, simple_loss=0.3087, pruned_loss=0.06448, over 17054.00 frames. ], tot_loss[loss=0.212, simple_loss=0.2905, pruned_loss=0.0668, over 2099406.42 frames. ], batch size: 53, lr: 8.85e-03, grad_scale: 1.0 2023-04-28 19:11:40,015 INFO [optim.py:368] (1/8) 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] (1/8) Epoch 8, batch 250, loss[loss=0.1888, simple_loss=0.2782, pruned_loss=0.04972, over 16829.00 frames. ], tot_loss[loss=0.211, simple_loss=0.2882, pruned_loss=0.06691, over 2368562.28 frames. ], batch size: 42, lr: 8.84e-03, grad_scale: 1.0 2023-04-28 19:11:56,861 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.1355, 1.6512, 2.3298, 2.8791, 2.7705, 2.9873, 1.7867, 3.0096], device='cuda:1'), covar=tensor([0.0079, 0.0294, 0.0176, 0.0139, 0.0136, 0.0098, 0.0292, 0.0070], device='cuda:1'), in_proj_covar=tensor([0.0136, 0.0155, 0.0138, 0.0138, 0.0144, 0.0098, 0.0152, 0.0090], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001], device='cuda:1') 2023-04-28 19:12:47,160 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 8, batch 300, loss[loss=0.1807, simple_loss=0.2739, pruned_loss=0.04379, over 17084.00 frames. ], tot_loss[loss=0.2072, simple_loss=0.2853, pruned_loss=0.06458, over 2590617.28 frames. ], batch size: 47, lr: 8.84e-03, grad_scale: 1.0 2023-04-28 19:13:38,970 INFO [zipformer.py:625] (1/8) 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] (1/8) 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 19:14:01,157 INFO [train.py:904] (1/8) Epoch 8, batch 350, loss[loss=0.2104, simple_loss=0.2915, pruned_loss=0.06464, over 16515.00 frames. ], tot_loss[loss=0.202, simple_loss=0.2808, pruned_loss=0.06164, over 2753935.48 frames. ], batch size: 68, lr: 8.84e-03, grad_scale: 1.0 2023-04-28 19:15:10,453 INFO [train.py:904] (1/8) Epoch 8, batch 400, loss[loss=0.2063, simple_loss=0.2874, pruned_loss=0.06266, over 17187.00 frames. ], tot_loss[loss=0.2008, simple_loss=0.2793, pruned_loss=0.06115, over 2874497.85 frames. ], batch size: 46, lr: 8.83e-03, grad_scale: 2.0 2023-04-28 19:15:19,824 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=4.93 vs. limit=5.0 2023-04-28 19:16:17,822 INFO [optim.py:368] (1/8) 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 19:16:20,153 INFO [train.py:904] (1/8) Epoch 8, batch 450, loss[loss=0.1611, simple_loss=0.2421, pruned_loss=0.04009, over 16200.00 frames. ], tot_loss[loss=0.1987, simple_loss=0.2774, pruned_loss=0.06002, over 2977086.99 frames. ], batch size: 36, lr: 8.83e-03, grad_scale: 2.0 2023-04-28 19:17:28,362 INFO [train.py:904] (1/8) Epoch 8, batch 500, loss[loss=0.2235, simple_loss=0.2967, pruned_loss=0.07519, over 16880.00 frames. ], tot_loss[loss=0.1981, simple_loss=0.2765, pruned_loss=0.05986, over 3046897.77 frames. ], batch size: 109, lr: 8.83e-03, grad_scale: 2.0 2023-04-28 19:17:40,369 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.4180, 3.5179, 3.6093, 1.7970, 3.7851, 3.8025, 2.9671, 2.8669], device='cuda:1'), covar=tensor([0.0659, 0.0116, 0.0149, 0.0995, 0.0062, 0.0113, 0.0323, 0.0350], device='cuda:1'), in_proj_covar=tensor([0.0140, 0.0092, 0.0081, 0.0137, 0.0067, 0.0087, 0.0116, 0.0124], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0002, 0.0002, 0.0003, 0.0003], device='cuda:1') 2023-04-28 19:18:33,694 INFO [optim.py:368] (1/8) 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 19:18:37,346 INFO [train.py:904] (1/8) Epoch 8, batch 550, loss[loss=0.2092, simple_loss=0.2816, pruned_loss=0.06836, over 16735.00 frames. ], tot_loss[loss=0.1968, simple_loss=0.2752, pruned_loss=0.05924, over 3112813.15 frames. ], batch size: 124, lr: 8.83e-03, grad_scale: 2.0 2023-04-28 19:18:56,800 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=71616.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 19:19:36,725 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.42 vs. limit=2.0 2023-04-28 19:19:41,827 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=71648.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 19:19:45,777 INFO [train.py:904] (1/8) Epoch 8, batch 600, loss[loss=0.2265, simple_loss=0.285, pruned_loss=0.08403, over 16867.00 frames. ], tot_loss[loss=0.198, simple_loss=0.2753, pruned_loss=0.0604, over 3145706.79 frames. ], batch size: 109, lr: 8.82e-03, grad_scale: 2.0 2023-04-28 19:20:19,047 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=71677.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 19:20:34,027 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=71686.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 19:20:43,560 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([5.2384, 5.6119, 5.2826, 5.3793, 4.9439, 4.8503, 5.0705, 5.6912], device='cuda:1'), covar=tensor([0.0906, 0.0757, 0.0994, 0.0581, 0.0794, 0.0750, 0.0852, 0.0782], device='cuda:1'), in_proj_covar=tensor([0.0474, 0.0602, 0.0502, 0.0406, 0.0385, 0.0400, 0.0503, 0.0444], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-28 19:20:45,906 INFO [zipformer.py:625] (1/8) 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] (1/8) 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] (1/8) Epoch 8, batch 650, loss[loss=0.2013, simple_loss=0.279, pruned_loss=0.06174, over 16905.00 frames. ], tot_loss[loss=0.1967, simple_loss=0.2736, pruned_loss=0.05992, over 3183242.89 frames. ], batch size: 116, lr: 8.82e-03, grad_scale: 2.0 2023-04-28 19:21:37,502 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=71734.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 19:21:43,219 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([5.3433, 5.6320, 5.1194, 5.6938, 5.1215, 4.8259, 5.3993, 5.7571], device='cuda:1'), covar=tensor([0.1921, 0.1428, 0.2073, 0.0882, 0.1459, 0.1218, 0.1448, 0.1396], device='cuda:1'), in_proj_covar=tensor([0.0479, 0.0608, 0.0507, 0.0410, 0.0387, 0.0404, 0.0507, 0.0449], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-28 19:22:01,894 INFO [train.py:904] (1/8) Epoch 8, batch 700, loss[loss=0.1935, simple_loss=0.2743, pruned_loss=0.05633, over 16678.00 frames. ], tot_loss[loss=0.1967, simple_loss=0.2736, pruned_loss=0.05991, over 3209622.78 frames. ], batch size: 62, lr: 8.82e-03, grad_scale: 2.0 2023-04-28 19:22:58,719 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.8230, 4.0870, 2.2620, 4.5533, 2.9174, 4.4431, 2.2693, 3.1343], device='cuda:1'), covar=tensor([0.0227, 0.0261, 0.1584, 0.0102, 0.0811, 0.0428, 0.1598, 0.0664], device='cuda:1'), in_proj_covar=tensor([0.0137, 0.0157, 0.0178, 0.0100, 0.0162, 0.0194, 0.0190, 0.0166], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-04-28 19:23:02,968 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.8678, 4.2605, 2.8541, 2.3835, 2.9026, 2.2732, 4.3838, 3.9779], device='cuda:1'), covar=tensor([0.2514, 0.0686, 0.1781, 0.2200, 0.2595, 0.1933, 0.0445, 0.0888], device='cuda:1'), in_proj_covar=tensor([0.0291, 0.0252, 0.0278, 0.0263, 0.0273, 0.0214, 0.0256, 0.0282], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-28 19:23:05,955 INFO [optim.py:368] (1/8) 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 19:23:08,648 INFO [train.py:904] (1/8) Epoch 8, batch 750, loss[loss=0.2263, simple_loss=0.2927, pruned_loss=0.07994, over 12506.00 frames. ], tot_loss[loss=0.1966, simple_loss=0.2738, pruned_loss=0.05967, over 3235432.07 frames. ], batch size: 246, lr: 8.81e-03, grad_scale: 2.0 2023-04-28 19:23:14,309 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.1404, 4.4019, 2.5753, 4.8005, 3.1116, 4.7042, 2.6864, 3.5520], device='cuda:1'), covar=tensor([0.0177, 0.0240, 0.1363, 0.0084, 0.0763, 0.0379, 0.1399, 0.0548], device='cuda:1'), in_proj_covar=tensor([0.0137, 0.0157, 0.0179, 0.0100, 0.0163, 0.0195, 0.0191, 0.0167], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-04-28 19:23:48,388 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.86 vs. limit=2.0 2023-04-28 19:24:17,945 INFO [train.py:904] (1/8) Epoch 8, batch 800, loss[loss=0.1811, simple_loss=0.2764, pruned_loss=0.04292, over 17171.00 frames. ], tot_loss[loss=0.1954, simple_loss=0.2733, pruned_loss=0.05872, over 3265059.86 frames. ], batch size: 46, lr: 8.81e-03, grad_scale: 4.0 2023-04-28 19:25:23,722 INFO [optim.py:368] (1/8) 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] (1/8) Epoch 8, batch 850, loss[loss=0.1967, simple_loss=0.2786, pruned_loss=0.0574, over 16757.00 frames. ], tot_loss[loss=0.1945, simple_loss=0.273, pruned_loss=0.05804, over 3280666.54 frames. ], batch size: 57, lr: 8.81e-03, grad_scale: 4.0 2023-04-28 19:25:49,767 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=71919.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 19:26:12,372 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.5204, 4.0989, 4.0485, 2.0517, 3.2177, 2.7952, 3.8428, 4.0100], device='cuda:1'), covar=tensor([0.0274, 0.0598, 0.0417, 0.1598, 0.0663, 0.0816, 0.0645, 0.0978], device='cuda:1'), in_proj_covar=tensor([0.0137, 0.0133, 0.0154, 0.0139, 0.0133, 0.0123, 0.0134, 0.0142], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-04-28 19:26:18,722 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([5.3815, 5.2943, 5.1689, 4.8352, 4.6677, 5.1808, 5.2476, 4.7709], device='cuda:1'), covar=tensor([0.0459, 0.0316, 0.0229, 0.0236, 0.1069, 0.0294, 0.0185, 0.0617], device='cuda:1'), in_proj_covar=tensor([0.0224, 0.0263, 0.0262, 0.0235, 0.0295, 0.0266, 0.0180, 0.0304], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-04-28 19:26:32,698 INFO [train.py:904] (1/8) Epoch 8, batch 900, loss[loss=0.1724, simple_loss=0.2636, pruned_loss=0.04059, over 16722.00 frames. ], tot_loss[loss=0.1918, simple_loss=0.2706, pruned_loss=0.05651, over 3285022.09 frames. ], batch size: 57, lr: 8.80e-03, grad_scale: 4.0 2023-04-28 19:27:00,594 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=71972.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 19:27:13,328 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=71980.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 19:27:38,431 INFO [optim.py:368] (1/8) 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 19:27:44,211 INFO [train.py:904] (1/8) Epoch 8, batch 950, loss[loss=0.2121, simple_loss=0.2783, pruned_loss=0.07301, over 16900.00 frames. ], tot_loss[loss=0.1922, simple_loss=0.2709, pruned_loss=0.05673, over 3287322.58 frames. ], batch size: 109, lr: 8.80e-03, grad_scale: 4.0 2023-04-28 19:28:47,232 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.6083, 4.9648, 4.7144, 4.7534, 4.4132, 4.3771, 4.4365, 5.0291], device='cuda:1'), covar=tensor([0.0853, 0.0852, 0.0886, 0.0552, 0.0762, 0.1014, 0.0837, 0.0794], device='cuda:1'), in_proj_covar=tensor([0.0477, 0.0611, 0.0508, 0.0409, 0.0390, 0.0403, 0.0508, 0.0451], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-28 19:28:52,161 INFO [train.py:904] (1/8) Epoch 8, batch 1000, loss[loss=0.2167, simple_loss=0.2739, pruned_loss=0.07979, over 16714.00 frames. ], tot_loss[loss=0.191, simple_loss=0.2694, pruned_loss=0.05629, over 3290132.30 frames. ], batch size: 124, lr: 8.80e-03, grad_scale: 4.0 2023-04-28 19:29:04,179 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.20 vs. limit=2.0 2023-04-28 19:29:58,311 INFO [optim.py:368] (1/8) 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] (1/8) Epoch 8, batch 1050, loss[loss=0.1966, simple_loss=0.2696, pruned_loss=0.06183, over 16477.00 frames. ], tot_loss[loss=0.1912, simple_loss=0.2695, pruned_loss=0.05651, over 3299320.72 frames. ], batch size: 68, lr: 8.79e-03, grad_scale: 4.0 2023-04-28 19:30:35,560 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.6483, 2.5400, 1.9449, 2.2200, 2.9790, 2.6722, 3.5945, 3.2918], device='cuda:1'), covar=tensor([0.0049, 0.0238, 0.0318, 0.0300, 0.0140, 0.0232, 0.0110, 0.0130], device='cuda:1'), in_proj_covar=tensor([0.0115, 0.0186, 0.0180, 0.0181, 0.0179, 0.0187, 0.0183, 0.0169], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-28 19:30:40,229 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=72129.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 19:30:40,460 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.21 vs. limit=2.0 2023-04-28 19:31:06,967 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.6163, 4.1080, 4.4312, 2.9982, 3.9151, 4.3719, 4.0114, 2.5350], device='cuda:1'), covar=tensor([0.0371, 0.0030, 0.0033, 0.0263, 0.0054, 0.0050, 0.0048, 0.0347], device='cuda:1'), in_proj_covar=tensor([0.0122, 0.0064, 0.0062, 0.0118, 0.0068, 0.0077, 0.0070, 0.0114], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:1') 2023-04-28 19:31:10,578 INFO [train.py:904] (1/8) Epoch 8, batch 1100, loss[loss=0.2018, simple_loss=0.2808, pruned_loss=0.06145, over 16541.00 frames. ], tot_loss[loss=0.1899, simple_loss=0.2686, pruned_loss=0.05559, over 3303245.91 frames. ], batch size: 68, lr: 8.79e-03, grad_scale: 4.0 2023-04-28 19:31:16,567 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.3680, 2.0305, 2.2483, 3.8650, 2.0251, 2.5401, 2.0626, 2.1829], device='cuda:1'), covar=tensor([0.0775, 0.2983, 0.1688, 0.0414, 0.3082, 0.1801, 0.2787, 0.2540], device='cuda:1'), in_proj_covar=tensor([0.0345, 0.0360, 0.0302, 0.0323, 0.0393, 0.0398, 0.0325, 0.0427], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-28 19:31:32,385 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=4.69 vs. limit=5.0 2023-04-28 19:31:50,920 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.8905, 4.8208, 4.7236, 4.4688, 4.3469, 4.7634, 4.7013, 4.4944], device='cuda:1'), covar=tensor([0.0488, 0.0393, 0.0217, 0.0232, 0.0899, 0.0336, 0.0307, 0.0566], device='cuda:1'), in_proj_covar=tensor([0.0228, 0.0267, 0.0266, 0.0240, 0.0299, 0.0271, 0.0182, 0.0308], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-04-28 19:32:02,994 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=72190.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 19:32:15,691 INFO [optim.py:368] (1/8) 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] (1/8) Epoch 8, batch 1150, loss[loss=0.2065, simple_loss=0.2869, pruned_loss=0.063, over 16836.00 frames. ], tot_loss[loss=0.1887, simple_loss=0.2679, pruned_loss=0.05475, over 3312176.04 frames. ], batch size: 42, lr: 8.79e-03, grad_scale: 4.0 2023-04-28 19:32:32,111 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.8700, 1.4924, 2.1516, 2.6724, 2.7077, 2.5810, 1.7291, 2.9194], device='cuda:1'), covar=tensor([0.0089, 0.0299, 0.0199, 0.0156, 0.0141, 0.0147, 0.0300, 0.0070], device='cuda:1'), in_proj_covar=tensor([0.0143, 0.0156, 0.0142, 0.0142, 0.0149, 0.0105, 0.0156, 0.0095], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001], device='cuda:1') 2023-04-28 19:33:26,639 INFO [train.py:904] (1/8) Epoch 8, batch 1200, loss[loss=0.1857, simple_loss=0.265, pruned_loss=0.05324, over 17225.00 frames. ], tot_loss[loss=0.1882, simple_loss=0.2674, pruned_loss=0.05453, over 3323449.09 frames. ], batch size: 44, lr: 8.79e-03, grad_scale: 8.0 2023-04-28 19:33:53,179 INFO [zipformer.py:625] (1/8) 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] (1/8) 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] (1/8) 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] (1/8) Epoch 8, batch 1250, loss[loss=0.1859, simple_loss=0.2611, pruned_loss=0.0554, over 15519.00 frames. ], tot_loss[loss=0.1884, simple_loss=0.2672, pruned_loss=0.05483, over 3325014.75 frames. ], batch size: 190, lr: 8.78e-03, grad_scale: 8.0 2023-04-28 19:34:58,451 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=72320.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 19:35:06,092 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=72326.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 19:35:41,550 INFO [train.py:904] (1/8) Epoch 8, batch 1300, loss[loss=0.1829, simple_loss=0.2678, pruned_loss=0.04901, over 17238.00 frames. ], tot_loss[loss=0.188, simple_loss=0.267, pruned_loss=0.05455, over 3329743.46 frames. ], batch size: 45, lr: 8.78e-03, grad_scale: 8.0 2023-04-28 19:35:44,327 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.87 vs. limit=2.0 2023-04-28 19:36:30,378 INFO [zipformer.py:625] (1/8) 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] (1/8) 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 19:36:52,126 INFO [train.py:904] (1/8) Epoch 8, batch 1350, loss[loss=0.1692, simple_loss=0.2682, pruned_loss=0.03515, over 17040.00 frames. ], tot_loss[loss=0.1886, simple_loss=0.2681, pruned_loss=0.05456, over 3336285.40 frames. ], batch size: 50, lr: 8.78e-03, grad_scale: 8.0 2023-04-28 19:37:40,924 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.2024, 4.2021, 4.1923, 3.6716, 4.1536, 1.7659, 3.9524, 3.8686], device='cuda:1'), covar=tensor([0.0089, 0.0077, 0.0126, 0.0284, 0.0072, 0.2116, 0.0120, 0.0180], device='cuda:1'), in_proj_covar=tensor([0.0114, 0.0103, 0.0155, 0.0147, 0.0118, 0.0168, 0.0140, 0.0146], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-28 19:38:01,922 INFO [train.py:904] (1/8) Epoch 8, batch 1400, loss[loss=0.224, simple_loss=0.2874, pruned_loss=0.08036, over 16747.00 frames. ], tot_loss[loss=0.1889, simple_loss=0.2684, pruned_loss=0.05472, over 3330395.73 frames. ], batch size: 124, lr: 8.77e-03, grad_scale: 8.0 2023-04-28 19:38:07,689 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.0216, 4.2737, 4.6020, 1.8683, 4.8316, 4.8828, 3.2790, 3.6337], device='cuda:1'), covar=tensor([0.0638, 0.0148, 0.0138, 0.1138, 0.0053, 0.0069, 0.0370, 0.0334], device='cuda:1'), in_proj_covar=tensor([0.0142, 0.0094, 0.0083, 0.0138, 0.0070, 0.0091, 0.0120, 0.0126], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0002, 0.0002, 0.0003, 0.0003], device='cuda:1') 2023-04-28 19:38:47,780 INFO [zipformer.py:625] (1/8) 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] (1/8) 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] (1/8) Epoch 8, batch 1450, loss[loss=0.1774, simple_loss=0.2569, pruned_loss=0.04892, over 16876.00 frames. ], tot_loss[loss=0.189, simple_loss=0.2681, pruned_loss=0.055, over 3333234.13 frames. ], batch size: 42, lr: 8.77e-03, grad_scale: 8.0 2023-04-28 19:40:17,121 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.84 vs. limit=2.0 2023-04-28 19:40:20,581 INFO [train.py:904] (1/8) Epoch 8, batch 1500, loss[loss=0.1754, simple_loss=0.2656, pruned_loss=0.04264, over 17181.00 frames. ], tot_loss[loss=0.1894, simple_loss=0.2682, pruned_loss=0.05526, over 3322704.71 frames. ], batch size: 46, lr: 8.77e-03, grad_scale: 8.0 2023-04-28 19:40:51,986 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=72575.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 19:41:08,340 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.96 vs. limit=2.0 2023-04-28 19:41:25,445 INFO [optim.py:368] (1/8) 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] (1/8) Epoch 8, batch 1550, loss[loss=0.2245, simple_loss=0.2799, pruned_loss=0.08454, over 16762.00 frames. ], tot_loss[loss=0.1912, simple_loss=0.2692, pruned_loss=0.05658, over 3330519.57 frames. ], batch size: 124, lr: 8.76e-03, grad_scale: 8.0 2023-04-28 19:41:58,082 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=72623.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 19:42:38,736 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.6243, 1.4865, 2.0350, 2.5266, 2.5606, 2.3274, 1.6119, 2.6496], device='cuda:1'), covar=tensor([0.0097, 0.0302, 0.0177, 0.0137, 0.0139, 0.0178, 0.0308, 0.0065], device='cuda:1'), in_proj_covar=tensor([0.0144, 0.0157, 0.0143, 0.0142, 0.0150, 0.0105, 0.0157, 0.0097], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001], device='cuda:1') 2023-04-28 19:42:39,324 INFO [train.py:904] (1/8) Epoch 8, batch 1600, loss[loss=0.2018, simple_loss=0.2761, pruned_loss=0.06372, over 16742.00 frames. ], tot_loss[loss=0.1939, simple_loss=0.2717, pruned_loss=0.05807, over 3321198.90 frames. ], batch size: 134, lr: 8.76e-03, grad_scale: 8.0 2023-04-28 19:42:47,059 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.7177, 4.0457, 4.2862, 3.1389, 3.7146, 4.2208, 3.9199, 2.6974], device='cuda:1'), covar=tensor([0.0315, 0.0046, 0.0029, 0.0203, 0.0065, 0.0042, 0.0040, 0.0270], device='cuda:1'), in_proj_covar=tensor([0.0121, 0.0065, 0.0062, 0.0117, 0.0068, 0.0076, 0.0069, 0.0114], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:1') 2023-04-28 19:43:20,447 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=72682.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 19:43:44,411 INFO [optim.py:368] (1/8) 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 19:43:47,315 INFO [train.py:904] (1/8) Epoch 8, batch 1650, loss[loss=0.219, simple_loss=0.2933, pruned_loss=0.07237, over 16506.00 frames. ], tot_loss[loss=0.195, simple_loss=0.2732, pruned_loss=0.05838, over 3329051.84 frames. ], batch size: 75, lr: 8.76e-03, grad_scale: 8.0 2023-04-28 19:44:58,258 INFO [train.py:904] (1/8) Epoch 8, batch 1700, loss[loss=0.2189, simple_loss=0.3034, pruned_loss=0.0672, over 16063.00 frames. ], tot_loss[loss=0.1964, simple_loss=0.2753, pruned_loss=0.05877, over 3334302.79 frames. ], batch size: 35, lr: 8.76e-03, grad_scale: 8.0 2023-04-28 19:45:00,667 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.0021, 4.1613, 2.4190, 4.5965, 2.9491, 4.6100, 2.4200, 3.2461], device='cuda:1'), covar=tensor([0.0169, 0.0273, 0.1363, 0.0130, 0.0757, 0.0336, 0.1379, 0.0596], device='cuda:1'), in_proj_covar=tensor([0.0140, 0.0163, 0.0181, 0.0107, 0.0165, 0.0202, 0.0191, 0.0170], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-04-28 19:45:24,562 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([5.3658, 5.2393, 5.1556, 4.7745, 4.6803, 5.1164, 5.2380, 4.7750], device='cuda:1'), covar=tensor([0.0434, 0.0283, 0.0222, 0.0241, 0.0985, 0.0377, 0.0235, 0.0619], device='cuda:1'), in_proj_covar=tensor([0.0233, 0.0274, 0.0273, 0.0244, 0.0305, 0.0276, 0.0186, 0.0316], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-04-28 19:45:44,129 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=72785.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 19:46:00,705 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.80 vs. limit=2.0 2023-04-28 19:46:05,113 INFO [optim.py:368] (1/8) 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] (1/8) Epoch 8, batch 1750, loss[loss=0.1789, simple_loss=0.2734, pruned_loss=0.04221, over 17209.00 frames. ], tot_loss[loss=0.1975, simple_loss=0.2769, pruned_loss=0.05903, over 3329848.37 frames. ], batch size: 46, lr: 8.75e-03, grad_scale: 8.0 2023-04-28 19:46:32,844 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=3.48 vs. limit=5.0 2023-04-28 19:46:50,155 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=72833.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 19:47:16,123 INFO [train.py:904] (1/8) Epoch 8, batch 1800, loss[loss=0.1933, simple_loss=0.2929, pruned_loss=0.04684, over 17099.00 frames. ], tot_loss[loss=0.1972, simple_loss=0.2771, pruned_loss=0.0587, over 3339637.67 frames. ], batch size: 47, lr: 8.75e-03, grad_scale: 4.0 2023-04-28 19:47:52,847 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.9655, 3.9054, 3.9067, 3.3807, 3.9209, 1.7131, 3.7261, 3.5495], device='cuda:1'), covar=tensor([0.0087, 0.0082, 0.0136, 0.0256, 0.0074, 0.2246, 0.0107, 0.0193], device='cuda:1'), in_proj_covar=tensor([0.0113, 0.0102, 0.0154, 0.0147, 0.0118, 0.0164, 0.0139, 0.0143], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-28 19:48:15,334 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=72894.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 19:48:24,988 INFO [optim.py:368] (1/8) 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] (1/8) Epoch 8, batch 1850, loss[loss=0.2133, simple_loss=0.2851, pruned_loss=0.07073, over 16431.00 frames. ], tot_loss[loss=0.1985, simple_loss=0.2785, pruned_loss=0.05924, over 3328102.04 frames. ], batch size: 146, lr: 8.75e-03, grad_scale: 4.0 2023-04-28 19:49:20,412 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.5905, 3.9201, 4.0932, 1.8524, 4.3352, 4.2734, 3.1187, 3.1385], device='cuda:1'), covar=tensor([0.0739, 0.0121, 0.0142, 0.1146, 0.0051, 0.0107, 0.0359, 0.0377], device='cuda:1'), in_proj_covar=tensor([0.0145, 0.0096, 0.0085, 0.0140, 0.0071, 0.0094, 0.0122, 0.0128], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0003, 0.0002, 0.0004, 0.0002, 0.0002, 0.0003, 0.0003], device='cuda:1') 2023-04-28 19:49:35,218 INFO [train.py:904] (1/8) Epoch 8, batch 1900, loss[loss=0.1667, simple_loss=0.2466, pruned_loss=0.04344, over 17015.00 frames. ], tot_loss[loss=0.1975, simple_loss=0.2774, pruned_loss=0.05878, over 3328322.83 frames. ], batch size: 41, lr: 8.74e-03, grad_scale: 4.0 2023-04-28 19:49:39,049 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=72955.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 19:49:51,981 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=72964.0, num_to_drop=1, layers_to_drop={0} 2023-04-28 19:50:11,191 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.1559, 2.1858, 2.3547, 4.8338, 2.1050, 2.9899, 2.3185, 2.4697], device='cuda:1'), covar=tensor([0.0585, 0.2890, 0.1746, 0.0263, 0.3295, 0.1694, 0.2474, 0.2848], device='cuda:1'), in_proj_covar=tensor([0.0350, 0.0366, 0.0307, 0.0326, 0.0397, 0.0406, 0.0329, 0.0433], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-28 19:50:16,651 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=72982.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 19:50:33,549 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.2285, 1.8026, 2.5075, 3.0937, 2.9386, 3.6708, 2.3299, 3.2992], device='cuda:1'), covar=tensor([0.0126, 0.0332, 0.0215, 0.0173, 0.0166, 0.0083, 0.0289, 0.0102], device='cuda:1'), in_proj_covar=tensor([0.0146, 0.0160, 0.0145, 0.0146, 0.0153, 0.0107, 0.0160, 0.0099], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001], device='cuda:1') 2023-04-28 19:50:41,095 INFO [optim.py:368] (1/8) 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] (1/8) Epoch 8, batch 1950, loss[loss=0.2253, simple_loss=0.2955, pruned_loss=0.0775, over 16765.00 frames. ], tot_loss[loss=0.1966, simple_loss=0.2772, pruned_loss=0.05802, over 3313752.48 frames. ], batch size: 124, lr: 8.74e-03, grad_scale: 4.0 2023-04-28 19:51:14,229 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=73025.0, num_to_drop=1, layers_to_drop={3} 2023-04-28 19:51:21,210 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=73030.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 19:51:50,194 INFO [train.py:904] (1/8) Epoch 8, batch 2000, loss[loss=0.2075, simple_loss=0.2955, pruned_loss=0.0598, over 17118.00 frames. ], tot_loss[loss=0.198, simple_loss=0.2777, pruned_loss=0.05914, over 3310478.17 frames. ], batch size: 48, lr: 8.74e-03, grad_scale: 8.0 2023-04-28 19:51:55,250 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.9305, 2.1639, 2.3331, 4.7192, 2.0645, 2.8227, 2.2719, 2.4347], device='cuda:1'), covar=tensor([0.0671, 0.3064, 0.1800, 0.0240, 0.3634, 0.1772, 0.2665, 0.3093], device='cuda:1'), in_proj_covar=tensor([0.0347, 0.0362, 0.0304, 0.0323, 0.0394, 0.0403, 0.0326, 0.0430], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-28 19:52:58,781 INFO [optim.py:368] (1/8) 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] (1/8) Epoch 8, batch 2050, loss[loss=0.2139, simple_loss=0.2806, pruned_loss=0.07363, over 16881.00 frames. ], tot_loss[loss=0.1973, simple_loss=0.2777, pruned_loss=0.05844, over 3303565.45 frames. ], batch size: 102, lr: 8.73e-03, grad_scale: 8.0 2023-04-28 19:53:07,735 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.8404, 3.9059, 4.1406, 3.0027, 3.7269, 4.1352, 3.8617, 2.4792], device='cuda:1'), covar=tensor([0.0299, 0.0066, 0.0026, 0.0236, 0.0058, 0.0051, 0.0043, 0.0294], device='cuda:1'), in_proj_covar=tensor([0.0120, 0.0066, 0.0063, 0.0117, 0.0068, 0.0078, 0.0071, 0.0114], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:1') 2023-04-28 19:54:09,330 INFO [train.py:904] (1/8) Epoch 8, batch 2100, loss[loss=0.2173, simple_loss=0.2876, pruned_loss=0.07355, over 16887.00 frames. ], tot_loss[loss=0.199, simple_loss=0.2788, pruned_loss=0.05956, over 3298996.96 frames. ], batch size: 109, lr: 8.73e-03, grad_scale: 8.0 2023-04-28 19:55:14,761 INFO [optim.py:368] (1/8) 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 19:55:16,488 INFO [train.py:904] (1/8) Epoch 8, batch 2150, loss[loss=0.2249, simple_loss=0.2983, pruned_loss=0.07576, over 16922.00 frames. ], tot_loss[loss=0.1994, simple_loss=0.2796, pruned_loss=0.05962, over 3309817.87 frames. ], batch size: 96, lr: 8.73e-03, grad_scale: 8.0 2023-04-28 19:55:22,685 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.9177, 1.6916, 2.2266, 2.7347, 2.7088, 2.7078, 1.8262, 2.9314], device='cuda:1'), covar=tensor([0.0085, 0.0282, 0.0197, 0.0151, 0.0133, 0.0145, 0.0282, 0.0072], device='cuda:1'), in_proj_covar=tensor([0.0146, 0.0159, 0.0146, 0.0147, 0.0152, 0.0107, 0.0159, 0.0099], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001], device='cuda:1') 2023-04-28 19:56:21,714 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 8, batch 2200, loss[loss=0.1829, simple_loss=0.2724, pruned_loss=0.04674, over 16837.00 frames. ], tot_loss[loss=0.1995, simple_loss=0.2801, pruned_loss=0.05945, over 3307685.98 frames. ], batch size: 42, lr: 8.73e-03, grad_scale: 8.0 2023-04-28 19:56:50,006 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.46 vs. limit=2.0 2023-04-28 19:57:20,263 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=73293.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 19:57:29,785 INFO [optim.py:368] (1/8) 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 19:57:31,974 INFO [train.py:904] (1/8) Epoch 8, batch 2250, loss[loss=0.1919, simple_loss=0.2594, pruned_loss=0.06225, over 16774.00 frames. ], tot_loss[loss=0.2004, simple_loss=0.2807, pruned_loss=0.06005, over 3305577.33 frames. ], batch size: 83, lr: 8.72e-03, grad_scale: 8.0 2023-04-28 19:57:37,974 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.2015, 2.4348, 2.4446, 4.9346, 2.1008, 3.1273, 2.4603, 2.5734], device='cuda:1'), covar=tensor([0.0635, 0.2901, 0.1816, 0.0248, 0.3790, 0.1958, 0.2589, 0.3231], device='cuda:1'), in_proj_covar=tensor([0.0350, 0.0366, 0.0308, 0.0328, 0.0397, 0.0409, 0.0330, 0.0436], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-28 19:57:56,913 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=73320.0, num_to_drop=1, layers_to_drop={0} 2023-04-28 19:58:15,615 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.26 vs. limit=2.0 2023-04-28 19:58:40,765 INFO [train.py:904] (1/8) Epoch 8, batch 2300, loss[loss=0.2015, simple_loss=0.2802, pruned_loss=0.06143, over 16163.00 frames. ], tot_loss[loss=0.2001, simple_loss=0.2803, pruned_loss=0.05993, over 3307888.45 frames. ], batch size: 164, lr: 8.72e-03, grad_scale: 8.0 2023-04-28 19:58:44,113 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=73354.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 19:59:48,860 INFO [optim.py:368] (1/8) Clipping_scale=2.0, grad-norm quartiles 1.976e+02 2.910e+02 3.420e+02 4.127e+02 1.105e+03, threshold=6.841e+02, percent-clipped=4.0 2023-04-28 19:59:49,986 INFO [train.py:904] (1/8) Epoch 8, batch 2350, loss[loss=0.2203, simple_loss=0.2879, pruned_loss=0.07633, over 16649.00 frames. ], tot_loss[loss=0.201, simple_loss=0.2814, pruned_loss=0.06025, over 3311051.10 frames. ], batch size: 89, lr: 8.72e-03, grad_scale: 8.0 2023-04-28 20:00:58,115 INFO [train.py:904] (1/8) Epoch 8, batch 2400, loss[loss=0.1941, simple_loss=0.2817, pruned_loss=0.05331, over 16697.00 frames. ], tot_loss[loss=0.2003, simple_loss=0.2814, pruned_loss=0.05961, over 3315198.09 frames. ], batch size: 57, lr: 8.71e-03, grad_scale: 8.0 2023-04-28 20:01:36,160 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=3.60 vs. limit=5.0 2023-04-28 20:01:37,274 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.86 vs. limit=2.0 2023-04-28 20:02:04,599 INFO [optim.py:368] (1/8) 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 20:02:05,790 INFO [train.py:904] (1/8) Epoch 8, batch 2450, loss[loss=0.2108, simple_loss=0.3033, pruned_loss=0.05914, over 16627.00 frames. ], tot_loss[loss=0.2014, simple_loss=0.2826, pruned_loss=0.06012, over 3309442.48 frames. ], batch size: 57, lr: 8.71e-03, grad_scale: 8.0 2023-04-28 20:03:12,522 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=73550.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 20:03:14,439 INFO [train.py:904] (1/8) Epoch 8, batch 2500, loss[loss=0.1887, simple_loss=0.283, pruned_loss=0.04724, over 17041.00 frames. ], tot_loss[loss=0.2003, simple_loss=0.282, pruned_loss=0.05931, over 3317046.16 frames. ], batch size: 50, lr: 8.71e-03, grad_scale: 8.0 2023-04-28 20:04:07,430 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=73590.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 20:04:18,253 INFO [zipformer.py:625] (1/8) 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] (1/8) 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 20:04:23,440 INFO [train.py:904] (1/8) Epoch 8, batch 2550, loss[loss=0.2034, simple_loss=0.3008, pruned_loss=0.05302, over 17066.00 frames. ], tot_loss[loss=0.2002, simple_loss=0.2821, pruned_loss=0.05908, over 3321248.73 frames. ], batch size: 53, lr: 8.70e-03, grad_scale: 8.0 2023-04-28 20:04:49,010 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=73620.0, num_to_drop=1, layers_to_drop={0} 2023-04-28 20:05:29,130 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=73649.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 20:05:32,245 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=73651.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 20:05:32,979 INFO [train.py:904] (1/8) Epoch 8, batch 2600, loss[loss=0.2077, simple_loss=0.2837, pruned_loss=0.06584, over 15558.00 frames. ], tot_loss[loss=0.1993, simple_loss=0.2813, pruned_loss=0.05865, over 3314380.64 frames. ], batch size: 191, lr: 8.70e-03, grad_scale: 8.0 2023-04-28 20:05:55,318 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=73668.0, num_to_drop=1, layers_to_drop={0} 2023-04-28 20:06:42,999 INFO [optim.py:368] (1/8) 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] (1/8) Epoch 8, batch 2650, loss[loss=0.1894, simple_loss=0.2828, pruned_loss=0.04794, over 17048.00 frames. ], tot_loss[loss=0.1993, simple_loss=0.2817, pruned_loss=0.05847, over 3317879.83 frames. ], batch size: 50, lr: 8.70e-03, grad_scale: 4.0 2023-04-28 20:07:23,351 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=73732.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 20:07:35,300 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.9469, 4.8597, 5.4264, 5.4117, 5.4500, 5.1154, 5.0461, 4.7877], device='cuda:1'), covar=tensor([0.0232, 0.0393, 0.0334, 0.0347, 0.0324, 0.0260, 0.0734, 0.0349], device='cuda:1'), in_proj_covar=tensor([0.0296, 0.0305, 0.0303, 0.0290, 0.0344, 0.0320, 0.0423, 0.0260], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:1') 2023-04-28 20:07:42,604 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.78 vs. limit=2.0 2023-04-28 20:07:50,956 INFO [train.py:904] (1/8) Epoch 8, batch 2700, loss[loss=0.2051, simple_loss=0.2805, pruned_loss=0.0649, over 16865.00 frames. ], tot_loss[loss=0.1987, simple_loss=0.2818, pruned_loss=0.05776, over 3329657.30 frames. ], batch size: 109, lr: 8.70e-03, grad_scale: 4.0 2023-04-28 20:07:55,761 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.3222, 1.8897, 2.5223, 3.0854, 2.9787, 3.6223, 2.3714, 3.4717], device='cuda:1'), covar=tensor([0.0090, 0.0284, 0.0173, 0.0160, 0.0152, 0.0074, 0.0229, 0.0077], device='cuda:1'), in_proj_covar=tensor([0.0148, 0.0160, 0.0144, 0.0149, 0.0155, 0.0109, 0.0160, 0.0099], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001], device='cuda:1') 2023-04-28 20:08:01,844 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.5634, 3.7323, 2.1510, 3.9037, 2.6758, 3.8871, 2.0851, 2.9893], device='cuda:1'), covar=tensor([0.0181, 0.0274, 0.1255, 0.0162, 0.0725, 0.0442, 0.1313, 0.0512], device='cuda:1'), in_proj_covar=tensor([0.0139, 0.0163, 0.0180, 0.0108, 0.0164, 0.0202, 0.0190, 0.0169], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-04-28 20:08:38,972 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([5.0791, 5.6565, 5.8906, 5.5863, 5.6396, 6.1883, 5.7073, 5.3736], device='cuda:1'), covar=tensor([0.0851, 0.1739, 0.1426, 0.1526, 0.2256, 0.0916, 0.1190, 0.2157], device='cuda:1'), in_proj_covar=tensor([0.0329, 0.0464, 0.0480, 0.0406, 0.0536, 0.0513, 0.0387, 0.0545], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-04-28 20:08:46,244 INFO [zipformer.py:625] (1/8) 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] (1/8) 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 20:08:57,127 INFO [train.py:904] (1/8) Epoch 8, batch 2750, loss[loss=0.2038, simple_loss=0.2804, pruned_loss=0.06364, over 15408.00 frames. ], tot_loss[loss=0.1985, simple_loss=0.2824, pruned_loss=0.0573, over 3335549.16 frames. ], batch size: 190, lr: 8.69e-03, grad_scale: 4.0 2023-04-28 20:09:58,644 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.91 vs. limit=2.0 2023-04-28 20:10:05,131 INFO [train.py:904] (1/8) Epoch 8, batch 2800, loss[loss=0.191, simple_loss=0.2627, pruned_loss=0.0596, over 16395.00 frames. ], tot_loss[loss=0.198, simple_loss=0.2815, pruned_loss=0.05723, over 3340765.19 frames. ], batch size: 146, lr: 8.69e-03, grad_scale: 8.0 2023-04-28 20:10:19,761 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=4.95 vs. limit=5.0 2023-04-28 20:11:14,610 INFO [optim.py:368] (1/8) 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] (1/8) Epoch 8, batch 2850, loss[loss=0.1996, simple_loss=0.271, pruned_loss=0.06406, over 16792.00 frames. ], tot_loss[loss=0.1984, simple_loss=0.2811, pruned_loss=0.05789, over 3330161.82 frames. ], batch size: 102, lr: 8.69e-03, grad_scale: 8.0 2023-04-28 20:12:01,593 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.66 vs. limit=2.0 2023-04-28 20:12:16,242 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=73946.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 20:12:20,813 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 8, batch 2900, loss[loss=0.1964, simple_loss=0.2765, pruned_loss=0.05812, over 16475.00 frames. ], tot_loss[loss=0.1979, simple_loss=0.28, pruned_loss=0.05793, over 3329713.66 frames. ], batch size: 68, lr: 8.68e-03, grad_scale: 8.0 2023-04-28 20:13:28,886 INFO [zipformer.py:625] (1/8) 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] (1/8) 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] (1/8) Epoch 8, batch 2950, loss[loss=0.2723, simple_loss=0.326, pruned_loss=0.1093, over 11642.00 frames. ], tot_loss[loss=0.1987, simple_loss=0.2791, pruned_loss=0.05913, over 3317130.93 frames. ], batch size: 246, lr: 8.68e-03, grad_scale: 8.0 2023-04-28 20:13:55,005 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.8832, 5.0417, 5.1627, 5.0782, 4.9981, 5.6095, 5.2148, 4.8230], device='cuda:1'), covar=tensor([0.1103, 0.1603, 0.1709, 0.1872, 0.2887, 0.1059, 0.1357, 0.2519], device='cuda:1'), in_proj_covar=tensor([0.0334, 0.0467, 0.0488, 0.0414, 0.0548, 0.0516, 0.0393, 0.0554], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-04-28 20:14:39,941 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.2611, 1.9871, 2.1455, 3.8703, 2.0228, 2.5523, 2.0925, 2.2492], device='cuda:1'), covar=tensor([0.0835, 0.2897, 0.1648, 0.0377, 0.3086, 0.1784, 0.2572, 0.2380], device='cuda:1'), in_proj_covar=tensor([0.0353, 0.0367, 0.0308, 0.0328, 0.0395, 0.0412, 0.0330, 0.0436], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-28 20:14:46,251 INFO [train.py:904] (1/8) Epoch 8, batch 3000, loss[loss=0.2267, simple_loss=0.2942, pruned_loss=0.07958, over 16856.00 frames. ], tot_loss[loss=0.201, simple_loss=0.2808, pruned_loss=0.06057, over 3304987.74 frames. ], batch size: 116, lr: 8.68e-03, grad_scale: 8.0 2023-04-28 20:14:46,251 INFO [train.py:929] (1/8) Computing validation loss 2023-04-28 20:14:55,852 INFO [train.py:938] (1/8) Epoch 8, validation: loss=0.1462, simple_loss=0.2525, pruned_loss=0.01995, over 944034.00 frames. 2023-04-28 20:14:55,853 INFO [train.py:939] (1/8) Maximum memory allocated so far is 17676MB 2023-04-28 20:15:13,440 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=74064.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 20:15:25,895 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.0295, 4.3790, 2.3515, 4.6942, 3.0587, 4.6743, 2.5722, 3.3154], device='cuda:1'), covar=tensor([0.0172, 0.0242, 0.1403, 0.0115, 0.0738, 0.0375, 0.1271, 0.0540], device='cuda:1'), in_proj_covar=tensor([0.0140, 0.0166, 0.0182, 0.0109, 0.0165, 0.0205, 0.0192, 0.0170], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-04-28 20:15:41,680 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.79 vs. limit=2.0 2023-04-28 20:15:45,911 INFO [zipformer.py:625] (1/8) 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] (1/8) 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 20:16:06,779 INFO [train.py:904] (1/8) Epoch 8, batch 3050, loss[loss=0.1913, simple_loss=0.2686, pruned_loss=0.05702, over 16817.00 frames. ], tot_loss[loss=0.2011, simple_loss=0.2809, pruned_loss=0.0606, over 3308285.37 frames. ], batch size: 83, lr: 8.68e-03, grad_scale: 8.0 2023-04-28 20:16:38,645 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=74125.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 20:17:13,362 INFO [train.py:904] (1/8) Epoch 8, batch 3100, loss[loss=0.1928, simple_loss=0.2802, pruned_loss=0.05268, over 17083.00 frames. ], tot_loss[loss=0.2012, simple_loss=0.2806, pruned_loss=0.06092, over 3307228.58 frames. ], batch size: 49, lr: 8.67e-03, grad_scale: 8.0 2023-04-28 20:18:21,290 INFO [optim.py:368] (1/8) 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] (1/8) Epoch 8, batch 3150, loss[loss=0.1631, simple_loss=0.2575, pruned_loss=0.03433, over 17256.00 frames. ], tot_loss[loss=0.1997, simple_loss=0.2797, pruned_loss=0.05986, over 3313987.66 frames. ], batch size: 52, lr: 8.67e-03, grad_scale: 8.0 2023-04-28 20:18:48,251 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=3.32 vs. limit=5.0 2023-04-28 20:19:00,154 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([5.4542, 5.7946, 5.4811, 5.5890, 5.2038, 4.9324, 5.2551, 5.8842], device='cuda:1'), covar=tensor([0.1068, 0.0839, 0.1094, 0.0577, 0.0803, 0.0653, 0.0956, 0.0862], device='cuda:1'), in_proj_covar=tensor([0.0499, 0.0637, 0.0523, 0.0427, 0.0398, 0.0412, 0.0529, 0.0473], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-28 20:19:14,443 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=74240.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 20:19:23,387 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=74246.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 20:19:32,369 INFO [train.py:904] (1/8) Epoch 8, batch 3200, loss[loss=0.189, simple_loss=0.2754, pruned_loss=0.05129, over 16539.00 frames. ], tot_loss[loss=0.1981, simple_loss=0.2782, pruned_loss=0.05896, over 3306174.20 frames. ], batch size: 68, lr: 8.67e-03, grad_scale: 8.0 2023-04-28 20:20:14,818 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.8765, 2.2117, 2.3548, 4.4767, 1.9683, 2.8827, 2.3159, 2.3926], device='cuda:1'), covar=tensor([0.0716, 0.3036, 0.1708, 0.0331, 0.3672, 0.1780, 0.2574, 0.2974], device='cuda:1'), in_proj_covar=tensor([0.0352, 0.0366, 0.0307, 0.0327, 0.0396, 0.0412, 0.0329, 0.0437], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-28 20:20:30,670 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=74294.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 20:20:40,612 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 8, batch 3250, loss[loss=0.2013, simple_loss=0.2798, pruned_loss=0.0614, over 16853.00 frames. ], tot_loss[loss=0.1985, simple_loss=0.2787, pruned_loss=0.05913, over 3308032.25 frames. ], batch size: 96, lr: 8.66e-03, grad_scale: 4.0 2023-04-28 20:20:42,358 INFO [optim.py:368] (1/8) 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 20:20:54,988 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.3860, 4.3599, 4.2748, 3.7939, 4.3302, 1.7308, 4.1152, 4.1005], device='cuda:1'), covar=tensor([0.0072, 0.0067, 0.0112, 0.0294, 0.0065, 0.2066, 0.0101, 0.0151], device='cuda:1'), in_proj_covar=tensor([0.0120, 0.0108, 0.0160, 0.0155, 0.0124, 0.0169, 0.0146, 0.0152], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-28 20:21:52,435 INFO [train.py:904] (1/8) Epoch 8, batch 3300, loss[loss=0.2234, simple_loss=0.2922, pruned_loss=0.07731, over 16700.00 frames. ], tot_loss[loss=0.1982, simple_loss=0.2791, pruned_loss=0.05864, over 3319731.76 frames. ], batch size: 134, lr: 8.66e-03, grad_scale: 4.0 2023-04-28 20:22:41,319 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 8, batch 3350, loss[loss=0.2319, simple_loss=0.2987, pruned_loss=0.08259, over 16455.00 frames. ], tot_loss[loss=0.1996, simple_loss=0.2805, pruned_loss=0.05935, over 3317006.88 frames. ], batch size: 146, lr: 8.66e-03, grad_scale: 4.0 2023-04-28 20:23:02,760 INFO [optim.py:368] (1/8) 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 20:23:04,440 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.2655, 2.0465, 1.5841, 1.9639, 2.4527, 2.2839, 2.4819, 2.5624], device='cuda:1'), covar=tensor([0.0107, 0.0238, 0.0313, 0.0272, 0.0117, 0.0181, 0.0151, 0.0149], device='cuda:1'), in_proj_covar=tensor([0.0122, 0.0185, 0.0181, 0.0181, 0.0182, 0.0186, 0.0189, 0.0173], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-28 20:23:27,973 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=74420.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 20:23:49,804 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=74436.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 20:24:09,842 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.8483, 2.2397, 1.5621, 1.8690, 2.8024, 2.6284, 3.0185, 2.8995], device='cuda:1'), covar=tensor([0.0113, 0.0247, 0.0389, 0.0303, 0.0124, 0.0182, 0.0165, 0.0139], device='cuda:1'), in_proj_covar=tensor([0.0123, 0.0185, 0.0182, 0.0182, 0.0182, 0.0187, 0.0190, 0.0174], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-28 20:24:11,180 INFO [train.py:904] (1/8) Epoch 8, batch 3400, loss[loss=0.2164, simple_loss=0.2889, pruned_loss=0.07199, over 16674.00 frames. ], tot_loss[loss=0.1996, simple_loss=0.2802, pruned_loss=0.0595, over 3316225.55 frames. ], batch size: 89, lr: 8.66e-03, grad_scale: 4.0 2023-04-28 20:24:20,427 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([1.7257, 3.3133, 2.9587, 5.0507, 4.3337, 4.7729, 1.5967, 3.4717], device='cuda:1'), covar=tensor([0.1345, 0.0535, 0.0941, 0.0128, 0.0222, 0.0248, 0.1413, 0.0618], device='cuda:1'), in_proj_covar=tensor([0.0147, 0.0153, 0.0172, 0.0121, 0.0202, 0.0206, 0.0171, 0.0170], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0003, 0.0004], device='cuda:1') 2023-04-28 20:25:15,929 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.8598, 4.8027, 4.6653, 4.1183, 4.7525, 2.0342, 4.5161, 4.6394], device='cuda:1'), covar=tensor([0.0064, 0.0061, 0.0117, 0.0306, 0.0077, 0.1830, 0.0118, 0.0120], device='cuda:1'), in_proj_covar=tensor([0.0120, 0.0107, 0.0159, 0.0155, 0.0125, 0.0168, 0.0145, 0.0152], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-28 20:25:21,712 INFO [train.py:904] (1/8) Epoch 8, batch 3450, loss[loss=0.2168, simple_loss=0.2781, pruned_loss=0.0777, over 16733.00 frames. ], tot_loss[loss=0.1987, simple_loss=0.2786, pruned_loss=0.05945, over 3313334.12 frames. ], batch size: 134, lr: 8.65e-03, grad_scale: 4.0 2023-04-28 20:25:22,415 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.18 vs. limit=2.0 2023-04-28 20:25:22,841 INFO [optim.py:368] (1/8) Clipping_scale=2.0, grad-norm quartiles 1.895e+02 2.611e+02 3.016e+02 3.656e+02 6.729e+02, threshold=6.032e+02, percent-clipped=3.0 2023-04-28 20:26:16,833 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.7457, 4.3134, 4.4029, 3.0355, 3.8121, 4.4128, 4.0395, 2.4837], device='cuda:1'), covar=tensor([0.0341, 0.0024, 0.0024, 0.0250, 0.0054, 0.0042, 0.0040, 0.0308], device='cuda:1'), in_proj_covar=tensor([0.0119, 0.0066, 0.0064, 0.0119, 0.0070, 0.0080, 0.0071, 0.0112], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:1') 2023-04-28 20:26:30,765 INFO [train.py:904] (1/8) Epoch 8, batch 3500, loss[loss=0.1703, simple_loss=0.2621, pruned_loss=0.03925, over 17121.00 frames. ], tot_loss[loss=0.1968, simple_loss=0.277, pruned_loss=0.05829, over 3311833.89 frames. ], batch size: 48, lr: 8.65e-03, grad_scale: 4.0 2023-04-28 20:27:05,368 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([5.0234, 4.9536, 4.8545, 4.5793, 4.4094, 4.9364, 4.8786, 4.5742], device='cuda:1'), covar=tensor([0.0529, 0.0449, 0.0257, 0.0242, 0.1044, 0.0346, 0.0362, 0.0611], device='cuda:1'), in_proj_covar=tensor([0.0241, 0.0286, 0.0283, 0.0253, 0.0317, 0.0287, 0.0194, 0.0325], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-04-28 20:27:31,774 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=74595.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 20:27:33,567 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=74596.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 20:27:42,854 INFO [train.py:904] (1/8) Epoch 8, batch 3550, loss[loss=0.1837, simple_loss=0.2566, pruned_loss=0.05538, over 15850.00 frames. ], tot_loss[loss=0.195, simple_loss=0.2754, pruned_loss=0.05734, over 3299461.17 frames. ], batch size: 35, lr: 8.65e-03, grad_scale: 4.0 2023-04-28 20:27:43,958 INFO [optim.py:368] (1/8) 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 20:28:26,515 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.5000, 4.5098, 4.6162, 4.5357, 4.3821, 5.0612, 4.6401, 4.2504], device='cuda:1'), covar=tensor([0.1305, 0.1651, 0.1353, 0.1678, 0.2410, 0.0989, 0.1267, 0.2288], device='cuda:1'), in_proj_covar=tensor([0.0329, 0.0465, 0.0478, 0.0406, 0.0537, 0.0512, 0.0387, 0.0544], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-04-28 20:28:27,835 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.4546, 2.2473, 1.6288, 1.9941, 2.6494, 2.3987, 2.8286, 2.7535], device='cuda:1'), covar=tensor([0.0107, 0.0208, 0.0316, 0.0282, 0.0124, 0.0215, 0.0135, 0.0139], device='cuda:1'), in_proj_covar=tensor([0.0123, 0.0185, 0.0180, 0.0181, 0.0182, 0.0185, 0.0189, 0.0172], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-28 20:28:39,753 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.81 vs. limit=2.0 2023-04-28 20:28:51,888 INFO [train.py:904] (1/8) Epoch 8, batch 3600, loss[loss=0.1858, simple_loss=0.2562, pruned_loss=0.05764, over 16836.00 frames. ], tot_loss[loss=0.1937, simple_loss=0.274, pruned_loss=0.05672, over 3302516.45 frames. ], batch size: 102, lr: 8.64e-03, grad_scale: 8.0 2023-04-28 20:28:56,987 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=74656.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 20:30:00,903 INFO [train.py:904] (1/8) Epoch 8, batch 3650, loss[loss=0.2046, simple_loss=0.2899, pruned_loss=0.05969, over 16645.00 frames. ], tot_loss[loss=0.194, simple_loss=0.2733, pruned_loss=0.05741, over 3296625.42 frames. ], batch size: 57, lr: 8.64e-03, grad_scale: 8.0 2023-04-28 20:30:02,112 INFO [optim.py:368] (1/8) 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 20:30:27,752 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 8, batch 3700, loss[loss=0.1993, simple_loss=0.2732, pruned_loss=0.06274, over 16568.00 frames. ], tot_loss[loss=0.1959, simple_loss=0.2727, pruned_loss=0.05952, over 3286014.52 frames. ], batch size: 146, lr: 8.64e-03, grad_scale: 8.0 2023-04-28 20:31:38,736 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=74768.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 20:32:29,720 INFO [train.py:904] (1/8) Epoch 8, batch 3750, loss[loss=0.2238, simple_loss=0.2867, pruned_loss=0.08045, over 16893.00 frames. ], tot_loss[loss=0.1975, simple_loss=0.2734, pruned_loss=0.06079, over 3275629.93 frames. ], batch size: 90, lr: 8.64e-03, grad_scale: 8.0 2023-04-28 20:32:30,688 INFO [optim.py:368] (1/8) 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 20:32:39,507 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.6400, 4.6304, 4.5359, 3.7819, 4.6143, 1.7722, 4.4202, 4.3578], device='cuda:1'), covar=tensor([0.0084, 0.0068, 0.0133, 0.0375, 0.0075, 0.2221, 0.0124, 0.0189], device='cuda:1'), in_proj_covar=tensor([0.0119, 0.0107, 0.0159, 0.0153, 0.0124, 0.0168, 0.0143, 0.0151], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-28 20:33:41,194 INFO [train.py:904] (1/8) Epoch 8, batch 3800, loss[loss=0.1958, simple_loss=0.2616, pruned_loss=0.06498, over 16904.00 frames. ], tot_loss[loss=0.1995, simple_loss=0.2745, pruned_loss=0.0623, over 3274046.84 frames. ], batch size: 90, lr: 8.63e-03, grad_scale: 8.0 2023-04-28 20:33:45,205 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.3720, 1.4729, 1.9904, 2.2994, 2.4003, 2.3159, 1.6144, 2.3988], device='cuda:1'), covar=tensor([0.0105, 0.0289, 0.0170, 0.0149, 0.0146, 0.0133, 0.0286, 0.0065], device='cuda:1'), in_proj_covar=tensor([0.0149, 0.0159, 0.0145, 0.0147, 0.0154, 0.0110, 0.0159, 0.0101], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001], device='cuda:1') 2023-04-28 20:34:04,434 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.5659, 3.6719, 2.6930, 2.1808, 2.5164, 2.0751, 3.5895, 3.3826], device='cuda:1'), covar=tensor([0.2221, 0.0609, 0.1349, 0.2091, 0.2165, 0.1768, 0.0495, 0.1037], device='cuda:1'), in_proj_covar=tensor([0.0291, 0.0258, 0.0280, 0.0268, 0.0290, 0.0215, 0.0259, 0.0290], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003], device='cuda:1') 2023-04-28 20:34:45,328 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=74896.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 20:34:51,952 INFO [train.py:904] (1/8) Epoch 8, batch 3850, loss[loss=0.2039, simple_loss=0.2754, pruned_loss=0.06623, over 16357.00 frames. ], tot_loss[loss=0.2007, simple_loss=0.2749, pruned_loss=0.06322, over 3270950.00 frames. ], batch size: 165, lr: 8.63e-03, grad_scale: 8.0 2023-04-28 20:34:53,139 INFO [optim.py:368] (1/8) 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 20:35:03,382 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=74909.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 20:35:30,422 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.3085, 4.0808, 4.3213, 4.4718, 4.5808, 4.1121, 4.2868, 4.5602], device='cuda:1'), covar=tensor([0.1076, 0.0822, 0.1119, 0.0522, 0.0437, 0.0992, 0.1308, 0.0498], device='cuda:1'), in_proj_covar=tensor([0.0481, 0.0591, 0.0736, 0.0603, 0.0453, 0.0453, 0.0468, 0.0517], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-28 20:35:52,591 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=74944.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 20:36:01,446 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=74951.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 20:36:02,298 INFO [train.py:904] (1/8) Epoch 8, batch 3900, loss[loss=0.1828, simple_loss=0.2551, pruned_loss=0.05524, over 16809.00 frames. ], tot_loss[loss=0.2007, simple_loss=0.2744, pruned_loss=0.06345, over 3270032.47 frames. ], batch size: 102, lr: 8.63e-03, grad_scale: 8.0 2023-04-28 20:36:20,400 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=74964.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 20:36:27,539 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=74970.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 20:36:47,964 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.8466, 3.3021, 2.9374, 1.8389, 2.5653, 2.3160, 3.3260, 3.3404], device='cuda:1'), covar=tensor([0.0294, 0.0622, 0.0668, 0.1687, 0.0880, 0.0955, 0.0598, 0.0798], device='cuda:1'), in_proj_covar=tensor([0.0139, 0.0139, 0.0153, 0.0138, 0.0133, 0.0123, 0.0134, 0.0150], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-04-28 20:37:12,341 INFO [train.py:904] (1/8) Epoch 8, batch 3950, loss[loss=0.1908, simple_loss=0.2588, pruned_loss=0.06134, over 16724.00 frames. ], tot_loss[loss=0.2013, simple_loss=0.2744, pruned_loss=0.06412, over 3277992.11 frames. ], batch size: 134, lr: 8.62e-03, grad_scale: 8.0 2023-04-28 20:37:14,093 INFO [optim.py:368] (1/8) Clipping_scale=2.0, grad-norm quartiles 1.920e+02 2.648e+02 3.131e+02 3.735e+02 8.073e+02, threshold=6.262e+02, percent-clipped=3.0 2023-04-28 20:37:46,142 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 8, batch 4000, loss[loss=0.1948, simple_loss=0.2706, pruned_loss=0.05949, over 16821.00 frames. ], tot_loss[loss=0.2015, simple_loss=0.2741, pruned_loss=0.06443, over 3278614.85 frames. ], batch size: 96, lr: 8.62e-03, grad_scale: 8.0 2023-04-28 20:38:38,756 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=4.88 vs. limit=5.0 2023-04-28 20:38:50,673 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([5.0423, 5.0540, 4.8300, 4.5977, 4.4782, 4.9153, 4.8472, 4.6093], device='cuda:1'), covar=tensor([0.0540, 0.0317, 0.0228, 0.0252, 0.0921, 0.0342, 0.0304, 0.0590], device='cuda:1'), in_proj_covar=tensor([0.0236, 0.0274, 0.0272, 0.0243, 0.0303, 0.0278, 0.0186, 0.0311], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-04-28 20:39:36,989 INFO [train.py:904] (1/8) Epoch 8, batch 4050, loss[loss=0.1787, simple_loss=0.2551, pruned_loss=0.05109, over 17255.00 frames. ], tot_loss[loss=0.2001, simple_loss=0.274, pruned_loss=0.06311, over 3264139.92 frames. ], batch size: 43, lr: 8.62e-03, grad_scale: 8.0 2023-04-28 20:39:38,166 INFO [optim.py:368] (1/8) Clipping_scale=2.0, grad-norm quartiles 1.540e+02 2.447e+02 2.780e+02 3.424e+02 6.417e+02, threshold=5.561e+02, percent-clipped=2.0 2023-04-28 20:40:49,054 INFO [train.py:904] (1/8) Epoch 8, batch 4100, loss[loss=0.205, simple_loss=0.2901, pruned_loss=0.05999, over 16738.00 frames. ], tot_loss[loss=0.1985, simple_loss=0.2744, pruned_loss=0.0613, over 3276093.55 frames. ], batch size: 83, lr: 8.62e-03, grad_scale: 8.0 2023-04-28 20:42:02,427 INFO [train.py:904] (1/8) Epoch 8, batch 4150, loss[loss=0.2316, simple_loss=0.3109, pruned_loss=0.07621, over 17021.00 frames. ], tot_loss[loss=0.2047, simple_loss=0.2817, pruned_loss=0.06385, over 3261885.86 frames. ], batch size: 53, lr: 8.61e-03, grad_scale: 8.0 2023-04-28 20:42:04,252 INFO [optim.py:368] (1/8) 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 20:42:38,174 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=4.75 vs. limit=5.0 2023-04-28 20:42:42,723 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=75228.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 20:43:19,308 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=75251.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 20:43:20,233 INFO [train.py:904] (1/8) Epoch 8, batch 4200, loss[loss=0.256, simple_loss=0.3227, pruned_loss=0.09462, over 11489.00 frames. ], tot_loss[loss=0.2112, simple_loss=0.2894, pruned_loss=0.06644, over 3217386.58 frames. ], batch size: 248, lr: 8.61e-03, grad_scale: 8.0 2023-04-28 20:43:40,423 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=75265.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 20:43:47,287 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=75270.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 20:43:58,024 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.4107, 2.9394, 2.6519, 2.3441, 2.3045, 2.2149, 2.8456, 2.8458], device='cuda:1'), covar=tensor([0.1752, 0.0646, 0.1019, 0.1464, 0.1564, 0.1430, 0.0358, 0.0712], device='cuda:1'), in_proj_covar=tensor([0.0285, 0.0251, 0.0272, 0.0261, 0.0283, 0.0210, 0.0254, 0.0281], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-28 20:44:16,434 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=75289.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 20:44:30,194 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 8, batch 4250, loss[loss=0.1943, simple_loss=0.2792, pruned_loss=0.05473, over 15322.00 frames. ], tot_loss[loss=0.2142, simple_loss=0.2933, pruned_loss=0.06758, over 3178750.06 frames. ], batch size: 190, lr: 8.61e-03, grad_scale: 8.0 2023-04-28 20:44:36,197 INFO [optim.py:368] (1/8) 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 20:45:01,373 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=75320.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 20:45:18,371 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=75331.0, num_to_drop=1, layers_to_drop={1} 2023-04-28 20:45:48,491 INFO [train.py:904] (1/8) Epoch 8, batch 4300, loss[loss=0.2138, simple_loss=0.303, pruned_loss=0.06231, over 16416.00 frames. ], tot_loss[loss=0.2134, simple_loss=0.2942, pruned_loss=0.06627, over 3184010.78 frames. ], batch size: 146, lr: 8.60e-03, grad_scale: 8.0 2023-04-28 20:46:31,311 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([5.1587, 5.1582, 5.0005, 4.8374, 4.6453, 5.0809, 4.9470, 4.7137], device='cuda:1'), covar=tensor([0.0417, 0.0166, 0.0198, 0.0162, 0.0692, 0.0192, 0.0228, 0.0444], device='cuda:1'), in_proj_covar=tensor([0.0218, 0.0256, 0.0256, 0.0228, 0.0283, 0.0258, 0.0174, 0.0289], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-28 20:47:02,655 INFO [train.py:904] (1/8) Epoch 8, batch 4350, loss[loss=0.2237, simple_loss=0.3054, pruned_loss=0.07103, over 17051.00 frames. ], tot_loss[loss=0.216, simple_loss=0.2972, pruned_loss=0.06737, over 3171287.44 frames. ], batch size: 55, lr: 8.60e-03, grad_scale: 8.0 2023-04-28 20:47:03,853 INFO [optim.py:368] (1/8) 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 20:48:17,344 INFO [train.py:904] (1/8) Epoch 8, batch 4400, loss[loss=0.2297, simple_loss=0.3171, pruned_loss=0.07118, over 16401.00 frames. ], tot_loss[loss=0.2182, simple_loss=0.2999, pruned_loss=0.0683, over 3182828.13 frames. ], batch size: 35, lr: 8.60e-03, grad_scale: 8.0 2023-04-28 20:48:24,175 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.5694, 3.1444, 3.0425, 1.6499, 2.5970, 2.0978, 3.0971, 3.1020], device='cuda:1'), covar=tensor([0.0246, 0.0556, 0.0505, 0.1804, 0.0736, 0.0893, 0.0591, 0.0816], device='cuda:1'), in_proj_covar=tensor([0.0137, 0.0136, 0.0153, 0.0138, 0.0132, 0.0122, 0.0132, 0.0146], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:1') 2023-04-28 20:49:26,866 INFO [train.py:904] (1/8) Epoch 8, batch 4450, loss[loss=0.2279, simple_loss=0.3061, pruned_loss=0.07482, over 16761.00 frames. ], tot_loss[loss=0.2205, simple_loss=0.303, pruned_loss=0.06905, over 3194497.13 frames. ], batch size: 39, lr: 8.60e-03, grad_scale: 8.0 2023-04-28 20:49:28,913 INFO [optim.py:368] (1/8) 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 20:50:38,184 INFO [train.py:904] (1/8) Epoch 8, batch 4500, loss[loss=0.2074, simple_loss=0.2893, pruned_loss=0.06273, over 16853.00 frames. ], tot_loss[loss=0.2201, simple_loss=0.3023, pruned_loss=0.06894, over 3199727.72 frames. ], batch size: 116, lr: 8.59e-03, grad_scale: 8.0 2023-04-28 20:50:57,699 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=75565.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 20:51:24,861 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=75584.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 20:51:51,099 INFO [train.py:904] (1/8) Epoch 8, batch 4550, loss[loss=0.2376, simple_loss=0.3182, pruned_loss=0.07848, over 16680.00 frames. ], tot_loss[loss=0.2202, simple_loss=0.3022, pruned_loss=0.06914, over 3209558.16 frames. ], batch size: 57, lr: 8.59e-03, grad_scale: 8.0 2023-04-28 20:51:52,275 INFO [optim.py:368] (1/8) 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 20:52:06,457 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=75613.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 20:52:14,746 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.1165, 1.8900, 2.1522, 3.6579, 1.8625, 2.3806, 2.0498, 2.0613], device='cuda:1'), covar=tensor([0.0809, 0.2821, 0.1626, 0.0365, 0.3402, 0.1772, 0.2568, 0.2838], device='cuda:1'), in_proj_covar=tensor([0.0344, 0.0367, 0.0304, 0.0321, 0.0395, 0.0412, 0.0327, 0.0435], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-28 20:52:17,103 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=75620.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 20:52:24,429 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=75626.0, num_to_drop=1, layers_to_drop={3} 2023-04-28 20:52:29,814 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.56 vs. limit=2.0 2023-04-28 20:52:41,977 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.8044, 3.8670, 4.2512, 4.1844, 4.2148, 3.8869, 3.9416, 3.8174], device='cuda:1'), covar=tensor([0.0266, 0.0371, 0.0275, 0.0383, 0.0355, 0.0312, 0.0746, 0.0472], device='cuda:1'), in_proj_covar=tensor([0.0282, 0.0287, 0.0286, 0.0276, 0.0327, 0.0304, 0.0408, 0.0248], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003], device='cuda:1') 2023-04-28 20:52:43,460 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.19 vs. limit=2.0 2023-04-28 20:53:02,628 INFO [train.py:904] (1/8) Epoch 8, batch 4600, loss[loss=0.2297, simple_loss=0.3207, pruned_loss=0.06934, over 17038.00 frames. ], tot_loss[loss=0.2204, simple_loss=0.3031, pruned_loss=0.06882, over 3213982.48 frames. ], batch size: 55, lr: 8.59e-03, grad_scale: 8.0 2023-04-28 20:53:25,889 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=75668.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 20:53:27,978 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.46 vs. limit=2.0 2023-04-28 20:53:42,258 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.24 vs. limit=2.0 2023-04-28 20:53:47,980 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([1.8004, 3.6398, 3.1767, 5.3665, 4.3832, 4.6304, 1.8719, 3.5295], device='cuda:1'), covar=tensor([0.1278, 0.0487, 0.0916, 0.0047, 0.0329, 0.0271, 0.1330, 0.0674], device='cuda:1'), in_proj_covar=tensor([0.0147, 0.0152, 0.0173, 0.0117, 0.0203, 0.0203, 0.0172, 0.0174], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0003, 0.0004], device='cuda:1') 2023-04-28 20:54:12,063 INFO [train.py:904] (1/8) Epoch 8, batch 4650, loss[loss=0.2165, simple_loss=0.2973, pruned_loss=0.06782, over 15348.00 frames. ], tot_loss[loss=0.219, simple_loss=0.3017, pruned_loss=0.06817, over 3211149.86 frames. ], batch size: 191, lr: 8.58e-03, grad_scale: 8.0 2023-04-28 20:54:13,264 INFO [optim.py:368] (1/8) Clipping_scale=2.0, grad-norm quartiles 1.478e+02 2.294e+02 2.810e+02 3.370e+02 9.861e+02, threshold=5.619e+02, percent-clipped=3.0 2023-04-28 20:54:47,565 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.9075, 3.4663, 3.3096, 1.8602, 2.8436, 2.0185, 3.5228, 3.4344], device='cuda:1'), covar=tensor([0.0200, 0.0577, 0.0562, 0.1872, 0.0780, 0.1069, 0.0527, 0.0872], device='cuda:1'), in_proj_covar=tensor([0.0135, 0.0135, 0.0152, 0.0138, 0.0132, 0.0121, 0.0132, 0.0145], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:1') 2023-04-28 20:55:17,919 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.7592, 4.7309, 4.5800, 4.3612, 4.1335, 4.6272, 4.5557, 4.2889], device='cuda:1'), covar=tensor([0.0487, 0.0417, 0.0219, 0.0197, 0.0936, 0.0372, 0.0254, 0.0532], device='cuda:1'), in_proj_covar=tensor([0.0214, 0.0251, 0.0252, 0.0225, 0.0280, 0.0253, 0.0173, 0.0285], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-28 20:55:23,489 INFO [train.py:904] (1/8) Epoch 8, batch 4700, loss[loss=0.2467, simple_loss=0.3263, pruned_loss=0.08351, over 11522.00 frames. ], tot_loss[loss=0.2165, simple_loss=0.2989, pruned_loss=0.06707, over 3197577.23 frames. ], batch size: 246, lr: 8.58e-03, grad_scale: 8.0 2023-04-28 20:56:00,647 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.3354, 3.2687, 2.4873, 2.0610, 2.4392, 2.0650, 3.2955, 3.1419], device='cuda:1'), covar=tensor([0.2630, 0.0810, 0.1594, 0.1958, 0.2032, 0.1757, 0.0572, 0.0888], device='cuda:1'), in_proj_covar=tensor([0.0291, 0.0253, 0.0277, 0.0265, 0.0284, 0.0212, 0.0258, 0.0282], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-28 20:56:31,924 INFO [train.py:904] (1/8) Epoch 8, batch 4750, loss[loss=0.1883, simple_loss=0.2664, pruned_loss=0.05511, over 16614.00 frames. ], tot_loss[loss=0.2125, simple_loss=0.2949, pruned_loss=0.06509, over 3203537.44 frames. ], batch size: 57, lr: 8.58e-03, grad_scale: 8.0 2023-04-28 20:56:33,066 INFO [optim.py:368] (1/8) 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 20:56:43,216 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.6671, 3.3210, 3.1069, 1.7349, 2.7697, 2.3558, 3.2424, 3.3027], device='cuda:1'), covar=tensor([0.0262, 0.0534, 0.0533, 0.1680, 0.0715, 0.0816, 0.0543, 0.0685], device='cuda:1'), in_proj_covar=tensor([0.0137, 0.0137, 0.0154, 0.0139, 0.0133, 0.0123, 0.0134, 0.0146], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-04-28 20:57:44,145 INFO [train.py:904] (1/8) Epoch 8, batch 4800, loss[loss=0.1952, simple_loss=0.2764, pruned_loss=0.05702, over 16507.00 frames. ], tot_loss[loss=0.2085, simple_loss=0.2912, pruned_loss=0.06285, over 3200630.27 frames. ], batch size: 68, lr: 8.58e-03, grad_scale: 8.0 2023-04-28 20:58:32,028 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 8, batch 4850, loss[loss=0.2161, simple_loss=0.3066, pruned_loss=0.06282, over 16708.00 frames. ], tot_loss[loss=0.2085, simple_loss=0.2927, pruned_loss=0.0622, over 3195786.33 frames. ], batch size: 134, lr: 8.57e-03, grad_scale: 8.0 2023-04-28 20:59:01,503 INFO [optim.py:368] (1/8) Clipping_scale=2.0, grad-norm quartiles 1.426e+02 2.327e+02 2.698e+02 3.138e+02 6.949e+02, threshold=5.395e+02, percent-clipped=1.0 2023-04-28 20:59:36,804 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=75926.0, num_to_drop=1, layers_to_drop={2} 2023-04-28 20:59:46,582 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=75932.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 20:59:53,816 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.4142, 3.1452, 2.9359, 1.7437, 2.6213, 2.2535, 3.0448, 3.1955], device='cuda:1'), covar=tensor([0.0256, 0.0556, 0.0564, 0.1685, 0.0753, 0.0880, 0.0602, 0.0573], device='cuda:1'), in_proj_covar=tensor([0.0136, 0.0134, 0.0152, 0.0139, 0.0131, 0.0122, 0.0133, 0.0145], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:1') 2023-04-28 21:00:17,705 INFO [train.py:904] (1/8) Epoch 8, batch 4900, loss[loss=0.1936, simple_loss=0.2897, pruned_loss=0.04869, over 15294.00 frames. ], tot_loss[loss=0.2071, simple_loss=0.292, pruned_loss=0.0611, over 3190667.23 frames. ], batch size: 190, lr: 8.57e-03, grad_scale: 8.0 2023-04-28 21:00:49,332 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=75974.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 21:01:34,531 INFO [train.py:904] (1/8) Epoch 8, batch 4950, loss[loss=0.2169, simple_loss=0.299, pruned_loss=0.0674, over 16435.00 frames. ], tot_loss[loss=0.2067, simple_loss=0.2917, pruned_loss=0.06083, over 3188146.31 frames. ], batch size: 68, lr: 8.57e-03, grad_scale: 8.0 2023-04-28 21:01:36,817 INFO [optim.py:368] (1/8) 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 21:02:45,330 INFO [train.py:904] (1/8) Epoch 8, batch 5000, loss[loss=0.1892, simple_loss=0.2808, pruned_loss=0.0488, over 17241.00 frames. ], tot_loss[loss=0.2072, simple_loss=0.2932, pruned_loss=0.06065, over 3192783.26 frames. ], batch size: 52, lr: 8.56e-03, grad_scale: 8.0 2023-04-28 21:03:55,755 INFO [train.py:904] (1/8) Epoch 8, batch 5050, loss[loss=0.2181, simple_loss=0.308, pruned_loss=0.06413, over 17049.00 frames. ], tot_loss[loss=0.2083, simple_loss=0.2944, pruned_loss=0.06113, over 3188560.83 frames. ], batch size: 53, lr: 8.56e-03, grad_scale: 8.0 2023-04-28 21:03:57,925 INFO [optim.py:368] (1/8) 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] (1/8) Epoch 8, batch 5100, loss[loss=0.2077, simple_loss=0.2953, pruned_loss=0.06004, over 16698.00 frames. ], tot_loss[loss=0.206, simple_loss=0.2921, pruned_loss=0.05997, over 3198374.66 frames. ], batch size: 134, lr: 8.56e-03, grad_scale: 8.0 2023-04-28 21:06:20,872 INFO [train.py:904] (1/8) Epoch 8, batch 5150, loss[loss=0.2242, simple_loss=0.3149, pruned_loss=0.0668, over 16578.00 frames. ], tot_loss[loss=0.2053, simple_loss=0.2921, pruned_loss=0.05931, over 3193460.93 frames. ], batch size: 57, lr: 8.56e-03, grad_scale: 8.0 2023-04-28 21:06:24,097 INFO [optim.py:368] (1/8) 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 21:07:33,721 INFO [train.py:904] (1/8) Epoch 8, batch 5200, loss[loss=0.1897, simple_loss=0.2734, pruned_loss=0.05304, over 16770.00 frames. ], tot_loss[loss=0.2038, simple_loss=0.2904, pruned_loss=0.05863, over 3201669.90 frames. ], batch size: 83, lr: 8.55e-03, grad_scale: 8.0 2023-04-28 21:07:59,819 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=76269.0, num_to_drop=1, layers_to_drop={1} 2023-04-28 21:08:36,419 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=76294.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 21:08:48,738 INFO [train.py:904] (1/8) Epoch 8, batch 5250, loss[loss=0.2439, simple_loss=0.3048, pruned_loss=0.09148, over 12555.00 frames. ], tot_loss[loss=0.202, simple_loss=0.2871, pruned_loss=0.05841, over 3194086.08 frames. ], batch size: 246, lr: 8.55e-03, grad_scale: 8.0 2023-04-28 21:08:51,147 INFO [optim.py:368] (1/8) 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 21:08:55,742 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.7981, 3.6641, 3.8004, 3.9707, 4.0302, 3.6640, 4.0014, 4.0630], device='cuda:1'), covar=tensor([0.1080, 0.0799, 0.1140, 0.0481, 0.0461, 0.1581, 0.0575, 0.0458], device='cuda:1'), in_proj_covar=tensor([0.0469, 0.0568, 0.0715, 0.0584, 0.0441, 0.0439, 0.0449, 0.0503], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-28 21:09:19,391 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=76323.0, num_to_drop=1, layers_to_drop={1} 2023-04-28 21:09:30,987 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=76330.0, num_to_drop=1, layers_to_drop={2} 2023-04-28 21:10:01,828 INFO [train.py:904] (1/8) Epoch 8, batch 5300, loss[loss=0.1675, simple_loss=0.2547, pruned_loss=0.04011, over 16370.00 frames. ], tot_loss[loss=0.199, simple_loss=0.2836, pruned_loss=0.05721, over 3188414.90 frames. ], batch size: 68, lr: 8.55e-03, grad_scale: 8.0 2023-04-28 21:10:06,458 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=76355.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 21:10:47,993 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=76384.0, num_to_drop=1, layers_to_drop={3} 2023-04-28 21:10:49,300 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.1282, 4.3811, 3.6717, 2.8143, 3.4206, 2.9597, 4.8842, 4.2156], device='cuda:1'), covar=tensor([0.2008, 0.0552, 0.1056, 0.1625, 0.1840, 0.1260, 0.0313, 0.0634], device='cuda:1'), in_proj_covar=tensor([0.0289, 0.0251, 0.0277, 0.0264, 0.0281, 0.0210, 0.0260, 0.0277], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-28 21:11:13,477 INFO [train.py:904] (1/8) Epoch 8, batch 5350, loss[loss=0.2052, simple_loss=0.2899, pruned_loss=0.06024, over 16805.00 frames. ], tot_loss[loss=0.1967, simple_loss=0.2814, pruned_loss=0.05602, over 3205308.88 frames. ], batch size: 83, lr: 8.54e-03, grad_scale: 8.0 2023-04-28 21:11:15,924 INFO [optim.py:368] (1/8) 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 21:11:50,314 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.39 vs. limit=2.0 2023-04-28 21:12:16,798 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.78 vs. limit=2.0 2023-04-28 21:12:26,519 INFO [train.py:904] (1/8) Epoch 8, batch 5400, loss[loss=0.2089, simple_loss=0.2997, pruned_loss=0.059, over 16552.00 frames. ], tot_loss[loss=0.1997, simple_loss=0.2849, pruned_loss=0.05729, over 3200435.59 frames. ], batch size: 75, lr: 8.54e-03, grad_scale: 8.0 2023-04-28 21:13:32,003 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([5.0983, 5.5852, 5.7555, 5.4734, 5.4674, 6.1310, 5.5578, 5.4255], device='cuda:1'), covar=tensor([0.0672, 0.1482, 0.1424, 0.1791, 0.2361, 0.0869, 0.1158, 0.2140], device='cuda:1'), in_proj_covar=tensor([0.0311, 0.0427, 0.0442, 0.0375, 0.0504, 0.0475, 0.0359, 0.0508], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-04-28 21:13:43,658 INFO [train.py:904] (1/8) Epoch 8, batch 5450, loss[loss=0.2183, simple_loss=0.2969, pruned_loss=0.06985, over 16709.00 frames. ], tot_loss[loss=0.2039, simple_loss=0.2888, pruned_loss=0.05952, over 3187818.27 frames. ], batch size: 57, lr: 8.54e-03, grad_scale: 8.0 2023-04-28 21:13:46,709 INFO [optim.py:368] (1/8) 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 21:14:33,560 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.85 vs. limit=2.0 2023-04-28 21:14:34,975 INFO [scaling.py:679] (1/8) 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] (1/8) attn_weights_entropy = tensor([3.3122, 3.4662, 1.6382, 3.7411, 2.4191, 3.7295, 1.8783, 2.5802], device='cuda:1'), covar=tensor([0.0219, 0.0311, 0.1782, 0.0104, 0.0801, 0.0387, 0.1626, 0.0708], device='cuda:1'), in_proj_covar=tensor([0.0138, 0.0158, 0.0184, 0.0101, 0.0165, 0.0196, 0.0190, 0.0167], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-04-28 21:15:01,741 INFO [train.py:904] (1/8) Epoch 8, batch 5500, loss[loss=0.2426, simple_loss=0.3215, pruned_loss=0.08183, over 16708.00 frames. ], tot_loss[loss=0.2148, simple_loss=0.298, pruned_loss=0.06579, over 3159455.73 frames. ], batch size: 89, lr: 8.54e-03, grad_scale: 8.0 2023-04-28 21:16:10,960 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([1.6500, 2.7077, 2.4179, 3.6830, 2.8090, 3.8367, 1.3649, 2.8110], device='cuda:1'), covar=tensor([0.1372, 0.0629, 0.1121, 0.0146, 0.0231, 0.0343, 0.1732, 0.0753], device='cuda:1'), in_proj_covar=tensor([0.0149, 0.0154, 0.0176, 0.0119, 0.0202, 0.0206, 0.0174, 0.0176], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0003, 0.0004], device='cuda:1') 2023-04-28 21:16:22,240 INFO [train.py:904] (1/8) Epoch 8, batch 5550, loss[loss=0.255, simple_loss=0.3354, pruned_loss=0.0873, over 16891.00 frames. ], tot_loss[loss=0.2243, simple_loss=0.3054, pruned_loss=0.0716, over 3140578.23 frames. ], batch size: 109, lr: 8.53e-03, grad_scale: 8.0 2023-04-28 21:16:26,052 INFO [optim.py:368] (1/8) 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 21:17:00,667 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=76625.0, num_to_drop=1, layers_to_drop={2} 2023-04-28 21:17:28,455 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.4122, 3.6454, 3.8455, 1.8573, 4.0754, 4.1274, 2.8953, 3.0581], device='cuda:1'), covar=tensor([0.0780, 0.0163, 0.0174, 0.1121, 0.0058, 0.0081, 0.0382, 0.0406], device='cuda:1'), in_proj_covar=tensor([0.0142, 0.0097, 0.0083, 0.0138, 0.0069, 0.0090, 0.0118, 0.0126], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0003, 0.0002, 0.0004, 0.0002, 0.0002, 0.0003, 0.0003], device='cuda:1') 2023-04-28 21:17:40,831 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=76650.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 21:17:43,809 INFO [train.py:904] (1/8) Epoch 8, batch 5600, loss[loss=0.296, simple_loss=0.3498, pruned_loss=0.1211, over 10864.00 frames. ], tot_loss[loss=0.232, simple_loss=0.311, pruned_loss=0.07649, over 3099957.03 frames. ], batch size: 246, lr: 8.53e-03, grad_scale: 8.0 2023-04-28 21:18:00,488 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.0771, 3.1159, 1.6592, 3.3188, 2.2172, 3.3428, 1.9446, 2.5452], device='cuda:1'), covar=tensor([0.0208, 0.0351, 0.1500, 0.0122, 0.0784, 0.0440, 0.1366, 0.0616], device='cuda:1'), in_proj_covar=tensor([0.0139, 0.0160, 0.0185, 0.0103, 0.0167, 0.0199, 0.0192, 0.0168], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-04-28 21:18:28,315 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 8, batch 5650, loss[loss=0.2279, simple_loss=0.3125, pruned_loss=0.07161, over 16859.00 frames. ], tot_loss[loss=0.2393, simple_loss=0.3162, pruned_loss=0.08119, over 3088875.64 frames. ], batch size: 42, lr: 8.53e-03, grad_scale: 8.0 2023-04-28 21:19:09,626 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.18 vs. limit=2.0 2023-04-28 21:19:10,210 INFO [optim.py:368] (1/8) Clipping_scale=2.0, grad-norm quartiles 2.391e+02 3.706e+02 4.487e+02 5.532e+02 1.181e+03, threshold=8.975e+02, percent-clipped=2.0 2023-04-28 21:20:27,956 INFO [train.py:904] (1/8) Epoch 8, batch 5700, loss[loss=0.3264, simple_loss=0.373, pruned_loss=0.1399, over 11053.00 frames. ], tot_loss[loss=0.2427, simple_loss=0.3182, pruned_loss=0.08358, over 3065641.81 frames. ], batch size: 248, lr: 8.53e-03, grad_scale: 8.0 2023-04-28 21:21:21,808 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.8334, 3.3211, 3.3701, 3.3308, 3.3323, 3.1903, 2.9401, 3.3092], device='cuda:1'), covar=tensor([0.0680, 0.0660, 0.0674, 0.0716, 0.0713, 0.0675, 0.1472, 0.0632], device='cuda:1'), in_proj_covar=tensor([0.0285, 0.0291, 0.0293, 0.0279, 0.0330, 0.0310, 0.0414, 0.0254], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003], device='cuda:1') 2023-04-28 21:21:49,368 INFO [train.py:904] (1/8) Epoch 8, batch 5750, loss[loss=0.232, simple_loss=0.3209, pruned_loss=0.07155, over 16866.00 frames. ], tot_loss[loss=0.2475, simple_loss=0.322, pruned_loss=0.08654, over 3010265.12 frames. ], batch size: 96, lr: 8.52e-03, grad_scale: 4.0 2023-04-28 21:21:54,072 INFO [optim.py:368] (1/8) 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 21:22:41,224 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.3251, 4.1252, 4.3147, 4.5246, 4.6018, 4.1457, 4.5629, 4.6235], device='cuda:1'), covar=tensor([0.1311, 0.0977, 0.1425, 0.0603, 0.0544, 0.0998, 0.0639, 0.0501], device='cuda:1'), in_proj_covar=tensor([0.0449, 0.0546, 0.0680, 0.0559, 0.0421, 0.0419, 0.0434, 0.0483], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-28 21:22:52,049 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=76839.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 21:23:12,556 INFO [train.py:904] (1/8) Epoch 8, batch 5800, loss[loss=0.2077, simple_loss=0.2975, pruned_loss=0.05894, over 16861.00 frames. ], tot_loss[loss=0.2451, simple_loss=0.3211, pruned_loss=0.08452, over 3028893.15 frames. ], batch size: 96, lr: 8.52e-03, grad_scale: 4.0 2023-04-28 21:24:30,502 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 8, batch 5850, loss[loss=0.2373, simple_loss=0.3056, pruned_loss=0.08446, over 11368.00 frames. ], tot_loss[loss=0.2408, simple_loss=0.3184, pruned_loss=0.08161, over 3046541.95 frames. ], batch size: 246, lr: 8.52e-03, grad_scale: 4.0 2023-04-28 21:24:37,986 INFO [optim.py:368] (1/8) 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 21:25:09,476 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=76925.0, num_to_drop=1, layers_to_drop={2} 2023-04-28 21:25:50,002 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([1.6174, 2.6286, 2.4153, 3.9949, 3.0494, 4.0226, 1.2852, 2.9774], device='cuda:1'), covar=tensor([0.1318, 0.0655, 0.1128, 0.0108, 0.0240, 0.0341, 0.1585, 0.0720], device='cuda:1'), in_proj_covar=tensor([0.0148, 0.0151, 0.0174, 0.0117, 0.0200, 0.0204, 0.0172, 0.0174], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0003, 0.0004], device='cuda:1') 2023-04-28 21:25:52,765 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 8, batch 5900, loss[loss=0.2828, simple_loss=0.3359, pruned_loss=0.1149, over 11638.00 frames. ], tot_loss[loss=0.2407, simple_loss=0.3183, pruned_loss=0.08161, over 3050159.30 frames. ], batch size: 246, lr: 8.51e-03, grad_scale: 4.0 2023-04-28 21:26:33,733 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=76973.0, num_to_drop=1, layers_to_drop={0} 2023-04-28 21:26:42,686 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=76979.0, num_to_drop=1, layers_to_drop={2} 2023-04-28 21:26:55,369 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.22 vs. limit=2.0 2023-04-28 21:27:02,918 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.2042, 4.4231, 2.5406, 5.0059, 3.2180, 4.9523, 2.8122, 3.3777], device='cuda:1'), covar=tensor([0.0142, 0.0216, 0.1326, 0.0082, 0.0616, 0.0271, 0.1094, 0.0486], device='cuda:1'), in_proj_covar=tensor([0.0139, 0.0160, 0.0184, 0.0103, 0.0166, 0.0198, 0.0191, 0.0167], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-04-28 21:27:10,469 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=76998.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 21:27:17,803 INFO [train.py:904] (1/8) Epoch 8, batch 5950, loss[loss=0.2208, simple_loss=0.3079, pruned_loss=0.06687, over 16559.00 frames. ], tot_loss[loss=0.2397, simple_loss=0.3193, pruned_loss=0.0801, over 3058534.70 frames. ], batch size: 75, lr: 8.51e-03, grad_scale: 4.0 2023-04-28 21:27:21,560 INFO [optim.py:368] (1/8) 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] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=77027.0, num_to_drop=1, layers_to_drop={0} 2023-04-28 21:28:33,566 INFO [train.py:904] (1/8) Epoch 8, batch 6000, loss[loss=0.2398, simple_loss=0.3129, pruned_loss=0.08331, over 15383.00 frames. ], tot_loss[loss=0.2387, simple_loss=0.3185, pruned_loss=0.07938, over 3073672.21 frames. ], batch size: 191, lr: 8.51e-03, grad_scale: 8.0 2023-04-28 21:28:33,566 INFO [train.py:929] (1/8) Computing validation loss 2023-04-28 21:28:44,113 INFO [train.py:938] (1/8) Epoch 8, validation: loss=0.1707, simple_loss=0.284, pruned_loss=0.02871, over 944034.00 frames. 2023-04-28 21:28:44,114 INFO [train.py:939] (1/8) Maximum memory allocated so far is 17676MB 2023-04-28 21:29:48,578 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=77095.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 21:30:00,325 INFO [train.py:904] (1/8) Epoch 8, batch 6050, loss[loss=0.233, simple_loss=0.3222, pruned_loss=0.07189, over 16578.00 frames. ], tot_loss[loss=0.2386, simple_loss=0.3175, pruned_loss=0.07982, over 3059081.86 frames. ], batch size: 62, lr: 8.51e-03, grad_scale: 8.0 2023-04-28 21:30:04,225 INFO [optim.py:368] (1/8) Clipping_scale=2.0, grad-norm quartiles 2.264e+02 3.621e+02 4.460e+02 5.601e+02 1.783e+03, threshold=8.919e+02, percent-clipped=7.0 2023-04-28 21:31:16,410 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=4.47 vs. limit=5.0 2023-04-28 21:31:19,254 INFO [train.py:904] (1/8) Epoch 8, batch 6100, loss[loss=0.2104, simple_loss=0.2993, pruned_loss=0.06075, over 16892.00 frames. ], tot_loss[loss=0.2366, simple_loss=0.3165, pruned_loss=0.0783, over 3076774.77 frames. ], batch size: 96, lr: 8.50e-03, grad_scale: 8.0 2023-04-28 21:31:22,685 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([1.7434, 1.7954, 1.4880, 1.5456, 1.8459, 1.6487, 1.7693, 1.9591], device='cuda:1'), covar=tensor([0.0079, 0.0155, 0.0224, 0.0219, 0.0114, 0.0166, 0.0114, 0.0107], device='cuda:1'), in_proj_covar=tensor([0.0108, 0.0179, 0.0176, 0.0178, 0.0175, 0.0178, 0.0175, 0.0162], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-28 21:31:26,091 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.3126, 3.4329, 3.7276, 1.6142, 3.9355, 3.9311, 2.8539, 2.7207], device='cuda:1'), covar=tensor([0.0814, 0.0178, 0.0110, 0.1339, 0.0051, 0.0085, 0.0391, 0.0538], device='cuda:1'), in_proj_covar=tensor([0.0143, 0.0097, 0.0083, 0.0141, 0.0070, 0.0092, 0.0118, 0.0127], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0003, 0.0002, 0.0004, 0.0002, 0.0002, 0.0003, 0.0003], device='cuda:1') 2023-04-28 21:31:26,114 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=77156.0, num_to_drop=1, layers_to_drop={0} 2023-04-28 21:31:46,029 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.7208, 4.5022, 4.7213, 4.9258, 5.0561, 4.5070, 4.9986, 5.0331], device='cuda:1'), covar=tensor([0.1317, 0.1158, 0.1358, 0.0538, 0.0433, 0.0818, 0.0496, 0.0464], device='cuda:1'), in_proj_covar=tensor([0.0464, 0.0568, 0.0705, 0.0578, 0.0436, 0.0433, 0.0449, 0.0498], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-28 21:31:48,525 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.9816, 3.5418, 3.4144, 1.9606, 2.8988, 2.4227, 3.4985, 3.5986], device='cuda:1'), covar=tensor([0.0246, 0.0591, 0.0530, 0.1680, 0.0741, 0.0846, 0.0545, 0.0733], device='cuda:1'), in_proj_covar=tensor([0.0139, 0.0135, 0.0157, 0.0141, 0.0135, 0.0125, 0.0136, 0.0147], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-04-28 21:31:53,388 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=2.00 vs. limit=2.0 2023-04-28 21:32:14,832 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([1.9721, 2.6961, 2.6270, 1.9911, 2.4398, 2.5761, 2.5959, 1.8851], device='cuda:1'), covar=tensor([0.0293, 0.0039, 0.0048, 0.0239, 0.0074, 0.0070, 0.0054, 0.0260], device='cuda:1'), in_proj_covar=tensor([0.0122, 0.0062, 0.0064, 0.0120, 0.0068, 0.0079, 0.0070, 0.0114], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:1') 2023-04-28 21:32:26,674 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=77195.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 21:32:37,680 INFO [train.py:904] (1/8) Epoch 8, batch 6150, loss[loss=0.2106, simple_loss=0.2956, pruned_loss=0.06282, over 16186.00 frames. ], tot_loss[loss=0.2349, simple_loss=0.3147, pruned_loss=0.07752, over 3088689.06 frames. ], batch size: 165, lr: 8.50e-03, grad_scale: 8.0 2023-04-28 21:32:42,675 INFO [optim.py:368] (1/8) 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 21:32:52,936 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=77211.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 21:33:59,283 INFO [train.py:904] (1/8) Epoch 8, batch 6200, loss[loss=0.2548, simple_loss=0.3161, pruned_loss=0.09673, over 11511.00 frames. ], tot_loss[loss=0.2325, simple_loss=0.3121, pruned_loss=0.07646, over 3088585.76 frames. ], batch size: 246, lr: 8.50e-03, grad_scale: 4.0 2023-04-28 21:34:29,793 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.63 vs. limit=2.0 2023-04-28 21:34:31,217 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=77272.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 21:34:33,153 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=77273.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 21:34:39,064 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 8, batch 6250, loss[loss=0.1969, simple_loss=0.2929, pruned_loss=0.05042, over 16744.00 frames. ], tot_loss[loss=0.2303, simple_loss=0.3104, pruned_loss=0.07512, over 3109724.47 frames. ], batch size: 89, lr: 8.50e-03, grad_scale: 4.0 2023-04-28 21:35:22,797 INFO [optim.py:368] (1/8) Clipping_scale=2.0, grad-norm quartiles 2.200e+02 3.064e+02 3.768e+02 4.806e+02 9.942e+02, threshold=7.536e+02, percent-clipped=2.0 2023-04-28 21:35:40,701 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.5433, 4.7307, 4.8579, 4.7974, 4.7529, 5.3437, 4.8046, 4.6267], device='cuda:1'), covar=tensor([0.1076, 0.1683, 0.1713, 0.1784, 0.2525, 0.0816, 0.1396, 0.2447], device='cuda:1'), in_proj_covar=tensor([0.0318, 0.0435, 0.0460, 0.0386, 0.0511, 0.0483, 0.0373, 0.0519], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-04-28 21:35:59,433 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([1.9361, 2.6945, 2.6691, 2.0165, 2.5054, 2.5766, 2.5736, 1.8693], device='cuda:1'), covar=tensor([0.0278, 0.0036, 0.0042, 0.0225, 0.0063, 0.0057, 0.0046, 0.0264], device='cuda:1'), in_proj_covar=tensor([0.0123, 0.0063, 0.0065, 0.0122, 0.0069, 0.0080, 0.0071, 0.0115], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:1') 2023-04-28 21:36:07,292 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=77334.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 21:36:13,020 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 8, batch 6300, loss[loss=0.2292, simple_loss=0.3076, pruned_loss=0.07541, over 16706.00 frames. ], tot_loss[loss=0.2306, simple_loss=0.3109, pruned_loss=0.07519, over 3102669.89 frames. ], batch size: 134, lr: 8.49e-03, grad_scale: 4.0 2023-04-28 21:37:14,946 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.5065, 2.0522, 2.1217, 4.1216, 1.9944, 2.6048, 2.1662, 2.1925], device='cuda:1'), covar=tensor([0.0772, 0.2927, 0.1843, 0.0304, 0.3397, 0.1866, 0.2617, 0.2818], device='cuda:1'), in_proj_covar=tensor([0.0344, 0.0365, 0.0304, 0.0318, 0.0398, 0.0407, 0.0326, 0.0430], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-28 21:37:54,077 INFO [train.py:904] (1/8) Epoch 8, batch 6350, loss[loss=0.2322, simple_loss=0.3096, pruned_loss=0.07736, over 16466.00 frames. ], tot_loss[loss=0.2341, simple_loss=0.313, pruned_loss=0.07761, over 3086284.24 frames. ], batch size: 68, lr: 8.49e-03, grad_scale: 4.0 2023-04-28 21:38:00,461 INFO [optim.py:368] (1/8) 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 21:39:10,452 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=77451.0, num_to_drop=1, layers_to_drop={2} 2023-04-28 21:39:11,216 INFO [train.py:904] (1/8) Epoch 8, batch 6400, loss[loss=0.2242, simple_loss=0.3014, pruned_loss=0.07353, over 16663.00 frames. ], tot_loss[loss=0.2354, simple_loss=0.3133, pruned_loss=0.07876, over 3081772.57 frames. ], batch size: 62, lr: 8.49e-03, grad_scale: 8.0 2023-04-28 21:39:53,867 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.3527, 4.3316, 4.8336, 4.7638, 4.7584, 4.4436, 4.3994, 4.2167], device='cuda:1'), covar=tensor([0.0290, 0.0468, 0.0320, 0.0390, 0.0401, 0.0281, 0.0961, 0.0475], device='cuda:1'), in_proj_covar=tensor([0.0291, 0.0296, 0.0298, 0.0285, 0.0334, 0.0314, 0.0421, 0.0256], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003], device='cuda:1') 2023-04-28 21:39:59,665 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([1.8053, 1.2919, 1.6581, 1.6881, 1.8222, 1.8556, 1.5691, 1.7842], device='cuda:1'), covar=tensor([0.0141, 0.0222, 0.0122, 0.0159, 0.0138, 0.0088, 0.0242, 0.0055], device='cuda:1'), in_proj_covar=tensor([0.0146, 0.0160, 0.0145, 0.0144, 0.0153, 0.0109, 0.0162, 0.0099], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001], device='cuda:1') 2023-04-28 21:40:17,191 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=77495.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 21:40:20,102 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.9430, 4.8912, 4.7623, 4.5293, 4.3081, 4.8077, 4.7624, 4.4312], device='cuda:1'), covar=tensor([0.0550, 0.0335, 0.0249, 0.0240, 0.0952, 0.0354, 0.0308, 0.0719], device='cuda:1'), in_proj_covar=tensor([0.0217, 0.0256, 0.0251, 0.0222, 0.0280, 0.0258, 0.0173, 0.0293], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-28 21:40:26,236 INFO [train.py:904] (1/8) Epoch 8, batch 6450, loss[loss=0.2215, simple_loss=0.3017, pruned_loss=0.07064, over 16212.00 frames. ], tot_loss[loss=0.2338, simple_loss=0.3125, pruned_loss=0.07755, over 3089153.47 frames. ], batch size: 165, lr: 8.48e-03, grad_scale: 8.0 2023-04-28 21:40:33,076 INFO [optim.py:368] (1/8) Clipping_scale=2.0, grad-norm quartiles 2.067e+02 3.422e+02 4.688e+02 6.041e+02 9.477e+02, threshold=9.377e+02, percent-clipped=7.0 2023-04-28 21:41:31,679 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=77543.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 21:41:45,632 INFO [train.py:904] (1/8) Epoch 8, batch 6500, loss[loss=0.2436, simple_loss=0.3287, pruned_loss=0.07922, over 16841.00 frames. ], tot_loss[loss=0.2317, simple_loss=0.3103, pruned_loss=0.07657, over 3092826.35 frames. ], batch size: 96, lr: 8.48e-03, grad_scale: 8.0 2023-04-28 21:42:09,168 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=77567.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 21:43:05,178 INFO [train.py:904] (1/8) Epoch 8, batch 6550, loss[loss=0.2539, simple_loss=0.339, pruned_loss=0.08438, over 16532.00 frames. ], tot_loss[loss=0.2358, simple_loss=0.3139, pruned_loss=0.07886, over 3071274.31 frames. ], batch size: 68, lr: 8.48e-03, grad_scale: 8.0 2023-04-28 21:43:11,237 INFO [optim.py:368] (1/8) 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 21:43:27,697 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=77616.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 21:43:47,288 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=77629.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 21:43:53,362 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=77633.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 21:44:22,368 INFO [train.py:904] (1/8) Epoch 8, batch 6600, loss[loss=0.2442, simple_loss=0.3248, pruned_loss=0.08178, over 16729.00 frames. ], tot_loss[loss=0.2377, simple_loss=0.3166, pruned_loss=0.07942, over 3084766.76 frames. ], batch size: 134, lr: 8.48e-03, grad_scale: 8.0 2023-04-28 21:45:00,776 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=77677.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 21:45:00,817 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=77677.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 21:45:21,643 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.8353, 5.1762, 4.8820, 4.8876, 4.5593, 4.5108, 4.6548, 5.2333], device='cuda:1'), covar=tensor([0.0785, 0.0658, 0.0895, 0.0575, 0.0697, 0.0754, 0.0764, 0.0700], device='cuda:1'), in_proj_covar=tensor([0.0468, 0.0585, 0.0496, 0.0399, 0.0370, 0.0383, 0.0486, 0.0428], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:1') 2023-04-28 21:45:38,844 INFO [train.py:904] (1/8) Epoch 8, batch 6650, loss[loss=0.3317, simple_loss=0.3694, pruned_loss=0.147, over 11711.00 frames. ], tot_loss[loss=0.2403, simple_loss=0.3179, pruned_loss=0.08134, over 3067585.97 frames. ], batch size: 248, lr: 8.47e-03, grad_scale: 8.0 2023-04-28 21:45:45,534 INFO [optim.py:368] (1/8) 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 21:46:35,393 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=77738.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 21:46:53,165 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=77751.0, num_to_drop=1, layers_to_drop={1} 2023-04-28 21:46:54,546 INFO [train.py:904] (1/8) Epoch 8, batch 6700, loss[loss=0.195, simple_loss=0.28, pruned_loss=0.055, over 16666.00 frames. ], tot_loss[loss=0.2394, simple_loss=0.3164, pruned_loss=0.08115, over 3071468.92 frames. ], batch size: 62, lr: 8.47e-03, grad_scale: 4.0 2023-04-28 21:48:06,139 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=77799.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 21:48:10,958 INFO [train.py:904] (1/8) Epoch 8, batch 6750, loss[loss=0.2498, simple_loss=0.3238, pruned_loss=0.08793, over 16176.00 frames. ], tot_loss[loss=0.2364, simple_loss=0.3138, pruned_loss=0.07952, over 3100989.06 frames. ], batch size: 165, lr: 8.47e-03, grad_scale: 4.0 2023-04-28 21:48:18,411 INFO [optim.py:368] (1/8) Clipping_scale=2.0, grad-norm quartiles 2.361e+02 3.344e+02 4.001e+02 5.088e+02 1.053e+03, threshold=8.003e+02, percent-clipped=2.0 2023-04-28 21:49:23,625 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.0045, 2.8971, 2.7498, 2.0435, 2.5755, 2.1788, 2.6851, 2.8821], device='cuda:1'), covar=tensor([0.0302, 0.0515, 0.0472, 0.1377, 0.0655, 0.0808, 0.0546, 0.0639], device='cuda:1'), in_proj_covar=tensor([0.0137, 0.0133, 0.0156, 0.0140, 0.0133, 0.0124, 0.0136, 0.0146], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-04-28 21:49:25,322 INFO [train.py:904] (1/8) Epoch 8, batch 6800, loss[loss=0.1967, simple_loss=0.2915, pruned_loss=0.05098, over 16819.00 frames. ], tot_loss[loss=0.2364, simple_loss=0.3138, pruned_loss=0.07951, over 3097833.14 frames. ], batch size: 102, lr: 8.47e-03, grad_scale: 8.0 2023-04-28 21:49:38,819 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.8686, 3.4941, 3.1915, 1.8444, 2.7276, 2.0874, 3.2832, 3.4265], device='cuda:1'), covar=tensor([0.0299, 0.0504, 0.0572, 0.1760, 0.0788, 0.1028, 0.0655, 0.0871], device='cuda:1'), in_proj_covar=tensor([0.0137, 0.0133, 0.0155, 0.0140, 0.0133, 0.0124, 0.0136, 0.0146], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-04-28 21:49:49,019 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=77867.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 21:50:42,754 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.1350, 3.9284, 4.1441, 4.3424, 4.4654, 3.9842, 4.3625, 4.4264], device='cuda:1'), covar=tensor([0.1290, 0.0958, 0.1382, 0.0605, 0.0496, 0.1149, 0.0694, 0.0512], device='cuda:1'), in_proj_covar=tensor([0.0463, 0.0572, 0.0705, 0.0582, 0.0442, 0.0439, 0.0462, 0.0505], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-28 21:50:44,158 INFO [train.py:904] (1/8) Epoch 8, batch 6850, loss[loss=0.2344, simple_loss=0.3281, pruned_loss=0.07031, over 16763.00 frames. ], tot_loss[loss=0.2367, simple_loss=0.315, pruned_loss=0.07923, over 3104383.80 frames. ], batch size: 134, lr: 8.46e-03, grad_scale: 4.0 2023-04-28 21:50:53,210 INFO [optim.py:368] (1/8) 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 21:51:03,700 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=77915.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 21:51:24,231 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=77929.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 21:51:29,897 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=77933.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 21:51:59,510 INFO [train.py:904] (1/8) Epoch 8, batch 6900, loss[loss=0.2464, simple_loss=0.325, pruned_loss=0.0839, over 16908.00 frames. ], tot_loss[loss=0.237, simple_loss=0.3174, pruned_loss=0.07833, over 3117573.02 frames. ], batch size: 109, lr: 8.46e-03, grad_scale: 4.0 2023-04-28 21:52:31,118 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=77972.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 21:52:39,050 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=77977.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 21:52:45,250 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 8, batch 6950, loss[loss=0.235, simple_loss=0.3168, pruned_loss=0.07665, over 16722.00 frames. ], tot_loss[loss=0.238, simple_loss=0.3181, pruned_loss=0.07896, over 3123953.03 frames. ], batch size: 83, lr: 8.46e-03, grad_scale: 4.0 2023-04-28 21:53:29,769 INFO [optim.py:368] (1/8) 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 21:54:01,564 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=4.92 vs. limit=5.0 2023-04-28 21:54:10,659 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 8, batch 7000, loss[loss=0.2273, simple_loss=0.3183, pruned_loss=0.06814, over 16834.00 frames. ], tot_loss[loss=0.2375, simple_loss=0.3178, pruned_loss=0.0786, over 3100153.65 frames. ], batch size: 116, lr: 8.45e-03, grad_scale: 4.0 2023-04-28 21:55:00,027 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.8332, 4.8622, 4.6921, 4.4867, 4.2562, 4.7473, 4.6649, 4.3976], device='cuda:1'), covar=tensor([0.0590, 0.0448, 0.0256, 0.0230, 0.0981, 0.0426, 0.0370, 0.0600], device='cuda:1'), in_proj_covar=tensor([0.0214, 0.0254, 0.0249, 0.0220, 0.0275, 0.0253, 0.0172, 0.0287], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-28 21:55:53,774 INFO [train.py:904] (1/8) Epoch 8, batch 7050, loss[loss=0.2323, simple_loss=0.3205, pruned_loss=0.07204, over 16213.00 frames. ], tot_loss[loss=0.2369, simple_loss=0.3181, pruned_loss=0.07781, over 3115160.39 frames. ], batch size: 165, lr: 8.45e-03, grad_scale: 4.0 2023-04-28 21:55:58,071 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.0395, 1.8877, 2.0998, 3.4839, 1.8968, 2.2706, 2.0861, 2.0122], device='cuda:1'), covar=tensor([0.0868, 0.2947, 0.1851, 0.0479, 0.3418, 0.1901, 0.2601, 0.2937], device='cuda:1'), in_proj_covar=tensor([0.0343, 0.0367, 0.0307, 0.0319, 0.0402, 0.0404, 0.0327, 0.0431], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-28 21:56:03,900 INFO [optim.py:368] (1/8) 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 21:56:19,332 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([1.8202, 2.6947, 2.6442, 1.8510, 2.4780, 2.6055, 2.5118, 1.7663], device='cuda:1'), covar=tensor([0.0314, 0.0042, 0.0042, 0.0272, 0.0077, 0.0064, 0.0058, 0.0300], device='cuda:1'), in_proj_covar=tensor([0.0123, 0.0062, 0.0064, 0.0122, 0.0069, 0.0080, 0.0071, 0.0114], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:1') 2023-04-28 21:57:11,208 INFO [train.py:904] (1/8) Epoch 8, batch 7100, loss[loss=0.2104, simple_loss=0.2957, pruned_loss=0.06257, over 16215.00 frames. ], tot_loss[loss=0.2368, simple_loss=0.3172, pruned_loss=0.07818, over 3097391.44 frames. ], batch size: 165, lr: 8.45e-03, grad_scale: 2.0 2023-04-28 21:57:52,037 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.2756, 1.4464, 1.8613, 2.2407, 2.3143, 2.4932, 1.5096, 2.4000], device='cuda:1'), covar=tensor([0.0149, 0.0344, 0.0221, 0.0206, 0.0187, 0.0112, 0.0349, 0.0081], device='cuda:1'), in_proj_covar=tensor([0.0142, 0.0161, 0.0144, 0.0143, 0.0151, 0.0108, 0.0158, 0.0099], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001], device='cuda:1') 2023-04-28 21:58:04,409 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.3946, 3.5743, 3.8071, 1.6867, 4.0872, 4.1015, 2.8925, 2.8167], device='cuda:1'), covar=tensor([0.0800, 0.0198, 0.0185, 0.1263, 0.0043, 0.0090, 0.0392, 0.0451], device='cuda:1'), in_proj_covar=tensor([0.0142, 0.0097, 0.0085, 0.0140, 0.0068, 0.0090, 0.0119, 0.0127], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0003, 0.0002, 0.0004, 0.0002, 0.0002, 0.0003, 0.0003], device='cuda:1') 2023-04-28 21:58:26,606 INFO [train.py:904] (1/8) Epoch 8, batch 7150, loss[loss=0.2128, simple_loss=0.3022, pruned_loss=0.06167, over 16869.00 frames. ], tot_loss[loss=0.2344, simple_loss=0.3145, pruned_loss=0.07715, over 3118735.27 frames. ], batch size: 96, lr: 8.45e-03, grad_scale: 2.0 2023-04-28 21:58:36,169 INFO [optim.py:368] (1/8) 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 21:59:09,638 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 8, batch 7200, loss[loss=0.1792, simple_loss=0.2658, pruned_loss=0.04631, over 16543.00 frames. ], tot_loss[loss=0.2313, simple_loss=0.3118, pruned_loss=0.07542, over 3100132.36 frames. ], batch size: 68, lr: 8.44e-03, grad_scale: 4.0 2023-04-28 22:00:09,928 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=78272.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 22:00:44,913 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=78293.0, num_to_drop=1, layers_to_drop={0} 2023-04-28 22:00:59,554 INFO [train.py:904] (1/8) Epoch 8, batch 7250, loss[loss=0.1993, simple_loss=0.274, pruned_loss=0.06234, over 16542.00 frames. ], tot_loss[loss=0.2287, simple_loss=0.3094, pruned_loss=0.07401, over 3101690.08 frames. ], batch size: 68, lr: 8.44e-03, grad_scale: 4.0 2023-04-28 22:01:07,714 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.1654, 3.5348, 3.4922, 1.9935, 3.0611, 2.5049, 3.6541, 3.6162], device='cuda:1'), covar=tensor([0.0250, 0.0523, 0.0526, 0.1650, 0.0681, 0.0849, 0.0517, 0.0692], device='cuda:1'), in_proj_covar=tensor([0.0138, 0.0133, 0.0156, 0.0141, 0.0133, 0.0124, 0.0135, 0.0146], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-04-28 22:01:10,048 INFO [optim.py:368] (1/8) 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 22:01:26,988 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=78320.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 22:01:47,677 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=78333.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 22:01:53,902 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([1.6421, 2.5691, 2.2021, 3.7917, 2.7045, 3.8704, 1.4366, 2.7580], device='cuda:1'), covar=tensor([0.1430, 0.0733, 0.1293, 0.0160, 0.0289, 0.0367, 0.1630, 0.0866], device='cuda:1'), in_proj_covar=tensor([0.0149, 0.0152, 0.0171, 0.0118, 0.0200, 0.0202, 0.0173, 0.0175], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0003, 0.0004], device='cuda:1') 2023-04-28 22:02:16,539 INFO [train.py:904] (1/8) Epoch 8, batch 7300, loss[loss=0.2981, simple_loss=0.3522, pruned_loss=0.122, over 11673.00 frames. ], tot_loss[loss=0.228, simple_loss=0.3087, pruned_loss=0.07367, over 3090445.29 frames. ], batch size: 247, lr: 8.44e-03, grad_scale: 4.0 2023-04-28 22:02:20,067 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.4619, 1.3129, 2.0655, 2.4263, 2.3397, 2.6480, 1.4333, 2.5507], device='cuda:1'), covar=tensor([0.0116, 0.0349, 0.0178, 0.0162, 0.0149, 0.0090, 0.0364, 0.0061], device='cuda:1'), in_proj_covar=tensor([0.0139, 0.0159, 0.0142, 0.0142, 0.0149, 0.0106, 0.0157, 0.0098], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001], device='cuda:1') 2023-04-28 22:03:02,230 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=78381.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 22:03:34,470 INFO [train.py:904] (1/8) Epoch 8, batch 7350, loss[loss=0.2219, simple_loss=0.3077, pruned_loss=0.06805, over 16821.00 frames. ], tot_loss[loss=0.2301, simple_loss=0.31, pruned_loss=0.07511, over 3070270.47 frames. ], batch size: 116, lr: 8.44e-03, grad_scale: 4.0 2023-04-28 22:03:39,138 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([1.8809, 2.9026, 3.0426, 1.6185, 3.3162, 3.3228, 2.4634, 2.4238], device='cuda:1'), covar=tensor([0.0945, 0.0270, 0.0244, 0.1194, 0.0061, 0.0111, 0.0490, 0.0463], device='cuda:1'), in_proj_covar=tensor([0.0139, 0.0097, 0.0082, 0.0137, 0.0066, 0.0088, 0.0118, 0.0124], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0003, 0.0002, 0.0004, 0.0002, 0.0002, 0.0003, 0.0003], device='cuda:1') 2023-04-28 22:03:45,281 INFO [optim.py:368] (1/8) 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 22:04:55,407 INFO [train.py:904] (1/8) Epoch 8, batch 7400, loss[loss=0.2638, simple_loss=0.3365, pruned_loss=0.09551, over 15130.00 frames. ], tot_loss[loss=0.2317, simple_loss=0.3116, pruned_loss=0.07585, over 3073116.49 frames. ], batch size: 190, lr: 8.43e-03, grad_scale: 4.0 2023-04-28 22:05:07,824 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=78460.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 22:06:13,356 INFO [train.py:904] (1/8) Epoch 8, batch 7450, loss[loss=0.2066, simple_loss=0.3049, pruned_loss=0.05418, over 16912.00 frames. ], tot_loss[loss=0.2328, simple_loss=0.3126, pruned_loss=0.07655, over 3079628.63 frames. ], batch size: 96, lr: 8.43e-03, grad_scale: 4.0 2023-04-28 22:06:26,556 INFO [optim.py:368] (1/8) 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 22:06:47,486 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=78521.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 22:07:30,840 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.7934, 3.8159, 3.2110, 2.2684, 2.7579, 2.3462, 4.1276, 3.6915], device='cuda:1'), covar=tensor([0.2250, 0.0661, 0.1216, 0.1778, 0.2019, 0.1537, 0.0372, 0.0750], device='cuda:1'), in_proj_covar=tensor([0.0294, 0.0251, 0.0274, 0.0261, 0.0277, 0.0210, 0.0258, 0.0274], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-28 22:07:34,683 INFO [train.py:904] (1/8) Epoch 8, batch 7500, loss[loss=0.2351, simple_loss=0.3118, pruned_loss=0.07926, over 16784.00 frames. ], tot_loss[loss=0.2328, simple_loss=0.3128, pruned_loss=0.0764, over 3077602.72 frames. ], batch size: 124, lr: 8.43e-03, grad_scale: 4.0 2023-04-28 22:07:42,119 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=78556.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 22:08:33,289 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=78588.0, num_to_drop=1, layers_to_drop={3} 2023-04-28 22:08:53,274 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.5937, 2.6482, 1.7040, 2.7732, 2.1318, 2.7646, 1.8556, 2.3761], device='cuda:1'), covar=tensor([0.0252, 0.0370, 0.1373, 0.0150, 0.0669, 0.0526, 0.1299, 0.0539], device='cuda:1'), in_proj_covar=tensor([0.0139, 0.0157, 0.0182, 0.0101, 0.0165, 0.0196, 0.0190, 0.0167], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-04-28 22:08:55,086 INFO [train.py:904] (1/8) Epoch 8, batch 7550, loss[loss=0.2162, simple_loss=0.2984, pruned_loss=0.06698, over 16783.00 frames. ], tot_loss[loss=0.2328, simple_loss=0.3119, pruned_loss=0.0768, over 3064034.37 frames. ], batch size: 124, lr: 8.42e-03, grad_scale: 4.0 2023-04-28 22:09:05,664 INFO [optim.py:368] (1/8) 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 22:09:19,415 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=78617.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 22:10:12,495 INFO [train.py:904] (1/8) Epoch 8, batch 7600, loss[loss=0.2335, simple_loss=0.3108, pruned_loss=0.07812, over 16591.00 frames. ], tot_loss[loss=0.2327, simple_loss=0.3114, pruned_loss=0.07706, over 3067529.06 frames. ], batch size: 68, lr: 8.42e-03, grad_scale: 8.0 2023-04-28 22:11:30,196 INFO [train.py:904] (1/8) Epoch 8, batch 7650, loss[loss=0.2233, simple_loss=0.3026, pruned_loss=0.07196, over 16681.00 frames. ], tot_loss[loss=0.2348, simple_loss=0.3127, pruned_loss=0.07845, over 3063144.30 frames. ], batch size: 89, lr: 8.42e-03, grad_scale: 8.0 2023-04-28 22:11:40,445 INFO [optim.py:368] (1/8) Clipping_scale=2.0, grad-norm quartiles 2.288e+02 3.310e+02 4.230e+02 5.150e+02 8.626e+02, threshold=8.460e+02, percent-clipped=2.0 2023-04-28 22:12:45,888 INFO [train.py:904] (1/8) Epoch 8, batch 7700, loss[loss=0.214, simple_loss=0.2953, pruned_loss=0.06635, over 16788.00 frames. ], tot_loss[loss=0.2352, simple_loss=0.3127, pruned_loss=0.07885, over 3055902.89 frames. ], batch size: 124, lr: 8.42e-03, grad_scale: 4.0 2023-04-28 22:14:03,977 INFO [train.py:904] (1/8) Epoch 8, batch 7750, loss[loss=0.2123, simple_loss=0.308, pruned_loss=0.05825, over 16898.00 frames. ], tot_loss[loss=0.2355, simple_loss=0.313, pruned_loss=0.07898, over 3047077.99 frames. ], batch size: 96, lr: 8.41e-03, grad_scale: 2.0 2023-04-28 22:14:09,728 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.6365, 4.5867, 4.4590, 3.7274, 4.4562, 1.5628, 4.2647, 4.2451], device='cuda:1'), covar=tensor([0.0059, 0.0055, 0.0111, 0.0315, 0.0075, 0.2231, 0.0093, 0.0151], device='cuda:1'), in_proj_covar=tensor([0.0110, 0.0096, 0.0143, 0.0140, 0.0114, 0.0162, 0.0128, 0.0133], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-28 22:14:15,933 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([5.0934, 5.1717, 5.7076, 5.6230, 5.6359, 5.2535, 5.2294, 4.9319], device='cuda:1'), covar=tensor([0.0293, 0.0355, 0.0328, 0.0435, 0.0434, 0.0283, 0.0842, 0.0389], device='cuda:1'), in_proj_covar=tensor([0.0296, 0.0300, 0.0302, 0.0291, 0.0343, 0.0319, 0.0422, 0.0262], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003], device='cuda:1') 2023-04-28 22:14:17,808 INFO [optim.py:368] (1/8) 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] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=78816.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 22:14:38,181 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.6501, 4.6144, 4.4633, 4.2775, 4.0619, 4.5226, 4.4020, 4.1709], device='cuda:1'), covar=tensor([0.0547, 0.0471, 0.0261, 0.0233, 0.0855, 0.0445, 0.0407, 0.0656], device='cuda:1'), in_proj_covar=tensor([0.0214, 0.0255, 0.0250, 0.0222, 0.0276, 0.0257, 0.0174, 0.0292], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-28 22:14:39,529 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([1.6914, 2.5101, 2.3023, 3.5925, 2.6160, 3.7193, 1.3241, 2.8587], device='cuda:1'), covar=tensor([0.1309, 0.0672, 0.1135, 0.0147, 0.0200, 0.0380, 0.1584, 0.0706], device='cuda:1'), in_proj_covar=tensor([0.0153, 0.0155, 0.0175, 0.0121, 0.0204, 0.0206, 0.0177, 0.0178], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-04-28 22:14:41,346 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.0911, 1.7348, 2.5400, 2.9459, 2.9095, 3.3819, 1.8865, 3.2138], device='cuda:1'), covar=tensor([0.0104, 0.0332, 0.0188, 0.0159, 0.0142, 0.0088, 0.0335, 0.0082], device='cuda:1'), in_proj_covar=tensor([0.0139, 0.0157, 0.0141, 0.0140, 0.0147, 0.0106, 0.0156, 0.0098], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001], device='cuda:1') 2023-04-28 22:15:19,809 INFO [train.py:904] (1/8) Epoch 8, batch 7800, loss[loss=0.2534, simple_loss=0.325, pruned_loss=0.09089, over 16712.00 frames. ], tot_loss[loss=0.2352, simple_loss=0.3131, pruned_loss=0.07866, over 3059205.03 frames. ], batch size: 134, lr: 8.41e-03, grad_scale: 2.0 2023-04-28 22:16:16,657 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=78888.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 22:16:21,753 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([1.9990, 2.3155, 2.3220, 2.9518, 2.2269, 3.2572, 1.6923, 2.7182], device='cuda:1'), covar=tensor([0.1087, 0.0518, 0.0877, 0.0142, 0.0149, 0.0401, 0.1233, 0.0626], device='cuda:1'), in_proj_covar=tensor([0.0152, 0.0155, 0.0175, 0.0121, 0.0203, 0.0205, 0.0175, 0.0178], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-04-28 22:16:36,905 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.4058, 4.1955, 4.4773, 4.6250, 4.7486, 4.3261, 4.6789, 4.6955], device='cuda:1'), covar=tensor([0.1356, 0.0943, 0.1196, 0.0516, 0.0476, 0.0759, 0.0553, 0.0507], device='cuda:1'), in_proj_covar=tensor([0.0467, 0.0571, 0.0701, 0.0581, 0.0445, 0.0430, 0.0464, 0.0507], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-28 22:16:38,225 INFO [train.py:904] (1/8) Epoch 8, batch 7850, loss[loss=0.2039, simple_loss=0.2963, pruned_loss=0.05577, over 16922.00 frames. ], tot_loss[loss=0.2363, simple_loss=0.3143, pruned_loss=0.07918, over 3035067.60 frames. ], batch size: 96, lr: 8.41e-03, grad_scale: 2.0 2023-04-28 22:16:50,958 INFO [optim.py:368] (1/8) 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 22:16:53,463 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=78912.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 22:17:18,999 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.6577, 4.9256, 4.6840, 4.6832, 4.4248, 4.3476, 4.4193, 4.9706], device='cuda:1'), covar=tensor([0.0897, 0.0765, 0.1014, 0.0623, 0.0768, 0.0935, 0.0902, 0.0796], device='cuda:1'), in_proj_covar=tensor([0.0462, 0.0577, 0.0496, 0.0398, 0.0361, 0.0387, 0.0485, 0.0432], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:1') 2023-04-28 22:17:20,809 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.1581, 4.1654, 4.6183, 4.5383, 4.5485, 4.2002, 4.2778, 4.0402], device='cuda:1'), covar=tensor([0.0308, 0.0415, 0.0368, 0.0426, 0.0397, 0.0349, 0.0817, 0.0507], device='cuda:1'), in_proj_covar=tensor([0.0296, 0.0300, 0.0302, 0.0294, 0.0343, 0.0318, 0.0422, 0.0264], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003], device='cuda:1') 2023-04-28 22:17:26,748 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.3684, 3.3326, 3.3831, 3.4868, 3.5128, 3.2458, 3.4587, 3.5388], device='cuda:1'), covar=tensor([0.0913, 0.0721, 0.0899, 0.0512, 0.0575, 0.2026, 0.0852, 0.0604], device='cuda:1'), in_proj_covar=tensor([0.0463, 0.0567, 0.0697, 0.0579, 0.0443, 0.0427, 0.0462, 0.0504], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-28 22:17:29,115 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 8, batch 7900, loss[loss=0.2365, simple_loss=0.3211, pruned_loss=0.07593, over 16360.00 frames. ], tot_loss[loss=0.2347, simple_loss=0.313, pruned_loss=0.07821, over 3056560.51 frames. ], batch size: 146, lr: 8.41e-03, grad_scale: 2.0 2023-04-28 22:18:11,334 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.40 vs. limit=2.0 2023-04-28 22:18:15,457 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=78965.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 22:18:24,932 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.8693, 4.1126, 3.0593, 2.4457, 2.9615, 2.4521, 4.4863, 3.9827], device='cuda:1'), covar=tensor([0.2401, 0.0673, 0.1482, 0.1847, 0.2374, 0.1562, 0.0351, 0.0807], device='cuda:1'), in_proj_covar=tensor([0.0295, 0.0252, 0.0274, 0.0263, 0.0278, 0.0211, 0.0258, 0.0276], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-28 22:18:49,743 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.36 vs. limit=2.0 2023-04-28 22:18:53,031 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([1.7949, 1.2749, 1.6088, 1.6824, 1.8774, 1.8109, 1.5810, 1.7519], device='cuda:1'), covar=tensor([0.0132, 0.0229, 0.0118, 0.0162, 0.0139, 0.0113, 0.0213, 0.0060], device='cuda:1'), in_proj_covar=tensor([0.0140, 0.0159, 0.0142, 0.0140, 0.0149, 0.0107, 0.0156, 0.0098], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001], device='cuda:1') 2023-04-28 22:19:13,438 INFO [train.py:904] (1/8) Epoch 8, batch 7950, loss[loss=0.244, simple_loss=0.3187, pruned_loss=0.08466, over 15343.00 frames. ], tot_loss[loss=0.2362, simple_loss=0.3138, pruned_loss=0.07928, over 3028496.07 frames. ], batch size: 190, lr: 8.40e-03, grad_scale: 2.0 2023-04-28 22:19:28,062 INFO [optim.py:368] (1/8) 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 22:19:52,423 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=79026.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 22:20:32,279 INFO [train.py:904] (1/8) Epoch 8, batch 8000, loss[loss=0.2893, simple_loss=0.3424, pruned_loss=0.1181, over 11513.00 frames. ], tot_loss[loss=0.2366, simple_loss=0.3143, pruned_loss=0.07949, over 3054684.94 frames. ], batch size: 248, lr: 8.40e-03, grad_scale: 4.0 2023-04-28 22:21:48,606 INFO [train.py:904] (1/8) Epoch 8, batch 8050, loss[loss=0.2884, simple_loss=0.3516, pruned_loss=0.1126, over 11707.00 frames. ], tot_loss[loss=0.2359, simple_loss=0.3141, pruned_loss=0.07885, over 3055732.62 frames. ], batch size: 247, lr: 8.40e-03, grad_scale: 4.0 2023-04-28 22:21:55,232 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.0308, 1.8485, 2.1375, 3.5275, 1.8818, 2.2623, 2.0551, 2.0073], device='cuda:1'), covar=tensor([0.0903, 0.3241, 0.1897, 0.0442, 0.3743, 0.2099, 0.2803, 0.2880], device='cuda:1'), in_proj_covar=tensor([0.0344, 0.0370, 0.0309, 0.0320, 0.0407, 0.0409, 0.0329, 0.0436], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-28 22:22:02,034 INFO [optim.py:368] (1/8) 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 22:22:11,217 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=79116.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 22:22:30,506 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.9760, 3.9865, 4.3971, 4.3433, 4.3428, 4.0105, 4.0797, 3.9811], device='cuda:1'), covar=tensor([0.0278, 0.0440, 0.0296, 0.0373, 0.0430, 0.0329, 0.0786, 0.0423], device='cuda:1'), in_proj_covar=tensor([0.0286, 0.0291, 0.0293, 0.0283, 0.0332, 0.0309, 0.0408, 0.0254], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003], device='cuda:1') 2023-04-28 22:23:05,315 INFO [train.py:904] (1/8) Epoch 8, batch 8100, loss[loss=0.2271, simple_loss=0.3053, pruned_loss=0.07447, over 16513.00 frames. ], tot_loss[loss=0.2349, simple_loss=0.3135, pruned_loss=0.07811, over 3057999.10 frames. ], batch size: 75, lr: 8.40e-03, grad_scale: 4.0 2023-04-28 22:23:15,695 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.85 vs. limit=2.0 2023-04-28 22:23:23,753 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=79164.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 22:23:28,373 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.3484, 4.2898, 4.2092, 2.9409, 4.2415, 1.4150, 3.9033, 3.9360], device='cuda:1'), covar=tensor([0.0127, 0.0099, 0.0174, 0.0693, 0.0111, 0.3008, 0.0160, 0.0302], device='cuda:1'), in_proj_covar=tensor([0.0111, 0.0097, 0.0145, 0.0143, 0.0116, 0.0164, 0.0130, 0.0135], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-28 22:24:22,973 INFO [train.py:904] (1/8) Epoch 8, batch 8150, loss[loss=0.1939, simple_loss=0.2756, pruned_loss=0.05608, over 16537.00 frames. ], tot_loss[loss=0.2336, simple_loss=0.3119, pruned_loss=0.07767, over 3064361.25 frames. ], batch size: 68, lr: 8.39e-03, grad_scale: 4.0 2023-04-28 22:24:36,887 INFO [optim.py:368] (1/8) 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 22:24:38,553 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 8, batch 8200, loss[loss=0.19, simple_loss=0.2721, pruned_loss=0.05392, over 17125.00 frames. ], tot_loss[loss=0.2302, simple_loss=0.3083, pruned_loss=0.07605, over 3068383.69 frames. ], batch size: 49, lr: 8.39e-03, grad_scale: 4.0 2023-04-28 22:25:55,508 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=79260.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 22:26:36,230 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=79285.0, num_to_drop=1, layers_to_drop={1} 2023-04-28 22:26:42,520 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.79 vs. limit=2.0 2023-04-28 22:27:04,307 INFO [train.py:904] (1/8) Epoch 8, batch 8250, loss[loss=0.1827, simple_loss=0.2654, pruned_loss=0.04998, over 12077.00 frames. ], tot_loss[loss=0.2274, simple_loss=0.3073, pruned_loss=0.07372, over 3055892.37 frames. ], batch size: 247, lr: 8.39e-03, grad_scale: 4.0 2023-04-28 22:27:19,441 INFO [optim.py:368] (1/8) 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 22:27:37,065 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=79321.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 22:27:44,893 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.66 vs. limit=2.0 2023-04-28 22:27:56,999 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=79333.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 22:28:09,484 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.3522, 4.6490, 4.4125, 4.4729, 4.1213, 4.1138, 4.2228, 4.6750], device='cuda:1'), covar=tensor([0.0991, 0.0901, 0.1037, 0.0638, 0.0802, 0.1260, 0.0853, 0.0838], device='cuda:1'), in_proj_covar=tensor([0.0463, 0.0574, 0.0492, 0.0398, 0.0360, 0.0387, 0.0482, 0.0430], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:1') 2023-04-28 22:28:17,636 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=79346.0, num_to_drop=1, layers_to_drop={2} 2023-04-28 22:28:26,472 INFO [train.py:904] (1/8) Epoch 8, batch 8300, loss[loss=0.1945, simple_loss=0.2939, pruned_loss=0.0476, over 16909.00 frames. ], tot_loss[loss=0.2223, simple_loss=0.3042, pruned_loss=0.07021, over 3059745.38 frames. ], batch size: 96, lr: 8.39e-03, grad_scale: 4.0 2023-04-28 22:28:27,053 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([1.7243, 1.2808, 1.5799, 1.6210, 1.7842, 1.7214, 1.6017, 1.5955], device='cuda:1'), covar=tensor([0.0131, 0.0235, 0.0114, 0.0162, 0.0159, 0.0120, 0.0233, 0.0062], device='cuda:1'), in_proj_covar=tensor([0.0140, 0.0157, 0.0141, 0.0139, 0.0148, 0.0107, 0.0156, 0.0098], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001], device='cuda:1') 2023-04-28 22:28:40,274 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.7560, 4.8025, 5.2095, 5.2017, 5.1269, 4.8639, 4.7360, 4.5816], device='cuda:1'), covar=tensor([0.0274, 0.0426, 0.0343, 0.0335, 0.0411, 0.0281, 0.1042, 0.0388], device='cuda:1'), in_proj_covar=tensor([0.0288, 0.0294, 0.0297, 0.0286, 0.0334, 0.0311, 0.0415, 0.0256], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003], device='cuda:1') 2023-04-28 22:28:46,913 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=79364.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 22:28:50,546 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.22 vs. limit=2.0 2023-04-28 22:29:04,991 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=79375.0, num_to_drop=1, layers_to_drop={0} 2023-04-28 22:29:22,581 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.4472, 4.4290, 4.2612, 3.7348, 4.3042, 1.7774, 4.1018, 4.1059], device='cuda:1'), covar=tensor([0.0054, 0.0048, 0.0099, 0.0234, 0.0062, 0.1959, 0.0085, 0.0126], device='cuda:1'), in_proj_covar=tensor([0.0107, 0.0093, 0.0140, 0.0138, 0.0112, 0.0160, 0.0126, 0.0130], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-28 22:29:36,118 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 8, batch 8350, loss[loss=0.2047, simple_loss=0.2985, pruned_loss=0.05542, over 16752.00 frames. ], tot_loss[loss=0.2189, simple_loss=0.3028, pruned_loss=0.06749, over 3076055.70 frames. ], batch size: 124, lr: 8.38e-03, grad_scale: 4.0 2023-04-28 22:30:02,872 INFO [optim.py:368] (1/8) 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 22:30:14,175 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.62 vs. limit=2.0 2023-04-28 22:30:26,148 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=79425.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 22:30:43,725 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=79436.0, num_to_drop=1, layers_to_drop={3} 2023-04-28 22:31:09,026 INFO [train.py:904] (1/8) Epoch 8, batch 8400, loss[loss=0.198, simple_loss=0.2925, pruned_loss=0.0517, over 16857.00 frames. ], tot_loss[loss=0.2152, simple_loss=0.3, pruned_loss=0.06518, over 3072351.80 frames. ], batch size: 102, lr: 8.38e-03, grad_scale: 8.0 2023-04-28 22:32:27,061 INFO [train.py:904] (1/8) Epoch 8, batch 8450, loss[loss=0.177, simple_loss=0.2722, pruned_loss=0.04093, over 16900.00 frames. ], tot_loss[loss=0.212, simple_loss=0.2979, pruned_loss=0.06309, over 3085043.40 frames. ], batch size: 102, lr: 8.38e-03, grad_scale: 8.0 2023-04-28 22:32:41,528 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.6622, 2.7704, 1.7351, 2.8988, 2.1521, 2.8793, 1.9784, 2.4567], device='cuda:1'), covar=tensor([0.0215, 0.0332, 0.1390, 0.0161, 0.0763, 0.0478, 0.1315, 0.0569], device='cuda:1'), in_proj_covar=tensor([0.0134, 0.0152, 0.0177, 0.0100, 0.0160, 0.0189, 0.0189, 0.0164], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0004, 0.0004, 0.0003], device='cuda:1') 2023-04-28 22:32:42,127 INFO [optim.py:368] (1/8) 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] (1/8) Epoch 8, batch 8500, loss[loss=0.1823, simple_loss=0.2713, pruned_loss=0.04666, over 16505.00 frames. ], tot_loss[loss=0.2069, simple_loss=0.2933, pruned_loss=0.06029, over 3080177.79 frames. ], batch size: 68, lr: 8.37e-03, grad_scale: 8.0 2023-04-28 22:34:10,246 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.5751, 4.6170, 4.3903, 4.2020, 4.0880, 4.5681, 4.3996, 4.2388], device='cuda:1'), covar=tensor([0.0526, 0.0368, 0.0252, 0.0226, 0.0796, 0.0315, 0.0332, 0.0570], device='cuda:1'), in_proj_covar=tensor([0.0208, 0.0248, 0.0246, 0.0218, 0.0268, 0.0249, 0.0171, 0.0282], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-28 22:34:28,857 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.20 vs. limit=2.0 2023-04-28 22:34:40,436 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=79585.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 22:34:53,467 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.4750, 4.4593, 4.2978, 3.7912, 4.3113, 1.5959, 4.1015, 4.1343], device='cuda:1'), covar=tensor([0.0064, 0.0070, 0.0127, 0.0274, 0.0079, 0.2214, 0.0113, 0.0156], device='cuda:1'), in_proj_covar=tensor([0.0109, 0.0096, 0.0142, 0.0137, 0.0113, 0.0163, 0.0128, 0.0131], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-28 22:35:09,512 INFO [train.py:904] (1/8) Epoch 8, batch 8550, loss[loss=0.2318, simple_loss=0.3228, pruned_loss=0.07043, over 16694.00 frames. ], tot_loss[loss=0.2045, simple_loss=0.2907, pruned_loss=0.05915, over 3050340.61 frames. ], batch size: 134, lr: 8.37e-03, grad_scale: 8.0 2023-04-28 22:35:26,481 INFO [optim.py:368] (1/8) 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 22:35:29,076 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.0381, 3.1627, 1.5709, 3.3728, 2.1997, 3.3189, 1.9075, 2.6521], device='cuda:1'), covar=tensor([0.0211, 0.0294, 0.1658, 0.0119, 0.0856, 0.0436, 0.1496, 0.0599], device='cuda:1'), in_proj_covar=tensor([0.0134, 0.0151, 0.0177, 0.0099, 0.0160, 0.0188, 0.0188, 0.0163], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0004, 0.0004, 0.0003], device='cuda:1') 2023-04-28 22:35:47,373 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=79621.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 22:36:28,518 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=79641.0, num_to_drop=1, layers_to_drop={1} 2023-04-28 22:36:36,801 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 8, batch 8600, loss[loss=0.1985, simple_loss=0.2921, pruned_loss=0.05245, over 16241.00 frames. ], tot_loss[loss=0.2039, simple_loss=0.2912, pruned_loss=0.05829, over 3054134.18 frames. ], batch size: 165, lr: 8.37e-03, grad_scale: 8.0 2023-04-28 22:36:59,411 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=79657.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 22:37:24,928 INFO [zipformer.py:625] (1/8) 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] (1/8) 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] (1/8) Epoch 8, batch 8650, loss[loss=0.1975, simple_loss=0.2888, pruned_loss=0.05308, over 15406.00 frames. ], tot_loss[loss=0.2013, simple_loss=0.2895, pruned_loss=0.05652, over 3056832.00 frames. ], batch size: 191, lr: 8.37e-03, grad_scale: 4.0 2023-04-28 22:38:50,263 INFO [optim.py:368] (1/8) 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 22:39:04,688 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=79718.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 22:39:09,490 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=79720.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 22:39:32,710 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=79731.0, num_to_drop=1, layers_to_drop={2} 2023-04-28 22:40:12,066 INFO [train.py:904] (1/8) Epoch 8, batch 8700, loss[loss=0.1908, simple_loss=0.2815, pruned_loss=0.05007, over 16662.00 frames. ], tot_loss[loss=0.1982, simple_loss=0.2863, pruned_loss=0.0551, over 3057834.62 frames. ], batch size: 134, lr: 8.36e-03, grad_scale: 4.0 2023-04-28 22:40:32,983 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=79762.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 22:40:54,512 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.51 vs. limit=2.0 2023-04-28 22:41:03,584 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.2359, 1.8824, 2.0433, 3.6975, 1.8331, 2.3567, 2.0478, 2.0391], device='cuda:1'), covar=tensor([0.0769, 0.3158, 0.1961, 0.0372, 0.3777, 0.1921, 0.2778, 0.2966], device='cuda:1'), in_proj_covar=tensor([0.0329, 0.0354, 0.0297, 0.0304, 0.0387, 0.0388, 0.0315, 0.0413], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-28 22:41:50,473 INFO [train.py:904] (1/8) Epoch 8, batch 8750, loss[loss=0.1939, simple_loss=0.2949, pruned_loss=0.04651, over 16888.00 frames. ], tot_loss[loss=0.1976, simple_loss=0.2859, pruned_loss=0.05465, over 3065630.19 frames. ], batch size: 116, lr: 8.36e-03, grad_scale: 4.0 2023-04-28 22:42:10,787 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.2256, 3.3471, 3.5783, 3.5567, 3.5896, 3.3823, 3.4057, 3.4225], device='cuda:1'), covar=tensor([0.0283, 0.0529, 0.0447, 0.0478, 0.0423, 0.0395, 0.0742, 0.0356], device='cuda:1'), in_proj_covar=tensor([0.0271, 0.0273, 0.0275, 0.0268, 0.0312, 0.0292, 0.0383, 0.0239], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0002], device='cuda:1') 2023-04-28 22:42:15,364 INFO [optim.py:368] (1/8) 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 22:42:43,954 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=79823.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 22:43:44,532 INFO [train.py:904] (1/8) Epoch 8, batch 8800, loss[loss=0.2058, simple_loss=0.2925, pruned_loss=0.0596, over 16936.00 frames. ], tot_loss[loss=0.1949, simple_loss=0.2836, pruned_loss=0.05306, over 3056491.08 frames. ], batch size: 109, lr: 8.36e-03, grad_scale: 8.0 2023-04-28 22:43:52,754 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.39 vs. limit=2.0 2023-04-28 22:44:20,004 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.23 vs. limit=2.0 2023-04-28 22:45:31,356 INFO [train.py:904] (1/8) Epoch 8, batch 8850, loss[loss=0.2014, simple_loss=0.3013, pruned_loss=0.0508, over 16666.00 frames. ], tot_loss[loss=0.1953, simple_loss=0.2859, pruned_loss=0.05233, over 3057807.63 frames. ], batch size: 134, lr: 8.36e-03, grad_scale: 8.0 2023-04-28 22:45:51,485 INFO [optim.py:368] (1/8) Clipping_scale=2.0, grad-norm quartiles 1.540e+02 2.661e+02 3.236e+02 3.873e+02 8.211e+02, threshold=6.471e+02, percent-clipped=3.0 2023-04-28 22:46:57,777 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=79941.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 22:46:57,946 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=79941.0, num_to_drop=1, layers_to_drop={0} 2023-04-28 22:47:07,139 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.6017, 1.5285, 1.8771, 2.5549, 2.4113, 2.5270, 1.8069, 2.6186], device='cuda:1'), covar=tensor([0.0098, 0.0324, 0.0217, 0.0140, 0.0167, 0.0109, 0.0298, 0.0080], device='cuda:1'), in_proj_covar=tensor([0.0139, 0.0156, 0.0142, 0.0137, 0.0144, 0.0103, 0.0154, 0.0094], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001], device='cuda:1') 2023-04-28 22:47:20,936 INFO [train.py:904] (1/8) Epoch 8, batch 8900, loss[loss=0.1924, simple_loss=0.2862, pruned_loss=0.04929, over 15560.00 frames. ], tot_loss[loss=0.1947, simple_loss=0.2864, pruned_loss=0.05148, over 3075173.42 frames. ], batch size: 191, lr: 8.35e-03, grad_scale: 8.0 2023-04-28 22:47:42,492 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.3995, 2.9802, 2.5569, 2.2057, 2.1243, 2.1351, 2.9967, 2.9301], device='cuda:1'), covar=tensor([0.2064, 0.0702, 0.1270, 0.1923, 0.1876, 0.1622, 0.0472, 0.0896], device='cuda:1'), in_proj_covar=tensor([0.0284, 0.0245, 0.0268, 0.0255, 0.0255, 0.0206, 0.0250, 0.0262], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-28 22:48:54,505 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=79989.0, num_to_drop=1, layers_to_drop={0} 2023-04-28 22:48:54,587 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=79989.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 22:48:58,063 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.4403, 4.2506, 4.4621, 4.5887, 4.7324, 4.2781, 4.7316, 4.7279], device='cuda:1'), covar=tensor([0.1179, 0.0852, 0.1203, 0.0546, 0.0468, 0.0807, 0.0376, 0.0412], device='cuda:1'), in_proj_covar=tensor([0.0437, 0.0541, 0.0662, 0.0551, 0.0421, 0.0414, 0.0438, 0.0481], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-28 22:49:13,176 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=4.08 vs. limit=5.0 2023-04-28 22:49:29,483 INFO [train.py:904] (1/8) Epoch 8, batch 8950, loss[loss=0.1805, simple_loss=0.2695, pruned_loss=0.04574, over 15297.00 frames. ], tot_loss[loss=0.1948, simple_loss=0.286, pruned_loss=0.05178, over 3084704.76 frames. ], batch size: 191, lr: 8.35e-03, grad_scale: 8.0 2023-04-28 22:49:50,484 INFO [optim.py:368] (1/8) 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 22:49:54,187 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=80013.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 22:50:08,334 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=80020.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 22:50:32,385 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=80031.0, num_to_drop=1, layers_to_drop={1} 2023-04-28 22:50:46,183 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=80037.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 22:51:17,194 INFO [train.py:904] (1/8) Epoch 8, batch 9000, loss[loss=0.1686, simple_loss=0.2588, pruned_loss=0.03915, over 16705.00 frames. ], tot_loss[loss=0.192, simple_loss=0.2826, pruned_loss=0.05065, over 3055400.65 frames. ], batch size: 83, lr: 8.35e-03, grad_scale: 8.0 2023-04-28 22:51:17,194 INFO [train.py:929] (1/8) Computing validation loss 2023-04-28 22:51:27,531 INFO [train.py:938] (1/8) Epoch 8, validation: loss=0.1608, simple_loss=0.265, pruned_loss=0.02828, over 944034.00 frames. 2023-04-28 22:51:27,532 INFO [train.py:939] (1/8) Maximum memory allocated so far is 17676MB 2023-04-28 22:52:04,336 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=80068.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 22:52:15,323 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.97 vs. limit=2.0 2023-04-28 22:52:26,509 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=80079.0, num_to_drop=1, layers_to_drop={0} 2023-04-28 22:53:14,313 INFO [train.py:904] (1/8) Epoch 8, batch 9050, loss[loss=0.1995, simple_loss=0.289, pruned_loss=0.05499, over 16991.00 frames. ], tot_loss[loss=0.1929, simple_loss=0.2835, pruned_loss=0.05111, over 3072014.78 frames. ], batch size: 55, lr: 8.35e-03, grad_scale: 8.0 2023-04-28 22:53:35,355 INFO [optim.py:368] (1/8) 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 22:53:47,906 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.82 vs. limit=2.0 2023-04-28 22:53:48,902 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=80118.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 22:54:06,926 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=4.71 vs. limit=5.0 2023-04-28 22:54:10,058 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.0007, 1.9249, 2.2239, 3.2344, 2.0945, 2.2699, 2.1944, 2.0031], device='cuda:1'), covar=tensor([0.0762, 0.3009, 0.1662, 0.0459, 0.3558, 0.2061, 0.2516, 0.3259], device='cuda:1'), in_proj_covar=tensor([0.0330, 0.0352, 0.0301, 0.0305, 0.0387, 0.0387, 0.0315, 0.0413], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-28 22:54:59,313 INFO [train.py:904] (1/8) Epoch 8, batch 9100, loss[loss=0.1863, simple_loss=0.2868, pruned_loss=0.04286, over 16876.00 frames. ], tot_loss[loss=0.1931, simple_loss=0.2831, pruned_loss=0.05158, over 3063638.73 frames. ], batch size: 96, lr: 8.34e-03, grad_scale: 8.0 2023-04-28 22:56:59,464 INFO [train.py:904] (1/8) Epoch 8, batch 9150, loss[loss=0.1885, simple_loss=0.2808, pruned_loss=0.04813, over 16305.00 frames. ], tot_loss[loss=0.1928, simple_loss=0.2831, pruned_loss=0.05124, over 3050917.83 frames. ], batch size: 166, lr: 8.34e-03, grad_scale: 8.0 2023-04-28 22:57:20,184 INFO [optim.py:368] (1/8) 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 22:57:46,771 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.1192, 2.8449, 2.8463, 2.0550, 2.5821, 2.1863, 2.6349, 2.9443], device='cuda:1'), covar=tensor([0.0330, 0.0649, 0.0459, 0.1426, 0.0688, 0.0869, 0.0667, 0.0694], device='cuda:1'), in_proj_covar=tensor([0.0134, 0.0124, 0.0149, 0.0136, 0.0128, 0.0121, 0.0130, 0.0135], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:1') 2023-04-28 22:58:26,807 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=80241.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 22:58:44,954 INFO [train.py:904] (1/8) Epoch 8, batch 9200, loss[loss=0.1538, simple_loss=0.2376, pruned_loss=0.03499, over 12475.00 frames. ], tot_loss[loss=0.1891, simple_loss=0.2786, pruned_loss=0.04981, over 3053788.19 frames. ], batch size: 247, lr: 8.34e-03, grad_scale: 8.0 2023-04-28 22:59:42,527 INFO [zipformer.py:625] (1/8) 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] (1/8) 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] (1/8) Epoch 8, batch 9250, loss[loss=0.1978, simple_loss=0.2842, pruned_loss=0.05567, over 16972.00 frames. ], tot_loss[loss=0.1897, simple_loss=0.279, pruned_loss=0.05023, over 3058447.65 frames. ], batch size: 109, lr: 8.34e-03, grad_scale: 4.0 2023-04-28 23:00:42,245 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.2738, 3.2367, 3.3204, 1.7045, 3.5879, 3.6140, 2.9170, 2.7928], device='cuda:1'), covar=tensor([0.0720, 0.0174, 0.0149, 0.1140, 0.0048, 0.0083, 0.0357, 0.0391], device='cuda:1'), in_proj_covar=tensor([0.0135, 0.0093, 0.0078, 0.0134, 0.0063, 0.0084, 0.0114, 0.0122], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003, 0.0003], device='cuda:1') 2023-04-28 23:00:42,872 INFO [optim.py:368] (1/8) 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 23:00:44,587 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=80313.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 23:00:45,124 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.75 vs. limit=2.0 2023-04-28 23:01:55,873 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 8, batch 9300, loss[loss=0.1756, simple_loss=0.2615, pruned_loss=0.04489, over 16805.00 frames. ], tot_loss[loss=0.1874, simple_loss=0.2767, pruned_loss=0.04908, over 3065460.02 frames. ], batch size: 124, lr: 8.33e-03, grad_scale: 4.0 2023-04-28 23:02:31,762 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=80361.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 23:03:42,284 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 8, batch 9350, loss[loss=0.1773, simple_loss=0.2617, pruned_loss=0.0465, over 12058.00 frames. ], tot_loss[loss=0.1872, simple_loss=0.2762, pruned_loss=0.04911, over 3043668.57 frames. ], batch size: 248, lr: 8.33e-03, grad_scale: 4.0 2023-04-28 23:04:22,275 INFO [optim.py:368] (1/8) 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 23:04:26,536 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=3.42 vs. limit=5.0 2023-04-28 23:04:34,173 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=80418.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 23:04:34,274 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.1580, 1.4787, 1.8573, 2.0637, 2.1372, 2.1881, 1.6512, 2.1606], device='cuda:1'), covar=tensor([0.0126, 0.0274, 0.0163, 0.0183, 0.0165, 0.0124, 0.0289, 0.0078], device='cuda:1'), in_proj_covar=tensor([0.0137, 0.0155, 0.0140, 0.0136, 0.0144, 0.0102, 0.0153, 0.0093], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001], device='cuda:1') 2023-04-28 23:05:40,898 INFO [train.py:904] (1/8) Epoch 8, batch 9400, loss[loss=0.2046, simple_loss=0.2959, pruned_loss=0.05667, over 16264.00 frames. ], tot_loss[loss=0.1871, simple_loss=0.2767, pruned_loss=0.04879, over 3061514.09 frames. ], batch size: 146, lr: 8.33e-03, grad_scale: 4.0 2023-04-28 23:05:46,192 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=80454.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 23:06:05,150 INFO [zipformer.py:625] (1/8) 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] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=80466.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 23:07:07,320 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=80494.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 23:07:19,863 INFO [train.py:904] (1/8) Epoch 8, batch 9450, loss[loss=0.2059, simple_loss=0.2928, pruned_loss=0.0595, over 12566.00 frames. ], tot_loss[loss=0.1882, simple_loss=0.2781, pruned_loss=0.0492, over 3045143.60 frames. ], batch size: 248, lr: 8.33e-03, grad_scale: 4.0 2023-04-28 23:07:38,811 INFO [optim.py:368] (1/8) 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 23:08:04,393 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=80524.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 23:08:39,807 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.6897, 3.8059, 2.9715, 2.1854, 2.6208, 2.2409, 4.0138, 3.5809], device='cuda:1'), covar=tensor([0.2360, 0.0623, 0.1355, 0.2120, 0.2061, 0.1676, 0.0392, 0.0780], device='cuda:1'), in_proj_covar=tensor([0.0280, 0.0242, 0.0265, 0.0254, 0.0245, 0.0204, 0.0246, 0.0258], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-28 23:08:58,868 INFO [train.py:904] (1/8) Epoch 8, batch 9500, loss[loss=0.1845, simple_loss=0.2857, pruned_loss=0.0417, over 16162.00 frames. ], tot_loss[loss=0.1874, simple_loss=0.2772, pruned_loss=0.04877, over 3049560.83 frames. ], batch size: 165, lr: 8.32e-03, grad_scale: 4.0 2023-04-28 23:09:08,310 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=80555.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 23:09:11,249 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.94 vs. limit=2.0 2023-04-28 23:10:46,572 INFO [train.py:904] (1/8) Epoch 8, batch 9550, loss[loss=0.1831, simple_loss=0.2748, pruned_loss=0.0457, over 16439.00 frames. ], tot_loss[loss=0.1865, simple_loss=0.2761, pruned_loss=0.04848, over 3043016.33 frames. ], batch size: 68, lr: 8.32e-03, grad_scale: 4.0 2023-04-28 23:11:10,119 INFO [optim.py:368] (1/8) 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 23:12:03,959 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 8, batch 9600, loss[loss=0.1789, simple_loss=0.2742, pruned_loss=0.04178, over 16538.00 frames. ], tot_loss[loss=0.1879, simple_loss=0.2773, pruned_loss=0.04926, over 3039283.30 frames. ], batch size: 68, lr: 8.32e-03, grad_scale: 8.0 2023-04-28 23:14:15,050 INFO [train.py:904] (1/8) Epoch 8, batch 9650, loss[loss=0.1741, simple_loss=0.2631, pruned_loss=0.04253, over 17159.00 frames. ], tot_loss[loss=0.1893, simple_loss=0.2792, pruned_loss=0.04964, over 3035913.96 frames. ], batch size: 46, lr: 8.31e-03, grad_scale: 8.0 2023-04-28 23:14:42,862 INFO [optim.py:368] (1/8) 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 23:15:29,955 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=80736.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 23:15:59,210 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=80749.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 23:15:59,466 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.1679, 1.9562, 2.1094, 3.5781, 1.8813, 2.3532, 2.0917, 2.1033], device='cuda:1'), covar=tensor([0.0776, 0.3171, 0.1950, 0.0388, 0.3845, 0.2016, 0.2737, 0.2937], device='cuda:1'), in_proj_covar=tensor([0.0333, 0.0355, 0.0305, 0.0310, 0.0391, 0.0389, 0.0317, 0.0415], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-28 23:16:03,265 INFO [train.py:904] (1/8) Epoch 8, batch 9700, loss[loss=0.2058, simple_loss=0.2929, pruned_loss=0.05939, over 16802.00 frames. ], tot_loss[loss=0.1887, simple_loss=0.2786, pruned_loss=0.0494, over 3042712.97 frames. ], batch size: 124, lr: 8.31e-03, grad_scale: 8.0 2023-04-28 23:16:17,855 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.2845, 1.9518, 2.0802, 3.7311, 1.8905, 2.3157, 2.0804, 2.0884], device='cuda:1'), covar=tensor([0.0744, 0.2984, 0.1980, 0.0343, 0.3694, 0.1954, 0.2821, 0.3114], device='cuda:1'), in_proj_covar=tensor([0.0332, 0.0354, 0.0304, 0.0309, 0.0389, 0.0388, 0.0316, 0.0414], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-28 23:16:46,551 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=80773.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 23:17:38,610 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=80797.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 23:17:46,349 INFO [train.py:904] (1/8) Epoch 8, batch 9750, loss[loss=0.1703, simple_loss=0.2592, pruned_loss=0.04075, over 16560.00 frames. ], tot_loss[loss=0.188, simple_loss=0.2773, pruned_loss=0.04933, over 3058597.81 frames. ], batch size: 68, lr: 8.31e-03, grad_scale: 8.0 2023-04-28 23:18:08,278 INFO [optim.py:368] (1/8) 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 23:18:20,464 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=80819.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 23:18:55,686 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=80834.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 23:18:57,932 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.56 vs. limit=2.0 2023-04-28 23:19:24,253 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=80850.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 23:19:26,321 INFO [train.py:904] (1/8) Epoch 8, batch 9800, loss[loss=0.1827, simple_loss=0.2828, pruned_loss=0.04137, over 17280.00 frames. ], tot_loss[loss=0.1873, simple_loss=0.2775, pruned_loss=0.04855, over 3054272.66 frames. ], batch size: 52, lr: 8.31e-03, grad_scale: 8.0 2023-04-28 23:19:37,857 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.3699, 3.2632, 2.6132, 2.0955, 2.2561, 2.0854, 3.2968, 3.0884], device='cuda:1'), covar=tensor([0.2464, 0.0688, 0.1388, 0.2034, 0.1945, 0.1629, 0.0439, 0.0870], device='cuda:1'), in_proj_covar=tensor([0.0279, 0.0239, 0.0265, 0.0252, 0.0242, 0.0202, 0.0243, 0.0254], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-28 23:19:47,604 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.7657, 4.1739, 3.1434, 2.3727, 2.9480, 2.4687, 4.3655, 3.7810], device='cuda:1'), covar=tensor([0.2303, 0.0479, 0.1258, 0.1891, 0.1944, 0.1496, 0.0288, 0.0706], device='cuda:1'), in_proj_covar=tensor([0.0280, 0.0239, 0.0265, 0.0252, 0.0243, 0.0202, 0.0243, 0.0255], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-28 23:21:11,977 INFO [train.py:904] (1/8) Epoch 8, batch 9850, loss[loss=0.1656, simple_loss=0.2603, pruned_loss=0.03547, over 16428.00 frames. ], tot_loss[loss=0.1876, simple_loss=0.279, pruned_loss=0.04811, over 3063567.89 frames. ], batch size: 68, lr: 8.30e-03, grad_scale: 8.0 2023-04-28 23:21:33,319 INFO [optim.py:368] (1/8) 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 23:22:36,942 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=80939.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 23:22:55,344 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.7624, 3.6952, 3.8252, 3.9296, 4.0235, 3.6607, 3.9860, 4.0659], device='cuda:1'), covar=tensor([0.1171, 0.0784, 0.1137, 0.0610, 0.0510, 0.1537, 0.0693, 0.0531], device='cuda:1'), in_proj_covar=tensor([0.0439, 0.0534, 0.0650, 0.0545, 0.0410, 0.0405, 0.0429, 0.0472], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:1') 2023-04-28 23:23:04,049 INFO [train.py:904] (1/8) Epoch 8, batch 9900, loss[loss=0.2052, simple_loss=0.2814, pruned_loss=0.0645, over 12690.00 frames. ], tot_loss[loss=0.1873, simple_loss=0.2788, pruned_loss=0.04788, over 3054775.07 frames. ], batch size: 247, lr: 8.30e-03, grad_scale: 8.0 2023-04-28 23:24:17,699 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.5131, 3.5139, 3.4843, 3.0954, 3.4568, 1.9014, 3.2789, 3.0139], device='cuda:1'), covar=tensor([0.0094, 0.0087, 0.0110, 0.0182, 0.0073, 0.1908, 0.0092, 0.0166], device='cuda:1'), in_proj_covar=tensor([0.0106, 0.0095, 0.0137, 0.0127, 0.0110, 0.0163, 0.0123, 0.0129], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0001, 0.0001, 0.0002, 0.0001, 0.0002], device='cuda:1') 2023-04-28 23:24:29,644 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=80987.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 23:24:54,501 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([5.3111, 5.6783, 5.4185, 5.4411, 5.0711, 4.9628, 5.1011, 5.6958], device='cuda:1'), covar=tensor([0.0933, 0.0762, 0.0821, 0.0459, 0.0711, 0.0602, 0.0840, 0.0834], device='cuda:1'), in_proj_covar=tensor([0.0453, 0.0576, 0.0472, 0.0392, 0.0354, 0.0382, 0.0474, 0.0429], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:1') 2023-04-28 23:25:03,336 INFO [train.py:904] (1/8) Epoch 8, batch 9950, loss[loss=0.1902, simple_loss=0.2806, pruned_loss=0.04989, over 12658.00 frames. ], tot_loss[loss=0.1888, simple_loss=0.281, pruned_loss=0.04831, over 3053818.37 frames. ], batch size: 248, lr: 8.30e-03, grad_scale: 8.0 2023-04-28 23:25:29,446 INFO [optim.py:368] (1/8) 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 23:26:37,726 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.7444, 4.1951, 4.2192, 3.0571, 3.9087, 4.2391, 3.9335, 2.5598], device='cuda:1'), covar=tensor([0.0321, 0.0019, 0.0019, 0.0214, 0.0039, 0.0038, 0.0025, 0.0287], device='cuda:1'), in_proj_covar=tensor([0.0120, 0.0059, 0.0062, 0.0116, 0.0067, 0.0075, 0.0066, 0.0112], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:1') 2023-04-28 23:26:48,504 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=3.28 vs. limit=5.0 2023-04-28 23:27:01,907 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 8, batch 10000, loss[loss=0.1987, simple_loss=0.2915, pruned_loss=0.05293, over 16334.00 frames. ], tot_loss[loss=0.1873, simple_loss=0.2794, pruned_loss=0.04765, over 3072472.03 frames. ], batch size: 146, lr: 8.30e-03, grad_scale: 8.0 2023-04-28 23:28:30,716 INFO [zipformer.py:625] (1/8) 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] (1/8) 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] (1/8) Epoch 8, batch 10050, loss[loss=0.1983, simple_loss=0.2958, pruned_loss=0.05042, over 16428.00 frames. ], tot_loss[loss=0.1875, simple_loss=0.2798, pruned_loss=0.04764, over 3069871.11 frames. ], batch size: 146, lr: 8.29e-03, grad_scale: 8.0 2023-04-28 23:29:08,275 INFO [optim.py:368] (1/8) Clipping_scale=2.0, grad-norm quartiles 1.565e+02 2.518e+02 2.989e+02 3.596e+02 8.754e+02, threshold=5.978e+02, percent-clipped=2.0 2023-04-28 23:29:15,158 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=81116.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 23:29:20,574 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=81119.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 23:29:39,811 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=81129.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 23:30:18,754 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 8, batch 10100, loss[loss=0.1694, simple_loss=0.255, pruned_loss=0.04191, over 12597.00 frames. ], tot_loss[loss=0.1879, simple_loss=0.2802, pruned_loss=0.0478, over 3064714.66 frames. ], batch size: 249, lr: 8.29e-03, grad_scale: 8.0 2023-04-28 23:30:51,850 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=81167.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 23:31:14,357 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=81177.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 23:31:37,456 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=81198.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 23:32:08,878 INFO [train.py:904] (1/8) Epoch 9, batch 0, loss[loss=0.3354, simple_loss=0.3625, pruned_loss=0.1542, over 16715.00 frames. ], tot_loss[loss=0.3354, simple_loss=0.3625, pruned_loss=0.1542, over 16715.00 frames. ], batch size: 83, lr: 7.85e-03, grad_scale: 8.0 2023-04-28 23:32:08,878 INFO [train.py:929] (1/8) Computing validation loss 2023-04-28 23:32:16,262 INFO [train.py:938] (1/8) Epoch 9, validation: loss=0.1602, simple_loss=0.2637, pruned_loss=0.02837, over 944034.00 frames. 2023-04-28 23:32:16,262 INFO [train.py:939] (1/8) Maximum memory allocated so far is 17676MB 2023-04-28 23:32:33,683 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.9381, 4.8855, 4.7104, 4.4371, 4.1440, 4.8905, 4.9971, 4.4341], device='cuda:1'), covar=tensor([0.0696, 0.0536, 0.0417, 0.0340, 0.1334, 0.0443, 0.0307, 0.0667], device='cuda:1'), in_proj_covar=tensor([0.0200, 0.0234, 0.0234, 0.0207, 0.0252, 0.0237, 0.0161, 0.0265], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002], device='cuda:1') 2023-04-28 23:32:36,783 INFO [optim.py:368] (1/8) 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 23:33:25,079 INFO [train.py:904] (1/8) Epoch 9, batch 50, loss[loss=0.2124, simple_loss=0.2819, pruned_loss=0.07143, over 16474.00 frames. ], tot_loss[loss=0.2188, simple_loss=0.2948, pruned_loss=0.07138, over 754047.84 frames. ], batch size: 146, lr: 7.85e-03, grad_scale: 1.0 2023-04-28 23:34:31,182 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=81299.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 23:34:34,094 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 9, batch 100, loss[loss=0.1711, simple_loss=0.263, pruned_loss=0.03958, over 17130.00 frames. ], tot_loss[loss=0.2134, simple_loss=0.2905, pruned_loss=0.06814, over 1311281.28 frames. ], batch size: 48, lr: 7.84e-03, grad_scale: 1.0 2023-04-28 23:34:47,348 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=4.96 vs. limit=5.0 2023-04-28 23:34:54,536 INFO [optim.py:368] (1/8) Clipping_scale=2.0, grad-norm quartiles 1.584e+02 2.676e+02 3.088e+02 4.301e+02 1.054e+03, threshold=6.177e+02, percent-clipped=2.0 2023-04-28 23:35:42,951 INFO [train.py:904] (1/8) Epoch 9, batch 150, loss[loss=0.2203, simple_loss=0.3064, pruned_loss=0.06711, over 17079.00 frames. ], tot_loss[loss=0.2088, simple_loss=0.2869, pruned_loss=0.06531, over 1757786.01 frames. ], batch size: 53, lr: 7.84e-03, grad_scale: 1.0 2023-04-28 23:35:55,887 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=81360.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 23:35:58,257 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=81362.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 23:36:40,007 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=81392.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 23:36:54,427 INFO [train.py:904] (1/8) Epoch 9, batch 200, loss[loss=0.1618, simple_loss=0.2456, pruned_loss=0.03905, over 16809.00 frames. ], tot_loss[loss=0.2084, simple_loss=0.2865, pruned_loss=0.06516, over 2104185.69 frames. ], batch size: 39, lr: 7.84e-03, grad_scale: 1.0 2023-04-28 23:37:13,043 INFO [optim.py:368] (1/8) 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 23:37:31,245 INFO [zipformer.py:625] (1/8) 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] (1/8) 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] (1/8) Epoch 9, batch 250, loss[loss=0.1753, simple_loss=0.2671, pruned_loss=0.04176, over 17155.00 frames. ], tot_loss[loss=0.2083, simple_loss=0.2854, pruned_loss=0.06567, over 2371953.24 frames. ], batch size: 49, lr: 7.84e-03, grad_scale: 1.0 2023-04-28 23:38:29,683 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=81472.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 23:38:36,876 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=81477.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 23:38:47,372 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.44 vs. limit=2.0 2023-04-28 23:39:10,399 INFO [train.py:904] (1/8) Epoch 9, batch 300, loss[loss=0.2248, simple_loss=0.3024, pruned_loss=0.07361, over 16640.00 frames. ], tot_loss[loss=0.2036, simple_loss=0.282, pruned_loss=0.06262, over 2592546.07 frames. ], batch size: 57, lr: 7.83e-03, grad_scale: 1.0 2023-04-28 23:39:29,675 INFO [optim.py:368] (1/8) Clipping_scale=2.0, grad-norm quartiles 1.655e+02 2.475e+02 2.919e+02 3.755e+02 7.155e+02, threshold=5.837e+02, percent-clipped=3.0 2023-04-28 23:39:32,606 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.4233, 4.1844, 4.4228, 4.6547, 4.7950, 4.3295, 4.6174, 4.7475], device='cuda:1'), covar=tensor([0.1284, 0.1122, 0.1420, 0.0643, 0.0520, 0.0927, 0.1681, 0.0531], device='cuda:1'), in_proj_covar=tensor([0.0488, 0.0598, 0.0738, 0.0607, 0.0454, 0.0450, 0.0479, 0.0525], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-28 23:39:32,752 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.7793, 3.8389, 2.9643, 2.3831, 2.7071, 2.3903, 3.9619, 3.6386], device='cuda:1'), covar=tensor([0.2159, 0.0639, 0.1343, 0.2124, 0.2136, 0.1626, 0.0427, 0.0983], device='cuda:1'), in_proj_covar=tensor([0.0293, 0.0253, 0.0278, 0.0266, 0.0264, 0.0215, 0.0257, 0.0276], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-28 23:40:17,832 INFO [train.py:904] (1/8) Epoch 9, batch 350, loss[loss=0.218, simple_loss=0.2785, pruned_loss=0.07868, over 16865.00 frames. ], tot_loss[loss=0.1999, simple_loss=0.2784, pruned_loss=0.06066, over 2751358.98 frames. ], batch size: 116, lr: 7.83e-03, grad_scale: 1.0 2023-04-28 23:40:44,001 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=5.09 vs. limit=5.0 2023-04-28 23:40:54,405 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.0255, 4.3064, 4.5067, 3.4203, 4.0270, 4.4063, 3.9762, 2.9510], device='cuda:1'), covar=tensor([0.0307, 0.0029, 0.0025, 0.0210, 0.0053, 0.0053, 0.0047, 0.0289], device='cuda:1'), in_proj_covar=tensor([0.0122, 0.0065, 0.0065, 0.0120, 0.0069, 0.0079, 0.0070, 0.0115], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:1') 2023-04-28 23:41:23,824 INFO [train.py:904] (1/8) Epoch 9, batch 400, loss[loss=0.1853, simple_loss=0.2763, pruned_loss=0.0471, over 17121.00 frames. ], tot_loss[loss=0.1988, simple_loss=0.2771, pruned_loss=0.06019, over 2888461.56 frames. ], batch size: 47, lr: 7.83e-03, grad_scale: 2.0 2023-04-28 23:41:43,586 INFO [optim.py:368] (1/8) Clipping_scale=2.0, grad-norm quartiles 1.573e+02 2.386e+02 2.843e+02 3.463e+02 6.249e+02, threshold=5.687e+02, percent-clipped=1.0 2023-04-28 23:42:33,320 INFO [train.py:904] (1/8) Epoch 9, batch 450, loss[loss=0.2016, simple_loss=0.2663, pruned_loss=0.06851, over 16855.00 frames. ], tot_loss[loss=0.1956, simple_loss=0.2744, pruned_loss=0.05842, over 2976194.29 frames. ], batch size: 96, lr: 7.83e-03, grad_scale: 2.0 2023-04-28 23:42:36,515 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=4.34 vs. limit=5.0 2023-04-28 23:42:37,319 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=81655.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 23:42:40,767 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=81657.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 23:42:42,225 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.6297, 2.7028, 2.1565, 2.5622, 3.0392, 2.8212, 3.5972, 3.2377], device='cuda:1'), covar=tensor([0.0055, 0.0286, 0.0339, 0.0287, 0.0166, 0.0240, 0.0123, 0.0154], device='cuda:1'), in_proj_covar=tensor([0.0118, 0.0190, 0.0185, 0.0186, 0.0185, 0.0189, 0.0186, 0.0173], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-28 23:42:58,086 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=81670.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 23:43:42,341 INFO [train.py:904] (1/8) Epoch 9, batch 500, loss[loss=0.1713, simple_loss=0.248, pruned_loss=0.04728, over 16320.00 frames. ], tot_loss[loss=0.1937, simple_loss=0.2729, pruned_loss=0.05726, over 3042474.12 frames. ], batch size: 36, lr: 7.82e-03, grad_scale: 2.0 2023-04-28 23:43:42,880 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.4698, 2.5888, 2.0625, 2.2691, 2.9546, 2.6801, 3.4043, 3.1844], device='cuda:1'), covar=tensor([0.0061, 0.0230, 0.0305, 0.0291, 0.0125, 0.0210, 0.0117, 0.0141], device='cuda:1'), in_proj_covar=tensor([0.0118, 0.0190, 0.0186, 0.0187, 0.0185, 0.0189, 0.0186, 0.0173], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-28 23:44:01,900 INFO [optim.py:368] (1/8) 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 23:44:22,620 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=81731.0, num_to_drop=1, layers_to_drop={1} 2023-04-28 23:44:40,813 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([5.4402, 5.3225, 5.1618, 4.7924, 4.7685, 5.2354, 5.2662, 4.8799], device='cuda:1'), covar=tensor([0.0472, 0.0382, 0.0233, 0.0237, 0.1047, 0.0363, 0.0172, 0.0646], device='cuda:1'), in_proj_covar=tensor([0.0228, 0.0267, 0.0263, 0.0235, 0.0290, 0.0271, 0.0181, 0.0304], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-28 23:44:47,756 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.31 vs. limit=2.0 2023-04-28 23:44:50,828 INFO [train.py:904] (1/8) Epoch 9, batch 550, loss[loss=0.194, simple_loss=0.2763, pruned_loss=0.05583, over 17273.00 frames. ], tot_loss[loss=0.1928, simple_loss=0.2719, pruned_loss=0.05689, over 3101014.99 frames. ], batch size: 52, lr: 7.82e-03, grad_scale: 2.0 2023-04-28 23:44:57,702 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([5.1392, 5.6006, 5.6705, 5.5473, 5.6299, 6.1188, 5.7416, 5.4994], device='cuda:1'), covar=tensor([0.0745, 0.1843, 0.2291, 0.2232, 0.3095, 0.1127, 0.1403, 0.2351], device='cuda:1'), in_proj_covar=tensor([0.0329, 0.0463, 0.0495, 0.0411, 0.0544, 0.0518, 0.0392, 0.0547], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-04-28 23:45:19,899 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=81772.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 23:45:26,176 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.7510, 2.6827, 2.1941, 2.5344, 3.0557, 2.9278, 3.5621, 3.2930], device='cuda:1'), covar=tensor([0.0060, 0.0262, 0.0355, 0.0281, 0.0168, 0.0231, 0.0157, 0.0166], device='cuda:1'), in_proj_covar=tensor([0.0120, 0.0192, 0.0188, 0.0188, 0.0187, 0.0191, 0.0189, 0.0175], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-28 23:45:59,672 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.18 vs. limit=2.0 2023-04-28 23:46:02,609 INFO [train.py:904] (1/8) Epoch 9, batch 600, loss[loss=0.2124, simple_loss=0.274, pruned_loss=0.07541, over 16827.00 frames. ], tot_loss[loss=0.1927, simple_loss=0.2712, pruned_loss=0.05705, over 3152021.75 frames. ], batch size: 96, lr: 7.82e-03, grad_scale: 2.0 2023-04-28 23:46:14,051 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=4.80 vs. limit=5.0 2023-04-28 23:46:17,722 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.0037, 2.6254, 2.7149, 1.8569, 2.8590, 2.8577, 2.4503, 2.4011], device='cuda:1'), covar=tensor([0.0762, 0.0199, 0.0207, 0.0964, 0.0087, 0.0178, 0.0426, 0.0448], device='cuda:1'), in_proj_covar=tensor([0.0144, 0.0097, 0.0084, 0.0145, 0.0069, 0.0094, 0.0121, 0.0130], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0003, 0.0002, 0.0004, 0.0002, 0.0002, 0.0003, 0.0003], device='cuda:1') 2023-04-28 23:46:21,575 INFO [optim.py:368] (1/8) 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 23:46:25,687 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.9181, 2.0744, 2.3464, 3.2018, 2.1344, 2.2422, 2.3016, 2.1535], device='cuda:1'), covar=tensor([0.0899, 0.2610, 0.1651, 0.0558, 0.3126, 0.1885, 0.2366, 0.2978], device='cuda:1'), in_proj_covar=tensor([0.0348, 0.0370, 0.0312, 0.0323, 0.0399, 0.0408, 0.0330, 0.0434], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-28 23:46:26,574 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 9, batch 650, loss[loss=0.1964, simple_loss=0.264, pruned_loss=0.06435, over 16813.00 frames. ], tot_loss[loss=0.1909, simple_loss=0.2697, pruned_loss=0.05606, over 3182331.48 frames. ], batch size: 96, lr: 7.82e-03, grad_scale: 2.0 2023-04-28 23:47:55,512 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.2207, 2.2690, 1.6322, 1.9728, 2.6228, 2.4498, 2.6982, 2.7682], device='cuda:1'), covar=tensor([0.0116, 0.0216, 0.0320, 0.0253, 0.0117, 0.0182, 0.0144, 0.0132], device='cuda:1'), in_proj_covar=tensor([0.0121, 0.0192, 0.0187, 0.0188, 0.0187, 0.0191, 0.0190, 0.0175], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-28 23:48:18,160 INFO [train.py:904] (1/8) Epoch 9, batch 700, loss[loss=0.1862, simple_loss=0.2716, pruned_loss=0.05042, over 16483.00 frames. ], tot_loss[loss=0.1908, simple_loss=0.2696, pruned_loss=0.05605, over 3223871.51 frames. ], batch size: 68, lr: 7.81e-03, grad_scale: 2.0 2023-04-28 23:48:37,201 INFO [optim.py:368] (1/8) 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] (1/8) Epoch 9, batch 750, loss[loss=0.1682, simple_loss=0.2671, pruned_loss=0.03468, over 17264.00 frames. ], tot_loss[loss=0.1908, simple_loss=0.2703, pruned_loss=0.05568, over 3249340.85 frames. ], batch size: 52, lr: 7.81e-03, grad_scale: 2.0 2023-04-28 23:49:29,053 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=81955.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 23:49:31,563 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=81957.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 23:49:36,441 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.8477, 2.9298, 2.6323, 2.6901, 3.2507, 3.0592, 3.6074, 3.3819], device='cuda:1'), covar=tensor([0.0049, 0.0218, 0.0265, 0.0243, 0.0156, 0.0195, 0.0168, 0.0142], device='cuda:1'), in_proj_covar=tensor([0.0121, 0.0192, 0.0187, 0.0189, 0.0187, 0.0191, 0.0190, 0.0175], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-28 23:50:05,419 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.2015, 2.0668, 2.4184, 2.9302, 2.8048, 3.4819, 2.3156, 3.2338], device='cuda:1'), covar=tensor([0.0122, 0.0297, 0.0225, 0.0186, 0.0192, 0.0106, 0.0263, 0.0101], device='cuda:1'), in_proj_covar=tensor([0.0149, 0.0164, 0.0150, 0.0148, 0.0156, 0.0112, 0.0163, 0.0102], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001], device='cuda:1') 2023-04-28 23:50:14,458 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=81987.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 23:50:31,622 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=2.05 vs. limit=2.0 2023-04-28 23:50:37,834 INFO [train.py:904] (1/8) Epoch 9, batch 800, loss[loss=0.1802, simple_loss=0.2698, pruned_loss=0.04533, over 17109.00 frames. ], tot_loss[loss=0.1905, simple_loss=0.27, pruned_loss=0.05547, over 3262593.96 frames. ], batch size: 47, lr: 7.81e-03, grad_scale: 4.0 2023-04-28 23:50:39,914 INFO [zipformer.py:625] (1/8) 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] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=82005.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 23:50:56,928 INFO [optim.py:368] (1/8) 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 23:51:10,863 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=82026.0, num_to_drop=1, layers_to_drop={1} 2023-04-28 23:51:34,753 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([5.2463, 5.5635, 5.3212, 5.3512, 5.0331, 4.8718, 5.0360, 5.6660], device='cuda:1'), covar=tensor([0.0953, 0.0883, 0.0970, 0.0570, 0.0651, 0.0773, 0.0801, 0.0792], device='cuda:1'), in_proj_covar=tensor([0.0495, 0.0638, 0.0533, 0.0435, 0.0393, 0.0410, 0.0527, 0.0468], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-28 23:51:40,206 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 9, batch 850, loss[loss=0.2056, simple_loss=0.28, pruned_loss=0.06567, over 15505.00 frames. ], tot_loss[loss=0.1896, simple_loss=0.2689, pruned_loss=0.05512, over 3280927.77 frames. ], batch size: 190, lr: 7.81e-03, grad_scale: 4.0 2023-04-28 23:51:49,583 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=82055.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 23:52:39,633 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=2.55 vs. limit=5.0 2023-04-28 23:52:54,647 INFO [train.py:904] (1/8) Epoch 9, batch 900, loss[loss=0.1915, simple_loss=0.2679, pruned_loss=0.0576, over 16843.00 frames. ], tot_loss[loss=0.1882, simple_loss=0.2677, pruned_loss=0.05438, over 3286232.91 frames. ], batch size: 90, lr: 7.80e-03, grad_scale: 4.0 2023-04-28 23:53:13,854 INFO [optim.py:368] (1/8) 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 23:53:14,409 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=82116.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 23:53:29,475 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.7908, 2.7629, 2.3162, 2.5891, 3.1204, 2.8893, 3.7055, 3.3374], device='cuda:1'), covar=tensor([0.0044, 0.0255, 0.0313, 0.0263, 0.0157, 0.0230, 0.0132, 0.0147], device='cuda:1'), in_proj_covar=tensor([0.0125, 0.0196, 0.0190, 0.0191, 0.0190, 0.0193, 0.0194, 0.0179], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-28 23:54:03,815 INFO [train.py:904] (1/8) Epoch 9, batch 950, loss[loss=0.2046, simple_loss=0.2754, pruned_loss=0.06697, over 16838.00 frames. ], tot_loss[loss=0.1886, simple_loss=0.2679, pruned_loss=0.05462, over 3290335.79 frames. ], batch size: 96, lr: 7.80e-03, grad_scale: 4.0 2023-04-28 23:55:04,941 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.7431, 4.7193, 4.6377, 4.1349, 4.6920, 2.0428, 4.3912, 4.4618], device='cuda:1'), covar=tensor([0.0085, 0.0084, 0.0137, 0.0301, 0.0075, 0.1984, 0.0137, 0.0144], device='cuda:1'), in_proj_covar=tensor([0.0120, 0.0108, 0.0156, 0.0149, 0.0124, 0.0175, 0.0141, 0.0147], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-28 23:55:11,099 INFO [train.py:904] (1/8) Epoch 9, batch 1000, loss[loss=0.1941, simple_loss=0.2588, pruned_loss=0.06468, over 16558.00 frames. ], tot_loss[loss=0.189, simple_loss=0.2674, pruned_loss=0.0553, over 3285126.78 frames. ], batch size: 75, lr: 7.80e-03, grad_scale: 4.0 2023-04-28 23:55:31,950 INFO [optim.py:368] (1/8) 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] (1/8) Epoch 9, batch 1050, loss[loss=0.2025, simple_loss=0.27, pruned_loss=0.06747, over 16789.00 frames. ], tot_loss[loss=0.1889, simple_loss=0.2679, pruned_loss=0.05496, over 3300826.41 frames. ], batch size: 124, lr: 7.80e-03, grad_scale: 4.0 2023-04-28 23:56:25,776 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.7927, 2.7286, 2.3161, 2.4631, 2.9678, 2.8116, 3.7474, 3.3577], device='cuda:1'), covar=tensor([0.0047, 0.0242, 0.0300, 0.0260, 0.0173, 0.0226, 0.0131, 0.0137], device='cuda:1'), in_proj_covar=tensor([0.0125, 0.0194, 0.0189, 0.0189, 0.0189, 0.0192, 0.0194, 0.0178], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-28 23:57:28,537 INFO [train.py:904] (1/8) Epoch 9, batch 1100, loss[loss=0.2028, simple_loss=0.2806, pruned_loss=0.06252, over 16586.00 frames. ], tot_loss[loss=0.1885, simple_loss=0.2674, pruned_loss=0.05481, over 3308530.22 frames. ], batch size: 68, lr: 7.80e-03, grad_scale: 4.0 2023-04-28 23:57:31,702 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.32 vs. limit=2.0 2023-04-28 23:57:47,216 INFO [optim.py:368] (1/8) Clipping_scale=2.0, grad-norm quartiles 1.669e+02 2.514e+02 3.056e+02 3.616e+02 1.290e+03, threshold=6.113e+02, percent-clipped=7.0 2023-04-28 23:58:01,829 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=82326.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 23:58:24,251 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=82343.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 23:58:27,998 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=82346.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 23:58:35,309 INFO [train.py:904] (1/8) Epoch 9, batch 1150, loss[loss=0.1729, simple_loss=0.2445, pruned_loss=0.0506, over 12099.00 frames. ], tot_loss[loss=0.1872, simple_loss=0.2664, pruned_loss=0.054, over 3297116.93 frames. ], batch size: 248, lr: 7.79e-03, grad_scale: 4.0 2023-04-28 23:58:59,582 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.7364, 4.0815, 4.2323, 2.9333, 3.5848, 3.9891, 3.6583, 2.1051], device='cuda:1'), covar=tensor([0.0378, 0.0051, 0.0044, 0.0303, 0.0095, 0.0150, 0.0143, 0.0457], device='cuda:1'), in_proj_covar=tensor([0.0121, 0.0066, 0.0065, 0.0120, 0.0070, 0.0080, 0.0071, 0.0113], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:1') 2023-04-28 23:59:04,894 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=82374.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 23:59:09,218 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.0459, 2.4914, 1.9468, 2.1627, 2.9522, 2.6815, 3.1976, 3.0499], device='cuda:1'), covar=tensor([0.0132, 0.0264, 0.0343, 0.0327, 0.0161, 0.0235, 0.0177, 0.0165], device='cuda:1'), in_proj_covar=tensor([0.0126, 0.0195, 0.0189, 0.0189, 0.0189, 0.0192, 0.0194, 0.0178], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-28 23:59:44,517 INFO [train.py:904] (1/8) Epoch 9, batch 1200, loss[loss=0.172, simple_loss=0.2479, pruned_loss=0.04804, over 16799.00 frames. ], tot_loss[loss=0.1861, simple_loss=0.2657, pruned_loss=0.05325, over 3303577.16 frames. ], batch size: 39, lr: 7.79e-03, grad_scale: 8.0 2023-04-28 23:59:50,741 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=82407.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 23:59:55,202 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=82411.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 00:00:02,690 INFO [optim.py:368] (1/8) 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 00:00:50,234 INFO [train.py:904] (1/8) Epoch 9, batch 1250, loss[loss=0.1711, simple_loss=0.2534, pruned_loss=0.04437, over 17003.00 frames. ], tot_loss[loss=0.1868, simple_loss=0.2655, pruned_loss=0.05401, over 3299998.06 frames. ], batch size: 41, lr: 7.79e-03, grad_scale: 4.0 2023-04-29 00:01:46,998 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=82493.0, num_to_drop=1, layers_to_drop={1} 2023-04-29 00:01:58,480 INFO [train.py:904] (1/8) Epoch 9, batch 1300, loss[loss=0.1849, simple_loss=0.2703, pruned_loss=0.04974, over 16485.00 frames. ], tot_loss[loss=0.1867, simple_loss=0.2656, pruned_loss=0.05389, over 3300752.41 frames. ], batch size: 68, lr: 7.79e-03, grad_scale: 4.0 2023-04-29 00:02:18,040 INFO [optim.py:368] (1/8) 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 00:03:05,221 INFO [train.py:904] (1/8) Epoch 9, batch 1350, loss[loss=0.1663, simple_loss=0.253, pruned_loss=0.03982, over 17220.00 frames. ], tot_loss[loss=0.1864, simple_loss=0.2657, pruned_loss=0.05351, over 3314304.80 frames. ], batch size: 46, lr: 7.78e-03, grad_scale: 4.0 2023-04-29 00:03:08,033 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=82554.0, num_to_drop=1, layers_to_drop={2} 2023-04-29 00:03:29,800 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=82569.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 00:04:12,577 INFO [train.py:904] (1/8) Epoch 9, batch 1400, loss[loss=0.1823, simple_loss=0.274, pruned_loss=0.04529, over 17112.00 frames. ], tot_loss[loss=0.1861, simple_loss=0.2663, pruned_loss=0.05298, over 3318634.42 frames. ], batch size: 48, lr: 7.78e-03, grad_scale: 4.0 2023-04-29 00:04:33,568 INFO [optim.py:368] (1/8) 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 00:04:37,573 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.0801, 3.0381, 3.1708, 1.7822, 3.3517, 3.3405, 2.6583, 2.5933], device='cuda:1'), covar=tensor([0.0815, 0.0210, 0.0200, 0.0990, 0.0072, 0.0141, 0.0437, 0.0437], device='cuda:1'), in_proj_covar=tensor([0.0141, 0.0097, 0.0084, 0.0139, 0.0068, 0.0095, 0.0119, 0.0127], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0003, 0.0002, 0.0004, 0.0002, 0.0002, 0.0003, 0.0003], device='cuda:1') 2023-04-29 00:04:53,109 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=82630.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 00:04:55,546 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.87 vs. limit=2.0 2023-04-29 00:05:09,606 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=82643.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 00:05:22,664 INFO [train.py:904] (1/8) Epoch 9, batch 1450, loss[loss=0.1954, simple_loss=0.2625, pruned_loss=0.06413, over 16675.00 frames. ], tot_loss[loss=0.1852, simple_loss=0.2651, pruned_loss=0.05269, over 3319640.36 frames. ], batch size: 134, lr: 7.78e-03, grad_scale: 4.0 2023-04-29 00:05:34,312 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.1254, 4.1299, 4.6100, 4.5625, 4.5974, 4.2603, 4.3105, 4.1463], device='cuda:1'), covar=tensor([0.0293, 0.0593, 0.0334, 0.0393, 0.0362, 0.0324, 0.0719, 0.0522], device='cuda:1'), in_proj_covar=tensor([0.0315, 0.0322, 0.0324, 0.0311, 0.0366, 0.0341, 0.0443, 0.0273], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:1') 2023-04-29 00:06:16,062 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 9, batch 1500, loss[loss=0.192, simple_loss=0.2687, pruned_loss=0.0577, over 16759.00 frames. ], tot_loss[loss=0.1859, simple_loss=0.2654, pruned_loss=0.05314, over 3320913.74 frames. ], batch size: 102, lr: 7.78e-03, grad_scale: 4.0 2023-04-29 00:06:29,995 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=82702.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 00:06:42,755 INFO [zipformer.py:625] (1/8) 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] (1/8) 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 00:06:52,112 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([5.0302, 4.7757, 5.0648, 5.2724, 5.4241, 4.7632, 5.3947, 5.3649], device='cuda:1'), covar=tensor([0.1265, 0.0978, 0.1353, 0.0521, 0.0470, 0.0671, 0.0437, 0.0456], device='cuda:1'), in_proj_covar=tensor([0.0527, 0.0641, 0.0807, 0.0659, 0.0495, 0.0488, 0.0514, 0.0571], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-29 00:07:39,175 INFO [train.py:904] (1/8) Epoch 9, batch 1550, loss[loss=0.1958, simple_loss=0.2888, pruned_loss=0.0514, over 17119.00 frames. ], tot_loss[loss=0.1882, simple_loss=0.2671, pruned_loss=0.05464, over 3312706.84 frames. ], batch size: 49, lr: 7.77e-03, grad_scale: 4.0 2023-04-29 00:07:49,786 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=82759.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 00:07:58,579 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.92 vs. limit=2.0 2023-04-29 00:08:12,352 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.6604, 4.2616, 4.3917, 2.4546, 3.6845, 2.9725, 4.2080, 4.2955], device='cuda:1'), covar=tensor([0.0233, 0.0530, 0.0365, 0.1403, 0.0544, 0.0788, 0.0456, 0.0770], device='cuda:1'), in_proj_covar=tensor([0.0143, 0.0140, 0.0156, 0.0142, 0.0134, 0.0125, 0.0136, 0.0150], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-04-29 00:08:19,719 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.3288, 2.0607, 2.1903, 4.0184, 2.0997, 2.6076, 2.1709, 2.2754], device='cuda:1'), covar=tensor([0.0918, 0.2978, 0.1903, 0.0373, 0.3108, 0.1869, 0.2808, 0.2549], device='cuda:1'), in_proj_covar=tensor([0.0353, 0.0374, 0.0316, 0.0326, 0.0399, 0.0420, 0.0335, 0.0441], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-29 00:08:48,744 INFO [train.py:904] (1/8) Epoch 9, batch 1600, loss[loss=0.2376, simple_loss=0.3133, pruned_loss=0.08091, over 12414.00 frames. ], tot_loss[loss=0.1904, simple_loss=0.2692, pruned_loss=0.05578, over 3308837.81 frames. ], batch size: 246, lr: 7.77e-03, grad_scale: 8.0 2023-04-29 00:09:09,709 INFO [optim.py:368] (1/8) 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 00:09:20,067 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.2166, 3.6397, 3.3774, 2.0328, 2.9356, 2.6265, 3.5896, 3.7616], device='cuda:1'), covar=tensor([0.0281, 0.0681, 0.0624, 0.1650, 0.0747, 0.0841, 0.0583, 0.0767], device='cuda:1'), in_proj_covar=tensor([0.0143, 0.0139, 0.0156, 0.0142, 0.0134, 0.0124, 0.0136, 0.0150], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-04-29 00:09:52,304 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=82849.0, num_to_drop=1, layers_to_drop={2} 2023-04-29 00:09:56,031 INFO [train.py:904] (1/8) Epoch 9, batch 1650, loss[loss=0.1817, simple_loss=0.2549, pruned_loss=0.05431, over 16227.00 frames. ], tot_loss[loss=0.1918, simple_loss=0.2712, pruned_loss=0.05625, over 3311229.44 frames. ], batch size: 36, lr: 7.77e-03, grad_scale: 8.0 2023-04-29 00:09:56,421 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.8404, 4.8125, 4.7087, 4.1323, 4.7651, 1.9853, 4.5039, 4.5757], device='cuda:1'), covar=tensor([0.0094, 0.0069, 0.0132, 0.0316, 0.0076, 0.2148, 0.0116, 0.0144], device='cuda:1'), in_proj_covar=tensor([0.0123, 0.0111, 0.0161, 0.0155, 0.0130, 0.0176, 0.0145, 0.0152], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-29 00:10:21,994 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([5.2325, 5.5683, 5.2455, 5.3877, 5.0191, 4.8968, 5.0468, 5.6546], device='cuda:1'), covar=tensor([0.1043, 0.0790, 0.1103, 0.0627, 0.0831, 0.0744, 0.0918, 0.0804], device='cuda:1'), in_proj_covar=tensor([0.0514, 0.0658, 0.0548, 0.0450, 0.0410, 0.0423, 0.0542, 0.0487], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-29 00:11:00,046 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.33 vs. limit=2.0 2023-04-29 00:11:05,716 INFO [train.py:904] (1/8) Epoch 9, batch 1700, loss[loss=0.1923, simple_loss=0.2699, pruned_loss=0.05735, over 16833.00 frames. ], tot_loss[loss=0.1938, simple_loss=0.2736, pruned_loss=0.05699, over 3305152.56 frames. ], batch size: 96, lr: 7.77e-03, grad_scale: 8.0 2023-04-29 00:11:24,536 INFO [optim.py:368] (1/8) 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 00:11:37,131 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=82925.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 00:11:53,693 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.2273, 2.4628, 2.0231, 2.2130, 2.8860, 2.6916, 3.2704, 3.1200], device='cuda:1'), covar=tensor([0.0091, 0.0286, 0.0352, 0.0301, 0.0168, 0.0229, 0.0154, 0.0142], device='cuda:1'), in_proj_covar=tensor([0.0131, 0.0196, 0.0189, 0.0193, 0.0192, 0.0196, 0.0198, 0.0181], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-29 00:12:09,340 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.7557, 4.2140, 2.9030, 2.2948, 2.8622, 2.1729, 4.2415, 3.7468], device='cuda:1'), covar=tensor([0.2601, 0.0600, 0.1651, 0.2142, 0.2468, 0.1903, 0.0470, 0.1012], device='cuda:1'), in_proj_covar=tensor([0.0291, 0.0254, 0.0278, 0.0267, 0.0276, 0.0215, 0.0257, 0.0285], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-29 00:12:13,189 INFO [train.py:904] (1/8) Epoch 9, batch 1750, loss[loss=0.1759, simple_loss=0.262, pruned_loss=0.04488, over 17245.00 frames. ], tot_loss[loss=0.1927, simple_loss=0.2733, pruned_loss=0.05607, over 3316481.30 frames. ], batch size: 45, lr: 7.76e-03, grad_scale: 8.0 2023-04-29 00:12:28,806 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=82963.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 00:13:19,615 INFO [train.py:904] (1/8) Epoch 9, batch 1800, loss[loss=0.2077, simple_loss=0.285, pruned_loss=0.06515, over 16500.00 frames. ], tot_loss[loss=0.1932, simple_loss=0.2744, pruned_loss=0.05606, over 3317705.74 frames. ], batch size: 146, lr: 7.76e-03, grad_scale: 8.0 2023-04-29 00:13:20,067 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=83002.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 00:13:40,257 INFO [optim.py:368] (1/8) 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 00:13:50,186 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=83024.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 00:14:11,352 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.5555, 4.3235, 4.5573, 4.7504, 4.8404, 4.3657, 4.6521, 4.7915], device='cuda:1'), covar=tensor([0.1110, 0.0830, 0.1096, 0.0454, 0.0449, 0.0913, 0.1416, 0.0444], device='cuda:1'), in_proj_covar=tensor([0.0521, 0.0636, 0.0800, 0.0648, 0.0487, 0.0488, 0.0508, 0.0564], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-29 00:14:13,637 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([5.3559, 5.8582, 5.6208, 5.6575, 5.1965, 5.1154, 5.3233, 5.9843], device='cuda:1'), covar=tensor([0.1196, 0.0807, 0.0907, 0.0611, 0.0790, 0.0620, 0.0857, 0.0775], device='cuda:1'), in_proj_covar=tensor([0.0507, 0.0647, 0.0539, 0.0443, 0.0403, 0.0413, 0.0532, 0.0478], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-29 00:14:19,461 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.8980, 3.9631, 3.8180, 3.6894, 3.3839, 3.9049, 3.6211, 3.6170], device='cuda:1'), covar=tensor([0.0631, 0.0501, 0.0321, 0.0254, 0.0838, 0.0401, 0.1003, 0.0610], device='cuda:1'), in_proj_covar=tensor([0.0242, 0.0290, 0.0284, 0.0254, 0.0307, 0.0290, 0.0196, 0.0326], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-04-29 00:14:26,662 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 9, batch 1850, loss[loss=0.2149, simple_loss=0.2923, pruned_loss=0.06872, over 16448.00 frames. ], tot_loss[loss=0.1951, simple_loss=0.2761, pruned_loss=0.05705, over 3308630.81 frames. ], batch size: 68, lr: 7.76e-03, grad_scale: 8.0 2023-04-29 00:15:37,693 INFO [train.py:904] (1/8) Epoch 9, batch 1900, loss[loss=0.1725, simple_loss=0.2495, pruned_loss=0.0478, over 16775.00 frames. ], tot_loss[loss=0.1943, simple_loss=0.2752, pruned_loss=0.05671, over 3300699.91 frames. ], batch size: 39, lr: 7.76e-03, grad_scale: 8.0 2023-04-29 00:15:59,152 INFO [optim.py:368] (1/8) 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 00:16:42,382 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=83149.0, num_to_drop=1, layers_to_drop={2} 2023-04-29 00:16:46,825 INFO [train.py:904] (1/8) Epoch 9, batch 1950, loss[loss=0.2029, simple_loss=0.2734, pruned_loss=0.06617, over 16735.00 frames. ], tot_loss[loss=0.1941, simple_loss=0.2755, pruned_loss=0.05633, over 3301719.38 frames. ], batch size: 124, lr: 7.76e-03, grad_scale: 8.0 2023-04-29 00:17:48,370 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=83196.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 00:17:49,286 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=83197.0, num_to_drop=1, layers_to_drop={0} 2023-04-29 00:17:55,081 INFO [train.py:904] (1/8) Epoch 9, batch 2000, loss[loss=0.2035, simple_loss=0.2865, pruned_loss=0.06024, over 16616.00 frames. ], tot_loss[loss=0.1937, simple_loss=0.2751, pruned_loss=0.05616, over 3296301.19 frames. ], batch size: 62, lr: 7.75e-03, grad_scale: 8.0 2023-04-29 00:18:17,512 INFO [optim.py:368] (1/8) 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 00:18:26,844 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=83225.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 00:19:04,138 INFO [train.py:904] (1/8) Epoch 9, batch 2050, loss[loss=0.2155, simple_loss=0.2835, pruned_loss=0.07369, over 16780.00 frames. ], tot_loss[loss=0.1925, simple_loss=0.2737, pruned_loss=0.05568, over 3305877.93 frames. ], batch size: 124, lr: 7.75e-03, grad_scale: 8.0 2023-04-29 00:19:12,804 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=83257.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 00:19:31,859 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.8468, 2.1278, 2.3919, 4.5558, 2.0053, 2.6785, 2.2246, 2.3343], device='cuda:1'), covar=tensor([0.0811, 0.3364, 0.1929, 0.0340, 0.3906, 0.2187, 0.2839, 0.3557], device='cuda:1'), in_proj_covar=tensor([0.0351, 0.0373, 0.0315, 0.0323, 0.0399, 0.0423, 0.0333, 0.0440], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-29 00:19:33,894 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=83273.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 00:19:48,975 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 9, batch 2100, loss[loss=0.1748, simple_loss=0.2611, pruned_loss=0.04426, over 17217.00 frames. ], tot_loss[loss=0.1936, simple_loss=0.2746, pruned_loss=0.0563, over 3312983.68 frames. ], batch size: 46, lr: 7.75e-03, grad_scale: 8.0 2023-04-29 00:20:25,152 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.9045, 3.2960, 3.0557, 1.9291, 2.7390, 2.2465, 3.4465, 3.4639], device='cuda:1'), covar=tensor([0.0218, 0.0718, 0.0577, 0.1654, 0.0749, 0.0898, 0.0511, 0.0747], device='cuda:1'), in_proj_covar=tensor([0.0142, 0.0140, 0.0155, 0.0141, 0.0133, 0.0124, 0.0135, 0.0150], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-04-29 00:20:35,021 INFO [optim.py:368] (1/8) 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 00:20:37,090 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=83319.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 00:21:05,396 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([5.4541, 5.4237, 5.1122, 4.6392, 5.2248, 2.0649, 5.0058, 5.2139], device='cuda:1'), covar=tensor([0.0047, 0.0046, 0.0129, 0.0283, 0.0058, 0.1898, 0.0087, 0.0116], device='cuda:1'), in_proj_covar=tensor([0.0123, 0.0110, 0.0160, 0.0153, 0.0128, 0.0174, 0.0144, 0.0152], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-29 00:21:11,823 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=83345.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 00:21:15,297 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.38 vs. limit=2.0 2023-04-29 00:21:20,735 INFO [train.py:904] (1/8) Epoch 9, batch 2150, loss[loss=0.196, simple_loss=0.2833, pruned_loss=0.05432, over 17126.00 frames. ], tot_loss[loss=0.1945, simple_loss=0.2756, pruned_loss=0.05673, over 3307882.15 frames. ], batch size: 48, lr: 7.75e-03, grad_scale: 8.0 2023-04-29 00:22:31,367 INFO [train.py:904] (1/8) Epoch 9, batch 2200, loss[loss=0.1764, simple_loss=0.2569, pruned_loss=0.04801, over 17014.00 frames. ], tot_loss[loss=0.1949, simple_loss=0.2763, pruned_loss=0.05675, over 3317814.50 frames. ], batch size: 41, lr: 7.74e-03, grad_scale: 8.0 2023-04-29 00:22:54,054 INFO [optim.py:368] (1/8) 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] (1/8) Epoch 9, batch 2250, loss[loss=0.2126, simple_loss=0.2826, pruned_loss=0.07131, over 16414.00 frames. ], tot_loss[loss=0.1963, simple_loss=0.2777, pruned_loss=0.05741, over 3313521.49 frames. ], batch size: 146, lr: 7.74e-03, grad_scale: 8.0 2023-04-29 00:24:00,964 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.9004, 2.1342, 2.2184, 4.7197, 2.0730, 2.8297, 2.3412, 2.4632], device='cuda:1'), covar=tensor([0.0730, 0.3217, 0.2021, 0.0287, 0.3686, 0.2167, 0.2638, 0.3226], device='cuda:1'), in_proj_covar=tensor([0.0354, 0.0376, 0.0316, 0.0324, 0.0401, 0.0425, 0.0334, 0.0442], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-29 00:24:23,949 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.16 vs. limit=2.0 2023-04-29 00:24:49,196 INFO [train.py:904] (1/8) Epoch 9, batch 2300, loss[loss=0.1698, simple_loss=0.2469, pruned_loss=0.04633, over 16804.00 frames. ], tot_loss[loss=0.1974, simple_loss=0.2786, pruned_loss=0.05811, over 3307324.19 frames. ], batch size: 39, lr: 7.74e-03, grad_scale: 8.0 2023-04-29 00:25:12,015 INFO [optim.py:368] (1/8) Clipping_scale=2.0, grad-norm quartiles 1.453e+02 2.674e+02 3.137e+02 4.026e+02 1.366e+03, threshold=6.274e+02, percent-clipped=5.0 2023-04-29 00:25:59,039 INFO [train.py:904] (1/8) Epoch 9, batch 2350, loss[loss=0.1944, simple_loss=0.2836, pruned_loss=0.0526, over 17017.00 frames. ], tot_loss[loss=0.1976, simple_loss=0.2787, pruned_loss=0.05826, over 3311313.10 frames. ], batch size: 55, lr: 7.74e-03, grad_scale: 8.0 2023-04-29 00:25:59,328 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=83552.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 00:27:06,421 INFO [train.py:904] (1/8) Epoch 9, batch 2400, loss[loss=0.1706, simple_loss=0.2557, pruned_loss=0.0427, over 16796.00 frames. ], tot_loss[loss=0.1967, simple_loss=0.2783, pruned_loss=0.05751, over 3323768.12 frames. ], batch size: 39, lr: 7.73e-03, grad_scale: 8.0 2023-04-29 00:27:21,098 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.89 vs. limit=2.0 2023-04-29 00:27:24,796 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=4.65 vs. limit=5.0 2023-04-29 00:27:29,672 INFO [optim.py:368] (1/8) 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 00:27:31,377 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=83619.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 00:27:59,097 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 9, batch 2450, loss[loss=0.2287, simple_loss=0.3104, pruned_loss=0.07355, over 17077.00 frames. ], tot_loss[loss=0.1968, simple_loss=0.2788, pruned_loss=0.05735, over 3318843.41 frames. ], batch size: 53, lr: 7.73e-03, grad_scale: 8.0 2023-04-29 00:28:16,634 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([1.6467, 3.0745, 2.6436, 4.8991, 4.1645, 4.5750, 1.4699, 3.0838], device='cuda:1'), covar=tensor([0.1331, 0.0621, 0.1092, 0.0148, 0.0236, 0.0339, 0.1440, 0.0770], device='cuda:1'), in_proj_covar=tensor([0.0147, 0.0153, 0.0174, 0.0128, 0.0199, 0.0210, 0.0172, 0.0175], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0003, 0.0004], device='cuda:1') 2023-04-29 00:28:22,605 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([5.1898, 5.0504, 5.0646, 4.6946, 4.6074, 5.0533, 5.1029, 4.6959], device='cuda:1'), covar=tensor([0.0485, 0.0316, 0.0220, 0.0238, 0.1034, 0.0304, 0.0254, 0.0557], device='cuda:1'), in_proj_covar=tensor([0.0245, 0.0295, 0.0286, 0.0258, 0.0311, 0.0293, 0.0196, 0.0329], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-04-29 00:28:25,715 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.6358, 4.1565, 4.2706, 2.1956, 3.3081, 2.7517, 3.9783, 4.1156], device='cuda:1'), covar=tensor([0.0249, 0.0562, 0.0372, 0.1614, 0.0675, 0.0782, 0.0596, 0.0854], device='cuda:1'), in_proj_covar=tensor([0.0143, 0.0140, 0.0156, 0.0140, 0.0134, 0.0123, 0.0135, 0.0151], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:1') 2023-04-29 00:28:34,504 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 9, batch 2500, loss[loss=0.2143, simple_loss=0.286, pruned_loss=0.07135, over 16868.00 frames. ], tot_loss[loss=0.1959, simple_loss=0.2782, pruned_loss=0.05681, over 3327398.35 frames. ], batch size: 116, lr: 7.73e-03, grad_scale: 8.0 2023-04-29 00:29:43,294 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.2160, 2.4792, 1.9645, 2.2638, 2.8899, 2.6376, 3.2567, 3.1062], device='cuda:1'), covar=tensor([0.0087, 0.0255, 0.0326, 0.0287, 0.0147, 0.0234, 0.0145, 0.0147], device='cuda:1'), in_proj_covar=tensor([0.0129, 0.0191, 0.0185, 0.0190, 0.0189, 0.0193, 0.0197, 0.0180], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-29 00:29:44,587 INFO [optim.py:368] (1/8) 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 00:30:28,783 INFO [train.py:904] (1/8) Epoch 9, batch 2550, loss[loss=0.1881, simple_loss=0.2654, pruned_loss=0.05538, over 16644.00 frames. ], tot_loss[loss=0.1964, simple_loss=0.2788, pruned_loss=0.05696, over 3323271.29 frames. ], batch size: 134, lr: 7.73e-03, grad_scale: 8.0 2023-04-29 00:30:34,583 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.4373, 3.4473, 3.4442, 2.8721, 3.4094, 1.9468, 3.2153, 2.9449], device='cuda:1'), covar=tensor([0.0108, 0.0093, 0.0125, 0.0223, 0.0074, 0.1990, 0.0119, 0.0190], device='cuda:1'), in_proj_covar=tensor([0.0124, 0.0110, 0.0161, 0.0154, 0.0129, 0.0175, 0.0146, 0.0152], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-29 00:30:41,788 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=83760.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 00:31:04,998 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.2063, 1.9899, 2.6442, 2.9885, 2.9489, 3.5041, 2.3525, 3.3678], device='cuda:1'), covar=tensor([0.0111, 0.0295, 0.0177, 0.0176, 0.0145, 0.0104, 0.0268, 0.0105], device='cuda:1'), in_proj_covar=tensor([0.0155, 0.0165, 0.0150, 0.0153, 0.0157, 0.0114, 0.0163, 0.0105], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001], device='cuda:1') 2023-04-29 00:31:37,367 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.0950, 4.3776, 3.3863, 2.5593, 3.0900, 2.5637, 4.6600, 3.9653], device='cuda:1'), covar=tensor([0.2044, 0.0629, 0.1314, 0.1959, 0.2315, 0.1601, 0.0334, 0.0974], device='cuda:1'), in_proj_covar=tensor([0.0291, 0.0256, 0.0278, 0.0268, 0.0281, 0.0214, 0.0262, 0.0289], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-29 00:31:38,616 INFO [train.py:904] (1/8) Epoch 9, batch 2600, loss[loss=0.1772, simple_loss=0.2547, pruned_loss=0.04989, over 17003.00 frames. ], tot_loss[loss=0.1957, simple_loss=0.2784, pruned_loss=0.05653, over 3330352.35 frames. ], batch size: 41, lr: 7.73e-03, grad_scale: 8.0 2023-04-29 00:31:42,359 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.3700, 4.2809, 4.4368, 4.3131, 4.2967, 4.9165, 4.4830, 4.1522], device='cuda:1'), covar=tensor([0.1611, 0.1976, 0.1910, 0.2200, 0.3109, 0.1325, 0.1438, 0.2545], device='cuda:1'), in_proj_covar=tensor([0.0337, 0.0476, 0.0505, 0.0418, 0.0556, 0.0531, 0.0402, 0.0560], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-04-29 00:31:59,379 INFO [optim.py:368] (1/8) Clipping_scale=2.0, grad-norm quartiles 1.575e+02 2.494e+02 3.064e+02 3.833e+02 6.863e+02, threshold=6.129e+02, percent-clipped=2.0 2023-04-29 00:32:04,253 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=83821.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 00:32:10,258 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=2.06 vs. limit=2.0 2023-04-29 00:32:45,547 INFO [train.py:904] (1/8) Epoch 9, batch 2650, loss[loss=0.1608, simple_loss=0.2553, pruned_loss=0.03316, over 17218.00 frames. ], tot_loss[loss=0.1955, simple_loss=0.2783, pruned_loss=0.05637, over 3329598.04 frames. ], batch size: 45, lr: 7.72e-03, grad_scale: 8.0 2023-04-29 00:32:45,925 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=83852.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 00:33:09,394 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.9582, 5.3434, 5.5026, 5.2764, 5.3006, 5.8918, 5.4799, 5.1839], device='cuda:1'), covar=tensor([0.0916, 0.1645, 0.1711, 0.1861, 0.2549, 0.1027, 0.1200, 0.2242], device='cuda:1'), in_proj_covar=tensor([0.0334, 0.0470, 0.0500, 0.0413, 0.0547, 0.0525, 0.0398, 0.0555], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-04-29 00:33:35,200 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.8915, 2.7117, 2.2458, 2.4736, 3.1095, 2.8943, 3.7329, 3.3628], device='cuda:1'), covar=tensor([0.0057, 0.0256, 0.0323, 0.0291, 0.0157, 0.0223, 0.0129, 0.0150], device='cuda:1'), in_proj_covar=tensor([0.0131, 0.0195, 0.0189, 0.0193, 0.0191, 0.0195, 0.0201, 0.0182], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-29 00:33:51,886 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 9, batch 2700, loss[loss=0.179, simple_loss=0.2587, pruned_loss=0.04967, over 16810.00 frames. ], tot_loss[loss=0.1943, simple_loss=0.2778, pruned_loss=0.05536, over 3328823.91 frames. ], batch size: 102, lr: 7.72e-03, grad_scale: 8.0 2023-04-29 00:34:17,451 INFO [optim.py:368] (1/8) 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 00:34:47,062 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=83940.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 00:34:58,147 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=83949.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 00:35:02,469 INFO [train.py:904] (1/8) Epoch 9, batch 2750, loss[loss=0.2076, simple_loss=0.2775, pruned_loss=0.06884, over 16832.00 frames. ], tot_loss[loss=0.1933, simple_loss=0.2777, pruned_loss=0.05451, over 3334313.68 frames. ], batch size: 116, lr: 7.72e-03, grad_scale: 8.0 2023-04-29 00:35:48,329 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=83985.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 00:35:51,693 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=83988.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 00:36:15,269 INFO [train.py:904] (1/8) Epoch 9, batch 2800, loss[loss=0.1583, simple_loss=0.2446, pruned_loss=0.03601, over 16788.00 frames. ], tot_loss[loss=0.1928, simple_loss=0.2773, pruned_loss=0.05415, over 3336034.40 frames. ], batch size: 39, lr: 7.72e-03, grad_scale: 8.0 2023-04-29 00:36:25,652 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=84010.0, num_to_drop=1, layers_to_drop={1} 2023-04-29 00:36:36,295 INFO [optim.py:368] (1/8) 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 00:37:16,006 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 9, batch 2850, loss[loss=0.1738, simple_loss=0.2535, pruned_loss=0.04708, over 17242.00 frames. ], tot_loss[loss=0.1928, simple_loss=0.2767, pruned_loss=0.05447, over 3334531.22 frames. ], batch size: 43, lr: 7.71e-03, grad_scale: 8.0 2023-04-29 00:37:25,325 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.84 vs. limit=2.0 2023-04-29 00:38:25,972 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=84097.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 00:38:32,592 INFO [train.py:904] (1/8) Epoch 9, batch 2900, loss[loss=0.2091, simple_loss=0.2728, pruned_loss=0.0727, over 16437.00 frames. ], tot_loss[loss=0.1943, simple_loss=0.2766, pruned_loss=0.05596, over 3320197.79 frames. ], batch size: 146, lr: 7.71e-03, grad_scale: 8.0 2023-04-29 00:38:52,488 INFO [zipformer.py:625] (1/8) 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] (1/8) 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 00:39:17,588 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.2980, 1.9247, 2.2077, 3.9434, 2.0890, 2.5050, 2.0324, 2.1610], device='cuda:1'), covar=tensor([0.0918, 0.3167, 0.1867, 0.0416, 0.3070, 0.1806, 0.2902, 0.2594], device='cuda:1'), in_proj_covar=tensor([0.0356, 0.0377, 0.0316, 0.0325, 0.0399, 0.0427, 0.0335, 0.0445], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-29 00:39:43,180 INFO [train.py:904] (1/8) Epoch 9, batch 2950, loss[loss=0.1841, simple_loss=0.2775, pruned_loss=0.0454, over 17119.00 frames. ], tot_loss[loss=0.1933, simple_loss=0.2751, pruned_loss=0.05577, over 3320892.04 frames. ], batch size: 49, lr: 7.71e-03, grad_scale: 8.0 2023-04-29 00:39:51,135 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=84158.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 00:40:21,148 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=4.75 vs. limit=5.0 2023-04-29 00:40:47,997 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=84199.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 00:40:50,425 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.0813, 2.2461, 2.3750, 4.7905, 2.2614, 2.9743, 2.3976, 2.6770], device='cuda:1'), covar=tensor([0.0631, 0.3099, 0.1878, 0.0274, 0.3335, 0.1932, 0.2529, 0.2988], device='cuda:1'), in_proj_covar=tensor([0.0357, 0.0378, 0.0316, 0.0326, 0.0400, 0.0428, 0.0337, 0.0447], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-29 00:40:52,809 INFO [train.py:904] (1/8) Epoch 9, batch 3000, loss[loss=0.2418, simple_loss=0.3089, pruned_loss=0.08733, over 12442.00 frames. ], tot_loss[loss=0.1943, simple_loss=0.2756, pruned_loss=0.05649, over 3320236.57 frames. ], batch size: 246, lr: 7.71e-03, grad_scale: 8.0 2023-04-29 00:40:52,809 INFO [train.py:929] (1/8) Computing validation loss 2023-04-29 00:41:02,059 INFO [train.py:938] (1/8) Epoch 9, validation: loss=0.1444, simple_loss=0.2507, pruned_loss=0.019, over 944034.00 frames. 2023-04-29 00:41:02,060 INFO [train.py:939] (1/8) Maximum memory allocated so far is 17676MB 2023-04-29 00:41:23,159 INFO [optim.py:368] (1/8) 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 00:41:33,528 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([5.0343, 4.9263, 4.9145, 4.6444, 4.5444, 4.8974, 4.7942, 4.5322], device='cuda:1'), covar=tensor([0.0532, 0.0457, 0.0213, 0.0219, 0.0872, 0.0357, 0.0293, 0.0651], device='cuda:1'), in_proj_covar=tensor([0.0246, 0.0297, 0.0285, 0.0260, 0.0312, 0.0296, 0.0196, 0.0331], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-04-29 00:41:47,886 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=84236.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 00:42:09,947 INFO [train.py:904] (1/8) Epoch 9, batch 3050, loss[loss=0.1599, simple_loss=0.2451, pruned_loss=0.03735, over 17202.00 frames. ], tot_loss[loss=0.1946, simple_loss=0.2753, pruned_loss=0.05694, over 3320249.43 frames. ], batch size: 45, lr: 7.70e-03, grad_scale: 8.0 2023-04-29 00:42:20,945 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=84260.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 00:42:41,832 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.1526, 4.1985, 4.5793, 4.5700, 4.5959, 4.2299, 4.2996, 4.1073], device='cuda:1'), covar=tensor([0.0324, 0.0483, 0.0387, 0.0405, 0.0406, 0.0359, 0.0753, 0.0551], device='cuda:1'), in_proj_covar=tensor([0.0319, 0.0326, 0.0329, 0.0314, 0.0372, 0.0347, 0.0458, 0.0279], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:1') 2023-04-29 00:42:45,160 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([5.4050, 5.7897, 5.5802, 5.6258, 5.2053, 4.9272, 5.3131, 5.9760], device='cuda:1'), covar=tensor([0.1235, 0.0971, 0.0997, 0.0569, 0.0858, 0.0679, 0.0867, 0.0781], device='cuda:1'), in_proj_covar=tensor([0.0524, 0.0664, 0.0554, 0.0451, 0.0409, 0.0421, 0.0546, 0.0491], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-29 00:43:11,179 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=84297.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 00:43:16,453 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.1726, 4.2390, 4.1057, 3.9796, 3.6619, 4.2022, 3.8902, 3.8488], device='cuda:1'), covar=tensor([0.0749, 0.0540, 0.0326, 0.0287, 0.1035, 0.0466, 0.0775, 0.0680], device='cuda:1'), in_proj_covar=tensor([0.0248, 0.0299, 0.0288, 0.0262, 0.0315, 0.0297, 0.0198, 0.0334], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-04-29 00:43:17,162 INFO [train.py:904] (1/8) Epoch 9, batch 3100, loss[loss=0.2151, simple_loss=0.2849, pruned_loss=0.07268, over 16402.00 frames. ], tot_loss[loss=0.1943, simple_loss=0.275, pruned_loss=0.05677, over 3326192.53 frames. ], batch size: 146, lr: 7.70e-03, grad_scale: 8.0 2023-04-29 00:43:22,113 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=84305.0, num_to_drop=1, layers_to_drop={1} 2023-04-29 00:43:39,988 INFO [optim.py:368] (1/8) 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 00:43:40,479 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.8848, 1.5808, 2.3085, 2.7250, 2.7694, 2.7326, 1.5535, 2.8562], device='cuda:1'), covar=tensor([0.0089, 0.0310, 0.0182, 0.0135, 0.0122, 0.0132, 0.0392, 0.0068], device='cuda:1'), in_proj_covar=tensor([0.0155, 0.0165, 0.0152, 0.0155, 0.0157, 0.0115, 0.0164, 0.0106], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001], device='cuda:1') 2023-04-29 00:44:06,176 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=84336.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 00:44:12,243 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=84341.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 00:44:28,369 INFO [train.py:904] (1/8) Epoch 9, batch 3150, loss[loss=0.2121, simple_loss=0.2819, pruned_loss=0.0712, over 16855.00 frames. ], tot_loss[loss=0.1932, simple_loss=0.2738, pruned_loss=0.05628, over 3325437.91 frames. ], batch size: 116, lr: 7.70e-03, grad_scale: 8.0 2023-04-29 00:45:30,267 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=84397.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 00:45:36,487 INFO [train.py:904] (1/8) Epoch 9, batch 3200, loss[loss=0.1801, simple_loss=0.2644, pruned_loss=0.04787, over 17231.00 frames. ], tot_loss[loss=0.1916, simple_loss=0.2724, pruned_loss=0.05544, over 3326875.51 frames. ], batch size: 44, lr: 7.70e-03, grad_scale: 8.0 2023-04-29 00:45:56,451 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=84416.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 00:45:56,482 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.5100, 3.7910, 4.0248, 2.6028, 3.6799, 3.9519, 3.7464, 1.9752], device='cuda:1'), covar=tensor([0.0406, 0.0173, 0.0049, 0.0321, 0.0070, 0.0097, 0.0070, 0.0447], device='cuda:1'), in_proj_covar=tensor([0.0124, 0.0069, 0.0067, 0.0123, 0.0074, 0.0084, 0.0075, 0.0115], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:1') 2023-04-29 00:45:59,092 INFO [optim.py:368] (1/8) Clipping_scale=2.0, grad-norm quartiles 1.715e+02 2.601e+02 3.159e+02 4.189e+02 1.012e+03, threshold=6.318e+02, percent-clipped=3.0 2023-04-29 00:46:45,891 INFO [train.py:904] (1/8) Epoch 9, batch 3250, loss[loss=0.2161, simple_loss=0.2925, pruned_loss=0.06984, over 16751.00 frames. ], tot_loss[loss=0.1928, simple_loss=0.2733, pruned_loss=0.0561, over 3323336.79 frames. ], batch size: 134, lr: 7.70e-03, grad_scale: 8.0 2023-04-29 00:46:46,621 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.62 vs. limit=2.0 2023-04-29 00:46:47,423 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=84453.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 00:47:02,629 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=84464.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 00:47:34,497 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.7573, 4.6650, 4.5825, 4.3766, 4.2584, 4.6539, 4.5377, 4.4107], device='cuda:1'), covar=tensor([0.0503, 0.0485, 0.0253, 0.0231, 0.0869, 0.0400, 0.0372, 0.0609], device='cuda:1'), in_proj_covar=tensor([0.0248, 0.0301, 0.0291, 0.0263, 0.0316, 0.0300, 0.0200, 0.0336], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-04-29 00:47:55,749 INFO [train.py:904] (1/8) Epoch 9, batch 3300, loss[loss=0.2332, simple_loss=0.3073, pruned_loss=0.07957, over 17060.00 frames. ], tot_loss[loss=0.1936, simple_loss=0.2746, pruned_loss=0.05631, over 3321957.09 frames. ], batch size: 55, lr: 7.69e-03, grad_scale: 8.0 2023-04-29 00:48:18,090 INFO [optim.py:368] (1/8) 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 00:48:27,961 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=84525.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 00:49:02,773 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.8113, 1.8053, 2.3930, 2.7571, 2.6151, 3.1221, 1.8386, 3.1627], device='cuda:1'), covar=tensor([0.0134, 0.0313, 0.0201, 0.0175, 0.0182, 0.0106, 0.0311, 0.0071], device='cuda:1'), in_proj_covar=tensor([0.0157, 0.0165, 0.0152, 0.0155, 0.0159, 0.0116, 0.0164, 0.0107], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001], device='cuda:1') 2023-04-29 00:49:04,557 INFO [train.py:904] (1/8) Epoch 9, batch 3350, loss[loss=0.1654, simple_loss=0.2554, pruned_loss=0.03774, over 17134.00 frames. ], tot_loss[loss=0.1953, simple_loss=0.2766, pruned_loss=0.05697, over 3311264.73 frames. ], batch size: 49, lr: 7.69e-03, grad_scale: 8.0 2023-04-29 00:49:10,236 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=84555.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 00:49:53,242 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=84586.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 00:50:01,509 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=84592.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 00:50:06,702 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.8435, 5.3150, 5.4235, 5.3654, 5.2311, 5.8880, 5.4019, 5.1540], device='cuda:1'), covar=tensor([0.0953, 0.1601, 0.1690, 0.1477, 0.2748, 0.0906, 0.1166, 0.2098], device='cuda:1'), in_proj_covar=tensor([0.0341, 0.0477, 0.0503, 0.0419, 0.0556, 0.0525, 0.0400, 0.0565], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-04-29 00:50:16,097 INFO [train.py:904] (1/8) Epoch 9, batch 3400, loss[loss=0.1675, simple_loss=0.2462, pruned_loss=0.04438, over 15828.00 frames. ], tot_loss[loss=0.1949, simple_loss=0.2765, pruned_loss=0.05662, over 3309009.81 frames. ], batch size: 35, lr: 7.69e-03, grad_scale: 8.0 2023-04-29 00:50:20,104 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=84605.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 00:50:38,674 INFO [optim.py:368] (1/8) 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 00:50:59,394 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=84633.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 00:51:10,705 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=84641.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 00:51:24,587 INFO [train.py:904] (1/8) Epoch 9, batch 3450, loss[loss=0.166, simple_loss=0.2488, pruned_loss=0.04164, over 15765.00 frames. ], tot_loss[loss=0.1929, simple_loss=0.2743, pruned_loss=0.05577, over 3315942.66 frames. ], batch size: 35, lr: 7.69e-03, grad_scale: 8.0 2023-04-29 00:51:26,678 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=84653.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 00:52:16,067 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=84689.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 00:52:20,244 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=84692.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 00:52:23,560 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=84694.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 00:52:35,104 INFO [train.py:904] (1/8) Epoch 9, batch 3500, loss[loss=0.1753, simple_loss=0.2564, pruned_loss=0.04715, over 15768.00 frames. ], tot_loss[loss=0.1919, simple_loss=0.2734, pruned_loss=0.05524, over 3321353.05 frames. ], batch size: 35, lr: 7.68e-03, grad_scale: 8.0 2023-04-29 00:52:56,975 INFO [optim.py:368] (1/8) 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 00:53:01,596 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.41 vs. limit=2.0 2023-04-29 00:53:15,090 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.9018, 2.2532, 2.3230, 4.6733, 2.0975, 2.8972, 2.3637, 2.5010], device='cuda:1'), covar=tensor([0.0744, 0.3222, 0.1990, 0.0308, 0.3637, 0.1869, 0.2741, 0.3125], device='cuda:1'), in_proj_covar=tensor([0.0360, 0.0379, 0.0318, 0.0329, 0.0404, 0.0430, 0.0339, 0.0447], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-29 00:53:37,792 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=4.88 vs. limit=5.0 2023-04-29 00:53:44,890 INFO [train.py:904] (1/8) Epoch 9, batch 3550, loss[loss=0.1605, simple_loss=0.2446, pruned_loss=0.03814, over 16091.00 frames. ], tot_loss[loss=0.1918, simple_loss=0.2733, pruned_loss=0.05509, over 3304521.93 frames. ], batch size: 36, lr: 7.68e-03, grad_scale: 8.0 2023-04-29 00:53:46,490 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=84753.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 00:54:14,876 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.28 vs. limit=2.0 2023-04-29 00:54:28,267 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=2.02 vs. limit=2.0 2023-04-29 00:54:32,964 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.7209, 1.6849, 2.2495, 2.6109, 2.6183, 2.6208, 1.7752, 2.8189], device='cuda:1'), covar=tensor([0.0113, 0.0296, 0.0200, 0.0184, 0.0146, 0.0148, 0.0303, 0.0078], device='cuda:1'), in_proj_covar=tensor([0.0156, 0.0165, 0.0151, 0.0154, 0.0158, 0.0115, 0.0163, 0.0106], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001], device='cuda:1') 2023-04-29 00:54:53,014 INFO [zipformer.py:625] (1/8) 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] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=84801.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 00:54:55,836 INFO [train.py:904] (1/8) Epoch 9, batch 3600, loss[loss=0.192, simple_loss=0.2732, pruned_loss=0.05537, over 17116.00 frames. ], tot_loss[loss=0.1901, simple_loss=0.2712, pruned_loss=0.05456, over 3313033.97 frames. ], batch size: 53, lr: 7.68e-03, grad_scale: 8.0 2023-04-29 00:55:17,925 INFO [optim.py:368] (1/8) 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 00:55:30,318 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.8184, 3.7405, 3.8516, 3.9980, 4.0630, 3.6901, 3.8468, 4.0587], device='cuda:1'), covar=tensor([0.1092, 0.0838, 0.1094, 0.0598, 0.0502, 0.1630, 0.1624, 0.0595], device='cuda:1'), in_proj_covar=tensor([0.0541, 0.0661, 0.0832, 0.0679, 0.0510, 0.0510, 0.0529, 0.0588], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-29 00:56:07,141 INFO [train.py:904] (1/8) Epoch 9, batch 3650, loss[loss=0.1579, simple_loss=0.2478, pruned_loss=0.03395, over 17225.00 frames. ], tot_loss[loss=0.1902, simple_loss=0.2703, pruned_loss=0.05506, over 3305962.27 frames. ], batch size: 45, lr: 7.68e-03, grad_scale: 8.0 2023-04-29 00:56:12,813 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=84855.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 00:56:21,132 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=84861.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 00:56:50,503 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=84881.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 00:57:07,465 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 9, batch 3700, loss[loss=0.1854, simple_loss=0.2609, pruned_loss=0.05496, over 11150.00 frames. ], tot_loss[loss=0.1907, simple_loss=0.2687, pruned_loss=0.05632, over 3279691.80 frames. ], batch size: 246, lr: 7.68e-03, grad_scale: 8.0 2023-04-29 00:57:23,506 INFO [zipformer.py:625] (1/8) 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] (1/8) 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 00:58:18,418 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=84940.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 00:58:34,907 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=84951.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 00:58:35,654 INFO [train.py:904] (1/8) Epoch 9, batch 3750, loss[loss=0.1862, simple_loss=0.259, pruned_loss=0.05671, over 16835.00 frames. ], tot_loss[loss=0.1925, simple_loss=0.2694, pruned_loss=0.05782, over 3252683.68 frames. ], batch size: 96, lr: 7.67e-03, grad_scale: 8.0 2023-04-29 00:59:28,295 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=84989.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 00:59:30,717 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.75 vs. limit=2.0 2023-04-29 00:59:34,000 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=84992.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 00:59:47,554 INFO [train.py:904] (1/8) Epoch 9, batch 3800, loss[loss=0.1921, simple_loss=0.2627, pruned_loss=0.0608, over 16896.00 frames. ], tot_loss[loss=0.1948, simple_loss=0.2709, pruned_loss=0.0593, over 3250686.76 frames. ], batch size: 109, lr: 7.67e-03, grad_scale: 8.0 2023-04-29 01:00:03,376 INFO [zipformer.py:625] (1/8) 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] (1/8) 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 01:00:43,935 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=85040.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 01:00:56,723 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([1.8877, 3.0324, 3.1928, 2.1127, 2.9194, 3.1726, 2.9033, 1.7115], device='cuda:1'), covar=tensor([0.0405, 0.0091, 0.0037, 0.0299, 0.0072, 0.0075, 0.0056, 0.0370], device='cuda:1'), in_proj_covar=tensor([0.0124, 0.0068, 0.0067, 0.0123, 0.0074, 0.0084, 0.0074, 0.0115], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:1') 2023-04-29 01:01:00,358 INFO [train.py:904] (1/8) Epoch 9, batch 3850, loss[loss=0.1922, simple_loss=0.2582, pruned_loss=0.0631, over 16879.00 frames. ], tot_loss[loss=0.1953, simple_loss=0.2709, pruned_loss=0.05981, over 3255011.55 frames. ], batch size: 96, lr: 7.67e-03, grad_scale: 8.0 2023-04-29 01:01:10,236 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.8293, 4.9940, 5.1703, 5.0714, 5.0844, 5.6430, 5.2119, 4.8828], device='cuda:1'), covar=tensor([0.1113, 0.1627, 0.1432, 0.1588, 0.2278, 0.0875, 0.1195, 0.2074], device='cuda:1'), in_proj_covar=tensor([0.0338, 0.0463, 0.0495, 0.0412, 0.0540, 0.0514, 0.0395, 0.0550], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-04-29 01:02:13,000 INFO [train.py:904] (1/8) Epoch 9, batch 3900, loss[loss=0.1815, simple_loss=0.2532, pruned_loss=0.05495, over 16792.00 frames. ], tot_loss[loss=0.1956, simple_loss=0.2703, pruned_loss=0.06041, over 3268368.34 frames. ], batch size: 90, lr: 7.67e-03, grad_scale: 8.0 2023-04-29 01:02:21,644 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=85107.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 01:02:23,018 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.0240, 2.7738, 2.7870, 1.9477, 2.5721, 2.0608, 2.7982, 2.8371], device='cuda:1'), covar=tensor([0.0250, 0.0640, 0.0487, 0.1577, 0.0737, 0.0888, 0.0471, 0.0550], device='cuda:1'), in_proj_covar=tensor([0.0143, 0.0143, 0.0156, 0.0141, 0.0135, 0.0125, 0.0135, 0.0153], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-04-29 01:02:32,590 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=85115.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 01:02:36,233 INFO [optim.py:368] (1/8) 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 01:03:24,357 INFO [train.py:904] (1/8) Epoch 9, batch 3950, loss[loss=0.1836, simple_loss=0.257, pruned_loss=0.05504, over 16802.00 frames. ], tot_loss[loss=0.1958, simple_loss=0.2697, pruned_loss=0.06096, over 3270259.19 frames. ], batch size: 102, lr: 7.66e-03, grad_scale: 8.0 2023-04-29 01:03:32,206 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=85156.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 01:03:50,049 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=85168.0, num_to_drop=1, layers_to_drop={0} 2023-04-29 01:04:00,732 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=85176.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 01:04:07,172 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=85181.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 01:04:37,099 INFO [train.py:904] (1/8) Epoch 9, batch 4000, loss[loss=0.1949, simple_loss=0.2687, pruned_loss=0.06058, over 16822.00 frames. ], tot_loss[loss=0.1956, simple_loss=0.2694, pruned_loss=0.06083, over 3279398.25 frames. ], batch size: 102, lr: 7.66e-03, grad_scale: 16.0 2023-04-29 01:05:00,773 INFO [optim.py:368] (1/8) Clipping_scale=2.0, grad-norm quartiles 1.397e+02 2.490e+02 3.071e+02 3.606e+02 5.108e+02, threshold=6.141e+02, percent-clipped=0.0 2023-04-29 01:05:17,723 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=85229.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 01:05:19,199 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([5.3920, 5.3732, 5.2255, 4.9548, 4.8450, 5.3220, 5.0643, 5.0124], device='cuda:1'), covar=tensor([0.0427, 0.0290, 0.0211, 0.0203, 0.0842, 0.0303, 0.0235, 0.0472], device='cuda:1'), in_proj_covar=tensor([0.0240, 0.0290, 0.0283, 0.0256, 0.0308, 0.0293, 0.0192, 0.0326], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-04-29 01:05:51,193 INFO [train.py:904] (1/8) Epoch 9, batch 4050, loss[loss=0.1843, simple_loss=0.2658, pruned_loss=0.05144, over 16467.00 frames. ], tot_loss[loss=0.1939, simple_loss=0.2691, pruned_loss=0.05933, over 3280139.18 frames. ], batch size: 68, lr: 7.66e-03, grad_scale: 16.0 2023-04-29 01:06:31,541 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([1.6549, 2.8586, 2.6174, 4.9394, 3.8432, 4.5379, 1.5525, 2.9924], device='cuda:1'), covar=tensor([0.1272, 0.0675, 0.1139, 0.0095, 0.0400, 0.0271, 0.1506, 0.0892], device='cuda:1'), in_proj_covar=tensor([0.0148, 0.0156, 0.0177, 0.0131, 0.0205, 0.0211, 0.0176, 0.0178], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-04-29 01:06:45,629 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=85289.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 01:07:04,621 INFO [train.py:904] (1/8) Epoch 9, batch 4100, loss[loss=0.1959, simple_loss=0.2913, pruned_loss=0.05019, over 16761.00 frames. ], tot_loss[loss=0.1935, simple_loss=0.2704, pruned_loss=0.05828, over 3272820.37 frames. ], batch size: 76, lr: 7.66e-03, grad_scale: 16.0 2023-04-29 01:07:12,415 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=85307.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 01:07:28,043 INFO [optim.py:368] (1/8) 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] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=85337.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 01:08:01,114 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.6516, 4.9500, 4.6820, 4.7397, 4.4011, 4.2595, 4.4471, 4.9943], device='cuda:1'), covar=tensor([0.0876, 0.0671, 0.0878, 0.0598, 0.0684, 0.1137, 0.0832, 0.0725], device='cuda:1'), in_proj_covar=tensor([0.0505, 0.0639, 0.0531, 0.0436, 0.0399, 0.0409, 0.0530, 0.0477], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-29 01:08:20,247 INFO [train.py:904] (1/8) Epoch 9, batch 4150, loss[loss=0.2727, simple_loss=0.3347, pruned_loss=0.1054, over 11070.00 frames. ], tot_loss[loss=0.201, simple_loss=0.2787, pruned_loss=0.0616, over 3249104.88 frames. ], batch size: 246, lr: 7.66e-03, grad_scale: 16.0 2023-04-29 01:08:49,077 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.2856, 3.4240, 1.8042, 3.6257, 2.4050, 3.5827, 1.9548, 2.6907], device='cuda:1'), covar=tensor([0.0196, 0.0292, 0.1684, 0.0100, 0.0744, 0.0394, 0.1427, 0.0647], device='cuda:1'), in_proj_covar=tensor([0.0141, 0.0163, 0.0183, 0.0116, 0.0164, 0.0205, 0.0190, 0.0167], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-04-29 01:08:57,256 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([1.7095, 1.7612, 1.5763, 1.4818, 1.8302, 1.6301, 1.7832, 1.9847], device='cuda:1'), covar=tensor([0.0088, 0.0173, 0.0237, 0.0215, 0.0109, 0.0167, 0.0106, 0.0128], device='cuda:1'), in_proj_covar=tensor([0.0131, 0.0192, 0.0189, 0.0188, 0.0189, 0.0192, 0.0196, 0.0180], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-29 01:09:37,264 INFO [train.py:904] (1/8) Epoch 9, batch 4200, loss[loss=0.2851, simple_loss=0.3473, pruned_loss=0.1114, over 11114.00 frames. ], tot_loss[loss=0.207, simple_loss=0.2863, pruned_loss=0.06381, over 3198810.09 frames. ], batch size: 248, lr: 7.65e-03, grad_scale: 16.0 2023-04-29 01:10:02,492 INFO [optim.py:368] (1/8) 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 01:10:13,722 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=2.97 vs. limit=5.0 2023-04-29 01:10:17,722 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=85428.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 01:10:52,278 INFO [train.py:904] (1/8) Epoch 9, batch 4250, loss[loss=0.2097, simple_loss=0.2992, pruned_loss=0.06008, over 16362.00 frames. ], tot_loss[loss=0.2072, simple_loss=0.289, pruned_loss=0.06275, over 3214645.54 frames. ], batch size: 146, lr: 7.65e-03, grad_scale: 16.0 2023-04-29 01:10:59,120 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=85456.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 01:11:08,379 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=85463.0, num_to_drop=1, layers_to_drop={0} 2023-04-29 01:11:20,273 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=85471.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 01:11:38,989 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=85483.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 01:11:48,339 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=85489.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 01:11:58,906 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.70 vs. limit=2.0 2023-04-29 01:12:07,381 INFO [train.py:904] (1/8) Epoch 9, batch 4300, loss[loss=0.2257, simple_loss=0.3136, pruned_loss=0.06896, over 16617.00 frames. ], tot_loss[loss=0.2068, simple_loss=0.2902, pruned_loss=0.06173, over 3210925.47 frames. ], batch size: 57, lr: 7.65e-03, grad_scale: 16.0 2023-04-29 01:12:11,406 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=85504.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 01:12:30,788 INFO [optim.py:368] (1/8) 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 01:12:49,229 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=85530.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 01:13:10,276 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 9, batch 4350, loss[loss=0.2186, simple_loss=0.2976, pruned_loss=0.0698, over 11761.00 frames. ], tot_loss[loss=0.2095, simple_loss=0.2933, pruned_loss=0.06286, over 3200979.69 frames. ], batch size: 248, lr: 7.65e-03, grad_scale: 16.0 2023-04-29 01:13:25,503 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.0878, 4.1949, 2.3479, 5.0479, 3.2096, 4.7948, 2.5131, 3.1157], device='cuda:1'), covar=tensor([0.0168, 0.0219, 0.1519, 0.0032, 0.0569, 0.0257, 0.1269, 0.0583], device='cuda:1'), in_proj_covar=tensor([0.0141, 0.0163, 0.0184, 0.0113, 0.0164, 0.0202, 0.0191, 0.0167], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-04-29 01:13:36,908 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.28 vs. limit=2.0 2023-04-29 01:13:58,198 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.7803, 3.9363, 2.2271, 4.5936, 2.9460, 4.4188, 2.4482, 3.0418], device='cuda:1'), covar=tensor([0.0184, 0.0269, 0.1576, 0.0046, 0.0677, 0.0334, 0.1259, 0.0619], device='cuda:1'), in_proj_covar=tensor([0.0141, 0.0163, 0.0183, 0.0113, 0.0164, 0.0201, 0.0191, 0.0167], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-04-29 01:13:58,445 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=3.46 vs. limit=5.0 2023-04-29 01:14:19,002 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=85591.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 01:14:34,737 INFO [train.py:904] (1/8) Epoch 9, batch 4400, loss[loss=0.2408, simple_loss=0.3109, pruned_loss=0.08534, over 11760.00 frames. ], tot_loss[loss=0.2122, simple_loss=0.2956, pruned_loss=0.06446, over 3175626.38 frames. ], batch size: 247, lr: 7.64e-03, grad_scale: 16.0 2023-04-29 01:14:38,188 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([5.0654, 4.8874, 5.1589, 5.3571, 5.4445, 4.8173, 5.4339, 5.4583], device='cuda:1'), covar=tensor([0.1310, 0.0898, 0.1128, 0.0388, 0.0363, 0.0574, 0.0363, 0.0356], device='cuda:1'), in_proj_covar=tensor([0.0492, 0.0600, 0.0742, 0.0604, 0.0465, 0.0468, 0.0479, 0.0534], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-29 01:14:41,579 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=85607.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 01:14:50,333 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.31 vs. limit=2.0 2023-04-29 01:14:56,972 INFO [optim.py:368] (1/8) 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 01:15:06,855 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 9, batch 4450, loss[loss=0.2323, simple_loss=0.3264, pruned_loss=0.06908, over 16392.00 frames. ], tot_loss[loss=0.2142, simple_loss=0.2988, pruned_loss=0.06481, over 3199359.06 frames. ], batch size: 146, lr: 7.64e-03, grad_scale: 16.0 2023-04-29 01:15:50,417 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=85655.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 01:16:01,676 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.4751, 3.9739, 3.4139, 3.8772, 3.4247, 3.5147, 3.4809, 3.9044], device='cuda:1'), covar=tensor([0.2878, 0.1385, 0.2421, 0.1158, 0.1729, 0.2976, 0.1937, 0.1785], device='cuda:1'), in_proj_covar=tensor([0.0492, 0.0621, 0.0519, 0.0428, 0.0388, 0.0404, 0.0517, 0.0466], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:1') 2023-04-29 01:16:33,348 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=85686.0, num_to_drop=1, layers_to_drop={3} 2023-04-29 01:16:56,349 INFO [train.py:904] (1/8) Epoch 9, batch 4500, loss[loss=0.2063, simple_loss=0.2845, pruned_loss=0.06408, over 16342.00 frames. ], tot_loss[loss=0.2152, simple_loss=0.2991, pruned_loss=0.06567, over 3197436.34 frames. ], batch size: 35, lr: 7.64e-03, grad_scale: 16.0 2023-04-29 01:17:08,448 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.3996, 3.4562, 3.7269, 1.6401, 4.0842, 4.0841, 3.0023, 2.9875], device='cuda:1'), covar=tensor([0.0810, 0.0190, 0.0184, 0.1227, 0.0044, 0.0068, 0.0351, 0.0418], device='cuda:1'), in_proj_covar=tensor([0.0139, 0.0095, 0.0084, 0.0137, 0.0067, 0.0095, 0.0118, 0.0126], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0003, 0.0002, 0.0004, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:1') 2023-04-29 01:17:20,314 INFO [optim.py:368] (1/8) 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 01:17:33,481 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.6343, 4.6930, 4.4740, 4.3489, 3.8907, 4.6570, 4.4409, 4.1616], device='cuda:1'), covar=tensor([0.0517, 0.0257, 0.0280, 0.0238, 0.1143, 0.0286, 0.0380, 0.0639], device='cuda:1'), in_proj_covar=tensor([0.0223, 0.0270, 0.0264, 0.0239, 0.0289, 0.0271, 0.0181, 0.0304], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-29 01:18:07,070 INFO [train.py:904] (1/8) Epoch 9, batch 4550, loss[loss=0.2517, simple_loss=0.3235, pruned_loss=0.08988, over 16296.00 frames. ], tot_loss[loss=0.2166, simple_loss=0.3, pruned_loss=0.06663, over 3216324.25 frames. ], batch size: 35, lr: 7.64e-03, grad_scale: 16.0 2023-04-29 01:18:23,712 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=85763.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 01:18:34,935 INFO [zipformer.py:625] (1/8) 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] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=85784.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 01:19:20,632 INFO [train.py:904] (1/8) Epoch 9, batch 4600, loss[loss=0.2209, simple_loss=0.3058, pruned_loss=0.06802, over 16530.00 frames. ], tot_loss[loss=0.217, simple_loss=0.3006, pruned_loss=0.06666, over 3225822.13 frames. ], batch size: 68, lr: 7.64e-03, grad_scale: 16.0 2023-04-29 01:19:32,652 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=85811.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 01:19:42,424 INFO [optim.py:368] (1/8) 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] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=85819.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 01:19:54,519 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.9674, 5.5643, 5.7107, 5.5045, 5.4498, 6.0767, 5.6787, 5.3426], device='cuda:1'), covar=tensor([0.0750, 0.1578, 0.1462, 0.1524, 0.2361, 0.0857, 0.0991, 0.2206], device='cuda:1'), in_proj_covar=tensor([0.0340, 0.0459, 0.0485, 0.0404, 0.0536, 0.0517, 0.0392, 0.0545], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-04-29 01:20:12,955 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=85839.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 01:20:16,619 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.2832, 1.8480, 2.5223, 3.1119, 2.9918, 3.5943, 1.8612, 3.3701], device='cuda:1'), covar=tensor([0.0088, 0.0314, 0.0186, 0.0135, 0.0141, 0.0095, 0.0326, 0.0083], device='cuda:1'), in_proj_covar=tensor([0.0148, 0.0161, 0.0147, 0.0151, 0.0155, 0.0113, 0.0161, 0.0104], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001], device='cuda:1') 2023-04-29 01:20:30,868 INFO [train.py:904] (1/8) Epoch 9, batch 4650, loss[loss=0.1686, simple_loss=0.257, pruned_loss=0.04014, over 16526.00 frames. ], tot_loss[loss=0.2157, simple_loss=0.299, pruned_loss=0.06614, over 3231303.49 frames. ], batch size: 75, lr: 7.63e-03, grad_scale: 16.0 2023-04-29 01:21:00,346 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.6955, 4.9928, 4.7659, 4.7630, 4.4423, 4.3503, 4.4808, 5.0649], device='cuda:1'), covar=tensor([0.0906, 0.0744, 0.0958, 0.0581, 0.0729, 0.0946, 0.0872, 0.0822], device='cuda:1'), in_proj_covar=tensor([0.0493, 0.0619, 0.0518, 0.0423, 0.0386, 0.0401, 0.0515, 0.0468], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:1') 2023-04-29 01:21:04,840 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=85875.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 01:21:20,345 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 9, batch 4700, loss[loss=0.1951, simple_loss=0.2789, pruned_loss=0.05564, over 16828.00 frames. ], tot_loss[loss=0.2122, simple_loss=0.2956, pruned_loss=0.06445, over 3227535.37 frames. ], batch size: 83, lr: 7.63e-03, grad_scale: 16.0 2023-04-29 01:21:50,572 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.8952, 4.6901, 4.9005, 5.1358, 5.2280, 4.6903, 5.2209, 5.2551], device='cuda:1'), covar=tensor([0.1244, 0.0929, 0.1220, 0.0444, 0.0422, 0.0636, 0.0363, 0.0388], device='cuda:1'), in_proj_covar=tensor([0.0484, 0.0593, 0.0737, 0.0600, 0.0459, 0.0461, 0.0472, 0.0526], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-29 01:22:06,316 INFO [optim.py:368] (1/8) Clipping_scale=2.0, grad-norm quartiles 1.424e+02 2.169e+02 2.505e+02 2.903e+02 6.033e+02, threshold=5.010e+02, percent-clipped=1.0 2023-04-29 01:22:25,132 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=85930.0, num_to_drop=1, layers_to_drop={0} 2023-04-29 01:22:32,848 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=85936.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 01:22:47,683 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.7428, 2.2315, 1.8657, 1.9785, 2.6101, 2.2394, 2.7405, 2.7943], device='cuda:1'), covar=tensor([0.0083, 0.0287, 0.0333, 0.0312, 0.0158, 0.0244, 0.0147, 0.0158], device='cuda:1'), in_proj_covar=tensor([0.0126, 0.0189, 0.0187, 0.0184, 0.0185, 0.0189, 0.0190, 0.0178], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-29 01:22:55,504 INFO [train.py:904] (1/8) Epoch 9, batch 4750, loss[loss=0.2004, simple_loss=0.2792, pruned_loss=0.06081, over 16660.00 frames. ], tot_loss[loss=0.2089, simple_loss=0.2922, pruned_loss=0.0628, over 3205658.45 frames. ], batch size: 134, lr: 7.63e-03, grad_scale: 16.0 2023-04-29 01:23:04,332 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=85957.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 01:23:39,952 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=85981.0, num_to_drop=1, layers_to_drop={2} 2023-04-29 01:23:54,589 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=85991.0, num_to_drop=1, layers_to_drop={0} 2023-04-29 01:24:13,772 INFO [train.py:904] (1/8) Epoch 9, batch 4800, loss[loss=0.2126, simple_loss=0.2844, pruned_loss=0.07042, over 16653.00 frames. ], tot_loss[loss=0.2046, simple_loss=0.288, pruned_loss=0.06061, over 3213771.09 frames. ], batch size: 57, lr: 7.63e-03, grad_scale: 16.0 2023-04-29 01:24:14,259 INFO [zipformer.py:625] (1/8) 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] (1/8) 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 01:24:38,424 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=86018.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 01:24:44,338 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.6846, 3.5917, 2.9013, 2.2026, 2.5457, 2.3073, 3.8320, 3.4372], device='cuda:1'), covar=tensor([0.2185, 0.0633, 0.1311, 0.2045, 0.1972, 0.1600, 0.0408, 0.0774], device='cuda:1'), in_proj_covar=tensor([0.0294, 0.0254, 0.0278, 0.0272, 0.0282, 0.0215, 0.0262, 0.0284], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-29 01:25:26,288 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([5.2357, 5.5002, 5.2337, 5.3118, 4.9869, 4.7714, 4.9818, 5.5512], device='cuda:1'), covar=tensor([0.0745, 0.0727, 0.0860, 0.0536, 0.0591, 0.0677, 0.0771, 0.0735], device='cuda:1'), in_proj_covar=tensor([0.0482, 0.0606, 0.0508, 0.0412, 0.0380, 0.0393, 0.0506, 0.0456], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:1') 2023-04-29 01:25:28,105 INFO [train.py:904] (1/8) Epoch 9, batch 4850, loss[loss=0.2182, simple_loss=0.3003, pruned_loss=0.06811, over 17025.00 frames. ], tot_loss[loss=0.2051, simple_loss=0.2891, pruned_loss=0.06055, over 3195730.30 frames. ], batch size: 53, lr: 7.62e-03, grad_scale: 16.0 2023-04-29 01:25:28,525 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=86052.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 01:25:45,147 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=86063.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 01:25:54,122 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=86069.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 01:26:16,388 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=86084.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 01:26:30,188 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.78 vs. limit=2.0 2023-04-29 01:26:42,269 INFO [train.py:904] (1/8) Epoch 9, batch 4900, loss[loss=0.1841, simple_loss=0.2731, pruned_loss=0.04753, over 12393.00 frames. ], tot_loss[loss=0.2039, simple_loss=0.2888, pruned_loss=0.05949, over 3183683.10 frames. ], batch size: 248, lr: 7.62e-03, grad_scale: 16.0 2023-04-29 01:26:59,155 INFO [zipformer.py:625] (1/8) 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] (1/8) 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 01:27:23,797 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=86130.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 01:27:25,908 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=86132.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 01:27:37,320 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=86139.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 01:27:55,674 INFO [train.py:904] (1/8) Epoch 9, batch 4950, loss[loss=0.2137, simple_loss=0.2948, pruned_loss=0.06634, over 16222.00 frames. ], tot_loss[loss=0.2038, simple_loss=0.2889, pruned_loss=0.05937, over 3191855.59 frames. ], batch size: 35, lr: 7.62e-03, grad_scale: 16.0 2023-04-29 01:28:12,549 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.1959, 1.8770, 2.4261, 3.0724, 2.9480, 3.5935, 1.8517, 3.4341], device='cuda:1'), covar=tensor([0.0103, 0.0329, 0.0207, 0.0171, 0.0152, 0.0080, 0.0322, 0.0056], device='cuda:1'), in_proj_covar=tensor([0.0148, 0.0162, 0.0147, 0.0151, 0.0155, 0.0113, 0.0161, 0.0103], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001], device='cuda:1') 2023-04-29 01:28:45,913 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=86186.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 01:28:46,988 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=86187.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 01:29:08,680 INFO [train.py:904] (1/8) Epoch 9, batch 5000, loss[loss=0.1996, simple_loss=0.2917, pruned_loss=0.05379, over 16764.00 frames. ], tot_loss[loss=0.2048, simple_loss=0.2906, pruned_loss=0.05948, over 3189909.86 frames. ], batch size: 83, lr: 7.62e-03, grad_scale: 16.0 2023-04-29 01:29:32,026 INFO [optim.py:368] (1/8) 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 01:29:49,843 INFO [zipformer.py:625] (1/8) 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] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=86234.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 01:30:11,782 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=86246.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 01:30:19,719 INFO [train.py:904] (1/8) Epoch 9, batch 5050, loss[loss=0.2474, simple_loss=0.3271, pruned_loss=0.08378, over 11755.00 frames. ], tot_loss[loss=0.205, simple_loss=0.2913, pruned_loss=0.05933, over 3184287.33 frames. ], batch size: 247, lr: 7.62e-03, grad_scale: 16.0 2023-04-29 01:30:47,381 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.1788, 4.0296, 4.2016, 4.4337, 4.5118, 4.1544, 4.4851, 4.5219], device='cuda:1'), covar=tensor([0.1233, 0.0914, 0.1212, 0.0467, 0.0437, 0.0839, 0.0522, 0.0436], device='cuda:1'), in_proj_covar=tensor([0.0492, 0.0600, 0.0744, 0.0608, 0.0462, 0.0465, 0.0482, 0.0535], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-29 01:31:01,279 INFO [zipformer.py:625] (1/8) 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] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=86286.0, num_to_drop=1, layers_to_drop={0} 2023-04-29 01:31:17,588 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.3627, 4.6553, 4.3964, 4.4663, 4.1418, 4.0661, 4.1620, 4.6322], device='cuda:1'), covar=tensor([0.0936, 0.0811, 0.0952, 0.0570, 0.0695, 0.1365, 0.0899, 0.0823], device='cuda:1'), in_proj_covar=tensor([0.0487, 0.0613, 0.0513, 0.0416, 0.0382, 0.0396, 0.0509, 0.0462], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:1') 2023-04-29 01:31:29,531 INFO [train.py:904] (1/8) Epoch 9, batch 5100, loss[loss=0.1934, simple_loss=0.2843, pruned_loss=0.05121, over 16288.00 frames. ], tot_loss[loss=0.2027, simple_loss=0.2888, pruned_loss=0.05829, over 3189607.95 frames. ], batch size: 165, lr: 7.61e-03, grad_scale: 16.0 2023-04-29 01:31:32,150 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.1990, 3.3237, 3.5839, 3.5329, 3.5277, 3.3411, 3.3467, 3.4197], device='cuda:1'), covar=tensor([0.0343, 0.0651, 0.0427, 0.0522, 0.0509, 0.0399, 0.0845, 0.0441], device='cuda:1'), in_proj_covar=tensor([0.0302, 0.0303, 0.0311, 0.0299, 0.0353, 0.0324, 0.0430, 0.0262], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003], device='cuda:1') 2023-04-29 01:31:36,839 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=86307.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 01:31:45,607 INFO [zipformer.py:625] (1/8) 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] (1/8) 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 01:32:08,054 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=86329.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 01:32:12,175 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.6707, 4.6475, 4.4623, 4.2172, 4.0415, 4.5722, 4.3963, 4.2505], device='cuda:1'), covar=tensor([0.0475, 0.0283, 0.0273, 0.0239, 0.0939, 0.0288, 0.0394, 0.0566], device='cuda:1'), in_proj_covar=tensor([0.0223, 0.0272, 0.0266, 0.0239, 0.0288, 0.0273, 0.0181, 0.0305], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-29 01:32:41,540 INFO [train.py:904] (1/8) Epoch 9, batch 5150, loss[loss=0.1909, simple_loss=0.2795, pruned_loss=0.05114, over 16869.00 frames. ], tot_loss[loss=0.2018, simple_loss=0.2885, pruned_loss=0.05754, over 3188907.63 frames. ], batch size: 116, lr: 7.61e-03, grad_scale: 16.0 2023-04-29 01:32:50,318 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=86358.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 01:33:54,430 INFO [train.py:904] (1/8) Epoch 9, batch 5200, loss[loss=0.1901, simple_loss=0.2707, pruned_loss=0.05476, over 16530.00 frames. ], tot_loss[loss=0.1998, simple_loss=0.2864, pruned_loss=0.05657, over 3212039.48 frames. ], batch size: 62, lr: 7.61e-03, grad_scale: 16.0 2023-04-29 01:34:02,835 INFO [zipformer.py:625] (1/8) 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] (1/8) Clipping_scale=2.0, grad-norm quartiles 1.533e+02 2.298e+02 2.670e+02 3.115e+02 5.719e+02, threshold=5.340e+02, percent-clipped=1.0 2023-04-29 01:34:27,691 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=86425.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 01:34:50,092 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=86440.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 01:34:57,907 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 9, batch 5250, loss[loss=0.21, simple_loss=0.3046, pruned_loss=0.05769, over 16753.00 frames. ], tot_loss[loss=0.1978, simple_loss=0.2839, pruned_loss=0.05584, over 3223463.11 frames. ], batch size: 76, lr: 7.61e-03, grad_scale: 16.0 2023-04-29 01:35:57,876 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.7155, 1.7120, 2.2538, 2.8199, 2.6989, 3.0337, 1.8735, 2.8786], device='cuda:1'), covar=tensor([0.0126, 0.0377, 0.0199, 0.0177, 0.0185, 0.0116, 0.0334, 0.0081], device='cuda:1'), in_proj_covar=tensor([0.0149, 0.0162, 0.0146, 0.0150, 0.0156, 0.0112, 0.0162, 0.0104], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001], device='cuda:1') 2023-04-29 01:36:17,589 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 9, batch 5300, loss[loss=0.1732, simple_loss=0.2559, pruned_loss=0.04524, over 16467.00 frames. ], tot_loss[loss=0.1945, simple_loss=0.2801, pruned_loss=0.05444, over 3218824.37 frames. ], batch size: 62, lr: 7.60e-03, grad_scale: 16.0 2023-04-29 01:36:25,609 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=86506.0, num_to_drop=1, layers_to_drop={0} 2023-04-29 01:36:42,024 INFO [optim.py:368] (1/8) 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 01:37:01,848 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=86531.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 01:37:32,903 INFO [train.py:904] (1/8) Epoch 9, batch 5350, loss[loss=0.2044, simple_loss=0.2841, pruned_loss=0.06236, over 12152.00 frames. ], tot_loss[loss=0.1933, simple_loss=0.2787, pruned_loss=0.05395, over 3208361.83 frames. ], batch size: 248, lr: 7.60e-03, grad_scale: 16.0 2023-04-29 01:38:00,983 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.9052, 4.2001, 4.0062, 4.0681, 3.7128, 3.7757, 3.8786, 4.1863], device='cuda:1'), covar=tensor([0.0914, 0.0795, 0.0890, 0.0552, 0.0670, 0.1525, 0.0710, 0.0812], device='cuda:1'), in_proj_covar=tensor([0.0486, 0.0613, 0.0510, 0.0414, 0.0383, 0.0394, 0.0506, 0.0460], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:1') 2023-04-29 01:38:11,779 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=86579.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 01:38:13,768 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=86580.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 01:38:23,169 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=86586.0, num_to_drop=1, layers_to_drop={0} 2023-04-29 01:38:45,877 INFO [train.py:904] (1/8) Epoch 9, batch 5400, loss[loss=0.1834, simple_loss=0.28, pruned_loss=0.0434, over 16920.00 frames. ], tot_loss[loss=0.1954, simple_loss=0.2813, pruned_loss=0.05474, over 3203603.73 frames. ], batch size: 96, lr: 7.60e-03, grad_scale: 16.0 2023-04-29 01:38:46,250 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=86602.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 01:39:01,678 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=86613.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 01:39:09,033 INFO [optim.py:368] (1/8) 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 01:39:16,527 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([1.5968, 2.7096, 2.4093, 4.4854, 3.1471, 4.0535, 1.3784, 2.9113], device='cuda:1'), covar=tensor([0.1304, 0.0674, 0.1172, 0.0105, 0.0243, 0.0352, 0.1530, 0.0813], device='cuda:1'), in_proj_covar=tensor([0.0151, 0.0154, 0.0176, 0.0127, 0.0199, 0.0206, 0.0174, 0.0177], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0003, 0.0004], device='cuda:1') 2023-04-29 01:39:31,534 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=86634.0, num_to_drop=1, layers_to_drop={1} 2023-04-29 01:39:41,854 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=86641.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 01:40:00,476 INFO [train.py:904] (1/8) Epoch 9, batch 5450, loss[loss=0.2349, simple_loss=0.3068, pruned_loss=0.08149, over 12152.00 frames. ], tot_loss[loss=0.198, simple_loss=0.2839, pruned_loss=0.0561, over 3179905.93 frames. ], batch size: 248, lr: 7.60e-03, grad_scale: 16.0 2023-04-29 01:40:10,953 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=86658.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 01:40:15,342 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=86661.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 01:41:18,953 INFO [train.py:904] (1/8) Epoch 9, batch 5500, loss[loss=0.2813, simple_loss=0.3407, pruned_loss=0.1109, over 11854.00 frames. ], tot_loss[loss=0.2075, simple_loss=0.2923, pruned_loss=0.06137, over 3166575.29 frames. ], batch size: 246, lr: 7.60e-03, grad_scale: 16.0 2023-04-29 01:41:24,318 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=86706.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 01:41:27,669 INFO [zipformer.py:625] (1/8) 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] (1/8) 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 01:41:54,494 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=86725.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 01:42:26,694 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.96 vs. limit=2.0 2023-04-29 01:42:36,877 INFO [train.py:904] (1/8) Epoch 9, batch 5550, loss[loss=0.3394, simple_loss=0.3802, pruned_loss=0.1493, over 11700.00 frames. ], tot_loss[loss=0.2181, simple_loss=0.3011, pruned_loss=0.06755, over 3154454.49 frames. ], batch size: 248, lr: 7.59e-03, grad_scale: 16.0 2023-04-29 01:42:43,064 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=86756.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 01:43:09,702 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=86773.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 01:43:44,650 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=86796.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 01:43:53,945 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=86801.0, num_to_drop=1, layers_to_drop={0} 2023-04-29 01:43:54,665 INFO [train.py:904] (1/8) Epoch 9, batch 5600, loss[loss=0.2272, simple_loss=0.3097, pruned_loss=0.07234, over 17023.00 frames. ], tot_loss[loss=0.2273, simple_loss=0.3076, pruned_loss=0.07349, over 3110411.09 frames. ], batch size: 50, lr: 7.59e-03, grad_scale: 8.0 2023-04-29 01:44:10,280 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.2335, 1.9376, 2.5280, 3.1738, 2.8911, 3.4867, 1.9597, 3.3835], device='cuda:1'), covar=tensor([0.0094, 0.0288, 0.0187, 0.0124, 0.0149, 0.0067, 0.0312, 0.0056], device='cuda:1'), in_proj_covar=tensor([0.0150, 0.0163, 0.0145, 0.0149, 0.0155, 0.0112, 0.0164, 0.0104], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001], device='cuda:1') 2023-04-29 01:44:14,025 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=86814.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 01:44:21,510 INFO [optim.py:368] (1/8) 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 01:45:17,291 INFO [train.py:904] (1/8) Epoch 9, batch 5650, loss[loss=0.2973, simple_loss=0.3574, pruned_loss=0.1186, over 15347.00 frames. ], tot_loss[loss=0.238, simple_loss=0.315, pruned_loss=0.08053, over 3046686.03 frames. ], batch size: 191, lr: 7.59e-03, grad_scale: 4.0 2023-04-29 01:45:36,404 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.42 vs. limit=2.0 2023-04-29 01:45:53,448 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=86875.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 01:45:58,527 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=86878.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 01:46:33,109 INFO [train.py:904] (1/8) Epoch 9, batch 5700, loss[loss=0.2035, simple_loss=0.2972, pruned_loss=0.05491, over 16332.00 frames. ], tot_loss[loss=0.2407, simple_loss=0.317, pruned_loss=0.08217, over 3043247.01 frames. ], batch size: 165, lr: 7.59e-03, grad_scale: 4.0 2023-04-29 01:46:33,460 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=86902.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 01:46:53,918 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.6270, 3.6461, 2.8982, 2.1808, 2.7681, 2.2499, 3.8430, 3.6068], device='cuda:1'), covar=tensor([0.2445, 0.0782, 0.1473, 0.1944, 0.1943, 0.1569, 0.0437, 0.0763], device='cuda:1'), in_proj_covar=tensor([0.0293, 0.0252, 0.0276, 0.0269, 0.0277, 0.0212, 0.0259, 0.0277], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-29 01:46:59,879 INFO [optim.py:368] (1/8) Clipping_scale=2.0, grad-norm quartiles 2.388e+02 3.612e+02 4.435e+02 5.746e+02 9.391e+02, threshold=8.870e+02, percent-clipped=0.0 2023-04-29 01:47:25,192 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=86936.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 01:47:30,354 INFO [zipformer.py:625] (1/8) 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] (1/8) 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] (1/8) Epoch 9, batch 5750, loss[loss=0.2457, simple_loss=0.3186, pruned_loss=0.08643, over 17032.00 frames. ], tot_loss[loss=0.2437, simple_loss=0.32, pruned_loss=0.08374, over 3024670.18 frames. ], batch size: 55, lr: 7.58e-03, grad_scale: 4.0 2023-04-29 01:48:03,845 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.2730, 4.2611, 4.4623, 2.6391, 4.9994, 4.9574, 3.3929, 3.8075], device='cuda:1'), covar=tensor([0.0577, 0.0145, 0.0189, 0.0933, 0.0031, 0.0067, 0.0329, 0.0321], device='cuda:1'), in_proj_covar=tensor([0.0140, 0.0097, 0.0084, 0.0138, 0.0066, 0.0095, 0.0119, 0.0127], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0003, 0.0002, 0.0004, 0.0002, 0.0002, 0.0003, 0.0003], device='cuda:1') 2023-04-29 01:49:03,009 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.8399, 4.2402, 3.1223, 2.3271, 3.2589, 2.4893, 4.4270, 4.0978], device='cuda:1'), covar=tensor([0.2537, 0.0673, 0.1506, 0.2085, 0.2039, 0.1627, 0.0447, 0.0708], device='cuda:1'), in_proj_covar=tensor([0.0296, 0.0253, 0.0277, 0.0270, 0.0279, 0.0214, 0.0262, 0.0278], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-29 01:49:12,515 INFO [train.py:904] (1/8) Epoch 9, batch 5800, loss[loss=0.2028, simple_loss=0.2968, pruned_loss=0.0544, over 16264.00 frames. ], tot_loss[loss=0.2411, simple_loss=0.3185, pruned_loss=0.08186, over 3022255.08 frames. ], batch size: 165, lr: 7.58e-03, grad_scale: 4.0 2023-04-29 01:49:13,789 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=87002.0, num_to_drop=1, layers_to_drop={0} 2023-04-29 01:49:40,720 INFO [optim.py:368] (1/8) 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 01:50:04,230 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.5133, 4.7695, 4.8635, 4.7206, 4.7249, 5.2966, 4.7786, 4.5215], device='cuda:1'), covar=tensor([0.1092, 0.1494, 0.1517, 0.1920, 0.2498, 0.0903, 0.1326, 0.2374], device='cuda:1'), in_proj_covar=tensor([0.0329, 0.0447, 0.0475, 0.0392, 0.0523, 0.0503, 0.0385, 0.0531], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-04-29 01:50:07,367 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.0085, 3.7548, 3.7731, 2.2929, 3.4388, 3.7395, 3.5640, 1.9198], device='cuda:1'), covar=tensor([0.0406, 0.0031, 0.0037, 0.0299, 0.0059, 0.0067, 0.0044, 0.0351], device='cuda:1'), in_proj_covar=tensor([0.0124, 0.0065, 0.0065, 0.0122, 0.0073, 0.0083, 0.0071, 0.0116], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:1') 2023-04-29 01:50:30,439 INFO [train.py:904] (1/8) Epoch 9, batch 5850, loss[loss=0.2298, simple_loss=0.3014, pruned_loss=0.07911, over 16646.00 frames. ], tot_loss[loss=0.2372, simple_loss=0.3152, pruned_loss=0.07959, over 3037350.32 frames. ], batch size: 57, lr: 7.58e-03, grad_scale: 4.0 2023-04-29 01:50:47,472 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=87063.0, num_to_drop=1, layers_to_drop={1} 2023-04-29 01:51:42,183 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=87096.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 01:51:50,008 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=87101.0, num_to_drop=1, layers_to_drop={1} 2023-04-29 01:51:50,871 INFO [train.py:904] (1/8) Epoch 9, batch 5900, loss[loss=0.2149, simple_loss=0.2957, pruned_loss=0.06703, over 16427.00 frames. ], tot_loss[loss=0.2361, simple_loss=0.3144, pruned_loss=0.07892, over 3047877.00 frames. ], batch size: 146, lr: 7.58e-03, grad_scale: 4.0 2023-04-29 01:52:22,906 INFO [optim.py:368] (1/8) 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 01:53:01,034 INFO [zipformer.py:625] (1/8) 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] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=87149.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 01:53:13,004 INFO [train.py:904] (1/8) Epoch 9, batch 5950, loss[loss=0.2578, simple_loss=0.3361, pruned_loss=0.08976, over 15204.00 frames. ], tot_loss[loss=0.235, simple_loss=0.3155, pruned_loss=0.07724, over 3064929.10 frames. ], batch size: 190, lr: 7.58e-03, grad_scale: 4.0 2023-04-29 01:53:21,250 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.47 vs. limit=2.0 2023-04-29 01:53:42,054 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=87170.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 01:54:31,279 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.19 vs. limit=2.0 2023-04-29 01:54:33,095 INFO [train.py:904] (1/8) Epoch 9, batch 6000, loss[loss=0.2334, simple_loss=0.3189, pruned_loss=0.07399, over 16754.00 frames. ], tot_loss[loss=0.234, simple_loss=0.3146, pruned_loss=0.07676, over 3056886.08 frames. ], batch size: 124, lr: 7.57e-03, grad_scale: 8.0 2023-04-29 01:54:33,095 INFO [train.py:929] (1/8) Computing validation loss 2023-04-29 01:54:44,301 INFO [train.py:938] (1/8) Epoch 9, validation: loss=0.1674, simple_loss=0.2809, pruned_loss=0.02692, over 944034.00 frames. 2023-04-29 01:54:44,302 INFO [train.py:939] (1/8) Maximum memory allocated so far is 17676MB 2023-04-29 01:55:11,800 INFO [optim.py:368] (1/8) 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 01:55:13,201 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.3131, 3.2873, 3.3348, 3.4735, 3.5107, 3.2010, 3.4716, 3.5463], device='cuda:1'), covar=tensor([0.1034, 0.0769, 0.0936, 0.0498, 0.0534, 0.2216, 0.0886, 0.0593], device='cuda:1'), in_proj_covar=tensor([0.0480, 0.0591, 0.0722, 0.0600, 0.0457, 0.0452, 0.0476, 0.0524], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-29 01:55:28,016 INFO [zipformer.py:625] (1/8) 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] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=87234.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 01:55:37,525 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=87236.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 01:56:03,542 INFO [train.py:904] (1/8) Epoch 9, batch 6050, loss[loss=0.2314, simple_loss=0.3098, pruned_loss=0.07651, over 15159.00 frames. ], tot_loss[loss=0.2329, simple_loss=0.3131, pruned_loss=0.07638, over 3064150.00 frames. ], batch size: 190, lr: 7.57e-03, grad_scale: 8.0 2023-04-29 01:56:52,086 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=87284.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 01:57:06,974 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 9, batch 6100, loss[loss=0.2199, simple_loss=0.3064, pruned_loss=0.06665, over 16341.00 frames. ], tot_loss[loss=0.2309, simple_loss=0.3119, pruned_loss=0.07489, over 3071998.75 frames. ], batch size: 146, lr: 7.57e-03, grad_scale: 8.0 2023-04-29 01:57:42,386 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.8825, 4.8467, 4.6787, 4.4787, 4.2777, 4.7281, 4.6795, 4.4062], device='cuda:1'), covar=tensor([0.0507, 0.0431, 0.0261, 0.0235, 0.0991, 0.0478, 0.0299, 0.0653], device='cuda:1'), in_proj_covar=tensor([0.0228, 0.0275, 0.0265, 0.0241, 0.0288, 0.0279, 0.0182, 0.0311], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-29 01:57:52,500 INFO [optim.py:368] (1/8) 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 01:58:42,518 INFO [train.py:904] (1/8) Epoch 9, batch 6150, loss[loss=0.2237, simple_loss=0.3104, pruned_loss=0.0685, over 16200.00 frames. ], tot_loss[loss=0.2303, simple_loss=0.3106, pruned_loss=0.07497, over 3052339.32 frames. ], batch size: 165, lr: 7.57e-03, grad_scale: 8.0 2023-04-29 01:58:52,888 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=87358.0, num_to_drop=1, layers_to_drop={3} 2023-04-29 01:58:56,272 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.26 vs. limit=2.0 2023-04-29 01:58:58,720 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=87362.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 01:59:38,441 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.4445, 3.2103, 2.6423, 2.1863, 2.3074, 2.0818, 3.2848, 3.1160], device='cuda:1'), covar=tensor([0.2382, 0.0698, 0.1465, 0.1885, 0.1807, 0.1659, 0.0454, 0.0840], device='cuda:1'), in_proj_covar=tensor([0.0300, 0.0256, 0.0283, 0.0274, 0.0285, 0.0216, 0.0266, 0.0281], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-29 02:00:00,630 INFO [train.py:904] (1/8) Epoch 9, batch 6200, loss[loss=0.2249, simple_loss=0.3033, pruned_loss=0.07318, over 16259.00 frames. ], tot_loss[loss=0.2279, simple_loss=0.3081, pruned_loss=0.07388, over 3076174.58 frames. ], batch size: 165, lr: 7.57e-03, grad_scale: 8.0 2023-04-29 02:00:28,300 INFO [optim.py:368] (1/8) 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 02:00:34,946 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=87423.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 02:01:18,824 INFO [train.py:904] (1/8) Epoch 9, batch 6250, loss[loss=0.2074, simple_loss=0.2992, pruned_loss=0.05773, over 16612.00 frames. ], tot_loss[loss=0.2266, simple_loss=0.3073, pruned_loss=0.07293, over 3097153.47 frames. ], batch size: 62, lr: 7.56e-03, grad_scale: 4.0 2023-04-29 02:01:32,125 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=87461.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 02:01:45,457 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=87470.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 02:02:12,707 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([1.7523, 1.6495, 1.5081, 1.5454, 1.8464, 1.6546, 1.7313, 1.9319], device='cuda:1'), covar=tensor([0.0108, 0.0183, 0.0263, 0.0226, 0.0123, 0.0175, 0.0125, 0.0126], device='cuda:1'), in_proj_covar=tensor([0.0124, 0.0189, 0.0187, 0.0186, 0.0187, 0.0190, 0.0189, 0.0177], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-29 02:02:33,526 INFO [train.py:904] (1/8) Epoch 9, batch 6300, loss[loss=0.2405, simple_loss=0.3188, pruned_loss=0.08116, over 16457.00 frames. ], tot_loss[loss=0.2272, simple_loss=0.308, pruned_loss=0.07316, over 3089408.09 frames. ], batch size: 68, lr: 7.56e-03, grad_scale: 4.0 2023-04-29 02:02:38,941 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=87505.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 02:02:59,940 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=87518.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 02:03:04,115 INFO [optim.py:368] (1/8) 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 02:03:06,301 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=87522.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 02:03:14,552 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.35 vs. limit=2.0 2023-04-29 02:03:24,757 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=87534.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 02:03:51,225 INFO [train.py:904] (1/8) Epoch 9, batch 6350, loss[loss=0.2191, simple_loss=0.296, pruned_loss=0.0711, over 16637.00 frames. ], tot_loss[loss=0.2292, simple_loss=0.3089, pruned_loss=0.07481, over 3072031.73 frames. ], batch size: 62, lr: 7.56e-03, grad_scale: 4.0 2023-04-29 02:04:12,979 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=87566.0, num_to_drop=1, layers_to_drop={3} 2023-04-29 02:04:37,474 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=87582.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 02:04:40,648 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=87584.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 02:04:44,418 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=87586.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 02:04:49,441 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=87589.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 02:05:08,847 INFO [train.py:904] (1/8) Epoch 9, batch 6400, loss[loss=0.2062, simple_loss=0.2819, pruned_loss=0.06528, over 16584.00 frames. ], tot_loss[loss=0.2315, simple_loss=0.3099, pruned_loss=0.07657, over 3052541.03 frames. ], batch size: 62, lr: 7.56e-03, grad_scale: 8.0 2023-04-29 02:05:13,454 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=87605.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 02:05:32,116 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.0058, 3.5296, 3.1939, 1.9135, 2.7617, 2.3117, 3.4514, 3.5152], device='cuda:1'), covar=tensor([0.0231, 0.0516, 0.0600, 0.1754, 0.0813, 0.0871, 0.0552, 0.0692], device='cuda:1'), in_proj_covar=tensor([0.0140, 0.0138, 0.0156, 0.0141, 0.0134, 0.0124, 0.0135, 0.0149], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-04-29 02:05:37,569 INFO [optim.py:368] (1/8) 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 02:06:04,169 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.3729, 3.4767, 1.9561, 3.8371, 2.5130, 3.7576, 1.9727, 2.7907], device='cuda:1'), covar=tensor([0.0253, 0.0415, 0.1687, 0.0124, 0.0822, 0.0655, 0.1598, 0.0718], device='cuda:1'), in_proj_covar=tensor([0.0140, 0.0160, 0.0184, 0.0111, 0.0162, 0.0199, 0.0191, 0.0167], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-04-29 02:06:14,974 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=87645.0, num_to_drop=1, layers_to_drop={0} 2023-04-29 02:06:21,608 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 9, batch 6450, loss[loss=0.2468, simple_loss=0.3049, pruned_loss=0.09433, over 11370.00 frames. ], tot_loss[loss=0.23, simple_loss=0.3096, pruned_loss=0.07519, over 3056931.15 frames. ], batch size: 247, lr: 7.55e-03, grad_scale: 4.0 2023-04-29 02:06:34,297 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=87658.0, num_to_drop=1, layers_to_drop={0} 2023-04-29 02:06:46,385 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=87666.0, num_to_drop=1, layers_to_drop={0} 2023-04-29 02:07:04,919 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.4189, 3.4098, 2.6737, 2.1792, 2.3674, 2.2164, 3.5318, 3.3262], device='cuda:1'), covar=tensor([0.2613, 0.0809, 0.1540, 0.2050, 0.2178, 0.1715, 0.0446, 0.0902], device='cuda:1'), in_proj_covar=tensor([0.0297, 0.0253, 0.0280, 0.0272, 0.0282, 0.0213, 0.0263, 0.0280], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-29 02:07:41,886 INFO [train.py:904] (1/8) Epoch 9, batch 6500, loss[loss=0.2038, simple_loss=0.2814, pruned_loss=0.06307, over 16949.00 frames. ], tot_loss[loss=0.2282, simple_loss=0.3072, pruned_loss=0.07457, over 3056894.49 frames. ], batch size: 41, lr: 7.55e-03, grad_scale: 2.0 2023-04-29 02:07:48,426 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=87706.0, num_to_drop=1, layers_to_drop={1} 2023-04-29 02:08:06,442 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=87718.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 02:08:13,251 INFO [optim.py:368] (1/8) Clipping_scale=2.0, grad-norm quartiles 2.348e+02 3.236e+02 4.021e+02 5.179e+02 1.078e+03, threshold=8.043e+02, percent-clipped=2.0 2023-04-29 02:09:00,842 INFO [train.py:904] (1/8) Epoch 9, batch 6550, loss[loss=0.22, simple_loss=0.3184, pruned_loss=0.06081, over 16637.00 frames. ], tot_loss[loss=0.2297, simple_loss=0.3099, pruned_loss=0.0747, over 3085018.00 frames. ], batch size: 68, lr: 7.55e-03, grad_scale: 2.0 2023-04-29 02:09:06,370 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([1.6620, 2.5299, 2.3242, 3.4107, 2.3059, 3.6724, 1.3045, 2.7957], device='cuda:1'), covar=tensor([0.1369, 0.0602, 0.1110, 0.0150, 0.0203, 0.0434, 0.1630, 0.0703], device='cuda:1'), in_proj_covar=tensor([0.0149, 0.0152, 0.0177, 0.0127, 0.0201, 0.0205, 0.0175, 0.0173], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0003, 0.0004], device='cuda:1') 2023-04-29 02:09:24,302 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.57 vs. limit=2.0 2023-04-29 02:09:26,128 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.53 vs. limit=2.0 2023-04-29 02:09:56,060 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.8816, 4.6919, 4.6265, 2.9238, 3.9256, 4.5177, 4.1377, 2.3903], device='cuda:1'), covar=tensor([0.0330, 0.0016, 0.0019, 0.0273, 0.0060, 0.0061, 0.0036, 0.0324], device='cuda:1'), in_proj_covar=tensor([0.0122, 0.0063, 0.0064, 0.0120, 0.0070, 0.0082, 0.0071, 0.0113], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:1') 2023-04-29 02:10:18,520 INFO [train.py:904] (1/8) Epoch 9, batch 6600, loss[loss=0.2197, simple_loss=0.3078, pruned_loss=0.06575, over 16718.00 frames. ], tot_loss[loss=0.2308, simple_loss=0.3118, pruned_loss=0.07485, over 3095232.03 frames. ], batch size: 57, lr: 7.55e-03, grad_scale: 2.0 2023-04-29 02:10:20,704 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([5.1223, 5.8537, 6.0984, 5.7807, 5.8505, 6.3232, 5.8935, 5.7589], device='cuda:1'), covar=tensor([0.0776, 0.1569, 0.1524, 0.1663, 0.2239, 0.0842, 0.1353, 0.2095], device='cuda:1'), in_proj_covar=tensor([0.0325, 0.0448, 0.0479, 0.0398, 0.0522, 0.0504, 0.0391, 0.0536], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-04-29 02:10:42,559 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=87817.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 02:10:51,646 INFO [optim.py:368] (1/8) Clipping_scale=2.0, grad-norm quartiles 2.077e+02 3.184e+02 4.033e+02 5.145e+02 1.254e+03, threshold=8.065e+02, percent-clipped=5.0 2023-04-29 02:11:31,463 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.2819, 1.9482, 1.9782, 3.7574, 1.8983, 2.4378, 2.0923, 2.0788], device='cuda:1'), covar=tensor([0.0810, 0.2998, 0.2130, 0.0409, 0.3524, 0.1971, 0.2795, 0.2903], device='cuda:1'), in_proj_covar=tensor([0.0352, 0.0374, 0.0314, 0.0320, 0.0406, 0.0426, 0.0338, 0.0443], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-29 02:11:36,975 INFO [train.py:904] (1/8) Epoch 9, batch 6650, loss[loss=0.2206, simple_loss=0.3019, pruned_loss=0.06969, over 16821.00 frames. ], tot_loss[loss=0.2309, simple_loss=0.3116, pruned_loss=0.07506, over 3108052.38 frames. ], batch size: 116, lr: 7.55e-03, grad_scale: 2.0 2023-04-29 02:11:50,958 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=87861.0, num_to_drop=1, layers_to_drop={0} 2023-04-29 02:12:28,578 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=87886.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 02:12:53,897 INFO [train.py:904] (1/8) Epoch 9, batch 6700, loss[loss=0.2148, simple_loss=0.2998, pruned_loss=0.06493, over 16837.00 frames. ], tot_loss[loss=0.2304, simple_loss=0.3107, pruned_loss=0.07501, over 3105590.49 frames. ], batch size: 116, lr: 7.54e-03, grad_scale: 2.0 2023-04-29 02:13:26,700 INFO [optim.py:368] (1/8) 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 02:13:35,835 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([5.6933, 5.9670, 5.6304, 5.6774, 5.3132, 5.1456, 5.3985, 6.0450], device='cuda:1'), covar=tensor([0.0888, 0.0759, 0.1009, 0.0603, 0.0741, 0.0659, 0.0978, 0.0780], device='cuda:1'), in_proj_covar=tensor([0.0498, 0.0626, 0.0524, 0.0428, 0.0389, 0.0407, 0.0522, 0.0469], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:1') 2023-04-29 02:13:43,771 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=87934.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 02:13:52,493 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=87940.0, num_to_drop=1, layers_to_drop={0} 2023-04-29 02:14:00,199 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 9, batch 6750, loss[loss=0.2452, simple_loss=0.3225, pruned_loss=0.08393, over 15437.00 frames. ], tot_loss[loss=0.2286, simple_loss=0.3086, pruned_loss=0.07431, over 3125968.87 frames. ], batch size: 191, lr: 7.54e-03, grad_scale: 2.0 2023-04-29 02:14:24,278 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=87961.0, num_to_drop=1, layers_to_drop={1} 2023-04-29 02:15:29,296 INFO [train.py:904] (1/8) Epoch 9, batch 6800, loss[loss=0.2288, simple_loss=0.3089, pruned_loss=0.07429, over 16417.00 frames. ], tot_loss[loss=0.2295, simple_loss=0.3094, pruned_loss=0.07479, over 3108746.47 frames. ], batch size: 35, lr: 7.54e-03, grad_scale: 4.0 2023-04-29 02:15:54,197 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=88018.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 02:16:02,235 INFO [optim.py:368] (1/8) 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 02:16:45,306 INFO [train.py:904] (1/8) Epoch 9, batch 6850, loss[loss=0.2782, simple_loss=0.338, pruned_loss=0.1092, over 11283.00 frames. ], tot_loss[loss=0.2305, simple_loss=0.3103, pruned_loss=0.07538, over 3102694.03 frames. ], batch size: 247, lr: 7.54e-03, grad_scale: 4.0 2023-04-29 02:17:06,648 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=88066.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 02:17:39,450 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=3.07 vs. limit=5.0 2023-04-29 02:17:59,065 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=88100.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 02:18:00,999 INFO [train.py:904] (1/8) Epoch 9, batch 6900, loss[loss=0.3139, simple_loss=0.3649, pruned_loss=0.1314, over 11403.00 frames. ], tot_loss[loss=0.2313, simple_loss=0.3128, pruned_loss=0.07491, over 3106896.24 frames. ], batch size: 247, lr: 7.54e-03, grad_scale: 4.0 2023-04-29 02:18:07,123 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.9155, 4.1499, 3.9277, 3.9416, 3.6860, 3.8333, 3.8513, 4.1257], device='cuda:1'), covar=tensor([0.0980, 0.0887, 0.1022, 0.0673, 0.0741, 0.1305, 0.0871, 0.0975], device='cuda:1'), in_proj_covar=tensor([0.0492, 0.0623, 0.0522, 0.0423, 0.0384, 0.0405, 0.0517, 0.0464], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:1') 2023-04-29 02:18:24,650 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=88117.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 02:18:33,104 INFO [optim.py:368] (1/8) 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 02:18:52,805 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.21 vs. limit=2.0 2023-04-29 02:19:18,075 INFO [train.py:904] (1/8) Epoch 9, batch 6950, loss[loss=0.221, simple_loss=0.2985, pruned_loss=0.07176, over 16663.00 frames. ], tot_loss[loss=0.234, simple_loss=0.3146, pruned_loss=0.07667, over 3107373.53 frames. ], batch size: 57, lr: 7.53e-03, grad_scale: 4.0 2023-04-29 02:19:32,124 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=88161.0, num_to_drop=1, layers_to_drop={2} 2023-04-29 02:19:32,205 INFO [zipformer.py:625] (1/8) 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] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=88165.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 02:20:15,221 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.1345, 3.2534, 1.6505, 3.4523, 2.2936, 3.4303, 1.8100, 2.5548], device='cuda:1'), covar=tensor([0.0255, 0.0351, 0.1762, 0.0149, 0.0940, 0.0547, 0.1673, 0.0756], device='cuda:1'), in_proj_covar=tensor([0.0142, 0.0161, 0.0186, 0.0112, 0.0166, 0.0202, 0.0195, 0.0170], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-04-29 02:20:23,550 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.90 vs. limit=2.0 2023-04-29 02:20:31,302 INFO [train.py:904] (1/8) Epoch 9, batch 7000, loss[loss=0.2083, simple_loss=0.301, pruned_loss=0.05781, over 17031.00 frames. ], tot_loss[loss=0.2332, simple_loss=0.3145, pruned_loss=0.07597, over 3109269.58 frames. ], batch size: 55, lr: 7.53e-03, grad_scale: 4.0 2023-04-29 02:20:43,255 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=88209.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 02:21:03,406 INFO [optim.py:368] (1/8) 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 02:21:29,013 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=88240.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 02:21:36,021 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 9, batch 7050, loss[loss=0.2376, simple_loss=0.3155, pruned_loss=0.07983, over 16687.00 frames. ], tot_loss[loss=0.2334, simple_loss=0.3151, pruned_loss=0.07581, over 3103551.51 frames. ], batch size: 76, lr: 7.53e-03, grad_scale: 4.0 2023-04-29 02:21:52,627 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.34 vs. limit=2.0 2023-04-29 02:22:00,338 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=88261.0, num_to_drop=1, layers_to_drop={0} 2023-04-29 02:22:40,451 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=88288.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 02:22:45,438 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=88291.0, num_to_drop=1, layers_to_drop={0} 2023-04-29 02:22:47,885 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=88293.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 02:23:02,082 INFO [train.py:904] (1/8) Epoch 9, batch 7100, loss[loss=0.2591, simple_loss=0.3126, pruned_loss=0.1028, over 11510.00 frames. ], tot_loss[loss=0.2319, simple_loss=0.3131, pruned_loss=0.07529, over 3097360.10 frames. ], batch size: 246, lr: 7.53e-03, grad_scale: 4.0 2023-04-29 02:23:12,558 INFO [zipformer.py:625] (1/8) 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] (1/8) 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 02:24:15,977 INFO [train.py:904] (1/8) Epoch 9, batch 7150, loss[loss=0.2013, simple_loss=0.2936, pruned_loss=0.05453, over 16537.00 frames. ], tot_loss[loss=0.2312, simple_loss=0.3119, pruned_loss=0.07526, over 3096153.70 frames. ], batch size: 68, lr: 7.52e-03, grad_scale: 4.0 2023-04-29 02:24:17,086 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=88352.0, num_to_drop=1, layers_to_drop={0} 2023-04-29 02:25:00,814 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.6169, 2.5975, 1.9004, 2.7538, 2.1658, 2.7158, 2.0255, 2.3409], device='cuda:1'), covar=tensor([0.0227, 0.0359, 0.1111, 0.0125, 0.0595, 0.0441, 0.1048, 0.0519], device='cuda:1'), in_proj_covar=tensor([0.0141, 0.0161, 0.0186, 0.0112, 0.0165, 0.0201, 0.0193, 0.0170], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-04-29 02:25:28,099 INFO [train.py:904] (1/8) Epoch 9, batch 7200, loss[loss=0.1861, simple_loss=0.2707, pruned_loss=0.05076, over 17225.00 frames. ], tot_loss[loss=0.2277, simple_loss=0.3092, pruned_loss=0.07308, over 3093896.69 frames. ], batch size: 45, lr: 7.52e-03, grad_scale: 8.0 2023-04-29 02:26:00,148 INFO [optim.py:368] (1/8) 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] (1/8) Epoch 9, batch 7250, loss[loss=0.2397, simple_loss=0.3082, pruned_loss=0.08561, over 11379.00 frames. ], tot_loss[loss=0.2254, simple_loss=0.3069, pruned_loss=0.07197, over 3070133.16 frames. ], batch size: 248, lr: 7.52e-03, grad_scale: 8.0 2023-04-29 02:26:53,242 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 9, batch 7300, loss[loss=0.2179, simple_loss=0.31, pruned_loss=0.0629, over 16757.00 frames. ], tot_loss[loss=0.2238, simple_loss=0.3055, pruned_loss=0.0711, over 3096274.40 frames. ], batch size: 102, lr: 7.52e-03, grad_scale: 8.0 2023-04-29 02:28:33,453 INFO [optim.py:368] (1/8) 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 02:28:50,690 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.2795, 3.2369, 3.2657, 3.3999, 3.4223, 3.1611, 3.3990, 3.4674], device='cuda:1'), covar=tensor([0.0911, 0.0751, 0.0832, 0.0442, 0.0589, 0.2363, 0.0825, 0.0621], device='cuda:1'), in_proj_covar=tensor([0.0474, 0.0585, 0.0723, 0.0587, 0.0455, 0.0453, 0.0474, 0.0524], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-29 02:29:01,023 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.7588, 3.6185, 3.8390, 3.6139, 3.7525, 4.1510, 3.8670, 3.6067], device='cuda:1'), covar=tensor([0.1828, 0.2255, 0.1716, 0.2370, 0.2635, 0.1573, 0.1392, 0.2269], device='cuda:1'), in_proj_covar=tensor([0.0332, 0.0450, 0.0483, 0.0400, 0.0527, 0.0513, 0.0390, 0.0543], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-04-29 02:29:06,120 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.3431, 4.4890, 4.4090, 2.8251, 3.9074, 4.3439, 3.9602, 2.7277], device='cuda:1'), covar=tensor([0.0413, 0.0020, 0.0023, 0.0275, 0.0050, 0.0063, 0.0044, 0.0266], device='cuda:1'), in_proj_covar=tensor([0.0125, 0.0063, 0.0065, 0.0123, 0.0071, 0.0083, 0.0072, 0.0116], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:1') 2023-04-29 02:29:14,123 INFO [train.py:904] (1/8) Epoch 9, batch 7350, loss[loss=0.2012, simple_loss=0.2922, pruned_loss=0.05512, over 16794.00 frames. ], tot_loss[loss=0.2242, simple_loss=0.3054, pruned_loss=0.07151, over 3079595.49 frames. ], batch size: 83, lr: 7.52e-03, grad_scale: 4.0 2023-04-29 02:30:02,895 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.9123, 2.3215, 2.4035, 4.6292, 2.1895, 2.8917, 2.5060, 2.6083], device='cuda:1'), covar=tensor([0.0777, 0.3021, 0.1828, 0.0335, 0.3517, 0.1919, 0.2583, 0.2685], device='cuda:1'), in_proj_covar=tensor([0.0353, 0.0379, 0.0316, 0.0321, 0.0409, 0.0431, 0.0340, 0.0444], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-29 02:30:27,926 INFO [train.py:904] (1/8) Epoch 9, batch 7400, loss[loss=0.2311, simple_loss=0.314, pruned_loss=0.07407, over 16308.00 frames. ], tot_loss[loss=0.2247, simple_loss=0.3064, pruned_loss=0.07149, over 3104873.34 frames. ], batch size: 165, lr: 7.51e-03, grad_scale: 4.0 2023-04-29 02:31:01,751 INFO [optim.py:368] (1/8) 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 02:31:03,609 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.2082, 3.2457, 1.7837, 3.4848, 2.4147, 3.4539, 1.9203, 2.5135], device='cuda:1'), covar=tensor([0.0201, 0.0334, 0.1707, 0.0136, 0.0829, 0.0518, 0.1475, 0.0703], device='cuda:1'), in_proj_covar=tensor([0.0141, 0.0161, 0.0187, 0.0112, 0.0166, 0.0201, 0.0193, 0.0169], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-04-29 02:31:34,130 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.9285, 5.2410, 4.9333, 4.9198, 4.6965, 4.6251, 4.6618, 5.2945], device='cuda:1'), covar=tensor([0.0992, 0.0764, 0.1017, 0.0710, 0.0749, 0.0779, 0.0950, 0.0855], device='cuda:1'), in_proj_covar=tensor([0.0488, 0.0617, 0.0520, 0.0420, 0.0382, 0.0400, 0.0512, 0.0465], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:1') 2023-04-29 02:31:37,865 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=88647.0, num_to_drop=1, layers_to_drop={3} 2023-04-29 02:31:44,621 INFO [train.py:904] (1/8) Epoch 9, batch 7450, loss[loss=0.2055, simple_loss=0.2953, pruned_loss=0.05788, over 16499.00 frames. ], tot_loss[loss=0.2272, simple_loss=0.3079, pruned_loss=0.07331, over 3097609.53 frames. ], batch size: 68, lr: 7.51e-03, grad_scale: 4.0 2023-04-29 02:33:03,515 INFO [train.py:904] (1/8) Epoch 9, batch 7500, loss[loss=0.2731, simple_loss=0.3304, pruned_loss=0.1079, over 11638.00 frames. ], tot_loss[loss=0.2273, simple_loss=0.3085, pruned_loss=0.07301, over 3088144.03 frames. ], batch size: 246, lr: 7.51e-03, grad_scale: 4.0 2023-04-29 02:33:03,987 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([5.2037, 5.8150, 5.9932, 5.7277, 5.7549, 6.2801, 5.8328, 5.6688], device='cuda:1'), covar=tensor([0.0671, 0.1569, 0.1559, 0.1607, 0.2043, 0.0888, 0.1175, 0.2166], device='cuda:1'), in_proj_covar=tensor([0.0326, 0.0447, 0.0480, 0.0396, 0.0524, 0.0510, 0.0387, 0.0536], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-04-29 02:33:36,929 INFO [optim.py:368] (1/8) 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] (1/8) Epoch 9, batch 7550, loss[loss=0.202, simple_loss=0.2856, pruned_loss=0.05925, over 17018.00 frames. ], tot_loss[loss=0.2273, simple_loss=0.3078, pruned_loss=0.07338, over 3079317.40 frames. ], batch size: 55, lr: 7.51e-03, grad_scale: 4.0 2023-04-29 02:34:23,648 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=88756.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 02:34:46,923 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.46 vs. limit=2.0 2023-04-29 02:35:04,591 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.47 vs. limit=2.0 2023-04-29 02:35:33,597 INFO [train.py:904] (1/8) Epoch 9, batch 7600, loss[loss=0.2219, simple_loss=0.2997, pruned_loss=0.07203, over 16917.00 frames. ], tot_loss[loss=0.2277, simple_loss=0.3076, pruned_loss=0.07394, over 3070938.55 frames. ], batch size: 109, lr: 7.51e-03, grad_scale: 8.0 2023-04-29 02:35:37,265 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=88804.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 02:35:49,044 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.17 vs. limit=2.0 2023-04-29 02:36:05,551 INFO [optim.py:368] (1/8) 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 02:36:21,653 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.17 vs. limit=2.0 2023-04-29 02:36:45,917 INFO [train.py:904] (1/8) Epoch 9, batch 7650, loss[loss=0.2222, simple_loss=0.3068, pruned_loss=0.06879, over 16836.00 frames. ], tot_loss[loss=0.2283, simple_loss=0.308, pruned_loss=0.0743, over 3080783.00 frames. ], batch size: 116, lr: 7.50e-03, grad_scale: 8.0 2023-04-29 02:37:05,491 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.46 vs. limit=2.0 2023-04-29 02:38:01,878 INFO [train.py:904] (1/8) Epoch 9, batch 7700, loss[loss=0.2191, simple_loss=0.3046, pruned_loss=0.06682, over 16243.00 frames. ], tot_loss[loss=0.2294, simple_loss=0.3083, pruned_loss=0.07522, over 3059762.45 frames. ], batch size: 165, lr: 7.50e-03, grad_scale: 8.0 2023-04-29 02:38:24,411 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.4732, 3.5149, 3.2232, 3.1474, 3.1312, 3.4145, 3.2161, 3.2592], device='cuda:1'), covar=tensor([0.0530, 0.0472, 0.0244, 0.0214, 0.0624, 0.0352, 0.1127, 0.0442], device='cuda:1'), in_proj_covar=tensor([0.0221, 0.0268, 0.0255, 0.0235, 0.0278, 0.0268, 0.0179, 0.0299], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-29 02:38:34,565 INFO [optim.py:368] (1/8) 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 02:39:08,864 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=88947.0, num_to_drop=1, layers_to_drop={1} 2023-04-29 02:39:16,828 INFO [train.py:904] (1/8) Epoch 9, batch 7750, loss[loss=0.2062, simple_loss=0.299, pruned_loss=0.05673, over 16891.00 frames. ], tot_loss[loss=0.2299, simple_loss=0.3089, pruned_loss=0.07541, over 3066781.10 frames. ], batch size: 96, lr: 7.50e-03, grad_scale: 8.0 2023-04-29 02:40:20,774 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=88995.0, num_to_drop=1, layers_to_drop={1} 2023-04-29 02:40:27,635 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.50 vs. limit=2.0 2023-04-29 02:40:29,342 INFO [train.py:904] (1/8) Epoch 9, batch 7800, loss[loss=0.2133, simple_loss=0.2989, pruned_loss=0.06386, over 16567.00 frames. ], tot_loss[loss=0.2311, simple_loss=0.3098, pruned_loss=0.07616, over 3064805.54 frames. ], batch size: 68, lr: 7.50e-03, grad_scale: 8.0 2023-04-29 02:41:02,672 INFO [optim.py:368] (1/8) 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 02:41:07,721 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=89027.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 02:41:44,457 INFO [train.py:904] (1/8) Epoch 9, batch 7850, loss[loss=0.2209, simple_loss=0.3073, pruned_loss=0.06721, over 16383.00 frames. ], tot_loss[loss=0.2308, simple_loss=0.3105, pruned_loss=0.07548, over 3076397.20 frames. ], batch size: 146, lr: 7.49e-03, grad_scale: 8.0 2023-04-29 02:42:38,974 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=89088.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 02:42:57,227 INFO [train.py:904] (1/8) Epoch 9, batch 7900, loss[loss=0.2264, simple_loss=0.3167, pruned_loss=0.06807, over 16871.00 frames. ], tot_loss[loss=0.2304, simple_loss=0.3102, pruned_loss=0.07523, over 3064017.98 frames. ], batch size: 102, lr: 7.49e-03, grad_scale: 8.0 2023-04-29 02:43:15,947 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.4188, 2.9149, 2.6838, 2.3720, 2.3100, 2.1984, 2.8615, 2.9068], device='cuda:1'), covar=tensor([0.1871, 0.0757, 0.1212, 0.1792, 0.1875, 0.1618, 0.0449, 0.0859], device='cuda:1'), in_proj_covar=tensor([0.0300, 0.0258, 0.0282, 0.0271, 0.0279, 0.0215, 0.0265, 0.0284], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-29 02:43:28,442 INFO [optim.py:368] (1/8) 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 02:44:12,923 INFO [train.py:904] (1/8) Epoch 9, batch 7950, loss[loss=0.2424, simple_loss=0.3221, pruned_loss=0.0814, over 16897.00 frames. ], tot_loss[loss=0.2307, simple_loss=0.3104, pruned_loss=0.07548, over 3070719.33 frames. ], batch size: 109, lr: 7.49e-03, grad_scale: 8.0 2023-04-29 02:44:23,035 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=4.75 vs. limit=5.0 2023-04-29 02:44:39,093 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=89169.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 02:45:26,762 INFO [train.py:904] (1/8) Epoch 9, batch 8000, loss[loss=0.2361, simple_loss=0.3134, pruned_loss=0.07941, over 15161.00 frames. ], tot_loss[loss=0.2317, simple_loss=0.3109, pruned_loss=0.07626, over 3065866.86 frames. ], batch size: 190, lr: 7.49e-03, grad_scale: 8.0 2023-04-29 02:45:59,128 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=3.11 vs. limit=5.0 2023-04-29 02:45:59,520 INFO [optim.py:368] (1/8) Clipping_scale=2.0, grad-norm quartiles 2.399e+02 3.668e+02 4.137e+02 4.605e+02 8.803e+02, threshold=8.275e+02, percent-clipped=3.0 2023-04-29 02:46:09,125 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=89230.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 02:46:40,072 INFO [train.py:904] (1/8) Epoch 9, batch 8050, loss[loss=0.2323, simple_loss=0.3134, pruned_loss=0.07556, over 16443.00 frames. ], tot_loss[loss=0.2315, simple_loss=0.3111, pruned_loss=0.07593, over 3067930.45 frames. ], batch size: 68, lr: 7.49e-03, grad_scale: 4.0 2023-04-29 02:47:55,815 INFO [train.py:904] (1/8) Epoch 9, batch 8100, loss[loss=0.2042, simple_loss=0.287, pruned_loss=0.06071, over 16485.00 frames. ], tot_loss[loss=0.2298, simple_loss=0.3101, pruned_loss=0.07475, over 3079748.34 frames. ], batch size: 75, lr: 7.48e-03, grad_scale: 4.0 2023-04-29 02:48:29,433 INFO [optim.py:368] (1/8) 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 02:49:10,973 INFO [train.py:904] (1/8) Epoch 9, batch 8150, loss[loss=0.1945, simple_loss=0.2796, pruned_loss=0.05471, over 17010.00 frames. ], tot_loss[loss=0.2269, simple_loss=0.3072, pruned_loss=0.07329, over 3088691.69 frames. ], batch size: 41, lr: 7.48e-03, grad_scale: 4.0 2023-04-29 02:49:14,579 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.0602, 2.6243, 2.6303, 1.9347, 2.8357, 2.8847, 2.4750, 2.3701], device='cuda:1'), covar=tensor([0.0670, 0.0162, 0.0186, 0.0871, 0.0075, 0.0152, 0.0388, 0.0411], device='cuda:1'), in_proj_covar=tensor([0.0146, 0.0100, 0.0088, 0.0142, 0.0068, 0.0096, 0.0120, 0.0129], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0003, 0.0002, 0.0004, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:1') 2023-04-29 02:49:19,359 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.4660, 4.4482, 4.2892, 4.0914, 3.9041, 4.3643, 4.2070, 4.0289], device='cuda:1'), covar=tensor([0.0558, 0.0440, 0.0296, 0.0278, 0.0982, 0.0426, 0.0471, 0.0688], device='cuda:1'), in_proj_covar=tensor([0.0228, 0.0275, 0.0262, 0.0243, 0.0284, 0.0275, 0.0181, 0.0308], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-29 02:49:45,539 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=89374.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 02:49:57,659 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.1440, 3.1848, 1.8500, 3.4623, 2.4023, 3.4910, 2.1032, 2.6966], device='cuda:1'), covar=tensor([0.0223, 0.0381, 0.1568, 0.0128, 0.0777, 0.0510, 0.1340, 0.0595], device='cuda:1'), in_proj_covar=tensor([0.0142, 0.0160, 0.0185, 0.0112, 0.0165, 0.0202, 0.0194, 0.0168], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-04-29 02:49:59,363 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=89383.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 02:50:28,005 INFO [train.py:904] (1/8) Epoch 9, batch 8200, loss[loss=0.2556, simple_loss=0.311, pruned_loss=0.1001, over 11476.00 frames. ], tot_loss[loss=0.224, simple_loss=0.3038, pruned_loss=0.07206, over 3093826.05 frames. ], batch size: 248, lr: 7.48e-03, grad_scale: 4.0 2023-04-29 02:51:05,334 INFO [zipformer.py:625] (1/8) 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] (1/8) Clipping_scale=2.0, grad-norm quartiles 1.992e+02 3.313e+02 4.005e+02 4.576e+02 8.683e+02, threshold=8.011e+02, percent-clipped=3.0 2023-04-29 02:51:06,830 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.7460, 3.3041, 3.1753, 1.9070, 2.8826, 2.2345, 3.2658, 3.3038], device='cuda:1'), covar=tensor([0.0279, 0.0597, 0.0531, 0.1784, 0.0748, 0.0970, 0.0668, 0.0674], device='cuda:1'), in_proj_covar=tensor([0.0142, 0.0139, 0.0157, 0.0142, 0.0134, 0.0125, 0.0136, 0.0149], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-04-29 02:51:20,835 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=89435.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 02:51:27,476 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=89439.0, num_to_drop=1, layers_to_drop={0} 2023-04-29 02:51:47,428 INFO [train.py:904] (1/8) Epoch 9, batch 8250, loss[loss=0.2062, simple_loss=0.2944, pruned_loss=0.05898, over 16753.00 frames. ], tot_loss[loss=0.2221, simple_loss=0.3033, pruned_loss=0.07042, over 3068222.32 frames. ], batch size: 124, lr: 7.48e-03, grad_scale: 2.0 2023-04-29 02:52:43,239 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=89486.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 02:53:06,347 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=89500.0, num_to_drop=1, layers_to_drop={0} 2023-04-29 02:53:08,808 INFO [train.py:904] (1/8) Epoch 9, batch 8300, loss[loss=0.1909, simple_loss=0.2718, pruned_loss=0.05501, over 12035.00 frames. ], tot_loss[loss=0.2171, simple_loss=0.3001, pruned_loss=0.06708, over 3059848.24 frames. ], batch size: 247, lr: 7.48e-03, grad_scale: 2.0 2023-04-29 02:53:21,163 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=89509.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 02:53:43,108 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.54 vs. limit=2.0 2023-04-29 02:53:47,008 INFO [zipformer.py:625] (1/8) 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] (1/8) 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] (1/8) Epoch 9, batch 8350, loss[loss=0.193, simple_loss=0.287, pruned_loss=0.04949, over 16345.00 frames. ], tot_loss[loss=0.2142, simple_loss=0.299, pruned_loss=0.06466, over 3067376.18 frames. ], batch size: 146, lr: 7.47e-03, grad_scale: 2.0 2023-04-29 02:55:01,142 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=89570.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 02:55:52,201 INFO [train.py:904] (1/8) Epoch 9, batch 8400, loss[loss=0.1882, simple_loss=0.2738, pruned_loss=0.05131, over 15261.00 frames. ], tot_loss[loss=0.2114, simple_loss=0.2966, pruned_loss=0.06305, over 3046219.89 frames. ], batch size: 190, lr: 7.47e-03, grad_scale: 4.0 2023-04-29 02:56:15,620 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.6155, 4.8805, 4.6994, 4.6849, 4.3941, 4.3967, 4.3745, 4.9855], device='cuda:1'), covar=tensor([0.0990, 0.1046, 0.1029, 0.0723, 0.0792, 0.0997, 0.0944, 0.0847], device='cuda:1'), in_proj_covar=tensor([0.0484, 0.0611, 0.0513, 0.0418, 0.0379, 0.0401, 0.0511, 0.0462], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:1') 2023-04-29 02:56:31,316 INFO [optim.py:368] (1/8) Clipping_scale=2.0, grad-norm quartiles 2.203e+02 2.904e+02 3.369e+02 3.930e+02 8.032e+02, threshold=6.737e+02, percent-clipped=2.0 2023-04-29 02:56:38,104 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([1.9425, 2.2912, 2.2102, 2.9063, 2.0445, 3.2847, 1.6187, 2.7573], device='cuda:1'), covar=tensor([0.1261, 0.0632, 0.1024, 0.0124, 0.0106, 0.0389, 0.1415, 0.0626], device='cuda:1'), in_proj_covar=tensor([0.0150, 0.0151, 0.0174, 0.0125, 0.0196, 0.0202, 0.0174, 0.0174], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0003, 0.0004], device='cuda:1') 2023-04-29 02:57:13,785 INFO [train.py:904] (1/8) Epoch 9, batch 8450, loss[loss=0.19, simple_loss=0.276, pruned_loss=0.05197, over 17193.00 frames. ], tot_loss[loss=0.2088, simple_loss=0.2948, pruned_loss=0.06145, over 3030809.30 frames. ], batch size: 46, lr: 7.47e-03, grad_scale: 4.0 2023-04-29 02:57:52,097 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.26 vs. limit=2.0 2023-04-29 02:57:59,660 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=89680.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 02:58:04,747 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=89683.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 02:58:34,964 INFO [train.py:904] (1/8) Epoch 9, batch 8500, loss[loss=0.1754, simple_loss=0.2697, pruned_loss=0.04051, over 16701.00 frames. ], tot_loss[loss=0.2041, simple_loss=0.2909, pruned_loss=0.05866, over 3046873.89 frames. ], batch size: 89, lr: 7.47e-03, grad_scale: 4.0 2023-04-29 02:59:14,097 INFO [optim.py:368] (1/8) 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 02:59:20,472 INFO [zipformer.py:625] (1/8) 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] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=89731.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 02:59:25,714 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.9966, 1.9648, 2.3348, 3.2180, 2.0871, 2.2346, 2.2196, 2.0272], device='cuda:1'), covar=tensor([0.0777, 0.3159, 0.1777, 0.0529, 0.3924, 0.2311, 0.2586, 0.3281], device='cuda:1'), in_proj_covar=tensor([0.0341, 0.0370, 0.0309, 0.0310, 0.0400, 0.0415, 0.0329, 0.0429], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-29 02:59:38,911 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=89741.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 02:59:58,592 INFO [train.py:904] (1/8) Epoch 9, batch 8550, loss[loss=0.195, simple_loss=0.2892, pruned_loss=0.05045, over 16809.00 frames. ], tot_loss[loss=0.2018, simple_loss=0.2884, pruned_loss=0.0576, over 3029840.23 frames. ], batch size: 83, lr: 7.47e-03, grad_scale: 4.0 2023-04-29 03:00:21,095 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.33 vs. limit=2.0 2023-04-29 03:00:54,502 INFO [zipformer.py:625] (1/8) 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] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=89795.0, num_to_drop=1, layers_to_drop={2} 2023-04-29 03:01:37,953 INFO [train.py:904] (1/8) Epoch 9, batch 8600, loss[loss=0.2192, simple_loss=0.3078, pruned_loss=0.06531, over 16220.00 frames. ], tot_loss[loss=0.2004, simple_loss=0.2886, pruned_loss=0.05609, over 3048703.23 frames. ], batch size: 165, lr: 7.46e-03, grad_scale: 4.0 2023-04-29 03:02:23,622 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.1445, 2.2940, 1.8749, 2.0260, 2.6893, 2.4725, 2.9758, 2.9291], device='cuda:1'), covar=tensor([0.0059, 0.0270, 0.0359, 0.0312, 0.0161, 0.0226, 0.0116, 0.0150], device='cuda:1'), in_proj_covar=tensor([0.0120, 0.0188, 0.0186, 0.0187, 0.0186, 0.0187, 0.0184, 0.0174], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-29 03:02:25,683 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=89825.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 03:02:26,424 INFO [optim.py:368] (1/8) Clipping_scale=2.0, grad-norm quartiles 1.775e+02 2.695e+02 3.313e+02 4.176e+02 8.025e+02, threshold=6.625e+02, percent-clipped=2.0 2023-04-29 03:03:15,781 INFO [train.py:904] (1/8) Epoch 9, batch 8650, loss[loss=0.1716, simple_loss=0.2734, pruned_loss=0.03491, over 15312.00 frames. ], tot_loss[loss=0.1977, simple_loss=0.2865, pruned_loss=0.05448, over 3035536.04 frames. ], batch size: 190, lr: 7.46e-03, grad_scale: 4.0 2023-04-29 03:03:47,288 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=89865.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 03:03:47,450 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=89865.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 03:03:49,625 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.6861, 2.0008, 2.1887, 4.2211, 2.0284, 2.5869, 2.1487, 2.2357], device='cuda:1'), covar=tensor([0.0666, 0.3460, 0.2176, 0.0303, 0.3880, 0.2041, 0.3106, 0.3296], device='cuda:1'), in_proj_covar=tensor([0.0337, 0.0367, 0.0308, 0.0307, 0.0398, 0.0411, 0.0327, 0.0427], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-29 03:03:57,763 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([1.8812, 3.4429, 2.8994, 5.0676, 3.9677, 4.7120, 1.8470, 3.4665], device='cuda:1'), covar=tensor([0.1350, 0.0540, 0.1002, 0.0108, 0.0161, 0.0243, 0.1344, 0.0564], device='cuda:1'), in_proj_covar=tensor([0.0150, 0.0150, 0.0174, 0.0124, 0.0192, 0.0202, 0.0173, 0.0174], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0003, 0.0004], device='cuda:1') 2023-04-29 03:04:05,356 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=89873.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 03:04:42,158 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.25 vs. limit=2.0 2023-04-29 03:05:02,347 INFO [train.py:904] (1/8) Epoch 9, batch 8700, loss[loss=0.2027, simple_loss=0.2776, pruned_loss=0.06393, over 12035.00 frames. ], tot_loss[loss=0.1945, simple_loss=0.2834, pruned_loss=0.05277, over 3046057.45 frames. ], batch size: 248, lr: 7.46e-03, grad_scale: 4.0 2023-04-29 03:05:45,062 INFO [optim.py:368] (1/8) 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 03:05:45,875 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=89926.0, num_to_drop=1, layers_to_drop={2} 2023-04-29 03:06:36,262 INFO [train.py:904] (1/8) Epoch 9, batch 8750, loss[loss=0.2163, simple_loss=0.3022, pruned_loss=0.06522, over 15295.00 frames. ], tot_loss[loss=0.1941, simple_loss=0.2832, pruned_loss=0.05257, over 3047696.38 frames. ], batch size: 190, lr: 7.46e-03, grad_scale: 4.0 2023-04-29 03:08:32,637 INFO [train.py:904] (1/8) Epoch 9, batch 8800, loss[loss=0.1607, simple_loss=0.2543, pruned_loss=0.03353, over 17120.00 frames. ], tot_loss[loss=0.1914, simple_loss=0.2809, pruned_loss=0.05094, over 3056378.52 frames. ], batch size: 49, lr: 7.46e-03, grad_scale: 8.0 2023-04-29 03:09:21,968 INFO [optim.py:368] (1/8) 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 03:09:31,734 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=90030.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 03:09:44,356 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=90036.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 03:09:53,391 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=4.87 vs. limit=5.0 2023-04-29 03:10:17,857 INFO [train.py:904] (1/8) Epoch 9, batch 8850, loss[loss=0.1585, simple_loss=0.2484, pruned_loss=0.03427, over 12310.00 frames. ], tot_loss[loss=0.1913, simple_loss=0.2828, pruned_loss=0.04995, over 3054594.48 frames. ], batch size: 248, lr: 7.45e-03, grad_scale: 8.0 2023-04-29 03:11:11,513 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=90078.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 03:11:18,193 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=90081.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 03:11:36,180 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.0247, 3.1098, 1.6107, 3.3104, 2.2924, 3.2835, 1.7738, 2.5027], device='cuda:1'), covar=tensor([0.0197, 0.0294, 0.1582, 0.0142, 0.0790, 0.0437, 0.1523, 0.0656], device='cuda:1'), in_proj_covar=tensor([0.0137, 0.0154, 0.0180, 0.0108, 0.0161, 0.0191, 0.0188, 0.0164], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0004, 0.0004, 0.0003], device='cuda:1') 2023-04-29 03:11:49,559 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=90095.0, num_to_drop=1, layers_to_drop={1} 2023-04-29 03:12:02,855 INFO [train.py:904] (1/8) Epoch 9, batch 8900, loss[loss=0.2058, simple_loss=0.2989, pruned_loss=0.05639, over 15366.00 frames. ], tot_loss[loss=0.1911, simple_loss=0.2832, pruned_loss=0.04954, over 3057646.32 frames. ], batch size: 191, lr: 7.45e-03, grad_scale: 4.0 2023-04-29 03:12:57,513 INFO [optim.py:368] (1/8) 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 03:13:03,559 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=4.29 vs. limit=5.0 2023-04-29 03:13:07,418 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=90129.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 03:13:10,461 INFO [zipformer.py:625] (1/8) 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] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=90143.0, num_to_drop=1, layers_to_drop={0} 2023-04-29 03:14:08,705 INFO [train.py:904] (1/8) Epoch 9, batch 8950, loss[loss=0.1671, simple_loss=0.2583, pruned_loss=0.03791, over 15279.00 frames. ], tot_loss[loss=0.192, simple_loss=0.2831, pruned_loss=0.05043, over 3053290.92 frames. ], batch size: 190, lr: 7.45e-03, grad_scale: 4.0 2023-04-29 03:14:38,328 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=90165.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 03:15:35,034 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 9, batch 9000, loss[loss=0.1819, simple_loss=0.2647, pruned_loss=0.04955, over 16346.00 frames. ], tot_loss[loss=0.1885, simple_loss=0.2796, pruned_loss=0.0487, over 3067200.04 frames. ], batch size: 146, lr: 7.45e-03, grad_scale: 4.0 2023-04-29 03:15:57,361 INFO [train.py:929] (1/8) Computing validation loss 2023-04-29 03:16:07,533 INFO [train.py:938] (1/8) Epoch 9, validation: loss=0.1581, simple_loss=0.2623, pruned_loss=0.02697, over 944034.00 frames. 2023-04-29 03:16:07,533 INFO [train.py:939] (1/8) Maximum memory allocated so far is 17676MB 2023-04-29 03:16:18,492 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.32 vs. limit=2.0 2023-04-29 03:16:31,271 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=90213.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 03:16:37,747 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.4051, 2.3104, 1.6494, 1.7568, 2.7215, 2.4004, 3.2684, 2.9896], device='cuda:1'), covar=tensor([0.0063, 0.0354, 0.0478, 0.0451, 0.0201, 0.0331, 0.0150, 0.0177], device='cuda:1'), in_proj_covar=tensor([0.0119, 0.0188, 0.0185, 0.0186, 0.0185, 0.0188, 0.0181, 0.0172], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-29 03:16:47,982 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=90221.0, num_to_drop=1, layers_to_drop={3} 2023-04-29 03:16:58,704 INFO [optim.py:368] (1/8) 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 03:17:08,274 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=90232.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 03:17:50,007 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.49 vs. limit=2.0 2023-04-29 03:17:51,878 INFO [train.py:904] (1/8) Epoch 9, batch 9050, loss[loss=0.1984, simple_loss=0.2764, pruned_loss=0.06019, over 16134.00 frames. ], tot_loss[loss=0.1893, simple_loss=0.2802, pruned_loss=0.04926, over 3079206.02 frames. ], batch size: 165, lr: 7.45e-03, grad_scale: 4.0 2023-04-29 03:18:52,444 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.1434, 4.1374, 3.9929, 3.7397, 3.7022, 4.1168, 3.8349, 3.7771], device='cuda:1'), covar=tensor([0.0515, 0.0480, 0.0308, 0.0288, 0.0782, 0.0441, 0.0607, 0.0588], device='cuda:1'), in_proj_covar=tensor([0.0215, 0.0262, 0.0252, 0.0233, 0.0269, 0.0262, 0.0173, 0.0292], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-29 03:19:19,183 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=90293.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 03:19:32,706 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.85 vs. limit=2.0 2023-04-29 03:19:36,843 INFO [train.py:904] (1/8) Epoch 9, batch 9100, loss[loss=0.1879, simple_loss=0.2849, pruned_loss=0.04543, over 16128.00 frames. ], tot_loss[loss=0.19, simple_loss=0.28, pruned_loss=0.04999, over 3078643.45 frames. ], batch size: 165, lr: 7.44e-03, grad_scale: 4.0 2023-04-29 03:20:34,112 INFO [optim.py:368] (1/8) 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 03:20:58,282 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=90336.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 03:21:13,706 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.9831, 1.9967, 2.2697, 3.2345, 2.1263, 2.2182, 2.1824, 2.0621], device='cuda:1'), covar=tensor([0.0811, 0.3117, 0.1848, 0.0476, 0.3608, 0.2285, 0.2700, 0.3276], device='cuda:1'), in_proj_covar=tensor([0.0334, 0.0362, 0.0308, 0.0306, 0.0394, 0.0405, 0.0326, 0.0420], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-29 03:21:33,817 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 9, batch 9150, loss[loss=0.1739, simple_loss=0.2673, pruned_loss=0.04024, over 17039.00 frames. ], tot_loss[loss=0.1894, simple_loss=0.28, pruned_loss=0.04944, over 3050622.75 frames. ], batch size: 55, lr: 7.44e-03, grad_scale: 4.0 2023-04-29 03:22:44,950 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=90384.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 03:23:22,748 INFO [train.py:904] (1/8) Epoch 9, batch 9200, loss[loss=0.158, simple_loss=0.242, pruned_loss=0.03698, over 12228.00 frames. ], tot_loss[loss=0.1853, simple_loss=0.2749, pruned_loss=0.04786, over 3062494.36 frames. ], batch size: 247, lr: 7.44e-03, grad_scale: 8.0 2023-04-29 03:23:42,319 INFO [zipformer.py:625] (1/8) 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] (1/8) 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 03:24:58,586 INFO [train.py:904] (1/8) Epoch 9, batch 9250, loss[loss=0.1507, simple_loss=0.2493, pruned_loss=0.026, over 16874.00 frames. ], tot_loss[loss=0.1858, simple_loss=0.2751, pruned_loss=0.04826, over 3043110.33 frames. ], batch size: 102, lr: 7.44e-03, grad_scale: 8.0 2023-04-29 03:25:14,807 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=90459.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 03:26:12,489 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=90486.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 03:26:38,477 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=90496.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 03:26:49,729 INFO [train.py:904] (1/8) Epoch 9, batch 9300, loss[loss=0.1793, simple_loss=0.2655, pruned_loss=0.04655, over 12160.00 frames. ], tot_loss[loss=0.1847, simple_loss=0.2739, pruned_loss=0.04773, over 3051196.37 frames. ], batch size: 248, lr: 7.43e-03, grad_scale: 8.0 2023-04-29 03:27:30,611 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.8447, 2.2130, 1.8273, 1.8745, 2.5342, 2.2390, 2.7367, 2.7540], device='cuda:1'), covar=tensor([0.0072, 0.0280, 0.0364, 0.0361, 0.0177, 0.0297, 0.0141, 0.0161], device='cuda:1'), in_proj_covar=tensor([0.0117, 0.0185, 0.0182, 0.0182, 0.0181, 0.0184, 0.0176, 0.0167], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-29 03:27:35,081 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=90520.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 03:27:36,964 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=90521.0, num_to_drop=1, layers_to_drop={2} 2023-04-29 03:27:46,913 INFO [optim.py:368] (1/8) 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 03:28:30,465 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.53 vs. limit=2.0 2023-04-29 03:28:35,632 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.82 vs. limit=2.0 2023-04-29 03:28:36,076 INFO [train.py:904] (1/8) Epoch 9, batch 9350, loss[loss=0.1855, simple_loss=0.2773, pruned_loss=0.04688, over 16730.00 frames. ], tot_loss[loss=0.1848, simple_loss=0.2741, pruned_loss=0.04776, over 3053775.17 frames. ], batch size: 134, lr: 7.43e-03, grad_scale: 8.0 2023-04-29 03:28:48,591 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=90557.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 03:29:13,849 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=90569.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 03:29:36,724 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.6431, 4.8632, 4.9975, 4.8886, 4.9469, 5.4222, 5.0238, 4.6522], device='cuda:1'), covar=tensor([0.0931, 0.1596, 0.1283, 0.1532, 0.1884, 0.0717, 0.1020, 0.2010], device='cuda:1'), in_proj_covar=tensor([0.0304, 0.0432, 0.0457, 0.0378, 0.0497, 0.0485, 0.0375, 0.0499], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002], device='cuda:1') 2023-04-29 03:29:50,763 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=90588.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 03:30:18,678 INFO [train.py:904] (1/8) Epoch 9, batch 9400, loss[loss=0.2139, simple_loss=0.3088, pruned_loss=0.05943, over 15509.00 frames. ], tot_loss[loss=0.1844, simple_loss=0.2739, pruned_loss=0.04748, over 3034343.92 frames. ], batch size: 193, lr: 7.43e-03, grad_scale: 8.0 2023-04-29 03:30:41,191 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.0109, 2.3371, 2.3510, 3.0938, 2.2853, 3.2992, 1.6586, 2.7942], device='cuda:1'), covar=tensor([0.1250, 0.0541, 0.0916, 0.0110, 0.0107, 0.0377, 0.1336, 0.0666], device='cuda:1'), in_proj_covar=tensor([0.0152, 0.0151, 0.0176, 0.0124, 0.0184, 0.0202, 0.0175, 0.0175], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0003, 0.0004], device='cuda:1') 2023-04-29 03:31:09,564 INFO [optim.py:368] (1/8) Clipping_scale=2.0, grad-norm quartiles 1.896e+02 2.474e+02 2.959e+02 3.714e+02 8.908e+02, threshold=5.918e+02, percent-clipped=5.0 2023-04-29 03:32:00,895 INFO [train.py:904] (1/8) Epoch 9, batch 9450, loss[loss=0.1772, simple_loss=0.2771, pruned_loss=0.03869, over 16443.00 frames. ], tot_loss[loss=0.1851, simple_loss=0.2751, pruned_loss=0.04749, over 3030663.51 frames. ], batch size: 68, lr: 7.43e-03, grad_scale: 8.0 2023-04-29 03:33:43,967 INFO [train.py:904] (1/8) Epoch 9, batch 9500, loss[loss=0.1757, simple_loss=0.2622, pruned_loss=0.04463, over 16652.00 frames. ], tot_loss[loss=0.1836, simple_loss=0.2741, pruned_loss=0.04659, over 3037919.24 frames. ], batch size: 62, lr: 7.43e-03, grad_scale: 8.0 2023-04-29 03:33:56,116 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.25 vs. limit=2.0 2023-04-29 03:33:58,232 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=90707.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 03:34:09,047 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.3579, 1.9681, 1.6726, 1.6359, 2.2829, 1.8964, 2.2149, 2.3728], device='cuda:1'), covar=tensor([0.0073, 0.0253, 0.0332, 0.0325, 0.0144, 0.0267, 0.0121, 0.0150], device='cuda:1'), in_proj_covar=tensor([0.0117, 0.0187, 0.0183, 0.0182, 0.0181, 0.0185, 0.0177, 0.0167], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-29 03:34:35,317 INFO [optim.py:368] (1/8) Clipping_scale=2.0, grad-norm quartiles 1.646e+02 2.298e+02 2.954e+02 3.660e+02 6.291e+02, threshold=5.908e+02, percent-clipped=3.0 2023-04-29 03:35:30,037 INFO [train.py:904] (1/8) Epoch 9, batch 9550, loss[loss=0.1835, simple_loss=0.2745, pruned_loss=0.0463, over 16759.00 frames. ], tot_loss[loss=0.1831, simple_loss=0.2732, pruned_loss=0.04652, over 3031987.28 frames. ], batch size: 83, lr: 7.42e-03, grad_scale: 8.0 2023-04-29 03:35:38,286 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.1523, 3.1863, 1.6102, 3.4770, 2.3087, 3.4335, 1.8643, 2.6602], device='cuda:1'), covar=tensor([0.0213, 0.0310, 0.1672, 0.0143, 0.0814, 0.0504, 0.1495, 0.0642], device='cuda:1'), in_proj_covar=tensor([0.0136, 0.0153, 0.0180, 0.0108, 0.0158, 0.0188, 0.0188, 0.0163], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0004, 0.0004, 0.0003], device='cuda:1') 2023-04-29 03:36:43,168 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.1560, 4.1507, 4.0206, 3.8138, 3.7052, 4.1339, 3.8929, 3.8485], device='cuda:1'), covar=tensor([0.0492, 0.0431, 0.0274, 0.0238, 0.0725, 0.0434, 0.0543, 0.0487], device='cuda:1'), in_proj_covar=tensor([0.0209, 0.0254, 0.0247, 0.0227, 0.0262, 0.0257, 0.0169, 0.0283], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002], device='cuda:1') 2023-04-29 03:36:43,177 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=90786.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 03:36:57,651 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.2583, 3.3259, 1.7886, 3.6272, 2.3678, 3.5347, 1.9103, 2.7244], device='cuda:1'), covar=tensor([0.0235, 0.0316, 0.1683, 0.0153, 0.0862, 0.0581, 0.1629, 0.0709], device='cuda:1'), in_proj_covar=tensor([0.0136, 0.0153, 0.0181, 0.0109, 0.0159, 0.0190, 0.0189, 0.0163], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0004, 0.0004, 0.0003], device='cuda:1') 2023-04-29 03:37:02,715 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.0139, 2.0687, 2.3690, 3.2066, 2.1535, 2.3309, 2.2906, 2.1183], device='cuda:1'), covar=tensor([0.0805, 0.2922, 0.1700, 0.0517, 0.3417, 0.1962, 0.2655, 0.2894], device='cuda:1'), in_proj_covar=tensor([0.0333, 0.0359, 0.0307, 0.0304, 0.0393, 0.0402, 0.0324, 0.0419], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-29 03:37:11,178 INFO [train.py:904] (1/8) Epoch 9, batch 9600, loss[loss=0.2145, simple_loss=0.3004, pruned_loss=0.06428, over 16953.00 frames. ], tot_loss[loss=0.1851, simple_loss=0.2751, pruned_loss=0.04758, over 3036329.12 frames. ], batch size: 109, lr: 7.42e-03, grad_scale: 8.0 2023-04-29 03:37:24,049 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.41 vs. limit=2.0 2023-04-29 03:37:37,096 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=90815.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 03:37:59,118 INFO [optim.py:368] (1/8) Clipping_scale=2.0, grad-norm quartiles 1.470e+02 2.542e+02 3.036e+02 4.023e+02 8.440e+02, threshold=6.073e+02, percent-clipped=4.0 2023-04-29 03:38:13,524 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.1505, 3.0692, 3.1654, 1.6605, 3.3560, 3.4047, 2.8496, 2.5572], device='cuda:1'), covar=tensor([0.0747, 0.0182, 0.0149, 0.1130, 0.0058, 0.0104, 0.0321, 0.0404], device='cuda:1'), in_proj_covar=tensor([0.0137, 0.0093, 0.0079, 0.0135, 0.0064, 0.0090, 0.0114, 0.0122], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0002, 0.0002, 0.0003, 0.0003], device='cuda:1') 2023-04-29 03:38:17,845 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=90834.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 03:38:55,260 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=90849.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 03:39:01,490 INFO [train.py:904] (1/8) Epoch 9, batch 9650, loss[loss=0.2096, simple_loss=0.2835, pruned_loss=0.06782, over 12301.00 frames. ], tot_loss[loss=0.1861, simple_loss=0.2767, pruned_loss=0.04771, over 3039242.73 frames. ], batch size: 248, lr: 7.42e-03, grad_scale: 8.0 2023-04-29 03:39:03,429 INFO [zipformer.py:625] (1/8) 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] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=90888.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 03:40:48,113 INFO [train.py:904] (1/8) Epoch 9, batch 9700, loss[loss=0.1641, simple_loss=0.2632, pruned_loss=0.0325, over 16898.00 frames. ], tot_loss[loss=0.185, simple_loss=0.2757, pruned_loss=0.04716, over 3055690.83 frames. ], batch size: 102, lr: 7.42e-03, grad_scale: 8.0 2023-04-29 03:41:04,596 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=90910.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 03:41:08,597 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([1.9085, 2.3411, 2.3476, 3.0962, 2.2663, 3.3109, 1.5627, 2.8026], device='cuda:1'), covar=tensor([0.1210, 0.0576, 0.0957, 0.0114, 0.0094, 0.0361, 0.1380, 0.0615], device='cuda:1'), in_proj_covar=tensor([0.0149, 0.0149, 0.0174, 0.0121, 0.0179, 0.0199, 0.0173, 0.0173], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0003, 0.0004], device='cuda:1') 2023-04-29 03:41:25,675 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.9948, 3.9823, 3.9034, 3.4444, 3.9187, 1.6706, 3.7570, 3.7318], device='cuda:1'), covar=tensor([0.0094, 0.0077, 0.0136, 0.0237, 0.0082, 0.2214, 0.0105, 0.0170], device='cuda:1'), in_proj_covar=tensor([0.0111, 0.0099, 0.0144, 0.0132, 0.0115, 0.0165, 0.0130, 0.0134], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0001, 0.0001, 0.0002, 0.0001, 0.0002], device='cuda:1') 2023-04-29 03:41:27,820 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.3442, 1.9913, 1.6380, 1.7409, 2.1986, 1.9555, 2.2091, 2.3284], device='cuda:1'), covar=tensor([0.0068, 0.0248, 0.0341, 0.0294, 0.0154, 0.0242, 0.0116, 0.0156], device='cuda:1'), in_proj_covar=tensor([0.0117, 0.0188, 0.0182, 0.0181, 0.0182, 0.0185, 0.0176, 0.0167], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-29 03:41:40,521 INFO [optim.py:368] (1/8) Clipping_scale=2.0, grad-norm quartiles 1.598e+02 2.339e+02 3.062e+02 3.716e+02 7.920e+02, threshold=6.123e+02, percent-clipped=1.0 2023-04-29 03:41:46,137 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=4.65 vs. limit=5.0 2023-04-29 03:41:49,910 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.67 vs. limit=2.0 2023-04-29 03:41:59,658 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=90936.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 03:42:23,234 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.53 vs. limit=2.0 2023-04-29 03:42:31,486 INFO [train.py:904] (1/8) Epoch 9, batch 9750, loss[loss=0.1855, simple_loss=0.2777, pruned_loss=0.04663, over 16927.00 frames. ], tot_loss[loss=0.1855, simple_loss=0.2754, pruned_loss=0.04782, over 3043325.81 frames. ], batch size: 109, lr: 7.42e-03, grad_scale: 8.0 2023-04-29 03:43:39,111 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.6478, 3.7649, 3.0126, 2.2387, 2.5425, 2.4268, 3.9843, 3.5205], device='cuda:1'), covar=tensor([0.2525, 0.0598, 0.1352, 0.2188, 0.2247, 0.1607, 0.0352, 0.0864], device='cuda:1'), in_proj_covar=tensor([0.0289, 0.0244, 0.0269, 0.0262, 0.0247, 0.0208, 0.0253, 0.0265], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-29 03:44:05,923 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.8552, 5.1183, 5.2644, 5.1301, 5.1701, 5.6749, 5.1846, 4.8942], device='cuda:1'), covar=tensor([0.0698, 0.1642, 0.1318, 0.1388, 0.1854, 0.0790, 0.1267, 0.2180], device='cuda:1'), in_proj_covar=tensor([0.0304, 0.0432, 0.0462, 0.0379, 0.0498, 0.0483, 0.0378, 0.0501], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002], device='cuda:1') 2023-04-29 03:44:10,866 INFO [train.py:904] (1/8) Epoch 9, batch 9800, loss[loss=0.1841, simple_loss=0.2847, pruned_loss=0.0418, over 16507.00 frames. ], tot_loss[loss=0.1845, simple_loss=0.276, pruned_loss=0.04648, over 3077235.31 frames. ], batch size: 68, lr: 7.41e-03, grad_scale: 8.0 2023-04-29 03:44:21,926 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=91007.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 03:44:57,913 INFO [optim.py:368] (1/8) Clipping_scale=2.0, grad-norm quartiles 1.670e+02 2.313e+02 2.751e+02 3.431e+02 5.847e+02, threshold=5.502e+02, percent-clipped=0.0 2023-04-29 03:45:50,610 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.8559, 2.3178, 2.3308, 4.6630, 2.1552, 2.8080, 2.3121, 2.6014], device='cuda:1'), covar=tensor([0.0679, 0.3014, 0.2012, 0.0251, 0.3557, 0.1932, 0.2831, 0.2948], device='cuda:1'), in_proj_covar=tensor([0.0336, 0.0363, 0.0309, 0.0306, 0.0395, 0.0403, 0.0326, 0.0420], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-29 03:45:57,961 INFO [train.py:904] (1/8) Epoch 9, batch 9850, loss[loss=0.182, simple_loss=0.2744, pruned_loss=0.0448, over 16506.00 frames. ], tot_loss[loss=0.1844, simple_loss=0.2766, pruned_loss=0.04605, over 3094917.41 frames. ], batch size: 68, lr: 7.41e-03, grad_scale: 8.0 2023-04-29 03:46:04,898 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=91055.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 03:47:48,700 INFO [train.py:904] (1/8) Epoch 9, batch 9900, loss[loss=0.1803, simple_loss=0.2835, pruned_loss=0.03855, over 16782.00 frames. ], tot_loss[loss=0.1847, simple_loss=0.277, pruned_loss=0.04617, over 3084873.26 frames. ], batch size: 83, lr: 7.41e-03, grad_scale: 8.0 2023-04-29 03:48:20,106 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=91115.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 03:48:47,933 INFO [optim.py:368] (1/8) Clipping_scale=2.0, grad-norm quartiles 1.713e+02 2.434e+02 3.232e+02 4.008e+02 9.044e+02, threshold=6.464e+02, percent-clipped=5.0 2023-04-29 03:49:47,991 INFO [train.py:904] (1/8) Epoch 9, batch 9950, loss[loss=0.1782, simple_loss=0.2749, pruned_loss=0.04077, over 16300.00 frames. ], tot_loss[loss=0.1854, simple_loss=0.2787, pruned_loss=0.04609, over 3081609.14 frames. ], batch size: 165, lr: 7.41e-03, grad_scale: 8.0 2023-04-29 03:49:49,172 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=91152.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 03:50:14,538 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=91163.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 03:50:44,587 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.8504, 4.3892, 4.0117, 2.1634, 3.3636, 2.8897, 4.2133, 4.3014], device='cuda:1'), covar=tensor([0.0178, 0.0427, 0.0561, 0.1636, 0.0643, 0.0795, 0.0488, 0.0655], device='cuda:1'), in_proj_covar=tensor([0.0137, 0.0129, 0.0152, 0.0139, 0.0130, 0.0122, 0.0130, 0.0140], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:1') 2023-04-29 03:50:44,615 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.0144, 4.6439, 3.5494, 2.6064, 3.2092, 2.8392, 4.7177, 4.2291], device='cuda:1'), covar=tensor([0.2278, 0.0399, 0.1346, 0.2302, 0.1794, 0.1346, 0.0366, 0.0616], device='cuda:1'), in_proj_covar=tensor([0.0285, 0.0242, 0.0266, 0.0258, 0.0243, 0.0205, 0.0250, 0.0262], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-29 03:51:44,349 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=91200.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 03:51:47,240 INFO [train.py:904] (1/8) Epoch 9, batch 10000, loss[loss=0.1482, simple_loss=0.2441, pruned_loss=0.02614, over 16884.00 frames. ], tot_loss[loss=0.184, simple_loss=0.2771, pruned_loss=0.0455, over 3082000.51 frames. ], batch size: 42, lr: 7.41e-03, grad_scale: 8.0 2023-04-29 03:51:53,872 INFO [zipformer.py:625] (1/8) 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] (1/8) 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] (1/8) Epoch 9, batch 10050, loss[loss=0.1894, simple_loss=0.2819, pruned_loss=0.04844, over 12039.00 frames. ], tot_loss[loss=0.1843, simple_loss=0.2775, pruned_loss=0.04555, over 3094326.77 frames. ], batch size: 246, lr: 7.40e-03, grad_scale: 8.0 2023-04-29 03:54:00,666 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([5.1163, 5.1120, 4.7020, 4.4943, 4.9059, 1.8145, 4.6168, 4.7422], device='cuda:1'), covar=tensor([0.0066, 0.0050, 0.0167, 0.0224, 0.0072, 0.2201, 0.0114, 0.0137], device='cuda:1'), in_proj_covar=tensor([0.0115, 0.0102, 0.0147, 0.0135, 0.0117, 0.0170, 0.0134, 0.0137], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-29 03:54:30,265 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 9, batch 10100, loss[loss=0.1648, simple_loss=0.2523, pruned_loss=0.03861, over 15282.00 frames. ], tot_loss[loss=0.1848, simple_loss=0.2776, pruned_loss=0.04604, over 3083426.02 frames. ], batch size: 191, lr: 7.40e-03, grad_scale: 8.0 2023-04-29 03:55:52,058 INFO [optim.py:368] (1/8) 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 03:56:14,365 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 10, batch 0, loss[loss=0.2032, simple_loss=0.2992, pruned_loss=0.05359, over 17043.00 frames. ], tot_loss[loss=0.2032, simple_loss=0.2992, pruned_loss=0.05359, over 17043.00 frames. ], batch size: 50, lr: 7.04e-03, grad_scale: 8.0 2023-04-29 03:56:45,260 INFO [train.py:929] (1/8) Computing validation loss 2023-04-29 03:56:50,592 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([6.2796, 6.6251, 6.3942, 6.5420, 6.1306, 6.1035, 6.1481, 6.6177], device='cuda:1'), covar=tensor([0.0626, 0.0578, 0.0495, 0.0347, 0.0667, 0.0236, 0.0558, 0.0589], device='cuda:1'), in_proj_covar=tensor([0.0467, 0.0596, 0.0485, 0.0407, 0.0373, 0.0388, 0.0496, 0.0443], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:1') 2023-04-29 03:56:52,894 INFO [train.py:938] (1/8) Epoch 10, validation: loss=0.158, simple_loss=0.2614, pruned_loss=0.02732, over 944034.00 frames. 2023-04-29 03:56:52,894 INFO [train.py:939] (1/8) Maximum memory allocated so far is 17676MB 2023-04-29 03:57:03,873 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.6508, 4.8104, 4.9454, 4.7393, 4.8082, 5.3487, 4.9789, 4.6178], device='cuda:1'), covar=tensor([0.1126, 0.1684, 0.1751, 0.1917, 0.2386, 0.1004, 0.1354, 0.2329], device='cuda:1'), in_proj_covar=tensor([0.0307, 0.0438, 0.0465, 0.0385, 0.0501, 0.0488, 0.0381, 0.0504], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002], device='cuda:1') 2023-04-29 03:58:02,502 INFO [train.py:904] (1/8) Epoch 10, batch 50, loss[loss=0.2043, simple_loss=0.2748, pruned_loss=0.06694, over 16692.00 frames. ], tot_loss[loss=0.216, simple_loss=0.2926, pruned_loss=0.06973, over 754660.32 frames. ], batch size: 124, lr: 7.04e-03, grad_scale: 2.0 2023-04-29 03:58:39,941 INFO [optim.py:368] (1/8) Clipping_scale=2.0, grad-norm quartiles 1.612e+02 2.984e+02 3.603e+02 4.569e+02 8.591e+02, threshold=7.207e+02, percent-clipped=1.0 2023-04-29 03:58:42,960 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.8023, 1.6957, 2.2116, 2.6711, 2.7017, 2.5705, 1.8921, 2.9665], device='cuda:1'), covar=tensor([0.0118, 0.0319, 0.0229, 0.0148, 0.0156, 0.0176, 0.0319, 0.0080], device='cuda:1'), in_proj_covar=tensor([0.0152, 0.0165, 0.0149, 0.0149, 0.0160, 0.0115, 0.0166, 0.0102], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001], device='cuda:1') 2023-04-29 03:59:08,809 INFO [train.py:904] (1/8) Epoch 10, batch 100, loss[loss=0.1819, simple_loss=0.2797, pruned_loss=0.04202, over 17115.00 frames. ], tot_loss[loss=0.2068, simple_loss=0.2856, pruned_loss=0.06401, over 1314970.04 frames. ], batch size: 47, lr: 7.03e-03, grad_scale: 1.0 2023-04-29 03:59:11,849 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.43 vs. limit=2.0 2023-04-29 03:59:45,957 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.9329, 4.1379, 2.6089, 4.7190, 3.0402, 4.6002, 2.3822, 3.3701], device='cuda:1'), covar=tensor([0.0220, 0.0299, 0.1360, 0.0121, 0.0800, 0.0402, 0.1502, 0.0604], device='cuda:1'), in_proj_covar=tensor([0.0141, 0.0159, 0.0185, 0.0114, 0.0164, 0.0195, 0.0193, 0.0168], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-04-29 04:00:16,894 INFO [train.py:904] (1/8) Epoch 10, batch 150, loss[loss=0.2302, simple_loss=0.2973, pruned_loss=0.08159, over 16491.00 frames. ], tot_loss[loss=0.2023, simple_loss=0.2823, pruned_loss=0.06113, over 1757115.24 frames. ], batch size: 146, lr: 7.03e-03, grad_scale: 1.0 2023-04-29 04:00:22,944 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=91505.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 04:00:34,732 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.1443, 3.4378, 3.2188, 2.1097, 2.7413, 2.2618, 3.5782, 3.4758], device='cuda:1'), covar=tensor([0.0217, 0.0597, 0.0636, 0.1534, 0.0744, 0.0935, 0.0462, 0.0771], device='cuda:1'), in_proj_covar=tensor([0.0138, 0.0131, 0.0154, 0.0140, 0.0131, 0.0123, 0.0131, 0.0143], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:1') 2023-04-29 04:00:46,617 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.2824, 3.8564, 4.2005, 2.8537, 3.6898, 4.0345, 3.8347, 2.4645], device='cuda:1'), covar=tensor([0.0418, 0.0087, 0.0025, 0.0272, 0.0073, 0.0059, 0.0058, 0.0325], device='cuda:1'), in_proj_covar=tensor([0.0127, 0.0067, 0.0066, 0.0124, 0.0074, 0.0081, 0.0073, 0.0118], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:1') 2023-04-29 04:00:56,392 INFO [optim.py:368] (1/8) 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 04:01:12,419 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=91542.0, num_to_drop=1, layers_to_drop={1} 2023-04-29 04:01:26,563 INFO [train.py:904] (1/8) Epoch 10, batch 200, loss[loss=0.223, simple_loss=0.2903, pruned_loss=0.07787, over 16773.00 frames. ], tot_loss[loss=0.2043, simple_loss=0.2833, pruned_loss=0.0627, over 2100305.68 frames. ], batch size: 124, lr: 7.03e-03, grad_scale: 1.0 2023-04-29 04:01:28,069 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=91553.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 04:01:34,974 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=4.02 vs. limit=5.0 2023-04-29 04:02:28,214 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.8044, 4.5052, 4.8384, 5.0156, 5.1706, 4.5357, 5.2119, 5.1728], device='cuda:1'), covar=tensor([0.1393, 0.1051, 0.1370, 0.0628, 0.0526, 0.0788, 0.0452, 0.0453], device='cuda:1'), in_proj_covar=tensor([0.0491, 0.0602, 0.0729, 0.0605, 0.0461, 0.0462, 0.0484, 0.0543], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-29 04:02:34,675 INFO [train.py:904] (1/8) Epoch 10, batch 250, loss[loss=0.1869, simple_loss=0.264, pruned_loss=0.05487, over 16735.00 frames. ], tot_loss[loss=0.2029, simple_loss=0.2817, pruned_loss=0.06208, over 2371860.33 frames. ], batch size: 134, lr: 7.03e-03, grad_scale: 1.0 2023-04-29 04:02:36,439 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=91603.0, num_to_drop=1, layers_to_drop={0} 2023-04-29 04:03:11,342 INFO [optim.py:368] (1/8) 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] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=91641.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 04:03:42,031 INFO [train.py:904] (1/8) Epoch 10, batch 300, loss[loss=0.1686, simple_loss=0.2531, pruned_loss=0.04204, over 16795.00 frames. ], tot_loss[loss=0.1995, simple_loss=0.2783, pruned_loss=0.06035, over 2587452.82 frames. ], batch size: 39, lr: 7.03e-03, grad_scale: 1.0 2023-04-29 04:04:51,283 INFO [train.py:904] (1/8) Epoch 10, batch 350, loss[loss=0.1767, simple_loss=0.2538, pruned_loss=0.04979, over 15444.00 frames. ], tot_loss[loss=0.1954, simple_loss=0.2746, pruned_loss=0.05808, over 2747630.39 frames. ], batch size: 190, lr: 7.02e-03, grad_scale: 1.0 2023-04-29 04:05:25,862 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.40 vs. limit=2.0 2023-04-29 04:05:28,631 INFO [optim.py:368] (1/8) 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] (1/8) Epoch 10, batch 400, loss[loss=0.2044, simple_loss=0.2775, pruned_loss=0.06568, over 15493.00 frames. ], tot_loss[loss=0.1936, simple_loss=0.2725, pruned_loss=0.05733, over 2871558.37 frames. ], batch size: 190, lr: 7.02e-03, grad_scale: 2.0 2023-04-29 04:07:11,414 INFO [train.py:904] (1/8) Epoch 10, batch 450, loss[loss=0.2113, simple_loss=0.2728, pruned_loss=0.07491, over 16877.00 frames. ], tot_loss[loss=0.1923, simple_loss=0.2712, pruned_loss=0.05677, over 2970235.69 frames. ], batch size: 116, lr: 7.02e-03, grad_scale: 2.0 2023-04-29 04:07:13,169 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.58 vs. limit=2.0 2023-04-29 04:07:50,758 INFO [optim.py:368] (1/8) Clipping_scale=2.0, grad-norm quartiles 1.774e+02 2.266e+02 2.904e+02 3.599e+02 6.333e+02, threshold=5.808e+02, percent-clipped=1.0 2023-04-29 04:08:20,249 INFO [train.py:904] (1/8) Epoch 10, batch 500, loss[loss=0.213, simple_loss=0.293, pruned_loss=0.06645, over 16470.00 frames. ], tot_loss[loss=0.1908, simple_loss=0.27, pruned_loss=0.05578, over 3050174.42 frames. ], batch size: 62, lr: 7.02e-03, grad_scale: 2.0 2023-04-29 04:09:10,739 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.98 vs. limit=2.0 2023-04-29 04:09:15,171 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=91891.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 04:09:23,882 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=91898.0, num_to_drop=1, layers_to_drop={0} 2023-04-29 04:09:29,472 INFO [train.py:904] (1/8) Epoch 10, batch 550, loss[loss=0.1527, simple_loss=0.236, pruned_loss=0.03469, over 16944.00 frames. ], tot_loss[loss=0.1902, simple_loss=0.2699, pruned_loss=0.05528, over 3112777.00 frames. ], batch size: 41, lr: 7.02e-03, grad_scale: 2.0 2023-04-29 04:09:41,788 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.1850, 2.3883, 1.7148, 1.9787, 2.7147, 2.3811, 3.2840, 2.9358], device='cuda:1'), covar=tensor([0.0114, 0.0336, 0.0475, 0.0406, 0.0232, 0.0359, 0.0185, 0.0215], device='cuda:1'), in_proj_covar=tensor([0.0132, 0.0198, 0.0193, 0.0193, 0.0195, 0.0198, 0.0198, 0.0183], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-29 04:10:07,829 INFO [optim.py:368] (1/8) 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 04:10:22,738 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=91941.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 04:10:38,350 INFO [train.py:904] (1/8) Epoch 10, batch 600, loss[loss=0.1984, simple_loss=0.2664, pruned_loss=0.06516, over 16770.00 frames. ], tot_loss[loss=0.1892, simple_loss=0.269, pruned_loss=0.05466, over 3149884.15 frames. ], batch size: 102, lr: 7.01e-03, grad_scale: 2.0 2023-04-29 04:10:38,904 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=91952.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 04:10:42,772 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=91955.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 04:11:18,726 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=91981.0, num_to_drop=1, layers_to_drop={1} 2023-04-29 04:11:30,054 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=91989.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 04:11:51,293 INFO [train.py:904] (1/8) Epoch 10, batch 650, loss[loss=0.2053, simple_loss=0.2791, pruned_loss=0.06578, over 16717.00 frames. ], tot_loss[loss=0.1882, simple_loss=0.2678, pruned_loss=0.05426, over 3184988.75 frames. ], batch size: 134, lr: 7.01e-03, grad_scale: 2.0 2023-04-29 04:11:59,919 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.90 vs. limit=2.0 2023-04-29 04:12:12,124 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=92016.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 04:12:31,534 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.84 vs. limit=2.0 2023-04-29 04:12:31,800 INFO [optim.py:368] (1/8) 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 04:12:49,500 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=92042.0, num_to_drop=1, layers_to_drop={1} 2023-04-29 04:13:03,065 INFO [train.py:904] (1/8) Epoch 10, batch 700, loss[loss=0.2319, simple_loss=0.3036, pruned_loss=0.08012, over 16270.00 frames. ], tot_loss[loss=0.1879, simple_loss=0.2678, pruned_loss=0.05402, over 3218337.33 frames. ], batch size: 165, lr: 7.01e-03, grad_scale: 2.0 2023-04-29 04:13:05,667 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.42 vs. limit=2.0 2023-04-29 04:14:12,200 INFO [train.py:904] (1/8) Epoch 10, batch 750, loss[loss=0.1722, simple_loss=0.252, pruned_loss=0.04621, over 15632.00 frames. ], tot_loss[loss=0.1876, simple_loss=0.2676, pruned_loss=0.05378, over 3242733.54 frames. ], batch size: 191, lr: 7.01e-03, grad_scale: 2.0 2023-04-29 04:14:31,707 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.88 vs. limit=2.0 2023-04-29 04:14:52,014 INFO [optim.py:368] (1/8) 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] (1/8) Epoch 10, batch 800, loss[loss=0.1728, simple_loss=0.2545, pruned_loss=0.04553, over 16495.00 frames. ], tot_loss[loss=0.1858, simple_loss=0.266, pruned_loss=0.05276, over 3257876.85 frames. ], batch size: 68, lr: 7.01e-03, grad_scale: 4.0 2023-04-29 04:16:27,560 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=92198.0, num_to_drop=1, layers_to_drop={0} 2023-04-29 04:16:32,826 INFO [train.py:904] (1/8) Epoch 10, batch 850, loss[loss=0.1839, simple_loss=0.2549, pruned_loss=0.0565, over 16915.00 frames. ], tot_loss[loss=0.185, simple_loss=0.2653, pruned_loss=0.05231, over 3274858.11 frames. ], batch size: 90, lr: 7.00e-03, grad_scale: 4.0 2023-04-29 04:16:56,521 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.5781, 4.0040, 4.2947, 2.8908, 3.7347, 4.0635, 3.7940, 2.4534], device='cuda:1'), covar=tensor([0.0371, 0.0056, 0.0029, 0.0290, 0.0067, 0.0084, 0.0063, 0.0345], device='cuda:1'), in_proj_covar=tensor([0.0127, 0.0069, 0.0067, 0.0125, 0.0076, 0.0084, 0.0075, 0.0120], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:1') 2023-04-29 04:17:10,136 INFO [optim.py:368] (1/8) 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] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=92246.0, num_to_drop=1, layers_to_drop={0} 2023-04-29 04:17:34,330 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=92247.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 04:17:41,160 INFO [train.py:904] (1/8) Epoch 10, batch 900, loss[loss=0.1743, simple_loss=0.2458, pruned_loss=0.05137, over 16454.00 frames. ], tot_loss[loss=0.1845, simple_loss=0.2647, pruned_loss=0.05218, over 3290379.38 frames. ], batch size: 146, lr: 7.00e-03, grad_scale: 4.0 2023-04-29 04:18:15,227 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.9742, 4.0842, 2.2101, 4.5668, 2.8671, 4.6137, 2.2590, 3.1396], device='cuda:1'), covar=tensor([0.0205, 0.0264, 0.1536, 0.0203, 0.0863, 0.0324, 0.1476, 0.0686], device='cuda:1'), in_proj_covar=tensor([0.0145, 0.0165, 0.0189, 0.0124, 0.0167, 0.0204, 0.0193, 0.0171], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-04-29 04:18:50,838 INFO [train.py:904] (1/8) Epoch 10, batch 950, loss[loss=0.1924, simple_loss=0.2588, pruned_loss=0.06295, over 16646.00 frames. ], tot_loss[loss=0.1848, simple_loss=0.265, pruned_loss=0.05231, over 3302640.40 frames. ], batch size: 134, lr: 7.00e-03, grad_scale: 4.0 2023-04-29 04:19:04,203 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=92311.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 04:19:27,849 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.8234, 4.0022, 2.1881, 4.2612, 2.8738, 4.2172, 2.1619, 3.0447], device='cuda:1'), covar=tensor([0.0195, 0.0278, 0.1465, 0.0211, 0.0699, 0.0525, 0.1431, 0.0612], device='cuda:1'), in_proj_covar=tensor([0.0145, 0.0164, 0.0188, 0.0124, 0.0166, 0.0203, 0.0192, 0.0170], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-04-29 04:19:29,763 INFO [optim.py:368] (1/8) 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 04:19:39,578 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=92337.0, num_to_drop=1, layers_to_drop={0} 2023-04-29 04:19:58,791 INFO [train.py:904] (1/8) Epoch 10, batch 1000, loss[loss=0.1935, simple_loss=0.2602, pruned_loss=0.06341, over 16481.00 frames. ], tot_loss[loss=0.1848, simple_loss=0.265, pruned_loss=0.05224, over 3302924.77 frames. ], batch size: 146, lr: 7.00e-03, grad_scale: 4.0 2023-04-29 04:20:47,679 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.16 vs. limit=2.0 2023-04-29 04:20:59,336 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.9138, 4.8730, 4.7716, 4.1533, 4.8285, 2.0465, 4.5525, 4.6995], device='cuda:1'), covar=tensor([0.0089, 0.0075, 0.0140, 0.0336, 0.0084, 0.2119, 0.0120, 0.0155], device='cuda:1'), in_proj_covar=tensor([0.0126, 0.0111, 0.0162, 0.0151, 0.0131, 0.0179, 0.0149, 0.0153], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-29 04:21:03,423 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.8230, 3.8557, 2.2539, 4.1358, 2.8183, 4.0414, 2.1719, 2.9286], device='cuda:1'), covar=tensor([0.0174, 0.0321, 0.1387, 0.0200, 0.0699, 0.0606, 0.1445, 0.0610], device='cuda:1'), in_proj_covar=tensor([0.0144, 0.0163, 0.0186, 0.0123, 0.0164, 0.0202, 0.0191, 0.0169], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0003, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-04-29 04:21:09,129 INFO [train.py:904] (1/8) Epoch 10, batch 1050, loss[loss=0.1916, simple_loss=0.2662, pruned_loss=0.05847, over 16518.00 frames. ], tot_loss[loss=0.1839, simple_loss=0.264, pruned_loss=0.05187, over 3309704.68 frames. ], batch size: 68, lr: 7.00e-03, grad_scale: 4.0 2023-04-29 04:21:48,359 INFO [optim.py:368] (1/8) 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] (1/8) Epoch 10, batch 1100, loss[loss=0.1591, simple_loss=0.244, pruned_loss=0.03706, over 16813.00 frames. ], tot_loss[loss=0.1829, simple_loss=0.2627, pruned_loss=0.05154, over 3306308.52 frames. ], batch size: 42, lr: 7.00e-03, grad_scale: 4.0 2023-04-29 04:23:28,211 INFO [train.py:904] (1/8) Epoch 10, batch 1150, loss[loss=0.1599, simple_loss=0.2345, pruned_loss=0.04265, over 16445.00 frames. ], tot_loss[loss=0.1819, simple_loss=0.2623, pruned_loss=0.05076, over 3302528.88 frames. ], batch size: 75, lr: 6.99e-03, grad_scale: 4.0 2023-04-29 04:24:08,393 INFO [optim.py:368] (1/8) 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 04:24:32,764 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 10, batch 1200, loss[loss=0.1743, simple_loss=0.2517, pruned_loss=0.04842, over 16922.00 frames. ], tot_loss[loss=0.1809, simple_loss=0.2615, pruned_loss=0.05019, over 3293967.53 frames. ], batch size: 90, lr: 6.99e-03, grad_scale: 8.0 2023-04-29 04:25:06,525 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.9049, 4.9303, 5.4851, 5.4641, 5.4311, 5.0544, 4.9902, 4.7171], device='cuda:1'), covar=tensor([0.0291, 0.0493, 0.0307, 0.0357, 0.0366, 0.0297, 0.0850, 0.0422], device='cuda:1'), in_proj_covar=tensor([0.0335, 0.0339, 0.0338, 0.0326, 0.0380, 0.0356, 0.0463, 0.0286], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:1') 2023-04-29 04:25:11,136 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.9691, 4.3925, 4.5738, 3.2490, 3.9567, 4.3585, 4.1502, 2.9490], device='cuda:1'), covar=tensor([0.0319, 0.0042, 0.0021, 0.0237, 0.0061, 0.0058, 0.0048, 0.0289], device='cuda:1'), in_proj_covar=tensor([0.0128, 0.0070, 0.0068, 0.0125, 0.0077, 0.0084, 0.0075, 0.0120], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:1') 2023-04-29 04:25:39,050 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=92595.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 04:25:47,778 INFO [train.py:904] (1/8) Epoch 10, batch 1250, loss[loss=0.1564, simple_loss=0.2378, pruned_loss=0.03749, over 16990.00 frames. ], tot_loss[loss=0.1821, simple_loss=0.2624, pruned_loss=0.05084, over 3302136.17 frames. ], batch size: 41, lr: 6.99e-03, grad_scale: 8.0 2023-04-29 04:26:01,768 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=92611.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 04:26:20,432 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=92625.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 04:26:27,272 INFO [optim.py:368] (1/8) 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 04:26:37,477 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=92637.0, num_to_drop=1, layers_to_drop={2} 2023-04-29 04:26:58,797 INFO [train.py:904] (1/8) Epoch 10, batch 1300, loss[loss=0.1736, simple_loss=0.2663, pruned_loss=0.04051, over 17196.00 frames. ], tot_loss[loss=0.1813, simple_loss=0.2619, pruned_loss=0.05039, over 3300986.28 frames. ], batch size: 46, lr: 6.99e-03, grad_scale: 8.0 2023-04-29 04:27:07,401 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=92659.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 04:27:18,793 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.9016, 1.7795, 2.2887, 2.8311, 2.7954, 2.9809, 2.0096, 2.9903], device='cuda:1'), covar=tensor([0.0129, 0.0349, 0.0231, 0.0182, 0.0175, 0.0143, 0.0320, 0.0093], device='cuda:1'), in_proj_covar=tensor([0.0158, 0.0170, 0.0155, 0.0157, 0.0165, 0.0121, 0.0170, 0.0111], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001], device='cuda:1') 2023-04-29 04:27:44,063 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=92685.0, num_to_drop=1, layers_to_drop={0} 2023-04-29 04:27:45,288 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=92686.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 04:28:07,567 INFO [train.py:904] (1/8) Epoch 10, batch 1350, loss[loss=0.1655, simple_loss=0.2597, pruned_loss=0.03567, over 17271.00 frames. ], tot_loss[loss=0.1818, simple_loss=0.2626, pruned_loss=0.05049, over 3306264.10 frames. ], batch size: 52, lr: 6.99e-03, grad_scale: 8.0 2023-04-29 04:28:34,049 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=92720.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 04:28:47,046 INFO [optim.py:368] (1/8) 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 04:29:05,935 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.72 vs. limit=2.0 2023-04-29 04:29:18,984 INFO [train.py:904] (1/8) Epoch 10, batch 1400, loss[loss=0.1407, simple_loss=0.2227, pruned_loss=0.02935, over 16819.00 frames. ], tot_loss[loss=0.1814, simple_loss=0.2622, pruned_loss=0.05027, over 3307170.72 frames. ], batch size: 39, lr: 6.98e-03, grad_scale: 8.0 2023-04-29 04:29:19,584 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.1336, 4.5615, 3.6957, 2.5522, 3.1552, 2.6331, 4.8079, 4.1155], device='cuda:1'), covar=tensor([0.2218, 0.0552, 0.1170, 0.2038, 0.2301, 0.1644, 0.0315, 0.0883], device='cuda:1'), in_proj_covar=tensor([0.0300, 0.0258, 0.0281, 0.0273, 0.0276, 0.0219, 0.0266, 0.0294], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-29 04:29:37,196 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([1.8329, 3.5187, 3.0250, 5.2334, 4.5124, 4.7576, 1.6934, 3.4883], device='cuda:1'), covar=tensor([0.1331, 0.0523, 0.1013, 0.0122, 0.0228, 0.0309, 0.1437, 0.0645], device='cuda:1'), in_proj_covar=tensor([0.0152, 0.0155, 0.0178, 0.0136, 0.0196, 0.0212, 0.0178, 0.0177], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-04-29 04:30:00,001 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=92781.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 04:30:14,811 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.9395, 4.0919, 2.2579, 4.7064, 2.8822, 4.5965, 2.4246, 3.3390], device='cuda:1'), covar=tensor([0.0217, 0.0325, 0.1488, 0.0152, 0.0781, 0.0372, 0.1320, 0.0596], device='cuda:1'), in_proj_covar=tensor([0.0148, 0.0167, 0.0189, 0.0125, 0.0167, 0.0206, 0.0195, 0.0172], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-04-29 04:30:25,822 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=4.88 vs. limit=5.0 2023-04-29 04:30:28,852 INFO [train.py:904] (1/8) Epoch 10, batch 1450, loss[loss=0.1623, simple_loss=0.2467, pruned_loss=0.03894, over 16973.00 frames. ], tot_loss[loss=0.1805, simple_loss=0.2611, pruned_loss=0.04993, over 3313646.20 frames. ], batch size: 41, lr: 6.98e-03, grad_scale: 8.0 2023-04-29 04:31:07,980 INFO [optim.py:368] (1/8) 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 04:31:09,672 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=92831.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 04:31:37,439 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.9226, 4.9494, 5.4599, 5.4512, 5.4215, 5.0693, 4.9995, 4.7049], device='cuda:1'), covar=tensor([0.0311, 0.0543, 0.0363, 0.0418, 0.0406, 0.0304, 0.0895, 0.0422], device='cuda:1'), in_proj_covar=tensor([0.0336, 0.0342, 0.0343, 0.0328, 0.0384, 0.0359, 0.0469, 0.0287], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:1') 2023-04-29 04:31:38,220 INFO [train.py:904] (1/8) Epoch 10, batch 1500, loss[loss=0.1792, simple_loss=0.2705, pruned_loss=0.04395, over 17169.00 frames. ], tot_loss[loss=0.1818, simple_loss=0.2617, pruned_loss=0.05102, over 3306416.95 frames. ], batch size: 46, lr: 6.98e-03, grad_scale: 8.0 2023-04-29 04:32:34,154 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=92892.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 04:32:45,558 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=92899.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 04:32:48,540 INFO [train.py:904] (1/8) Epoch 10, batch 1550, loss[loss=0.2168, simple_loss=0.2782, pruned_loss=0.07766, over 16714.00 frames. ], tot_loss[loss=0.1832, simple_loss=0.2628, pruned_loss=0.05177, over 3311911.46 frames. ], batch size: 124, lr: 6.98e-03, grad_scale: 8.0 2023-04-29 04:32:56,755 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.0040, 1.9387, 2.3899, 2.7900, 2.8235, 2.9328, 2.0198, 3.1442], device='cuda:1'), covar=tensor([0.0109, 0.0289, 0.0199, 0.0184, 0.0149, 0.0159, 0.0293, 0.0079], device='cuda:1'), in_proj_covar=tensor([0.0155, 0.0167, 0.0152, 0.0155, 0.0163, 0.0120, 0.0168, 0.0108], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001], device='cuda:1') 2023-04-29 04:33:25,411 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([5.1752, 4.8653, 5.0966, 5.4015, 5.5795, 4.8301, 5.5096, 5.4938], device='cuda:1'), covar=tensor([0.1327, 0.1060, 0.1582, 0.0635, 0.0451, 0.0716, 0.0414, 0.0457], device='cuda:1'), in_proj_covar=tensor([0.0540, 0.0662, 0.0812, 0.0678, 0.0505, 0.0508, 0.0527, 0.0600], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-29 04:33:26,145 INFO [optim.py:368] (1/8) 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 04:33:27,727 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=92931.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 04:33:45,330 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.72 vs. limit=2.0 2023-04-29 04:33:52,821 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([1.7924, 2.9372, 2.5402, 4.2433, 3.6643, 4.1572, 1.5426, 3.0625], device='cuda:1'), covar=tensor([0.1242, 0.0524, 0.1030, 0.0128, 0.0173, 0.0342, 0.1342, 0.0644], device='cuda:1'), in_proj_covar=tensor([0.0152, 0.0154, 0.0177, 0.0136, 0.0195, 0.0211, 0.0176, 0.0176], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0003, 0.0004], device='cuda:1') 2023-04-29 04:33:56,442 INFO [train.py:904] (1/8) Epoch 10, batch 1600, loss[loss=0.1788, simple_loss=0.253, pruned_loss=0.0523, over 16756.00 frames. ], tot_loss[loss=0.1855, simple_loss=0.2655, pruned_loss=0.05278, over 3314243.49 frames. ], batch size: 83, lr: 6.98e-03, grad_scale: 8.0 2023-04-29 04:34:07,751 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=92960.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 04:34:29,039 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=92975.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 04:34:37,285 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=92981.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 04:34:52,602 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=92992.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 04:34:58,127 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.85 vs. limit=2.0 2023-04-29 04:35:06,232 INFO [train.py:904] (1/8) Epoch 10, batch 1650, loss[loss=0.1724, simple_loss=0.266, pruned_loss=0.03935, over 17121.00 frames. ], tot_loss[loss=0.1858, simple_loss=0.2658, pruned_loss=0.05289, over 3321968.61 frames. ], batch size: 49, lr: 6.97e-03, grad_scale: 8.0 2023-04-29 04:35:18,211 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=93010.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 04:35:45,199 INFO [optim.py:368] (1/8) 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 04:35:53,161 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=93036.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 04:35:53,174 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=93036.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 04:35:54,887 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.63 vs. limit=2.0 2023-04-29 04:36:15,639 INFO [train.py:904] (1/8) Epoch 10, batch 1700, loss[loss=0.211, simple_loss=0.2904, pruned_loss=0.06579, over 16829.00 frames. ], tot_loss[loss=0.1883, simple_loss=0.2687, pruned_loss=0.05391, over 3317667.74 frames. ], batch size: 96, lr: 6.97e-03, grad_scale: 8.0 2023-04-29 04:36:42,615 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=93071.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 04:36:48,265 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=93076.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 04:37:17,812 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=93097.0, num_to_drop=1, layers_to_drop={3} 2023-04-29 04:37:23,692 INFO [train.py:904] (1/8) Epoch 10, batch 1750, loss[loss=0.1692, simple_loss=0.2531, pruned_loss=0.04263, over 17185.00 frames. ], tot_loss[loss=0.1884, simple_loss=0.2694, pruned_loss=0.05366, over 3317826.77 frames. ], batch size: 44, lr: 6.97e-03, grad_scale: 8.0 2023-04-29 04:37:42,855 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=93115.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 04:37:46,185 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([5.0705, 4.8059, 5.0553, 5.2818, 5.4946, 4.8377, 5.4509, 5.4334], device='cuda:1'), covar=tensor([0.1413, 0.1062, 0.1521, 0.0665, 0.0480, 0.0613, 0.0428, 0.0502], device='cuda:1'), in_proj_covar=tensor([0.0536, 0.0653, 0.0804, 0.0668, 0.0500, 0.0503, 0.0520, 0.0593], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-29 04:37:56,752 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.73 vs. limit=2.0 2023-04-29 04:38:01,414 INFO [optim.py:368] (1/8) Clipping_scale=2.0, grad-norm quartiles 1.528e+02 2.469e+02 2.889e+02 3.646e+02 7.131e+02, threshold=5.778e+02, percent-clipped=4.0 2023-04-29 04:38:32,562 INFO [train.py:904] (1/8) Epoch 10, batch 1800, loss[loss=0.2572, simple_loss=0.3232, pruned_loss=0.09561, over 11882.00 frames. ], tot_loss[loss=0.1892, simple_loss=0.2708, pruned_loss=0.05379, over 3314829.70 frames. ], batch size: 246, lr: 6.97e-03, grad_scale: 8.0 2023-04-29 04:39:06,717 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=93176.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 04:39:20,429 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 10, batch 1850, loss[loss=0.1867, simple_loss=0.2788, pruned_loss=0.04726, over 17288.00 frames. ], tot_loss[loss=0.1894, simple_loss=0.2713, pruned_loss=0.05375, over 3324304.86 frames. ], batch size: 52, lr: 6.97e-03, grad_scale: 8.0 2023-04-29 04:39:56,096 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.4459, 3.6723, 3.8413, 1.7999, 3.9450, 3.9901, 3.1765, 2.8024], device='cuda:1'), covar=tensor([0.0761, 0.0128, 0.0110, 0.1132, 0.0068, 0.0108, 0.0353, 0.0401], device='cuda:1'), in_proj_covar=tensor([0.0146, 0.0099, 0.0087, 0.0141, 0.0070, 0.0101, 0.0122, 0.0129], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0003, 0.0002, 0.0004, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:1') 2023-04-29 04:40:21,084 INFO [optim.py:368] (1/8) Clipping_scale=2.0, grad-norm quartiles 1.739e+02 2.500e+02 2.909e+02 3.446e+02 8.007e+02, threshold=5.817e+02, percent-clipped=2.0 2023-04-29 04:40:39,261 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.4957, 3.4777, 3.3877, 2.7921, 3.4071, 1.9713, 3.1306, 2.8448], device='cuda:1'), covar=tensor([0.0114, 0.0095, 0.0146, 0.0210, 0.0079, 0.2076, 0.0115, 0.0193], device='cuda:1'), in_proj_covar=tensor([0.0131, 0.0116, 0.0168, 0.0158, 0.0136, 0.0181, 0.0154, 0.0159], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-29 04:40:52,062 INFO [train.py:904] (1/8) Epoch 10, batch 1900, loss[loss=0.2069, simple_loss=0.2753, pruned_loss=0.06922, over 16689.00 frames. ], tot_loss[loss=0.1895, simple_loss=0.271, pruned_loss=0.05401, over 3324351.83 frames. ], batch size: 124, lr: 6.97e-03, grad_scale: 8.0 2023-04-29 04:40:56,694 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=93255.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 04:41:33,303 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=93281.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 04:41:41,113 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=93287.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 04:41:58,938 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.4344, 3.6290, 3.9745, 2.7126, 3.5719, 3.9082, 3.7199, 2.5073], device='cuda:1'), covar=tensor([0.0360, 0.0158, 0.0034, 0.0280, 0.0074, 0.0070, 0.0056, 0.0314], device='cuda:1'), in_proj_covar=tensor([0.0125, 0.0069, 0.0067, 0.0123, 0.0076, 0.0083, 0.0074, 0.0118], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:1') 2023-04-29 04:42:02,724 INFO [train.py:904] (1/8) Epoch 10, batch 1950, loss[loss=0.1984, simple_loss=0.2868, pruned_loss=0.05498, over 16706.00 frames. ], tot_loss[loss=0.1892, simple_loss=0.2712, pruned_loss=0.05361, over 3313290.48 frames. ], batch size: 57, lr: 6.96e-03, grad_scale: 8.0 2023-04-29 04:42:39,968 INFO [zipformer.py:625] (1/8) 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] (1/8) 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] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=93331.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 04:43:12,557 INFO [train.py:904] (1/8) Epoch 10, batch 2000, loss[loss=0.1653, simple_loss=0.2519, pruned_loss=0.03935, over 16872.00 frames. ], tot_loss[loss=0.1888, simple_loss=0.2709, pruned_loss=0.05333, over 3305229.50 frames. ], batch size: 42, lr: 6.96e-03, grad_scale: 8.0 2023-04-29 04:43:31,505 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=93366.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 04:43:44,586 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=93376.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 04:44:01,513 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=4.88 vs. limit=5.0 2023-04-29 04:44:08,369 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=93392.0, num_to_drop=1, layers_to_drop={1} 2023-04-29 04:44:21,503 INFO [train.py:904] (1/8) Epoch 10, batch 2050, loss[loss=0.2004, simple_loss=0.2726, pruned_loss=0.06412, over 16856.00 frames. ], tot_loss[loss=0.1895, simple_loss=0.272, pruned_loss=0.05348, over 3299076.71 frames. ], batch size: 116, lr: 6.96e-03, grad_scale: 4.0 2023-04-29 04:44:51,420 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=93424.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 04:45:00,990 INFO [optim.py:368] (1/8) Clipping_scale=2.0, grad-norm quartiles 1.855e+02 2.531e+02 2.855e+02 3.295e+02 5.942e+02, threshold=5.709e+02, percent-clipped=0.0 2023-04-29 04:45:29,946 INFO [train.py:904] (1/8) Epoch 10, batch 2100, loss[loss=0.1671, simple_loss=0.2546, pruned_loss=0.03979, over 16846.00 frames. ], tot_loss[loss=0.1903, simple_loss=0.2725, pruned_loss=0.0541, over 3293316.49 frames. ], batch size: 42, lr: 6.96e-03, grad_scale: 4.0 2023-04-29 04:45:56,192 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=93471.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 04:46:18,620 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=93487.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 04:46:40,125 INFO [train.py:904] (1/8) Epoch 10, batch 2150, loss[loss=0.1941, simple_loss=0.2769, pruned_loss=0.05564, over 16459.00 frames. ], tot_loss[loss=0.191, simple_loss=0.2729, pruned_loss=0.05459, over 3295386.04 frames. ], batch size: 68, lr: 6.96e-03, grad_scale: 4.0 2023-04-29 04:47:18,313 INFO [optim.py:368] (1/8) 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 04:47:18,993 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.9558, 2.2618, 2.4296, 4.8580, 2.2412, 2.7656, 2.3980, 2.5407], device='cuda:1'), covar=tensor([0.0813, 0.3346, 0.2189, 0.0285, 0.3602, 0.2297, 0.2875, 0.3191], device='cuda:1'), in_proj_covar=tensor([0.0363, 0.0388, 0.0325, 0.0326, 0.0410, 0.0440, 0.0348, 0.0456], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-29 04:47:23,865 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.24 vs. limit=2.0 2023-04-29 04:47:24,632 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 10, batch 2200, loss[loss=0.2217, simple_loss=0.2871, pruned_loss=0.07821, over 16706.00 frames. ], tot_loss[loss=0.1906, simple_loss=0.2729, pruned_loss=0.05417, over 3308889.02 frames. ], batch size: 134, lr: 6.95e-03, grad_scale: 4.0 2023-04-29 04:47:51,940 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=93555.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 04:48:35,541 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=93587.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 04:48:44,285 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.88 vs. limit=2.0 2023-04-29 04:48:54,025 INFO [train.py:904] (1/8) Epoch 10, batch 2250, loss[loss=0.2114, simple_loss=0.2974, pruned_loss=0.06271, over 16692.00 frames. ], tot_loss[loss=0.1914, simple_loss=0.2733, pruned_loss=0.05474, over 3315601.88 frames. ], batch size: 57, lr: 6.95e-03, grad_scale: 4.0 2023-04-29 04:48:55,410 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=93603.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 04:49:21,507 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=3.79 vs. limit=5.0 2023-04-29 04:49:33,830 INFO [optim.py:368] (1/8) 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 04:49:34,847 INFO [zipformer.py:625] (1/8) 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] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=93635.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 04:50:04,001 INFO [train.py:904] (1/8) Epoch 10, batch 2300, loss[loss=0.1525, simple_loss=0.247, pruned_loss=0.029, over 16870.00 frames. ], tot_loss[loss=0.1912, simple_loss=0.2732, pruned_loss=0.05462, over 3322553.49 frames. ], batch size: 42, lr: 6.95e-03, grad_scale: 4.0 2023-04-29 04:50:22,491 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=93666.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 04:50:33,963 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.4314, 2.1660, 2.2528, 4.2284, 2.1382, 2.6641, 2.1940, 2.3834], device='cuda:1'), covar=tensor([0.0948, 0.2829, 0.1942, 0.0366, 0.3194, 0.1960, 0.2891, 0.2412], device='cuda:1'), in_proj_covar=tensor([0.0362, 0.0386, 0.0326, 0.0325, 0.0409, 0.0440, 0.0348, 0.0455], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-29 04:50:39,710 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=93679.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 04:50:52,822 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.9807, 5.3590, 5.5287, 5.3489, 5.3299, 5.9719, 5.5423, 5.2815], device='cuda:1'), covar=tensor([0.0872, 0.2053, 0.1600, 0.1864, 0.2592, 0.0892, 0.1169, 0.2170], device='cuda:1'), in_proj_covar=tensor([0.0346, 0.0498, 0.0521, 0.0428, 0.0576, 0.0552, 0.0419, 0.0578], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-04-29 04:50:57,916 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=93692.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 04:51:09,768 INFO [train.py:904] (1/8) Epoch 10, batch 2350, loss[loss=0.2243, simple_loss=0.3162, pruned_loss=0.06622, over 16755.00 frames. ], tot_loss[loss=0.1911, simple_loss=0.2733, pruned_loss=0.05447, over 3330573.77 frames. ], batch size: 57, lr: 6.95e-03, grad_scale: 4.0 2023-04-29 04:51:27,698 INFO [zipformer.py:625] (1/8) 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] (1/8) 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 04:52:03,151 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 10, batch 2400, loss[loss=0.205, simple_loss=0.2931, pruned_loss=0.05846, over 16700.00 frames. ], tot_loss[loss=0.1924, simple_loss=0.2747, pruned_loss=0.05509, over 3332752.08 frames. ], batch size: 62, lr: 6.95e-03, grad_scale: 8.0 2023-04-29 04:52:40,800 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=93770.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 04:52:42,519 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=93771.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 04:53:26,813 INFO [train.py:904] (1/8) Epoch 10, batch 2450, loss[loss=0.2465, simple_loss=0.3212, pruned_loss=0.08594, over 11985.00 frames. ], tot_loss[loss=0.1923, simple_loss=0.2752, pruned_loss=0.05473, over 3332425.81 frames. ], batch size: 246, lr: 6.94e-03, grad_scale: 8.0 2023-04-29 04:53:33,037 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([5.1935, 5.0665, 5.0172, 4.6255, 4.5632, 5.0345, 5.0871, 4.6522], device='cuda:1'), covar=tensor([0.0535, 0.0441, 0.0249, 0.0299, 0.1144, 0.0448, 0.0256, 0.0694], device='cuda:1'), in_proj_covar=tensor([0.0251, 0.0311, 0.0296, 0.0274, 0.0323, 0.0310, 0.0205, 0.0349], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-04-29 04:53:37,450 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.9858, 5.0040, 5.4683, 5.4580, 5.4852, 5.1236, 5.0928, 4.7971], device='cuda:1'), covar=tensor([0.0281, 0.0392, 0.0338, 0.0397, 0.0364, 0.0281, 0.0809, 0.0348], device='cuda:1'), in_proj_covar=tensor([0.0330, 0.0339, 0.0342, 0.0326, 0.0381, 0.0357, 0.0463, 0.0286], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:1') 2023-04-29 04:53:45,357 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([5.2168, 5.2146, 5.0190, 4.4447, 5.0621, 2.0334, 4.8170, 5.0091], device='cuda:1'), covar=tensor([0.0064, 0.0053, 0.0123, 0.0347, 0.0077, 0.2107, 0.0116, 0.0149], device='cuda:1'), in_proj_covar=tensor([0.0130, 0.0117, 0.0167, 0.0158, 0.0136, 0.0180, 0.0154, 0.0158], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-29 04:53:49,969 INFO [zipformer.py:625] (1/8) 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] (1/8) 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 04:54:05,484 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=93831.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 04:54:34,536 INFO [train.py:904] (1/8) Epoch 10, batch 2500, loss[loss=0.1825, simple_loss=0.2795, pruned_loss=0.04271, over 17102.00 frames. ], tot_loss[loss=0.1911, simple_loss=0.2746, pruned_loss=0.05384, over 3330715.82 frames. ], batch size: 49, lr: 6.94e-03, grad_scale: 8.0 2023-04-29 04:55:43,655 INFO [train.py:904] (1/8) Epoch 10, batch 2550, loss[loss=0.2285, simple_loss=0.3101, pruned_loss=0.07351, over 15521.00 frames. ], tot_loss[loss=0.1924, simple_loss=0.2754, pruned_loss=0.05468, over 3325717.74 frames. ], batch size: 190, lr: 6.94e-03, grad_scale: 8.0 2023-04-29 04:55:50,271 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=4.73 vs. limit=5.0 2023-04-29 04:56:23,962 INFO [optim.py:368] (1/8) 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 04:56:42,177 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([5.3711, 5.3274, 5.1184, 4.5927, 5.1805, 2.0635, 4.9079, 5.1712], device='cuda:1'), covar=tensor([0.0060, 0.0055, 0.0124, 0.0336, 0.0069, 0.2081, 0.0102, 0.0135], device='cuda:1'), in_proj_covar=tensor([0.0132, 0.0118, 0.0168, 0.0160, 0.0138, 0.0181, 0.0155, 0.0161], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-29 04:56:52,858 INFO [train.py:904] (1/8) Epoch 10, batch 2600, loss[loss=0.1517, simple_loss=0.2446, pruned_loss=0.02943, over 17203.00 frames. ], tot_loss[loss=0.1911, simple_loss=0.2748, pruned_loss=0.05372, over 3325804.76 frames. ], batch size: 44, lr: 6.94e-03, grad_scale: 8.0 2023-04-29 04:57:32,336 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=93981.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 04:58:03,851 INFO [train.py:904] (1/8) Epoch 10, batch 2650, loss[loss=0.1777, simple_loss=0.2654, pruned_loss=0.04504, over 16786.00 frames. ], tot_loss[loss=0.1911, simple_loss=0.2751, pruned_loss=0.05354, over 3327970.33 frames. ], batch size: 39, lr: 6.94e-03, grad_scale: 8.0 2023-04-29 04:58:43,800 INFO [optim.py:368] (1/8) Clipping_scale=2.0, grad-norm quartiles 1.633e+02 2.229e+02 2.744e+02 3.271e+02 8.724e+02, threshold=5.488e+02, percent-clipped=1.0 2023-04-29 04:59:00,259 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 10, batch 2700, loss[loss=0.1658, simple_loss=0.262, pruned_loss=0.03484, over 17018.00 frames. ], tot_loss[loss=0.1903, simple_loss=0.2752, pruned_loss=0.05271, over 3326851.78 frames. ], batch size: 50, lr: 6.94e-03, grad_scale: 8.0 2023-04-29 04:59:14,748 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.50 vs. limit=2.0 2023-04-29 04:59:16,270 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.25 vs. limit=2.0 2023-04-29 04:59:47,251 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.74 vs. limit=2.0 2023-04-29 04:59:58,732 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([5.2219, 5.0366, 5.2566, 5.4731, 5.6272, 4.8593, 5.5712, 5.5891], device='cuda:1'), covar=tensor([0.1224, 0.0943, 0.1220, 0.0491, 0.0401, 0.0721, 0.0483, 0.0413], device='cuda:1'), in_proj_covar=tensor([0.0540, 0.0661, 0.0812, 0.0670, 0.0499, 0.0519, 0.0530, 0.0599], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-29 05:00:01,165 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.2344, 3.2957, 1.8568, 3.5069, 2.4026, 3.4470, 1.8485, 2.6732], device='cuda:1'), covar=tensor([0.0213, 0.0389, 0.1294, 0.0198, 0.0733, 0.0609, 0.1379, 0.0530], device='cuda:1'), in_proj_covar=tensor([0.0148, 0.0167, 0.0187, 0.0130, 0.0168, 0.0210, 0.0196, 0.0172], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-04-29 05:00:23,305 INFO [train.py:904] (1/8) Epoch 10, batch 2750, loss[loss=0.179, simple_loss=0.273, pruned_loss=0.0425, over 17059.00 frames. ], tot_loss[loss=0.1901, simple_loss=0.2751, pruned_loss=0.05254, over 3324615.69 frames. ], batch size: 50, lr: 6.93e-03, grad_scale: 8.0 2023-04-29 05:00:55,374 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=94126.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 05:01:00,212 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([5.0876, 5.6910, 5.7950, 5.5884, 5.6042, 6.1439, 5.7058, 5.4751], device='cuda:1'), covar=tensor([0.0835, 0.1729, 0.1620, 0.1997, 0.2624, 0.0896, 0.1341, 0.2394], device='cuda:1'), in_proj_covar=tensor([0.0345, 0.0490, 0.0515, 0.0426, 0.0566, 0.0546, 0.0417, 0.0570], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-04-29 05:01:01,088 INFO [optim.py:368] (1/8) Clipping_scale=2.0, grad-norm quartiles 1.563e+02 2.410e+02 2.958e+02 3.419e+02 5.641e+02, threshold=5.917e+02, percent-clipped=1.0 2023-04-29 05:01:12,706 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=4.97 vs. limit=5.0 2023-04-29 05:01:18,055 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.3040, 3.9138, 3.8764, 2.1817, 3.2344, 2.5363, 3.8968, 3.8793], device='cuda:1'), covar=tensor([0.0286, 0.0690, 0.0510, 0.1566, 0.0697, 0.0861, 0.0589, 0.0972], device='cuda:1'), in_proj_covar=tensor([0.0143, 0.0144, 0.0157, 0.0142, 0.0135, 0.0124, 0.0136, 0.0155], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-04-29 05:01:29,686 INFO [train.py:904] (1/8) Epoch 10, batch 2800, loss[loss=0.1553, simple_loss=0.2402, pruned_loss=0.03524, over 16831.00 frames. ], tot_loss[loss=0.1904, simple_loss=0.2751, pruned_loss=0.05289, over 3321651.71 frames. ], batch size: 39, lr: 6.93e-03, grad_scale: 8.0 2023-04-29 05:02:39,399 INFO [train.py:904] (1/8) Epoch 10, batch 2850, loss[loss=0.2186, simple_loss=0.2819, pruned_loss=0.07763, over 16672.00 frames. ], tot_loss[loss=0.1895, simple_loss=0.2739, pruned_loss=0.05253, over 3322809.20 frames. ], batch size: 89, lr: 6.93e-03, grad_scale: 8.0 2023-04-29 05:03:08,983 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=94223.0, num_to_drop=1, layers_to_drop={1} 2023-04-29 05:03:20,117 INFO [optim.py:368] (1/8) 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 05:03:49,058 INFO [train.py:904] (1/8) Epoch 10, batch 2900, loss[loss=0.2804, simple_loss=0.3288, pruned_loss=0.116, over 12017.00 frames. ], tot_loss[loss=0.1885, simple_loss=0.2724, pruned_loss=0.05231, over 3327663.74 frames. ], batch size: 247, lr: 6.93e-03, grad_scale: 8.0 2023-04-29 05:03:56,165 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.19 vs. limit=2.0 2023-04-29 05:04:19,852 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.9622, 5.3278, 5.0366, 5.0762, 4.7553, 4.6528, 4.8325, 5.4270], device='cuda:1'), covar=tensor([0.0971, 0.0854, 0.1122, 0.0688, 0.0795, 0.0957, 0.1035, 0.0839], device='cuda:1'), in_proj_covar=tensor([0.0529, 0.0678, 0.0557, 0.0462, 0.0424, 0.0431, 0.0558, 0.0513], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-29 05:04:33,871 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=94284.0, num_to_drop=1, layers_to_drop={1} 2023-04-29 05:04:51,587 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.4822, 4.5060, 4.4235, 4.2383, 3.8112, 4.5362, 4.3703, 4.1277], device='cuda:1'), covar=tensor([0.0777, 0.0669, 0.0358, 0.0328, 0.1319, 0.0466, 0.0540, 0.0705], device='cuda:1'), in_proj_covar=tensor([0.0255, 0.0314, 0.0297, 0.0275, 0.0327, 0.0312, 0.0205, 0.0352], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-04-29 05:04:58,055 INFO [train.py:904] (1/8) Epoch 10, batch 2950, loss[loss=0.196, simple_loss=0.2643, pruned_loss=0.06388, over 16691.00 frames. ], tot_loss[loss=0.1897, simple_loss=0.2724, pruned_loss=0.05354, over 3315695.97 frames. ], batch size: 124, lr: 6.93e-03, grad_scale: 8.0 2023-04-29 05:05:39,543 INFO [optim.py:368] (1/8) 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 05:05:49,272 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=94337.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 05:06:03,109 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=94348.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 05:06:08,014 INFO [train.py:904] (1/8) Epoch 10, batch 3000, loss[loss=0.1839, simple_loss=0.2755, pruned_loss=0.04614, over 17119.00 frames. ], tot_loss[loss=0.1912, simple_loss=0.2734, pruned_loss=0.05448, over 3313010.05 frames. ], batch size: 47, lr: 6.92e-03, grad_scale: 8.0 2023-04-29 05:06:08,015 INFO [train.py:929] (1/8) Computing validation loss 2023-04-29 05:06:17,138 INFO [train.py:938] (1/8) Epoch 10, validation: loss=0.1426, simple_loss=0.2488, pruned_loss=0.01818, over 944034.00 frames. 2023-04-29 05:06:17,138 INFO [train.py:939] (1/8) Maximum memory allocated so far is 17676MB 2023-04-29 05:07:26,707 INFO [train.py:904] (1/8) Epoch 10, batch 3050, loss[loss=0.1834, simple_loss=0.2661, pruned_loss=0.05037, over 17183.00 frames. ], tot_loss[loss=0.1917, simple_loss=0.2737, pruned_loss=0.05483, over 3306513.67 frames. ], batch size: 46, lr: 6.92e-03, grad_scale: 8.0 2023-04-29 05:07:36,823 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=94409.0, num_to_drop=1, layers_to_drop={3} 2023-04-29 05:07:57,829 INFO [zipformer.py:625] (1/8) 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] (1/8) 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 05:08:28,962 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([5.3612, 5.3973, 5.1734, 4.7953, 4.6847, 5.2897, 5.2990, 4.8292], device='cuda:1'), covar=tensor([0.0577, 0.0356, 0.0269, 0.0273, 0.1190, 0.0361, 0.0212, 0.0653], device='cuda:1'), in_proj_covar=tensor([0.0254, 0.0312, 0.0296, 0.0274, 0.0324, 0.0312, 0.0204, 0.0348], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-04-29 05:08:33,248 INFO [train.py:904] (1/8) Epoch 10, batch 3100, loss[loss=0.1543, simple_loss=0.2357, pruned_loss=0.03642, over 16978.00 frames. ], tot_loss[loss=0.1914, simple_loss=0.2731, pruned_loss=0.05484, over 3303349.21 frames. ], batch size: 41, lr: 6.92e-03, grad_scale: 8.0 2023-04-29 05:09:04,481 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=94474.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 05:09:43,504 INFO [train.py:904] (1/8) Epoch 10, batch 3150, loss[loss=0.184, simple_loss=0.2742, pruned_loss=0.0469, over 16619.00 frames. ], tot_loss[loss=0.1909, simple_loss=0.2722, pruned_loss=0.05479, over 3304625.53 frames. ], batch size: 62, lr: 6.92e-03, grad_scale: 8.0 2023-04-29 05:10:23,683 INFO [optim.py:368] (1/8) 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 05:10:36,354 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.53 vs. limit=2.0 2023-04-29 05:10:52,169 INFO [train.py:904] (1/8) Epoch 10, batch 3200, loss[loss=0.1612, simple_loss=0.2484, pruned_loss=0.037, over 17228.00 frames. ], tot_loss[loss=0.1896, simple_loss=0.2713, pruned_loss=0.05397, over 3311808.30 frames. ], batch size: 44, lr: 6.92e-03, grad_scale: 8.0 2023-04-29 05:11:12,425 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=4.85 vs. limit=5.0 2023-04-29 05:11:32,212 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=94579.0, num_to_drop=1, layers_to_drop={0} 2023-04-29 05:12:04,574 INFO [train.py:904] (1/8) Epoch 10, batch 3250, loss[loss=0.2192, simple_loss=0.2926, pruned_loss=0.07286, over 16367.00 frames. ], tot_loss[loss=0.1901, simple_loss=0.2716, pruned_loss=0.05434, over 3321080.25 frames. ], batch size: 165, lr: 6.92e-03, grad_scale: 8.0 2023-04-29 05:12:44,901 INFO [optim.py:368] (1/8) 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 05:12:53,012 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 10, batch 3300, loss[loss=0.1899, simple_loss=0.2739, pruned_loss=0.053, over 16253.00 frames. ], tot_loss[loss=0.1905, simple_loss=0.2725, pruned_loss=0.05427, over 3326796.93 frames. ], batch size: 165, lr: 6.91e-03, grad_scale: 8.0 2023-04-29 05:14:01,715 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=94685.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 05:14:10,728 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=3.22 vs. limit=5.0 2023-04-29 05:14:24,568 INFO [train.py:904] (1/8) Epoch 10, batch 3350, loss[loss=0.175, simple_loss=0.2628, pruned_loss=0.04361, over 17183.00 frames. ], tot_loss[loss=0.1905, simple_loss=0.2732, pruned_loss=0.05389, over 3314654.59 frames. ], batch size: 46, lr: 6.91e-03, grad_scale: 8.0 2023-04-29 05:14:28,588 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=94704.0, num_to_drop=1, layers_to_drop={3} 2023-04-29 05:14:57,907 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([5.4435, 5.4034, 5.2969, 4.9784, 4.8524, 5.3678, 5.2058, 4.9060], device='cuda:1'), covar=tensor([0.0527, 0.0352, 0.0230, 0.0227, 0.1069, 0.0364, 0.0252, 0.0618], device='cuda:1'), in_proj_covar=tensor([0.0257, 0.0316, 0.0300, 0.0278, 0.0327, 0.0316, 0.0207, 0.0352], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-04-29 05:15:05,006 INFO [optim.py:368] (1/8) Clipping_scale=2.0, grad-norm quartiles 1.683e+02 2.488e+02 2.932e+02 3.912e+02 8.438e+02, threshold=5.863e+02, percent-clipped=4.0 2023-04-29 05:15:35,789 INFO [train.py:904] (1/8) Epoch 10, batch 3400, loss[loss=0.1874, simple_loss=0.2649, pruned_loss=0.05492, over 16869.00 frames. ], tot_loss[loss=0.1909, simple_loss=0.2732, pruned_loss=0.05431, over 3321325.28 frames. ], batch size: 116, lr: 6.91e-03, grad_scale: 8.0 2023-04-29 05:16:01,061 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.93 vs. limit=2.0 2023-04-29 05:16:44,907 INFO [train.py:904] (1/8) Epoch 10, batch 3450, loss[loss=0.1811, simple_loss=0.275, pruned_loss=0.04362, over 17157.00 frames. ], tot_loss[loss=0.1897, simple_loss=0.2716, pruned_loss=0.05397, over 3321525.17 frames. ], batch size: 46, lr: 6.91e-03, grad_scale: 8.0 2023-04-29 05:16:57,655 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.1088, 3.9429, 4.0951, 4.2757, 4.3769, 3.9197, 4.1240, 4.3356], device='cuda:1'), covar=tensor([0.1167, 0.0903, 0.1325, 0.0613, 0.0518, 0.1395, 0.1721, 0.0573], device='cuda:1'), in_proj_covar=tensor([0.0545, 0.0671, 0.0831, 0.0686, 0.0510, 0.0528, 0.0538, 0.0605], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-29 05:17:26,304 INFO [optim.py:368] (1/8) 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 05:17:49,963 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([5.0185, 5.5919, 5.8080, 5.5917, 5.5573, 6.1554, 5.8620, 5.5636], device='cuda:1'), covar=tensor([0.0862, 0.1895, 0.1735, 0.1919, 0.2977, 0.0938, 0.1088, 0.2271], device='cuda:1'), in_proj_covar=tensor([0.0348, 0.0498, 0.0525, 0.0432, 0.0573, 0.0551, 0.0424, 0.0585], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-04-29 05:17:56,543 INFO [train.py:904] (1/8) Epoch 10, batch 3500, loss[loss=0.1467, simple_loss=0.2403, pruned_loss=0.02651, over 17237.00 frames. ], tot_loss[loss=0.1887, simple_loss=0.2703, pruned_loss=0.05353, over 3310489.90 frames. ], batch size: 43, lr: 6.91e-03, grad_scale: 8.0 2023-04-29 05:18:35,814 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([5.2006, 5.1450, 4.9641, 4.4076, 5.0511, 1.8583, 4.7039, 4.9399], device='cuda:1'), covar=tensor([0.0068, 0.0060, 0.0144, 0.0361, 0.0073, 0.2325, 0.0122, 0.0165], device='cuda:1'), in_proj_covar=tensor([0.0134, 0.0121, 0.0170, 0.0164, 0.0140, 0.0181, 0.0160, 0.0162], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-29 05:18:35,817 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=94879.0, num_to_drop=1, layers_to_drop={2} 2023-04-29 05:19:06,951 INFO [train.py:904] (1/8) Epoch 10, batch 3550, loss[loss=0.2033, simple_loss=0.2652, pruned_loss=0.07071, over 16825.00 frames. ], tot_loss[loss=0.1876, simple_loss=0.2693, pruned_loss=0.05294, over 3321982.54 frames. ], batch size: 124, lr: 6.90e-03, grad_scale: 8.0 2023-04-29 05:19:41,450 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=94927.0, num_to_drop=1, layers_to_drop={0} 2023-04-29 05:19:47,557 INFO [optim.py:368] (1/8) 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 05:20:17,532 INFO [train.py:904] (1/8) Epoch 10, batch 3600, loss[loss=0.1498, simple_loss=0.2352, pruned_loss=0.03226, over 16964.00 frames. ], tot_loss[loss=0.1865, simple_loss=0.2683, pruned_loss=0.05233, over 3314219.70 frames. ], batch size: 41, lr: 6.90e-03, grad_scale: 8.0 2023-04-29 05:20:17,955 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.7057, 3.8717, 2.1317, 4.1237, 2.8683, 4.1452, 2.4222, 3.0993], device='cuda:1'), covar=tensor([0.0203, 0.0323, 0.1486, 0.0207, 0.0747, 0.0439, 0.1193, 0.0580], device='cuda:1'), in_proj_covar=tensor([0.0146, 0.0164, 0.0184, 0.0130, 0.0165, 0.0208, 0.0191, 0.0170], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0003, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-04-29 05:20:46,814 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([5.1647, 5.0900, 4.9672, 4.3469, 5.0054, 1.8520, 4.7894, 5.0492], device='cuda:1'), covar=tensor([0.0090, 0.0078, 0.0142, 0.0411, 0.0091, 0.2308, 0.0130, 0.0169], device='cuda:1'), in_proj_covar=tensor([0.0135, 0.0122, 0.0171, 0.0164, 0.0140, 0.0182, 0.0160, 0.0162], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-29 05:21:28,700 INFO [train.py:904] (1/8) Epoch 10, batch 3650, loss[loss=0.1672, simple_loss=0.2399, pruned_loss=0.04724, over 16652.00 frames. ], tot_loss[loss=0.1859, simple_loss=0.2668, pruned_loss=0.05245, over 3296736.51 frames. ], batch size: 89, lr: 6.90e-03, grad_scale: 8.0 2023-04-29 05:21:31,105 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.18 vs. limit=2.0 2023-04-29 05:21:32,928 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=95004.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 05:21:45,767 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.98 vs. limit=2.0 2023-04-29 05:22:10,213 INFO [optim.py:368] (1/8) 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 05:22:22,164 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.6096, 3.9586, 4.1597, 2.2061, 3.2158, 2.6086, 3.8816, 3.9352], device='cuda:1'), covar=tensor([0.0244, 0.0681, 0.0451, 0.1681, 0.0734, 0.0860, 0.0594, 0.0954], device='cuda:1'), in_proj_covar=tensor([0.0144, 0.0146, 0.0155, 0.0143, 0.0135, 0.0124, 0.0137, 0.0156], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:1') 2023-04-29 05:22:43,025 INFO [train.py:904] (1/8) Epoch 10, batch 3700, loss[loss=0.1861, simple_loss=0.2639, pruned_loss=0.05414, over 16239.00 frames. ], tot_loss[loss=0.1874, simple_loss=0.2661, pruned_loss=0.05436, over 3284109.56 frames. ], batch size: 165, lr: 6.90e-03, grad_scale: 8.0 2023-04-29 05:22:43,300 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=95052.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 05:23:39,651 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.94 vs. limit=2.0 2023-04-29 05:23:56,092 INFO [train.py:904] (1/8) Epoch 10, batch 3750, loss[loss=0.2282, simple_loss=0.2888, pruned_loss=0.08385, over 16716.00 frames. ], tot_loss[loss=0.1898, simple_loss=0.2675, pruned_loss=0.05606, over 3267692.05 frames. ], batch size: 134, lr: 6.90e-03, grad_scale: 8.0 2023-04-29 05:23:57,928 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.6380, 3.4699, 2.7933, 2.1982, 2.3830, 2.0954, 3.5754, 3.2583], device='cuda:1'), covar=tensor([0.2185, 0.0648, 0.1388, 0.2196, 0.2260, 0.1867, 0.0477, 0.1150], device='cuda:1'), in_proj_covar=tensor([0.0301, 0.0260, 0.0281, 0.0276, 0.0285, 0.0220, 0.0269, 0.0301], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:1') 2023-04-29 05:24:00,763 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=95105.0, num_to_drop=1, layers_to_drop={1} 2023-04-29 05:24:38,276 INFO [optim.py:368] (1/8) 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 05:25:00,910 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([1.7946, 3.1736, 2.4851, 4.4932, 3.6579, 4.3483, 1.5922, 2.9528], device='cuda:1'), covar=tensor([0.1348, 0.0576, 0.1168, 0.0140, 0.0262, 0.0311, 0.1453, 0.0831], device='cuda:1'), in_proj_covar=tensor([0.0152, 0.0158, 0.0179, 0.0142, 0.0203, 0.0213, 0.0177, 0.0177], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0003, 0.0004], device='cuda:1') 2023-04-29 05:25:07,897 INFO [train.py:904] (1/8) Epoch 10, batch 3800, loss[loss=0.2036, simple_loss=0.2769, pruned_loss=0.06517, over 16865.00 frames. ], tot_loss[loss=0.1914, simple_loss=0.2682, pruned_loss=0.05732, over 3261322.59 frames. ], batch size: 124, lr: 6.90e-03, grad_scale: 8.0 2023-04-29 05:25:28,681 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=95166.0, num_to_drop=1, layers_to_drop={0} 2023-04-29 05:25:57,911 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([1.6488, 2.6285, 2.4704, 4.0826, 3.5237, 4.1074, 1.3049, 3.0199], device='cuda:1'), covar=tensor([0.1254, 0.0585, 0.1054, 0.0137, 0.0180, 0.0271, 0.1392, 0.0646], device='cuda:1'), in_proj_covar=tensor([0.0152, 0.0159, 0.0180, 0.0142, 0.0204, 0.0213, 0.0177, 0.0178], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0003, 0.0004], device='cuda:1') 2023-04-29 05:26:20,914 INFO [train.py:904] (1/8) Epoch 10, batch 3850, loss[loss=0.2294, simple_loss=0.3026, pruned_loss=0.07806, over 12821.00 frames. ], tot_loss[loss=0.1929, simple_loss=0.269, pruned_loss=0.05843, over 3252202.70 frames. ], batch size: 247, lr: 6.89e-03, grad_scale: 8.0 2023-04-29 05:26:22,478 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=3.90 vs. limit=5.0 2023-04-29 05:27:00,982 INFO [optim.py:368] (1/8) 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 05:27:31,951 INFO [train.py:904] (1/8) Epoch 10, batch 3900, loss[loss=0.1652, simple_loss=0.2402, pruned_loss=0.04508, over 16734.00 frames. ], tot_loss[loss=0.1926, simple_loss=0.2681, pruned_loss=0.0586, over 3265547.57 frames. ], batch size: 83, lr: 6.89e-03, grad_scale: 8.0 2023-04-29 05:27:43,983 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=95260.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 05:27:56,143 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=95268.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 05:28:45,299 INFO [train.py:904] (1/8) Epoch 10, batch 3950, loss[loss=0.2048, simple_loss=0.2854, pruned_loss=0.06207, over 15445.00 frames. ], tot_loss[loss=0.1934, simple_loss=0.2683, pruned_loss=0.05923, over 3270213.10 frames. ], batch size: 190, lr: 6.89e-03, grad_scale: 8.0 2023-04-29 05:29:05,180 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=2.75 vs. limit=5.0 2023-04-29 05:29:12,989 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=95321.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 05:29:23,111 INFO [zipformer.py:625] (1/8) 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] (1/8) 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 05:29:38,137 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=95339.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 05:29:56,326 INFO [train.py:904] (1/8) Epoch 10, batch 4000, loss[loss=0.1709, simple_loss=0.2487, pruned_loss=0.04656, over 16847.00 frames. ], tot_loss[loss=0.1941, simple_loss=0.2684, pruned_loss=0.05993, over 3267619.30 frames. ], batch size: 116, lr: 6.89e-03, grad_scale: 8.0 2023-04-29 05:30:01,101 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.7799, 2.4365, 2.3596, 4.7891, 2.2223, 2.8764, 2.3980, 2.6667], device='cuda:1'), covar=tensor([0.0808, 0.2929, 0.1929, 0.0254, 0.3340, 0.1886, 0.2729, 0.2574], device='cuda:1'), in_proj_covar=tensor([0.0365, 0.0392, 0.0326, 0.0328, 0.0411, 0.0450, 0.0354, 0.0465], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-29 05:30:24,799 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.8137, 2.3636, 1.8082, 2.0924, 2.6932, 2.3737, 2.9049, 2.9997], device='cuda:1'), covar=tensor([0.0091, 0.0251, 0.0380, 0.0351, 0.0149, 0.0253, 0.0116, 0.0124], device='cuda:1'), in_proj_covar=tensor([0.0145, 0.0200, 0.0197, 0.0197, 0.0201, 0.0203, 0.0210, 0.0191], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-29 05:31:05,673 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 10, batch 4050, loss[loss=0.1683, simple_loss=0.2567, pruned_loss=0.04002, over 17114.00 frames. ], tot_loss[loss=0.1926, simple_loss=0.2682, pruned_loss=0.05849, over 3267279.13 frames. ], batch size: 48, lr: 6.89e-03, grad_scale: 16.0 2023-04-29 05:31:22,072 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([1.6512, 2.9234, 2.9250, 4.9821, 3.6742, 4.4495, 1.5315, 3.0685], device='cuda:1'), covar=tensor([0.1449, 0.0764, 0.1139, 0.0126, 0.0525, 0.0302, 0.1657, 0.0882], device='cuda:1'), in_proj_covar=tensor([0.0153, 0.0158, 0.0180, 0.0142, 0.0204, 0.0212, 0.0176, 0.0179], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0003, 0.0004], device='cuda:1') 2023-04-29 05:31:49,145 INFO [optim.py:368] (1/8) 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] (1/8) Epoch 10, batch 4100, loss[loss=0.2187, simple_loss=0.3089, pruned_loss=0.0643, over 16838.00 frames. ], tot_loss[loss=0.192, simple_loss=0.2693, pruned_loss=0.0574, over 3275449.57 frames. ], batch size: 116, lr: 6.88e-03, grad_scale: 16.0 2023-04-29 05:32:34,841 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=95461.0, num_to_drop=1, layers_to_drop={1} 2023-04-29 05:32:49,408 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([5.0624, 5.8487, 6.1179, 5.7166, 5.8075, 6.3163, 5.9836, 5.6607], device='cuda:1'), covar=tensor([0.0737, 0.1179, 0.1368, 0.1368, 0.2015, 0.0786, 0.1030, 0.1672], device='cuda:1'), in_proj_covar=tensor([0.0343, 0.0488, 0.0512, 0.0423, 0.0556, 0.0537, 0.0416, 0.0570], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-04-29 05:33:33,930 INFO [train.py:904] (1/8) Epoch 10, batch 4150, loss[loss=0.2179, simple_loss=0.3078, pruned_loss=0.06401, over 16739.00 frames. ], tot_loss[loss=0.1993, simple_loss=0.2773, pruned_loss=0.06066, over 3246875.22 frames. ], batch size: 76, lr: 6.88e-03, grad_scale: 16.0 2023-04-29 05:33:52,726 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.5283, 3.5210, 2.7918, 2.0416, 2.6608, 2.2519, 3.8020, 3.3292], device='cuda:1'), covar=tensor([0.2440, 0.0851, 0.1478, 0.2190, 0.2065, 0.1673, 0.0447, 0.1032], device='cuda:1'), in_proj_covar=tensor([0.0298, 0.0258, 0.0282, 0.0276, 0.0288, 0.0219, 0.0269, 0.0300], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:1') 2023-04-29 05:34:00,781 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.34 vs. limit=2.0 2023-04-29 05:34:17,107 INFO [optim.py:368] (1/8) 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 05:34:43,977 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.1194, 3.3948, 3.5517, 3.5104, 3.5077, 3.2859, 3.3762, 3.3575], device='cuda:1'), covar=tensor([0.0461, 0.0553, 0.0434, 0.0494, 0.0535, 0.0494, 0.0791, 0.0513], device='cuda:1'), in_proj_covar=tensor([0.0325, 0.0331, 0.0333, 0.0318, 0.0378, 0.0354, 0.0456, 0.0281], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:1') 2023-04-29 05:34:49,630 INFO [train.py:904] (1/8) Epoch 10, batch 4200, loss[loss=0.2302, simple_loss=0.323, pruned_loss=0.06873, over 16789.00 frames. ], tot_loss[loss=0.2054, simple_loss=0.2849, pruned_loss=0.06294, over 3208671.15 frames. ], batch size: 89, lr: 6.88e-03, grad_scale: 16.0 2023-04-29 05:34:50,570 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.55 vs. limit=2.0 2023-04-29 05:35:32,877 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.5386, 4.8010, 4.5601, 4.5923, 4.2899, 4.2858, 4.1659, 4.8558], device='cuda:1'), covar=tensor([0.0832, 0.0749, 0.0900, 0.0601, 0.0723, 0.1063, 0.1013, 0.0737], device='cuda:1'), in_proj_covar=tensor([0.0508, 0.0646, 0.0534, 0.0444, 0.0407, 0.0417, 0.0538, 0.0495], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:1') 2023-04-29 05:35:34,947 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.8421, 3.4807, 3.3410, 1.8558, 2.9284, 2.3386, 3.3179, 3.3238], device='cuda:1'), covar=tensor([0.0285, 0.0526, 0.0499, 0.1808, 0.0685, 0.0852, 0.0653, 0.0769], device='cuda:1'), in_proj_covar=tensor([0.0141, 0.0142, 0.0153, 0.0140, 0.0133, 0.0121, 0.0134, 0.0152], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:1') 2023-04-29 05:36:04,061 INFO [train.py:904] (1/8) Epoch 10, batch 4250, loss[loss=0.1918, simple_loss=0.2931, pruned_loss=0.0452, over 16868.00 frames. ], tot_loss[loss=0.2077, simple_loss=0.2886, pruned_loss=0.06338, over 3171877.32 frames. ], batch size: 96, lr: 6.88e-03, grad_scale: 8.0 2023-04-29 05:36:24,748 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=95616.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 05:36:26,056 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=95617.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 05:36:37,902 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=95624.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 05:36:49,130 INFO [optim.py:368] (1/8) 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 05:37:19,468 INFO [train.py:904] (1/8) Epoch 10, batch 4300, loss[loss=0.2264, simple_loss=0.3238, pruned_loss=0.06445, over 17250.00 frames. ], tot_loss[loss=0.2072, simple_loss=0.29, pruned_loss=0.06223, over 3184183.89 frames. ], batch size: 43, lr: 6.88e-03, grad_scale: 8.0 2023-04-29 05:37:44,421 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.7790, 5.1207, 5.3319, 5.0658, 5.1060, 5.6675, 5.2289, 4.8356], device='cuda:1'), covar=tensor([0.0800, 0.1387, 0.1468, 0.1494, 0.2147, 0.0793, 0.1141, 0.2104], device='cuda:1'), in_proj_covar=tensor([0.0340, 0.0480, 0.0508, 0.0418, 0.0550, 0.0533, 0.0411, 0.0566], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-04-29 05:37:57,353 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.9598, 5.0143, 4.8076, 4.5292, 4.3905, 4.9378, 4.7367, 4.5183], device='cuda:1'), covar=tensor([0.0481, 0.0252, 0.0223, 0.0235, 0.0974, 0.0220, 0.0266, 0.0540], device='cuda:1'), in_proj_covar=tensor([0.0238, 0.0296, 0.0283, 0.0259, 0.0305, 0.0292, 0.0194, 0.0328], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-29 05:37:59,679 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=95678.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 05:38:24,126 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=95695.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 05:38:28,732 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.2409, 2.9104, 2.7262, 1.8316, 2.5544, 2.0904, 2.7659, 3.0018], device='cuda:1'), covar=tensor([0.0282, 0.0555, 0.0519, 0.1703, 0.0820, 0.0911, 0.0581, 0.0607], device='cuda:1'), in_proj_covar=tensor([0.0142, 0.0143, 0.0155, 0.0141, 0.0134, 0.0123, 0.0135, 0.0152], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:1') 2023-04-29 05:38:33,664 INFO [train.py:904] (1/8) Epoch 10, batch 4350, loss[loss=0.2066, simple_loss=0.2963, pruned_loss=0.05842, over 16537.00 frames. ], tot_loss[loss=0.2104, simple_loss=0.2937, pruned_loss=0.06352, over 3158516.45 frames. ], batch size: 68, lr: 6.88e-03, grad_scale: 8.0 2023-04-29 05:39:18,685 INFO [optim.py:368] (1/8) 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 05:39:19,263 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.1860, 2.3701, 1.9617, 2.2014, 2.7498, 2.3855, 3.0885, 2.9622], device='cuda:1'), covar=tensor([0.0061, 0.0229, 0.0331, 0.0278, 0.0152, 0.0261, 0.0115, 0.0145], device='cuda:1'), in_proj_covar=tensor([0.0140, 0.0194, 0.0193, 0.0191, 0.0195, 0.0198, 0.0201, 0.0186], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-29 05:39:49,467 INFO [train.py:904] (1/8) Epoch 10, batch 4400, loss[loss=0.2122, simple_loss=0.2994, pruned_loss=0.06255, over 16797.00 frames. ], tot_loss[loss=0.2124, simple_loss=0.2957, pruned_loss=0.06451, over 3149420.56 frames. ], batch size: 124, lr: 6.87e-03, grad_scale: 8.0 2023-04-29 05:40:02,125 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=95761.0, num_to_drop=1, layers_to_drop={2} 2023-04-29 05:40:51,143 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.0180, 2.5427, 2.6509, 1.8195, 2.7913, 2.8473, 2.4029, 2.3893], device='cuda:1'), covar=tensor([0.0701, 0.0219, 0.0183, 0.0901, 0.0086, 0.0152, 0.0416, 0.0399], device='cuda:1'), in_proj_covar=tensor([0.0146, 0.0101, 0.0088, 0.0140, 0.0070, 0.0100, 0.0122, 0.0130], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0003, 0.0002, 0.0004, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:1') 2023-04-29 05:41:01,592 INFO [train.py:904] (1/8) Epoch 10, batch 4450, loss[loss=0.2255, simple_loss=0.3094, pruned_loss=0.07076, over 16410.00 frames. ], tot_loss[loss=0.2149, simple_loss=0.2987, pruned_loss=0.06555, over 3148058.47 frames. ], batch size: 35, lr: 6.87e-03, grad_scale: 8.0 2023-04-29 05:41:12,948 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=95809.0, num_to_drop=1, layers_to_drop={0} 2023-04-29 05:41:17,779 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=95812.0, num_to_drop=1, layers_to_drop={0} 2023-04-29 05:41:46,152 INFO [optim.py:368] (1/8) 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 05:42:08,798 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=95847.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 05:42:15,001 INFO [train.py:904] (1/8) Epoch 10, batch 4500, loss[loss=0.213, simple_loss=0.2909, pruned_loss=0.06756, over 16938.00 frames. ], tot_loss[loss=0.2159, simple_loss=0.2992, pruned_loss=0.0663, over 3163930.97 frames. ], batch size: 109, lr: 6.87e-03, grad_scale: 8.0 2023-04-29 05:42:16,727 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.9637, 5.4194, 5.7569, 5.4496, 5.4807, 6.0945, 5.6719, 5.3029], device='cuda:1'), covar=tensor([0.0815, 0.1499, 0.1524, 0.1566, 0.2083, 0.0805, 0.1100, 0.1943], device='cuda:1'), in_proj_covar=tensor([0.0339, 0.0473, 0.0498, 0.0412, 0.0543, 0.0526, 0.0406, 0.0558], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-04-29 05:42:45,919 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=95873.0, num_to_drop=1, layers_to_drop={0} 2023-04-29 05:43:27,217 INFO [train.py:904] (1/8) Epoch 10, batch 4550, loss[loss=0.2056, simple_loss=0.3023, pruned_loss=0.05448, over 16842.00 frames. ], tot_loss[loss=0.2163, simple_loss=0.2995, pruned_loss=0.06661, over 3186395.25 frames. ], batch size: 83, lr: 6.87e-03, grad_scale: 8.0 2023-04-29 05:43:35,536 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=95908.0, num_to_drop=1, layers_to_drop={1} 2023-04-29 05:43:47,875 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=95916.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 05:43:59,565 INFO [zipformer.py:625] (1/8) 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] (1/8) Clipping_scale=2.0, grad-norm quartiles 1.676e+02 2.125e+02 2.355e+02 2.832e+02 4.769e+02, threshold=4.710e+02, percent-clipped=0.0 2023-04-29 05:44:39,213 INFO [train.py:904] (1/8) Epoch 10, batch 4600, loss[loss=0.1799, simple_loss=0.2726, pruned_loss=0.04355, over 16771.00 frames. ], tot_loss[loss=0.2163, simple_loss=0.2998, pruned_loss=0.06641, over 3190505.27 frames. ], batch size: 89, lr: 6.87e-03, grad_scale: 8.0 2023-04-29 05:44:57,859 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=95964.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 05:45:09,543 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=95972.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 05:45:11,441 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=95973.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 05:45:20,985 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=95980.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 05:45:43,285 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=95995.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 05:45:55,175 INFO [train.py:904] (1/8) Epoch 10, batch 4650, loss[loss=0.1873, simple_loss=0.2727, pruned_loss=0.05092, over 16793.00 frames. ], tot_loss[loss=0.2151, simple_loss=0.2984, pruned_loss=0.06586, over 3207482.17 frames. ], batch size: 89, lr: 6.87e-03, grad_scale: 8.0 2023-04-29 05:46:07,898 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.78 vs. limit=2.0 2023-04-29 05:46:40,629 INFO [optim.py:368] (1/8) 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 05:46:55,263 INFO [zipformer.py:625] (1/8) 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] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=96043.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 05:47:10,307 INFO [train.py:904] (1/8) Epoch 10, batch 4700, loss[loss=0.2034, simple_loss=0.2824, pruned_loss=0.06217, over 16288.00 frames. ], tot_loss[loss=0.2119, simple_loss=0.2955, pruned_loss=0.06416, over 3223724.89 frames. ], batch size: 35, lr: 6.86e-03, grad_scale: 8.0 2023-04-29 05:47:52,273 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=96080.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 05:48:23,997 INFO [train.py:904] (1/8) Epoch 10, batch 4750, loss[loss=0.1862, simple_loss=0.2759, pruned_loss=0.04825, over 16955.00 frames. ], tot_loss[loss=0.2078, simple_loss=0.2913, pruned_loss=0.06215, over 3216070.57 frames. ], batch size: 90, lr: 6.86e-03, grad_scale: 8.0 2023-04-29 05:49:08,953 INFO [optim.py:368] (1/8) 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 05:49:22,675 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=96141.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 05:49:38,023 INFO [train.py:904] (1/8) Epoch 10, batch 4800, loss[loss=0.2048, simple_loss=0.2972, pruned_loss=0.05614, over 15500.00 frames. ], tot_loss[loss=0.2037, simple_loss=0.2875, pruned_loss=0.05991, over 3214502.53 frames. ], batch size: 190, lr: 6.86e-03, grad_scale: 8.0 2023-04-29 05:49:58,168 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=96165.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 05:50:03,062 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=96168.0, num_to_drop=1, layers_to_drop={0} 2023-04-29 05:50:54,658 INFO [train.py:904] (1/8) Epoch 10, batch 4850, loss[loss=0.2391, simple_loss=0.3135, pruned_loss=0.08237, over 12110.00 frames. ], tot_loss[loss=0.2038, simple_loss=0.2881, pruned_loss=0.05976, over 3198398.47 frames. ], batch size: 246, lr: 6.86e-03, grad_scale: 8.0 2023-04-29 05:50:56,264 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=96203.0, num_to_drop=1, layers_to_drop={2} 2023-04-29 05:51:32,235 INFO [zipformer.py:625] (1/8) 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] (1/8) 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] (1/8) Epoch 10, batch 4900, loss[loss=0.1812, simple_loss=0.2683, pruned_loss=0.04708, over 12236.00 frames. ], tot_loss[loss=0.2013, simple_loss=0.2868, pruned_loss=0.05791, over 3190423.18 frames. ], batch size: 246, lr: 6.86e-03, grad_scale: 8.0 2023-04-29 05:52:33,338 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([5.5913, 5.9522, 5.5306, 5.6643, 5.2150, 5.1469, 5.3165, 6.0381], device='cuda:1'), covar=tensor([0.0963, 0.0643, 0.1116, 0.0634, 0.0790, 0.0584, 0.0909, 0.0727], device='cuda:1'), in_proj_covar=tensor([0.0494, 0.0629, 0.0521, 0.0430, 0.0392, 0.0406, 0.0518, 0.0477], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:1') 2023-04-29 05:52:42,388 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=96273.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 05:53:14,955 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.3723, 2.9129, 2.6251, 2.1469, 2.1971, 2.1378, 2.9780, 2.8154], device='cuda:1'), covar=tensor([0.2128, 0.0842, 0.1375, 0.1974, 0.1824, 0.1586, 0.0562, 0.0963], device='cuda:1'), in_proj_covar=tensor([0.0298, 0.0255, 0.0277, 0.0272, 0.0280, 0.0215, 0.0264, 0.0289], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-29 05:53:24,549 INFO [train.py:904] (1/8) Epoch 10, batch 4950, loss[loss=0.2187, simple_loss=0.3038, pruned_loss=0.06677, over 16687.00 frames. ], tot_loss[loss=0.2009, simple_loss=0.2866, pruned_loss=0.05755, over 3189979.81 frames. ], batch size: 134, lr: 6.85e-03, grad_scale: 8.0 2023-04-29 05:53:38,961 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.3992, 3.3048, 3.3908, 3.5070, 3.5443, 3.2620, 3.5106, 3.5881], device='cuda:1'), covar=tensor([0.0935, 0.0815, 0.0907, 0.0495, 0.0479, 0.2235, 0.0790, 0.0509], device='cuda:1'), in_proj_covar=tensor([0.0498, 0.0619, 0.0755, 0.0629, 0.0469, 0.0486, 0.0492, 0.0559], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-29 05:53:51,588 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=96321.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 05:53:56,290 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=96324.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 05:54:05,870 INFO [optim.py:368] (1/8) 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] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=96336.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 05:54:33,260 INFO [train.py:904] (1/8) Epoch 10, batch 5000, loss[loss=0.2128, simple_loss=0.3079, pruned_loss=0.05885, over 16454.00 frames. ], tot_loss[loss=0.2018, simple_loss=0.2883, pruned_loss=0.05768, over 3212260.57 frames. ], batch size: 146, lr: 6.85e-03, grad_scale: 8.0 2023-04-29 05:55:20,994 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=96385.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 05:55:27,994 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=96390.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 05:55:44,796 INFO [train.py:904] (1/8) Epoch 10, batch 5050, loss[loss=0.1909, simple_loss=0.2784, pruned_loss=0.05165, over 16428.00 frames. ], tot_loss[loss=0.2011, simple_loss=0.2884, pruned_loss=0.05694, over 3220063.96 frames. ], batch size: 68, lr: 6.85e-03, grad_scale: 8.0 2023-04-29 05:56:27,889 INFO [optim.py:368] (1/8) 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 05:56:33,444 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=96436.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 05:56:52,270 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.6461, 3.7072, 2.8426, 2.2000, 2.5876, 2.2962, 3.9255, 3.5026], device='cuda:1'), covar=tensor([0.2452, 0.0705, 0.1559, 0.2142, 0.2115, 0.1668, 0.0432, 0.0859], device='cuda:1'), in_proj_covar=tensor([0.0301, 0.0259, 0.0280, 0.0275, 0.0284, 0.0218, 0.0269, 0.0293], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-29 05:56:54,598 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 10, batch 5100, loss[loss=0.176, simple_loss=0.2675, pruned_loss=0.04226, over 16841.00 frames. ], tot_loss[loss=0.1988, simple_loss=0.286, pruned_loss=0.0558, over 3226008.59 frames. ], batch size: 96, lr: 6.85e-03, grad_scale: 8.0 2023-04-29 05:57:20,121 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=96468.0, num_to_drop=1, layers_to_drop={1} 2023-04-29 05:57:40,200 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.3233, 3.2982, 2.4781, 2.0355, 2.2713, 2.0896, 3.2361, 3.1101], device='cuda:1'), covar=tensor([0.2724, 0.0806, 0.1789, 0.2512, 0.2238, 0.1950, 0.0728, 0.1146], device='cuda:1'), in_proj_covar=tensor([0.0301, 0.0257, 0.0279, 0.0275, 0.0283, 0.0217, 0.0268, 0.0292], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-29 05:58:08,611 INFO [train.py:904] (1/8) Epoch 10, batch 5150, loss[loss=0.1899, simple_loss=0.2752, pruned_loss=0.05228, over 16110.00 frames. ], tot_loss[loss=0.1979, simple_loss=0.2857, pruned_loss=0.05507, over 3216433.14 frames. ], batch size: 35, lr: 6.85e-03, grad_scale: 8.0 2023-04-29 05:58:10,537 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=96503.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 05:58:29,579 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=96516.0, num_to_drop=1, layers_to_drop={1} 2023-04-29 05:58:36,670 INFO [zipformer.py:625] (1/8) 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] (1/8) Clipping_scale=2.0, grad-norm quartiles 1.417e+02 2.088e+02 2.479e+02 2.937e+02 7.130e+02, threshold=4.958e+02, percent-clipped=1.0 2023-04-29 05:58:59,582 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.7098, 4.6889, 4.5918, 4.3186, 4.1757, 4.6362, 4.5098, 4.3449], device='cuda:1'), covar=tensor([0.0566, 0.0455, 0.0228, 0.0234, 0.0990, 0.0414, 0.0348, 0.0630], device='cuda:1'), in_proj_covar=tensor([0.0233, 0.0290, 0.0278, 0.0253, 0.0301, 0.0291, 0.0187, 0.0322], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-29 05:59:08,220 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.1388, 3.6340, 2.9703, 1.6778, 2.5140, 1.9661, 3.4713, 3.7441], device='cuda:1'), covar=tensor([0.0193, 0.0593, 0.0753, 0.2134, 0.1058, 0.1237, 0.0601, 0.0581], device='cuda:1'), in_proj_covar=tensor([0.0142, 0.0142, 0.0157, 0.0142, 0.0134, 0.0124, 0.0135, 0.0151], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:1') 2023-04-29 05:59:21,922 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=96551.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 05:59:22,745 INFO [train.py:904] (1/8) Epoch 10, batch 5200, loss[loss=0.1905, simple_loss=0.2805, pruned_loss=0.05023, over 16421.00 frames. ], tot_loss[loss=0.1961, simple_loss=0.2839, pruned_loss=0.05417, over 3228721.98 frames. ], batch size: 146, lr: 6.85e-03, grad_scale: 8.0 2023-04-29 06:00:13,471 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=2.04 vs. limit=2.0 2023-04-29 06:00:25,804 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.22 vs. limit=2.0 2023-04-29 06:00:35,355 INFO [train.py:904] (1/8) Epoch 10, batch 5250, loss[loss=0.182, simple_loss=0.2746, pruned_loss=0.0447, over 16306.00 frames. ], tot_loss[loss=0.1948, simple_loss=0.2816, pruned_loss=0.05401, over 3235567.58 frames. ], batch size: 165, lr: 6.84e-03, grad_scale: 4.0 2023-04-29 06:01:21,023 INFO [optim.py:368] (1/8) 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 06:01:22,501 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.2292, 3.8538, 3.3901, 2.1574, 2.9888, 2.8795, 3.4632, 3.8757], device='cuda:1'), covar=tensor([0.0345, 0.0602, 0.0596, 0.1654, 0.0833, 0.0691, 0.0840, 0.0719], device='cuda:1'), in_proj_covar=tensor([0.0142, 0.0141, 0.0156, 0.0141, 0.0134, 0.0123, 0.0135, 0.0150], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:1') 2023-04-29 06:01:26,344 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=96636.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 06:01:27,849 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=2.36 vs. limit=5.0 2023-04-29 06:01:48,988 INFO [train.py:904] (1/8) Epoch 10, batch 5300, loss[loss=0.1781, simple_loss=0.2566, pruned_loss=0.04984, over 16406.00 frames. ], tot_loss[loss=0.1927, simple_loss=0.2788, pruned_loss=0.05327, over 3224605.12 frames. ], batch size: 146, lr: 6.84e-03, grad_scale: 4.0 2023-04-29 06:02:30,151 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=96680.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 06:02:36,971 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=96684.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 06:03:03,535 INFO [train.py:904] (1/8) Epoch 10, batch 5350, loss[loss=0.2299, simple_loss=0.314, pruned_loss=0.07293, over 15435.00 frames. ], tot_loss[loss=0.1917, simple_loss=0.2778, pruned_loss=0.05284, over 3215173.93 frames. ], batch size: 190, lr: 6.84e-03, grad_scale: 4.0 2023-04-29 06:03:20,657 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.4929, 3.6750, 1.8685, 3.9834, 2.5835, 3.8984, 2.1619, 2.8123], device='cuda:1'), covar=tensor([0.0192, 0.0263, 0.1674, 0.0080, 0.0785, 0.0357, 0.1399, 0.0656], device='cuda:1'), in_proj_covar=tensor([0.0147, 0.0162, 0.0187, 0.0119, 0.0166, 0.0201, 0.0191, 0.0170], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-04-29 06:03:48,671 INFO [optim.py:368] (1/8) 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 06:03:52,908 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=96736.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 06:04:08,474 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=96746.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 06:04:15,625 INFO [train.py:904] (1/8) Epoch 10, batch 5400, loss[loss=0.1898, simple_loss=0.2825, pruned_loss=0.04859, over 16468.00 frames. ], tot_loss[loss=0.1949, simple_loss=0.2813, pruned_loss=0.05424, over 3211570.67 frames. ], batch size: 68, lr: 6.84e-03, grad_scale: 4.0 2023-04-29 06:04:45,227 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.17 vs. limit=2.0 2023-04-29 06:04:51,426 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=96776.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 06:05:02,544 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 10, batch 5450, loss[loss=0.2289, simple_loss=0.3155, pruned_loss=0.07119, over 16735.00 frames. ], tot_loss[loss=0.198, simple_loss=0.2842, pruned_loss=0.05594, over 3202694.94 frames. ], batch size: 83, lr: 6.84e-03, grad_scale: 4.0 2023-04-29 06:06:02,265 INFO [zipformer.py:625] (1/8) 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] (1/8) 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 06:06:27,178 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=96837.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 06:06:49,547 INFO [train.py:904] (1/8) Epoch 10, batch 5500, loss[loss=0.2964, simple_loss=0.3511, pruned_loss=0.1208, over 11842.00 frames. ], tot_loss[loss=0.2069, simple_loss=0.2919, pruned_loss=0.06096, over 3192462.08 frames. ], batch size: 247, lr: 6.83e-03, grad_scale: 4.0 2023-04-29 06:07:06,645 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.3660, 3.2973, 3.3208, 3.4842, 3.5045, 3.2673, 3.4611, 3.5483], device='cuda:1'), covar=tensor([0.1120, 0.0838, 0.1161, 0.0559, 0.0638, 0.2055, 0.0893, 0.0653], device='cuda:1'), in_proj_covar=tensor([0.0503, 0.0625, 0.0765, 0.0635, 0.0479, 0.0487, 0.0498, 0.0566], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-29 06:07:17,927 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=96869.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 06:07:37,143 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.3325, 4.6433, 4.3842, 4.4046, 4.1577, 4.1111, 4.1934, 4.6445], device='cuda:1'), covar=tensor([0.0909, 0.0775, 0.0933, 0.0685, 0.0726, 0.1222, 0.0938, 0.0941], device='cuda:1'), in_proj_covar=tensor([0.0505, 0.0641, 0.0533, 0.0443, 0.0404, 0.0414, 0.0532, 0.0492], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:1') 2023-04-29 06:08:08,494 INFO [train.py:904] (1/8) Epoch 10, batch 5550, loss[loss=0.2712, simple_loss=0.3466, pruned_loss=0.09793, over 16142.00 frames. ], tot_loss[loss=0.2164, simple_loss=0.2995, pruned_loss=0.06667, over 3167451.92 frames. ], batch size: 165, lr: 6.83e-03, grad_scale: 4.0 2023-04-29 06:08:47,566 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.71 vs. limit=2.0 2023-04-29 06:09:01,355 INFO [optim.py:368] (1/8) 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 06:09:09,859 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.3494, 2.0177, 2.1605, 3.9257, 2.0046, 2.5248, 2.1517, 2.2721], device='cuda:1'), covar=tensor([0.0821, 0.3042, 0.1988, 0.0374, 0.3381, 0.1964, 0.2648, 0.2763], device='cuda:1'), in_proj_covar=tensor([0.0356, 0.0381, 0.0318, 0.0317, 0.0403, 0.0435, 0.0344, 0.0447], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-29 06:09:28,005 INFO [train.py:904] (1/8) Epoch 10, batch 5600, loss[loss=0.3593, simple_loss=0.3891, pruned_loss=0.1647, over 11223.00 frames. ], tot_loss[loss=0.2235, simple_loss=0.3046, pruned_loss=0.07118, over 3117138.49 frames. ], batch size: 248, lr: 6.83e-03, grad_scale: 2.0 2023-04-29 06:10:15,388 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=96980.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 06:10:22,786 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.37 vs. limit=2.0 2023-04-29 06:10:51,169 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=97000.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 06:10:53,952 INFO [train.py:904] (1/8) Epoch 10, batch 5650, loss[loss=0.2356, simple_loss=0.3119, pruned_loss=0.07965, over 16586.00 frames. ], tot_loss[loss=0.2327, simple_loss=0.3113, pruned_loss=0.07705, over 3064412.99 frames. ], batch size: 62, lr: 6.83e-03, grad_scale: 2.0 2023-04-29 06:11:01,823 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.6305, 3.7342, 4.0826, 4.0206, 4.0627, 3.7721, 3.8092, 3.8180], device='cuda:1'), covar=tensor([0.0400, 0.0607, 0.0383, 0.0491, 0.0481, 0.0448, 0.0919, 0.0485], device='cuda:1'), in_proj_covar=tensor([0.0314, 0.0322, 0.0324, 0.0309, 0.0371, 0.0345, 0.0442, 0.0277], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003], device='cuda:1') 2023-04-29 06:11:22,070 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([5.3819, 5.7046, 5.4300, 5.4176, 5.0767, 4.9348, 5.1430, 5.8611], device='cuda:1'), covar=tensor([0.0816, 0.0742, 0.0827, 0.0652, 0.0720, 0.0670, 0.0880, 0.0710], device='cuda:1'), in_proj_covar=tensor([0.0498, 0.0634, 0.0526, 0.0439, 0.0397, 0.0407, 0.0524, 0.0484], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:1') 2023-04-29 06:11:33,820 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=97028.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 06:11:43,932 INFO [optim.py:368] (1/8) 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 06:12:02,171 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 10, batch 5700, loss[loss=0.2841, simple_loss=0.3377, pruned_loss=0.1153, over 11267.00 frames. ], tot_loss[loss=0.2345, simple_loss=0.3124, pruned_loss=0.07826, over 3074927.80 frames. ], batch size: 248, lr: 6.83e-03, grad_scale: 2.0 2023-04-29 06:12:16,566 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([1.8077, 1.3161, 1.6740, 1.6309, 1.8721, 1.9447, 1.5649, 1.8377], device='cuda:1'), covar=tensor([0.0177, 0.0242, 0.0130, 0.0210, 0.0156, 0.0115, 0.0252, 0.0076], device='cuda:1'), in_proj_covar=tensor([0.0157, 0.0165, 0.0151, 0.0154, 0.0161, 0.0118, 0.0167, 0.0109], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001], device='cuda:1') 2023-04-29 06:12:25,254 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=97061.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 06:13:16,339 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=97094.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 06:13:29,501 INFO [train.py:904] (1/8) Epoch 10, batch 5750, loss[loss=0.2784, simple_loss=0.3274, pruned_loss=0.1147, over 11128.00 frames. ], tot_loss[loss=0.2376, simple_loss=0.3149, pruned_loss=0.08008, over 3044259.67 frames. ], batch size: 247, lr: 6.83e-03, grad_scale: 2.0 2023-04-29 06:13:57,569 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.0670, 2.5703, 2.6014, 1.9228, 2.7635, 2.8358, 2.3714, 2.3758], device='cuda:1'), covar=tensor([0.0681, 0.0191, 0.0196, 0.0843, 0.0087, 0.0178, 0.0410, 0.0395], device='cuda:1'), in_proj_covar=tensor([0.0140, 0.0097, 0.0085, 0.0136, 0.0068, 0.0095, 0.0119, 0.0126], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0003, 0.0002, 0.0004, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:1') 2023-04-29 06:14:17,784 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=97132.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 06:14:22,109 INFO [optim.py:368] (1/8) 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 06:14:49,726 INFO [train.py:904] (1/8) Epoch 10, batch 5800, loss[loss=0.2304, simple_loss=0.3249, pruned_loss=0.0679, over 15528.00 frames. ], tot_loss[loss=0.2368, simple_loss=0.3148, pruned_loss=0.07936, over 3022814.02 frames. ], batch size: 191, lr: 6.82e-03, grad_scale: 2.0 2023-04-29 06:15:32,866 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.6905, 4.7112, 4.5374, 4.2746, 4.1712, 4.5914, 4.4626, 4.2671], device='cuda:1'), covar=tensor([0.0566, 0.0466, 0.0259, 0.0244, 0.0913, 0.0415, 0.0365, 0.0620], device='cuda:1'), in_proj_covar=tensor([0.0235, 0.0292, 0.0277, 0.0253, 0.0299, 0.0290, 0.0187, 0.0322], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-29 06:15:43,440 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.4796, 3.9706, 3.9096, 2.5731, 3.6032, 3.9019, 3.6874, 1.9936], device='cuda:1'), covar=tensor([0.0367, 0.0027, 0.0031, 0.0309, 0.0057, 0.0074, 0.0048, 0.0378], device='cuda:1'), in_proj_covar=tensor([0.0126, 0.0066, 0.0068, 0.0124, 0.0077, 0.0086, 0.0074, 0.0117], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:1') 2023-04-29 06:16:07,864 INFO [train.py:904] (1/8) Epoch 10, batch 5850, loss[loss=0.2415, simple_loss=0.3143, pruned_loss=0.08439, over 15372.00 frames. ], tot_loss[loss=0.2329, simple_loss=0.3125, pruned_loss=0.07662, over 3057245.25 frames. ], batch size: 190, lr: 6.82e-03, grad_scale: 2.0 2023-04-29 06:17:00,835 INFO [optim.py:368] (1/8) 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 06:17:28,940 INFO [train.py:904] (1/8) Epoch 10, batch 5900, loss[loss=0.2252, simple_loss=0.3111, pruned_loss=0.06966, over 16520.00 frames. ], tot_loss[loss=0.2321, simple_loss=0.3121, pruned_loss=0.07605, over 3070840.02 frames. ], batch size: 75, lr: 6.82e-03, grad_scale: 2.0 2023-04-29 06:17:38,776 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.1419, 4.1116, 3.9956, 3.1906, 4.0930, 1.5194, 3.7648, 3.7333], device='cuda:1'), covar=tensor([0.0115, 0.0092, 0.0155, 0.0436, 0.0097, 0.2862, 0.0143, 0.0239], device='cuda:1'), in_proj_covar=tensor([0.0121, 0.0109, 0.0155, 0.0150, 0.0126, 0.0168, 0.0142, 0.0145], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-29 06:17:45,081 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=97260.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 06:18:32,643 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=2.97 vs. limit=5.0 2023-04-29 06:18:49,538 INFO [train.py:904] (1/8) Epoch 10, batch 5950, loss[loss=0.215, simple_loss=0.2987, pruned_loss=0.06568, over 16835.00 frames. ], tot_loss[loss=0.2313, simple_loss=0.3131, pruned_loss=0.07473, over 3081500.03 frames. ], batch size: 96, lr: 6.82e-03, grad_scale: 2.0 2023-04-29 06:19:02,573 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([5.0178, 4.9564, 4.7466, 4.1929, 4.9118, 1.9458, 4.6156, 4.7436], device='cuda:1'), covar=tensor([0.0060, 0.0050, 0.0117, 0.0290, 0.0055, 0.2014, 0.0085, 0.0114], device='cuda:1'), in_proj_covar=tensor([0.0120, 0.0109, 0.0155, 0.0149, 0.0126, 0.0167, 0.0141, 0.0144], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-29 06:19:15,847 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.8733, 4.7992, 5.3448, 5.3099, 5.3257, 4.9589, 4.9505, 4.6388], device='cuda:1'), covar=tensor([0.0281, 0.0489, 0.0357, 0.0418, 0.0398, 0.0323, 0.0848, 0.0418], device='cuda:1'), in_proj_covar=tensor([0.0315, 0.0320, 0.0323, 0.0311, 0.0373, 0.0347, 0.0444, 0.0279], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003], device='cuda:1') 2023-04-29 06:19:16,407 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.72 vs. limit=2.0 2023-04-29 06:19:20,364 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=97321.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 06:19:40,990 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.3840, 3.3845, 1.9455, 3.7199, 2.4870, 3.7112, 1.9472, 2.6443], device='cuda:1'), covar=tensor([0.0218, 0.0373, 0.1627, 0.0136, 0.0821, 0.0547, 0.1645, 0.0745], device='cuda:1'), in_proj_covar=tensor([0.0145, 0.0162, 0.0186, 0.0120, 0.0166, 0.0201, 0.0190, 0.0170], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-04-29 06:19:41,580 INFO [optim.py:368] (1/8) 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 06:19:52,234 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=97342.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 06:20:00,726 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.62 vs. limit=2.0 2023-04-29 06:20:01,449 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 10, batch 6000, loss[loss=0.2236, simple_loss=0.304, pruned_loss=0.07165, over 17041.00 frames. ], tot_loss[loss=0.2306, simple_loss=0.3124, pruned_loss=0.07445, over 3084270.07 frames. ], batch size: 55, lr: 6.82e-03, grad_scale: 4.0 2023-04-29 06:20:09,122 INFO [train.py:929] (1/8) Computing validation loss 2023-04-29 06:20:23,728 INFO [train.py:938] (1/8) Epoch 10, validation: loss=0.165, simple_loss=0.2783, pruned_loss=0.02583, over 944034.00 frames. 2023-04-29 06:20:23,729 INFO [train.py:939] (1/8) Maximum memory allocated so far is 17676MB 2023-04-29 06:20:30,460 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=97356.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 06:20:57,489 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=97374.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 06:21:42,482 INFO [train.py:904] (1/8) Epoch 10, batch 6050, loss[loss=0.2025, simple_loss=0.2961, pruned_loss=0.05443, over 16559.00 frames. ], tot_loss[loss=0.2283, simple_loss=0.31, pruned_loss=0.07336, over 3075093.93 frames. ], batch size: 62, lr: 6.82e-03, grad_scale: 2.0 2023-04-29 06:21:44,951 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=97403.0, num_to_drop=1, layers_to_drop={1} 2023-04-29 06:21:52,170 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=97408.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 06:22:29,751 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=97432.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 06:22:34,968 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=97435.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 06:22:35,671 INFO [optim.py:368] (1/8) 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 06:23:02,108 INFO [train.py:904] (1/8) Epoch 10, batch 6100, loss[loss=0.2189, simple_loss=0.3073, pruned_loss=0.06529, over 16354.00 frames. ], tot_loss[loss=0.2259, simple_loss=0.3087, pruned_loss=0.07151, over 3101321.07 frames. ], batch size: 146, lr: 6.81e-03, grad_scale: 2.0 2023-04-29 06:23:48,674 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=97480.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 06:24:15,939 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=4.76 vs. limit=5.0 2023-04-29 06:24:20,429 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([1.5189, 2.8064, 2.5393, 4.2075, 3.2209, 4.0189, 1.4104, 2.8364], device='cuda:1'), covar=tensor([0.1426, 0.0648, 0.1117, 0.0124, 0.0322, 0.0385, 0.1609, 0.0857], device='cuda:1'), in_proj_covar=tensor([0.0152, 0.0156, 0.0177, 0.0136, 0.0200, 0.0206, 0.0178, 0.0178], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0003, 0.0004], device='cuda:1') 2023-04-29 06:24:23,702 INFO [train.py:904] (1/8) Epoch 10, batch 6150, loss[loss=0.2154, simple_loss=0.292, pruned_loss=0.06938, over 16328.00 frames. ], tot_loss[loss=0.2244, simple_loss=0.3064, pruned_loss=0.0712, over 3082645.31 frames. ], batch size: 35, lr: 6.81e-03, grad_scale: 2.0 2023-04-29 06:24:44,627 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.7634, 5.1017, 5.3644, 5.1438, 5.1525, 5.6877, 5.2150, 5.0534], device='cuda:1'), covar=tensor([0.0975, 0.1582, 0.1499, 0.1682, 0.2315, 0.0910, 0.1223, 0.2197], device='cuda:1'), in_proj_covar=tensor([0.0345, 0.0476, 0.0506, 0.0413, 0.0550, 0.0539, 0.0411, 0.0564], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-04-29 06:25:17,520 INFO [optim.py:368] (1/8) 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 06:25:37,851 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.42 vs. limit=2.0 2023-04-29 06:25:41,180 INFO [train.py:904] (1/8) Epoch 10, batch 6200, loss[loss=0.2619, simple_loss=0.3189, pruned_loss=0.1025, over 11407.00 frames. ], tot_loss[loss=0.2237, simple_loss=0.3052, pruned_loss=0.07108, over 3085353.01 frames. ], batch size: 247, lr: 6.81e-03, grad_scale: 2.0 2023-04-29 06:25:46,619 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=97555.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 06:25:54,777 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.4068, 4.1681, 4.2350, 4.5953, 4.7031, 4.2840, 4.7539, 4.7154], device='cuda:1'), covar=tensor([0.1568, 0.1311, 0.2060, 0.0859, 0.0910, 0.1031, 0.0790, 0.0781], device='cuda:1'), in_proj_covar=tensor([0.0502, 0.0621, 0.0758, 0.0635, 0.0484, 0.0481, 0.0501, 0.0567], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-29 06:25:55,094 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=4.79 vs. limit=5.0 2023-04-29 06:26:57,938 INFO [train.py:904] (1/8) Epoch 10, batch 6250, loss[loss=0.2164, simple_loss=0.3183, pruned_loss=0.05724, over 16659.00 frames. ], tot_loss[loss=0.2233, simple_loss=0.3048, pruned_loss=0.0709, over 3084930.47 frames. ], batch size: 62, lr: 6.81e-03, grad_scale: 2.0 2023-04-29 06:27:18,738 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=97616.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 06:27:18,856 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=97616.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 06:27:47,811 INFO [optim.py:368] (1/8) 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] (1/8) Epoch 10, batch 6300, loss[loss=0.1887, simple_loss=0.2764, pruned_loss=0.05057, over 16450.00 frames. ], tot_loss[loss=0.2225, simple_loss=0.3048, pruned_loss=0.07012, over 3111812.46 frames. ], batch size: 75, lr: 6.81e-03, grad_scale: 2.0 2023-04-29 06:28:18,576 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=97656.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 06:28:42,508 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.48 vs. limit=2.0 2023-04-29 06:28:48,412 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=97674.0, num_to_drop=1, layers_to_drop={1} 2023-04-29 06:29:20,173 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.8211, 3.8122, 2.1623, 4.3683, 2.7807, 4.3327, 2.2489, 2.9726], device='cuda:1'), covar=tensor([0.0195, 0.0325, 0.1591, 0.0085, 0.0764, 0.0328, 0.1476, 0.0701], device='cuda:1'), in_proj_covar=tensor([0.0144, 0.0162, 0.0185, 0.0120, 0.0164, 0.0201, 0.0190, 0.0170], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-04-29 06:29:24,485 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=97698.0, num_to_drop=1, layers_to_drop={1} 2023-04-29 06:29:30,239 INFO [train.py:904] (1/8) Epoch 10, batch 6350, loss[loss=0.2773, simple_loss=0.3315, pruned_loss=0.1116, over 11509.00 frames. ], tot_loss[loss=0.2232, simple_loss=0.3048, pruned_loss=0.07083, over 3116220.54 frames. ], batch size: 248, lr: 6.81e-03, grad_scale: 2.0 2023-04-29 06:29:31,922 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=97703.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 06:29:33,047 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=97704.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 06:30:02,455 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.85 vs. limit=2.0 2023-04-29 06:30:13,649 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=97730.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 06:30:21,602 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=97735.0, num_to_drop=1, layers_to_drop={2} 2023-04-29 06:30:22,845 INFO [optim.py:368] (1/8) 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 06:30:46,248 INFO [train.py:904] (1/8) Epoch 10, batch 6400, loss[loss=0.2006, simple_loss=0.2792, pruned_loss=0.06098, over 15435.00 frames. ], tot_loss[loss=0.2253, simple_loss=0.3058, pruned_loss=0.07245, over 3099380.69 frames. ], batch size: 190, lr: 6.80e-03, grad_scale: 4.0 2023-04-29 06:31:36,461 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.4399, 4.4279, 4.3232, 4.1009, 3.9555, 4.3865, 4.1789, 4.0452], device='cuda:1'), covar=tensor([0.0719, 0.0663, 0.0288, 0.0295, 0.0905, 0.0468, 0.0620, 0.0808], device='cuda:1'), in_proj_covar=tensor([0.0237, 0.0296, 0.0278, 0.0255, 0.0300, 0.0290, 0.0189, 0.0323], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-29 06:31:58,657 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([1.7656, 1.3649, 1.6265, 1.5818, 1.7866, 1.8869, 1.4978, 1.8202], device='cuda:1'), covar=tensor([0.0162, 0.0240, 0.0129, 0.0192, 0.0155, 0.0110, 0.0271, 0.0074], device='cuda:1'), in_proj_covar=tensor([0.0155, 0.0165, 0.0150, 0.0154, 0.0161, 0.0118, 0.0169, 0.0110], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001], device='cuda:1') 2023-04-29 06:32:00,789 INFO [train.py:904] (1/8) Epoch 10, batch 6450, loss[loss=0.1934, simple_loss=0.2823, pruned_loss=0.05218, over 16649.00 frames. ], tot_loss[loss=0.2247, simple_loss=0.3055, pruned_loss=0.07189, over 3082824.70 frames. ], batch size: 62, lr: 6.80e-03, grad_scale: 4.0 2023-04-29 06:32:57,255 INFO [optim.py:368] (1/8) 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 06:33:21,802 INFO [train.py:904] (1/8) Epoch 10, batch 6500, loss[loss=0.2261, simple_loss=0.3094, pruned_loss=0.07136, over 16778.00 frames. ], tot_loss[loss=0.2227, simple_loss=0.3033, pruned_loss=0.07106, over 3090575.12 frames. ], batch size: 124, lr: 6.80e-03, grad_scale: 4.0 2023-04-29 06:34:16,941 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=97888.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 06:34:31,984 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.26 vs. limit=2.0 2023-04-29 06:34:33,124 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.8338, 5.1728, 4.5878, 4.9661, 4.6623, 4.4892, 4.6701, 5.1642], device='cuda:1'), covar=tensor([0.1941, 0.1366, 0.2333, 0.1392, 0.1653, 0.1460, 0.2148, 0.1746], device='cuda:1'), in_proj_covar=tensor([0.0512, 0.0639, 0.0535, 0.0447, 0.0400, 0.0418, 0.0533, 0.0484], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:1') 2023-04-29 06:34:36,129 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.1254, 4.1141, 4.4941, 4.4945, 4.4415, 4.1955, 4.1785, 4.1012], device='cuda:1'), covar=tensor([0.0287, 0.0562, 0.0476, 0.0365, 0.0426, 0.0416, 0.0763, 0.0437], device='cuda:1'), in_proj_covar=tensor([0.0314, 0.0321, 0.0321, 0.0309, 0.0370, 0.0342, 0.0442, 0.0276], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003], device='cuda:1') 2023-04-29 06:34:41,158 INFO [train.py:904] (1/8) Epoch 10, batch 6550, loss[loss=0.2243, simple_loss=0.3251, pruned_loss=0.06174, over 16548.00 frames. ], tot_loss[loss=0.2249, simple_loss=0.3064, pruned_loss=0.07167, over 3085760.45 frames. ], batch size: 75, lr: 6.80e-03, grad_scale: 4.0 2023-04-29 06:34:55,678 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=97911.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 06:35:04,979 INFO [zipformer.py:625] (1/8) 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] (1/8) Clipping_scale=2.0, grad-norm quartiles 2.197e+02 3.236e+02 3.952e+02 5.112e+02 9.277e+02, threshold=7.905e+02, percent-clipped=2.0 2023-04-29 06:35:56,303 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 10, batch 6600, loss[loss=0.2416, simple_loss=0.3109, pruned_loss=0.08611, over 11477.00 frames. ], tot_loss[loss=0.226, simple_loss=0.3083, pruned_loss=0.07184, over 3098792.78 frames. ], batch size: 247, lr: 6.80e-03, grad_scale: 4.0 2023-04-29 06:36:19,346 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=97964.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 06:37:14,711 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 10, batch 6650, loss[loss=0.2057, simple_loss=0.2945, pruned_loss=0.05846, over 16917.00 frames. ], tot_loss[loss=0.2265, simple_loss=0.3082, pruned_loss=0.07239, over 3098667.90 frames. ], batch size: 96, lr: 6.79e-03, grad_scale: 4.0 2023-04-29 06:37:24,128 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=98003.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 06:38:05,316 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=98030.0, num_to_drop=1, layers_to_drop={0} 2023-04-29 06:38:05,437 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=98030.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 06:38:14,887 INFO [optim.py:368] (1/8) 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 06:38:26,280 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([1.9828, 2.3005, 2.2597, 2.9036, 2.1709, 3.2318, 1.7288, 2.6817], device='cuda:1'), covar=tensor([0.1041, 0.0523, 0.0976, 0.0115, 0.0126, 0.0340, 0.1309, 0.0628], device='cuda:1'), in_proj_covar=tensor([0.0153, 0.0157, 0.0179, 0.0136, 0.0202, 0.0205, 0.0180, 0.0178], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-04-29 06:38:29,249 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.6961, 4.9885, 5.1686, 4.9619, 4.9594, 5.5418, 5.0762, 4.8568], device='cuda:1'), covar=tensor([0.0942, 0.1529, 0.1526, 0.1708, 0.2443, 0.0891, 0.1299, 0.2178], device='cuda:1'), in_proj_covar=tensor([0.0340, 0.0473, 0.0505, 0.0409, 0.0548, 0.0538, 0.0412, 0.0563], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-04-29 06:38:31,003 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=98046.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 06:38:38,423 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=98051.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 06:38:39,793 INFO [train.py:904] (1/8) Epoch 10, batch 6700, loss[loss=0.293, simple_loss=0.3447, pruned_loss=0.1207, over 11744.00 frames. ], tot_loss[loss=0.2247, simple_loss=0.3059, pruned_loss=0.07178, over 3116264.86 frames. ], batch size: 249, lr: 6.79e-03, grad_scale: 4.0 2023-04-29 06:39:05,148 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.37 vs. limit=2.0 2023-04-29 06:39:20,423 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=98078.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 06:39:28,798 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.4973, 3.5866, 2.9013, 2.1292, 2.4205, 2.2807, 3.8687, 3.3268], device='cuda:1'), covar=tensor([0.2588, 0.0699, 0.1441, 0.2145, 0.2189, 0.1682, 0.0424, 0.0938], device='cuda:1'), in_proj_covar=tensor([0.0302, 0.0256, 0.0279, 0.0276, 0.0281, 0.0217, 0.0265, 0.0288], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-29 06:39:57,939 INFO [train.py:904] (1/8) Epoch 10, batch 6750, loss[loss=0.1998, simple_loss=0.2835, pruned_loss=0.05802, over 16468.00 frames. ], tot_loss[loss=0.2245, simple_loss=0.3048, pruned_loss=0.07208, over 3109977.71 frames. ], batch size: 68, lr: 6.79e-03, grad_scale: 4.0 2023-04-29 06:40:04,393 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=98106.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 06:40:13,079 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.5796, 4.7757, 5.0316, 4.8809, 4.9530, 5.4554, 4.9887, 4.7479], device='cuda:1'), covar=tensor([0.0997, 0.1652, 0.1473, 0.1737, 0.2041, 0.0867, 0.1343, 0.2416], device='cuda:1'), in_proj_covar=tensor([0.0340, 0.0474, 0.0506, 0.0409, 0.0547, 0.0537, 0.0412, 0.0561], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-04-29 06:40:17,230 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=3.54 vs. limit=5.0 2023-04-29 06:40:27,503 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([5.4106, 5.8070, 5.4219, 5.5586, 5.0562, 4.9478, 5.1995, 5.8516], device='cuda:1'), covar=tensor([0.0950, 0.0722, 0.0994, 0.0656, 0.0694, 0.0706, 0.1082, 0.0827], device='cuda:1'), in_proj_covar=tensor([0.0510, 0.0638, 0.0534, 0.0445, 0.0399, 0.0415, 0.0535, 0.0484], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:1') 2023-04-29 06:40:49,800 INFO [optim.py:368] (1/8) 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 06:41:15,032 INFO [train.py:904] (1/8) Epoch 10, batch 6800, loss[loss=0.2493, simple_loss=0.3182, pruned_loss=0.09013, over 11891.00 frames. ], tot_loss[loss=0.2251, simple_loss=0.3054, pruned_loss=0.07242, over 3104508.15 frames. ], batch size: 246, lr: 6.79e-03, grad_scale: 8.0 2023-04-29 06:41:38,571 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 10, batch 6850, loss[loss=0.2096, simple_loss=0.3088, pruned_loss=0.05515, over 16485.00 frames. ], tot_loss[loss=0.2262, simple_loss=0.3065, pruned_loss=0.0729, over 3092373.67 frames. ], batch size: 62, lr: 6.79e-03, grad_scale: 8.0 2023-04-29 06:42:47,557 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=98211.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 06:43:24,606 INFO [optim.py:368] (1/8) 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 06:43:36,644 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.47 vs. limit=2.0 2023-04-29 06:43:37,275 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=98244.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 06:43:49,044 INFO [train.py:904] (1/8) Epoch 10, batch 6900, loss[loss=0.2697, simple_loss=0.3403, pruned_loss=0.09955, over 15374.00 frames. ], tot_loss[loss=0.2265, simple_loss=0.3086, pruned_loss=0.07216, over 3086851.49 frames. ], batch size: 191, lr: 6.79e-03, grad_scale: 8.0 2023-04-29 06:43:53,113 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=4.90 vs. limit=5.0 2023-04-29 06:44:01,150 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=98259.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 06:45:09,158 INFO [train.py:904] (1/8) Epoch 10, batch 6950, loss[loss=0.2045, simple_loss=0.2923, pruned_loss=0.05833, over 17205.00 frames. ], tot_loss[loss=0.2291, simple_loss=0.3104, pruned_loss=0.07392, over 3089839.41 frames. ], batch size: 44, lr: 6.78e-03, grad_scale: 8.0 2023-04-29 06:45:17,525 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.7968, 3.6964, 3.8383, 3.9768, 4.0398, 3.6288, 3.9617, 4.0528], device='cuda:1'), covar=tensor([0.1269, 0.0953, 0.1208, 0.0578, 0.0537, 0.1599, 0.0760, 0.0602], device='cuda:1'), in_proj_covar=tensor([0.0504, 0.0623, 0.0760, 0.0634, 0.0487, 0.0483, 0.0503, 0.0568], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-29 06:45:39,750 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([1.6666, 2.7955, 2.5691, 4.4406, 3.3712, 4.1551, 1.5048, 2.9087], device='cuda:1'), covar=tensor([0.1295, 0.0636, 0.1111, 0.0134, 0.0266, 0.0334, 0.1465, 0.0807], device='cuda:1'), in_proj_covar=tensor([0.0152, 0.0155, 0.0179, 0.0136, 0.0201, 0.0204, 0.0177, 0.0177], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0003, 0.0004], device='cuda:1') 2023-04-29 06:45:54,171 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=98330.0, num_to_drop=1, layers_to_drop={1} 2023-04-29 06:46:01,761 INFO [optim.py:368] (1/8) 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 06:46:27,404 INFO [train.py:904] (1/8) Epoch 10, batch 7000, loss[loss=0.2252, simple_loss=0.3156, pruned_loss=0.06734, over 16242.00 frames. ], tot_loss[loss=0.2285, simple_loss=0.3105, pruned_loss=0.07328, over 3079472.60 frames. ], batch size: 165, lr: 6.78e-03, grad_scale: 8.0 2023-04-29 06:46:45,994 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.5426, 1.5948, 2.0765, 2.3742, 2.5759, 2.6171, 1.6387, 2.5679], device='cuda:1'), covar=tensor([0.0139, 0.0378, 0.0231, 0.0229, 0.0191, 0.0148, 0.0406, 0.0106], device='cuda:1'), in_proj_covar=tensor([0.0156, 0.0166, 0.0149, 0.0155, 0.0163, 0.0119, 0.0169, 0.0109], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001], device='cuda:1') 2023-04-29 06:47:07,453 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=98378.0, num_to_drop=1, layers_to_drop={1} 2023-04-29 06:47:43,137 INFO [train.py:904] (1/8) Epoch 10, batch 7050, loss[loss=0.2153, simple_loss=0.3038, pruned_loss=0.06338, over 16795.00 frames. ], tot_loss[loss=0.2291, simple_loss=0.3114, pruned_loss=0.07341, over 3078886.67 frames. ], batch size: 83, lr: 6.78e-03, grad_scale: 8.0 2023-04-29 06:47:56,570 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=98410.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 06:48:34,410 INFO [optim.py:368] (1/8) 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] (1/8) Epoch 10, batch 7100, loss[loss=0.2143, simple_loss=0.2945, pruned_loss=0.06708, over 16761.00 frames. ], tot_loss[loss=0.2292, simple_loss=0.3107, pruned_loss=0.07383, over 3068709.68 frames. ], batch size: 83, lr: 6.78e-03, grad_scale: 8.0 2023-04-29 06:49:14,092 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=98462.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 06:49:27,975 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=98471.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 06:49:56,549 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.1554, 4.1810, 4.3902, 4.2670, 4.3107, 4.8041, 4.3914, 4.2073], device='cuda:1'), covar=tensor([0.1691, 0.1949, 0.1929, 0.1842, 0.2605, 0.1078, 0.1510, 0.2449], device='cuda:1'), in_proj_covar=tensor([0.0342, 0.0477, 0.0512, 0.0415, 0.0548, 0.0540, 0.0416, 0.0565], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-04-29 06:50:09,113 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.8747, 4.8698, 4.6801, 4.4033, 4.3116, 4.7494, 4.6973, 4.4200], device='cuda:1'), covar=tensor([0.0541, 0.0335, 0.0244, 0.0241, 0.0984, 0.0354, 0.0314, 0.0591], device='cuda:1'), in_proj_covar=tensor([0.0229, 0.0288, 0.0268, 0.0247, 0.0292, 0.0282, 0.0185, 0.0312], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-29 06:50:12,812 INFO [train.py:904] (1/8) Epoch 10, batch 7150, loss[loss=0.2443, simple_loss=0.32, pruned_loss=0.08426, over 16642.00 frames. ], tot_loss[loss=0.2286, simple_loss=0.3095, pruned_loss=0.07383, over 3069855.12 frames. ], batch size: 62, lr: 6.78e-03, grad_scale: 8.0 2023-04-29 06:50:25,348 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.8746, 4.8822, 4.7015, 4.4517, 4.3542, 4.7799, 4.6579, 4.4676], device='cuda:1'), covar=tensor([0.0545, 0.0384, 0.0234, 0.0222, 0.0901, 0.0403, 0.0296, 0.0536], device='cuda:1'), in_proj_covar=tensor([0.0230, 0.0288, 0.0269, 0.0247, 0.0293, 0.0282, 0.0185, 0.0313], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-29 06:51:03,784 INFO [optim.py:368] (1/8) 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 06:51:15,765 INFO [zipformer.py:625] (1/8) 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] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=98544.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 06:51:29,368 INFO [train.py:904] (1/8) Epoch 10, batch 7200, loss[loss=0.2137, simple_loss=0.2904, pruned_loss=0.06851, over 11602.00 frames. ], tot_loss[loss=0.2248, simple_loss=0.3064, pruned_loss=0.07166, over 3058271.66 frames. ], batch size: 247, lr: 6.78e-03, grad_scale: 8.0 2023-04-29 06:51:32,411 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([1.6957, 2.4741, 2.2918, 3.2247, 2.3609, 3.6573, 1.3589, 2.6677], device='cuda:1'), covar=tensor([0.1325, 0.0643, 0.1163, 0.0146, 0.0169, 0.0352, 0.1592, 0.0780], device='cuda:1'), in_proj_covar=tensor([0.0154, 0.0157, 0.0180, 0.0137, 0.0202, 0.0206, 0.0178, 0.0179], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0003, 0.0004], device='cuda:1') 2023-04-29 06:51:48,938 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.2494, 3.2197, 3.5315, 1.5965, 3.7276, 3.8068, 2.6945, 2.8969], device='cuda:1'), covar=tensor([0.0792, 0.0222, 0.0203, 0.1257, 0.0050, 0.0094, 0.0427, 0.0413], device='cuda:1'), in_proj_covar=tensor([0.0143, 0.0100, 0.0087, 0.0140, 0.0069, 0.0097, 0.0121, 0.0129], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0003, 0.0002, 0.0004, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:1') 2023-04-29 06:52:32,648 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=98592.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 06:52:43,319 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.33 vs. limit=2.0 2023-04-29 06:52:49,206 INFO [train.py:904] (1/8) Epoch 10, batch 7250, loss[loss=0.2013, simple_loss=0.284, pruned_loss=0.05933, over 16770.00 frames. ], tot_loss[loss=0.2221, simple_loss=0.3039, pruned_loss=0.07019, over 3049867.97 frames. ], batch size: 124, lr: 6.77e-03, grad_scale: 2.0 2023-04-29 06:52:51,614 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.5519, 4.5938, 4.5781, 3.0171, 3.9025, 4.4672, 4.0608, 2.7234], device='cuda:1'), covar=tensor([0.0419, 0.0017, 0.0025, 0.0298, 0.0059, 0.0065, 0.0046, 0.0288], device='cuda:1'), in_proj_covar=tensor([0.0127, 0.0066, 0.0068, 0.0126, 0.0077, 0.0087, 0.0075, 0.0119], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:1') 2023-04-29 06:52:53,542 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=98604.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 06:52:54,874 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.8901, 3.8986, 2.0837, 4.5352, 2.8890, 4.4081, 2.2975, 2.9091], device='cuda:1'), covar=tensor([0.0172, 0.0278, 0.1615, 0.0079, 0.0689, 0.0403, 0.1348, 0.0659], device='cuda:1'), in_proj_covar=tensor([0.0145, 0.0161, 0.0186, 0.0117, 0.0165, 0.0201, 0.0190, 0.0169], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-04-29 06:53:45,147 INFO [optim.py:368] (1/8) 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 06:54:03,556 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.9546, 2.7105, 2.7182, 1.9805, 2.5734, 2.1691, 2.6148, 2.9237], device='cuda:1'), covar=tensor([0.0320, 0.0697, 0.0514, 0.1617, 0.0741, 0.0860, 0.0627, 0.0615], device='cuda:1'), in_proj_covar=tensor([0.0143, 0.0143, 0.0159, 0.0144, 0.0136, 0.0127, 0.0138, 0.0153], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-04-29 06:54:05,881 INFO [train.py:904] (1/8) Epoch 10, batch 7300, loss[loss=0.2365, simple_loss=0.3174, pruned_loss=0.0778, over 15530.00 frames. ], tot_loss[loss=0.222, simple_loss=0.3038, pruned_loss=0.07005, over 3071959.96 frames. ], batch size: 190, lr: 6.77e-03, grad_scale: 2.0 2023-04-29 06:54:13,216 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.2476, 4.1517, 4.3227, 4.4801, 4.5609, 4.1657, 4.5336, 4.5739], device='cuda:1'), covar=tensor([0.1447, 0.0808, 0.1179, 0.0490, 0.0449, 0.0952, 0.0585, 0.0507], device='cuda:1'), in_proj_covar=tensor([0.0499, 0.0616, 0.0753, 0.0627, 0.0479, 0.0477, 0.0501, 0.0564], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-29 06:55:23,864 INFO [train.py:904] (1/8) Epoch 10, batch 7350, loss[loss=0.2255, simple_loss=0.3077, pruned_loss=0.07164, over 15257.00 frames. ], tot_loss[loss=0.2221, simple_loss=0.3037, pruned_loss=0.07021, over 3081894.27 frames. ], batch size: 190, lr: 6.77e-03, grad_scale: 2.0 2023-04-29 06:56:18,567 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.3579, 2.8477, 2.6379, 2.2057, 2.2502, 2.1696, 2.8671, 2.7799], device='cuda:1'), covar=tensor([0.2083, 0.0656, 0.1296, 0.1792, 0.1801, 0.1721, 0.0481, 0.0893], device='cuda:1'), in_proj_covar=tensor([0.0304, 0.0257, 0.0283, 0.0276, 0.0280, 0.0217, 0.0266, 0.0289], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-29 06:56:19,126 INFO [optim.py:368] (1/8) 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 06:56:41,001 INFO [train.py:904] (1/8) Epoch 10, batch 7400, loss[loss=0.2158, simple_loss=0.3013, pruned_loss=0.06515, over 16492.00 frames. ], tot_loss[loss=0.2229, simple_loss=0.3042, pruned_loss=0.07079, over 3075544.60 frames. ], batch size: 75, lr: 6.77e-03, grad_scale: 2.0 2023-04-29 06:56:57,543 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=98762.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 06:57:03,066 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 10, batch 7450, loss[loss=0.2184, simple_loss=0.3088, pruned_loss=0.06402, over 16814.00 frames. ], tot_loss[loss=0.2245, simple_loss=0.3054, pruned_loss=0.07181, over 3074072.79 frames. ], batch size: 102, lr: 6.77e-03, grad_scale: 2.0 2023-04-29 06:58:14,249 INFO [zipformer.py:625] (1/8) 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] (1/8) 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 06:59:19,899 INFO [train.py:904] (1/8) Epoch 10, batch 7500, loss[loss=0.2434, simple_loss=0.3135, pruned_loss=0.08668, over 16152.00 frames. ], tot_loss[loss=0.2248, simple_loss=0.3061, pruned_loss=0.07173, over 3060173.40 frames. ], batch size: 165, lr: 6.77e-03, grad_scale: 2.0 2023-04-29 06:59:58,717 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.35 vs. limit=2.0 2023-04-29 07:00:20,009 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=98890.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 07:00:20,064 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.5638, 3.1441, 3.0800, 1.8135, 2.7870, 2.2010, 3.1616, 3.2813], device='cuda:1'), covar=tensor([0.0263, 0.0688, 0.0557, 0.1881, 0.0765, 0.0910, 0.0657, 0.0801], device='cuda:1'), in_proj_covar=tensor([0.0142, 0.0141, 0.0157, 0.0143, 0.0135, 0.0126, 0.0136, 0.0151], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-04-29 07:00:27,375 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=98895.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 07:00:34,264 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=98899.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 07:00:38,278 INFO [train.py:904] (1/8) Epoch 10, batch 7550, loss[loss=0.2078, simple_loss=0.293, pruned_loss=0.06128, over 16742.00 frames. ], tot_loss[loss=0.2243, simple_loss=0.3051, pruned_loss=0.07173, over 3057608.50 frames. ], batch size: 134, lr: 6.76e-03, grad_scale: 2.0 2023-04-29 07:01:06,620 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.75 vs. limit=2.0 2023-04-29 07:01:15,637 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.0204, 2.3418, 2.3367, 2.7654, 2.1479, 3.2050, 1.7673, 2.6930], device='cuda:1'), covar=tensor([0.1049, 0.0523, 0.0849, 0.0131, 0.0120, 0.0344, 0.1198, 0.0607], device='cuda:1'), in_proj_covar=tensor([0.0153, 0.0155, 0.0177, 0.0134, 0.0199, 0.0204, 0.0176, 0.0176], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0003, 0.0004], device='cuda:1') 2023-04-29 07:01:32,281 INFO [optim.py:368] (1/8) 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 07:01:53,300 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=98951.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 07:01:53,976 INFO [train.py:904] (1/8) Epoch 10, batch 7600, loss[loss=0.1959, simple_loss=0.279, pruned_loss=0.05643, over 17042.00 frames. ], tot_loss[loss=0.2234, simple_loss=0.304, pruned_loss=0.07139, over 3061527.26 frames. ], batch size: 53, lr: 6.76e-03, grad_scale: 4.0 2023-04-29 07:02:01,637 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=98956.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 07:02:24,349 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 10, batch 7650, loss[loss=0.2096, simple_loss=0.2997, pruned_loss=0.05974, over 16815.00 frames. ], tot_loss[loss=0.2237, simple_loss=0.304, pruned_loss=0.0717, over 3068023.79 frames. ], batch size: 83, lr: 6.76e-03, grad_scale: 4.0 2023-04-29 07:03:53,689 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.7799, 3.5866, 3.8210, 3.6053, 3.7463, 4.1820, 3.8829, 3.6218], device='cuda:1'), covar=tensor([0.1873, 0.2347, 0.2129, 0.2414, 0.2613, 0.1583, 0.1519, 0.2658], device='cuda:1'), in_proj_covar=tensor([0.0344, 0.0479, 0.0516, 0.0416, 0.0547, 0.0541, 0.0420, 0.0569], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-04-29 07:03:59,317 INFO [zipformer.py:625] (1/8) 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] (1/8) 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 07:04:28,329 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=99051.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 07:04:29,003 INFO [train.py:904] (1/8) Epoch 10, batch 7700, loss[loss=0.2421, simple_loss=0.3323, pruned_loss=0.07596, over 16207.00 frames. ], tot_loss[loss=0.2251, simple_loss=0.3049, pruned_loss=0.07259, over 3077060.74 frames. ], batch size: 165, lr: 6.76e-03, grad_scale: 4.0 2023-04-29 07:04:50,084 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=99066.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 07:05:43,695 INFO [train.py:904] (1/8) Epoch 10, batch 7750, loss[loss=0.2497, simple_loss=0.3245, pruned_loss=0.08745, over 15474.00 frames. ], tot_loss[loss=0.224, simple_loss=0.3044, pruned_loss=0.07179, over 3089281.55 frames. ], batch size: 191, lr: 6.76e-03, grad_scale: 4.0 2023-04-29 07:05:51,126 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.5359, 4.5185, 4.3890, 3.7073, 4.4651, 1.5381, 4.1385, 4.1595], device='cuda:1'), covar=tensor([0.0091, 0.0072, 0.0158, 0.0351, 0.0079, 0.2487, 0.0124, 0.0175], device='cuda:1'), in_proj_covar=tensor([0.0122, 0.0107, 0.0155, 0.0150, 0.0126, 0.0173, 0.0142, 0.0144], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-29 07:05:59,701 INFO [zipformer.py:625] (1/8) 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] (1/8) 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] (1/8) Clipping_scale=2.0, grad-norm quartiles 2.181e+02 3.321e+02 3.885e+02 5.264e+02 1.269e+03, threshold=7.770e+02, percent-clipped=1.0 2023-04-29 07:06:59,330 INFO [train.py:904] (1/8) Epoch 10, batch 7800, loss[loss=0.2229, simple_loss=0.3146, pruned_loss=0.06558, over 16633.00 frames. ], tot_loss[loss=0.2258, simple_loss=0.3059, pruned_loss=0.07282, over 3086595.31 frames. ], batch size: 76, lr: 6.76e-03, grad_scale: 4.0 2023-04-29 07:07:34,693 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.21 vs. limit=2.0 2023-04-29 07:07:57,237 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.9974, 3.5601, 3.3574, 1.9314, 2.8075, 2.2491, 3.2849, 3.6711], device='cuda:1'), covar=tensor([0.0334, 0.0659, 0.0573, 0.1864, 0.0900, 0.0970, 0.0825, 0.0785], device='cuda:1'), in_proj_covar=tensor([0.0142, 0.0142, 0.0157, 0.0142, 0.0136, 0.0126, 0.0136, 0.0152], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-04-29 07:08:08,294 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.64 vs. limit=2.0 2023-04-29 07:08:12,878 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=99199.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 07:08:16,573 INFO [train.py:904] (1/8) Epoch 10, batch 7850, loss[loss=0.2117, simple_loss=0.3034, pruned_loss=0.05996, over 16191.00 frames. ], tot_loss[loss=0.2257, simple_loss=0.3066, pruned_loss=0.07238, over 3104859.54 frames. ], batch size: 165, lr: 6.75e-03, grad_scale: 4.0 2023-04-29 07:08:38,121 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.8244, 1.7713, 2.2340, 2.8604, 2.8160, 3.0706, 1.7834, 3.0328], device='cuda:1'), covar=tensor([0.0144, 0.0369, 0.0239, 0.0178, 0.0180, 0.0123, 0.0379, 0.0099], device='cuda:1'), in_proj_covar=tensor([0.0154, 0.0167, 0.0150, 0.0154, 0.0163, 0.0117, 0.0169, 0.0108], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001], device='cuda:1') 2023-04-29 07:08:57,397 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=99229.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 07:09:10,221 INFO [optim.py:368] (1/8) Clipping_scale=2.0, grad-norm quartiles 2.079e+02 2.953e+02 3.751e+02 4.670e+02 9.934e+02, threshold=7.502e+02, percent-clipped=3.0 2023-04-29 07:09:20,895 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.58 vs. limit=2.0 2023-04-29 07:09:22,397 INFO [zipformer.py:625] (1/8) 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] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=99247.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 07:09:29,679 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 10, batch 7900, loss[loss=0.2089, simple_loss=0.303, pruned_loss=0.05744, over 16637.00 frames. ], tot_loss[loss=0.2244, simple_loss=0.3058, pruned_loss=0.0715, over 3099494.25 frames. ], batch size: 62, lr: 6.75e-03, grad_scale: 4.0 2023-04-29 07:09:55,542 INFO [zipformer.py:625] (1/8) 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] (1/8) 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] (1/8) Epoch 10, batch 7950, loss[loss=0.2589, simple_loss=0.3195, pruned_loss=0.09919, over 11775.00 frames. ], tot_loss[loss=0.2249, simple_loss=0.3062, pruned_loss=0.07178, over 3099286.62 frames. ], batch size: 246, lr: 6.75e-03, grad_scale: 4.0 2023-04-29 07:11:05,199 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.8466, 2.5800, 2.0004, 2.4776, 3.0498, 2.8371, 3.6441, 3.3989], device='cuda:1'), covar=tensor([0.0053, 0.0268, 0.0413, 0.0293, 0.0182, 0.0253, 0.0167, 0.0157], device='cuda:1'), in_proj_covar=tensor([0.0131, 0.0194, 0.0193, 0.0191, 0.0192, 0.0195, 0.0196, 0.0181], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-29 07:11:13,146 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.2245, 3.3771, 3.6006, 3.5846, 3.5907, 3.3710, 3.3970, 3.4390], device='cuda:1'), covar=tensor([0.0376, 0.0558, 0.0408, 0.0435, 0.0458, 0.0465, 0.0851, 0.0531], device='cuda:1'), in_proj_covar=tensor([0.0316, 0.0322, 0.0325, 0.0311, 0.0373, 0.0344, 0.0444, 0.0281], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003], device='cuda:1') 2023-04-29 07:11:22,402 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=99324.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 07:11:27,240 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=99327.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 07:11:30,489 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=99329.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 07:11:43,783 INFO [optim.py:368] (1/8) 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 07:11:44,140 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.8804, 5.3309, 5.5715, 5.3986, 5.4123, 5.9358, 5.4214, 5.2308], device='cuda:1'), covar=tensor([0.0908, 0.1523, 0.1405, 0.1706, 0.2194, 0.0802, 0.1215, 0.2244], device='cuda:1'), in_proj_covar=tensor([0.0339, 0.0471, 0.0510, 0.0410, 0.0539, 0.0535, 0.0412, 0.0562], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-04-29 07:12:06,035 INFO [train.py:904] (1/8) Epoch 10, batch 8000, loss[loss=0.211, simple_loss=0.299, pruned_loss=0.06145, over 16707.00 frames. ], tot_loss[loss=0.2271, simple_loss=0.3074, pruned_loss=0.07342, over 3076375.74 frames. ], batch size: 134, lr: 6.75e-03, grad_scale: 8.0 2023-04-29 07:12:34,147 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.8868, 5.3580, 5.5839, 5.4309, 5.4150, 5.9371, 5.4054, 5.1981], device='cuda:1'), covar=tensor([0.0810, 0.1460, 0.1516, 0.1459, 0.1975, 0.0871, 0.1207, 0.2080], device='cuda:1'), in_proj_covar=tensor([0.0338, 0.0470, 0.0507, 0.0408, 0.0537, 0.0533, 0.0410, 0.0560], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-04-29 07:12:57,264 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 10, batch 8050, loss[loss=0.1973, simple_loss=0.2938, pruned_loss=0.05037, over 16764.00 frames. ], tot_loss[loss=0.2267, simple_loss=0.3073, pruned_loss=0.07303, over 3087170.48 frames. ], batch size: 102, lr: 6.75e-03, grad_scale: 8.0 2023-04-29 07:13:29,868 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=99407.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 07:13:50,392 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=99420.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 07:14:18,153 INFO [optim.py:368] (1/8) 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] (1/8) Epoch 10, batch 8100, loss[loss=0.2145, simple_loss=0.2962, pruned_loss=0.06643, over 16106.00 frames. ], tot_loss[loss=0.2247, simple_loss=0.3059, pruned_loss=0.07173, over 3097817.28 frames. ], batch size: 35, lr: 6.75e-03, grad_scale: 8.0 2023-04-29 07:15:03,987 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=99468.0, num_to_drop=1, layers_to_drop={0} 2023-04-29 07:15:06,349 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.6936, 4.6945, 4.5127, 4.3390, 4.1584, 4.6182, 4.4779, 4.2743], device='cuda:1'), covar=tensor([0.0526, 0.0327, 0.0255, 0.0239, 0.0918, 0.0383, 0.0365, 0.0599], device='cuda:1'), in_proj_covar=tensor([0.0236, 0.0297, 0.0274, 0.0254, 0.0297, 0.0290, 0.0190, 0.0320], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-29 07:15:23,068 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 10, batch 8150, loss[loss=0.1904, simple_loss=0.2709, pruned_loss=0.05489, over 16484.00 frames. ], tot_loss[loss=0.222, simple_loss=0.3033, pruned_loss=0.07033, over 3110063.65 frames. ], batch size: 68, lr: 6.74e-03, grad_scale: 8.0 2023-04-29 07:16:22,913 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.16 vs. limit=2.0 2023-04-29 07:16:35,916 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=99529.0, num_to_drop=1, layers_to_drop={3} 2023-04-29 07:16:44,655 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=4.87 vs. limit=5.0 2023-04-29 07:16:49,692 INFO [optim.py:368] (1/8) 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 07:17:00,819 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=99546.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 07:17:08,899 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=99551.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 07:17:09,685 INFO [train.py:904] (1/8) Epoch 10, batch 8200, loss[loss=0.2079, simple_loss=0.2963, pruned_loss=0.05973, over 16832.00 frames. ], tot_loss[loss=0.2199, simple_loss=0.3011, pruned_loss=0.06939, over 3121400.84 frames. ], batch size: 102, lr: 6.74e-03, grad_scale: 4.0 2023-04-29 07:17:35,590 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=99568.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 07:17:47,717 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.9469, 4.2079, 3.9713, 4.0495, 3.6880, 3.8243, 3.8617, 4.1654], device='cuda:1'), covar=tensor([0.0976, 0.0857, 0.1001, 0.0728, 0.0780, 0.1414, 0.0894, 0.0995], device='cuda:1'), in_proj_covar=tensor([0.0514, 0.0644, 0.0541, 0.0449, 0.0403, 0.0426, 0.0537, 0.0489], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:1') 2023-04-29 07:18:04,717 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=99585.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 07:18:11,203 INFO [zipformer.py:625] (1/8) 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] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=99594.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 07:18:28,355 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 10, batch 8250, loss[loss=0.1707, simple_loss=0.2518, pruned_loss=0.04478, over 11823.00 frames. ], tot_loss[loss=0.218, simple_loss=0.3001, pruned_loss=0.06794, over 3080046.51 frames. ], batch size: 248, lr: 6.74e-03, grad_scale: 4.0 2023-04-29 07:19:07,440 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.98 vs. limit=2.0 2023-04-29 07:19:10,184 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=99624.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 07:19:16,073 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=99627.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 07:19:19,878 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=99629.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 07:19:36,139 INFO [optim.py:368] (1/8) 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 07:19:40,670 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.87 vs. limit=2.0 2023-04-29 07:19:55,535 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=99650.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 07:19:57,956 INFO [train.py:904] (1/8) Epoch 10, batch 8300, loss[loss=0.2031, simple_loss=0.3078, pruned_loss=0.04919, over 16862.00 frames. ], tot_loss[loss=0.213, simple_loss=0.2968, pruned_loss=0.06458, over 3057403.85 frames. ], batch size: 102, lr: 6.74e-03, grad_scale: 4.0 2023-04-29 07:20:37,235 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=99675.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 07:20:44,582 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=99680.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 07:20:44,735 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=99680.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 07:21:21,642 INFO [train.py:904] (1/8) Epoch 10, batch 8350, loss[loss=0.2309, simple_loss=0.3035, pruned_loss=0.07912, over 11861.00 frames. ], tot_loss[loss=0.21, simple_loss=0.295, pruned_loss=0.0625, over 3029213.10 frames. ], batch size: 248, lr: 6.74e-03, grad_scale: 4.0 2023-04-29 07:21:29,850 INFO [zipformer.py:625] (1/8) 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] (1/8) Clipping_scale=2.0, grad-norm quartiles 1.480e+02 2.332e+02 2.884e+02 3.614e+02 8.033e+02, threshold=5.769e+02, percent-clipped=2.0 2023-04-29 07:22:26,022 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=99741.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 07:22:43,904 INFO [train.py:904] (1/8) Epoch 10, batch 8400, loss[loss=0.1921, simple_loss=0.2841, pruned_loss=0.05004, over 17028.00 frames. ], tot_loss[loss=0.2066, simple_loss=0.2921, pruned_loss=0.06048, over 3013768.43 frames. ], batch size: 53, lr: 6.74e-03, grad_scale: 8.0 2023-04-29 07:22:49,362 INFO [zipformer.py:625] (1/8) 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] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=99776.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 07:24:06,249 INFO [train.py:904] (1/8) Epoch 10, batch 8450, loss[loss=0.1809, simple_loss=0.2684, pruned_loss=0.04668, over 17051.00 frames. ], tot_loss[loss=0.2037, simple_loss=0.2905, pruned_loss=0.05848, over 3030292.84 frames. ], batch size: 55, lr: 6.73e-03, grad_scale: 8.0 2023-04-29 07:24:18,074 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.28 vs. limit=2.0 2023-04-29 07:24:39,121 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.2663, 2.5805, 2.0575, 2.1644, 2.8794, 2.5523, 3.0566, 3.0841], device='cuda:1'), covar=tensor([0.0097, 0.0241, 0.0340, 0.0328, 0.0176, 0.0238, 0.0161, 0.0152], device='cuda:1'), in_proj_covar=tensor([0.0129, 0.0189, 0.0188, 0.0186, 0.0189, 0.0191, 0.0189, 0.0177], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-29 07:24:42,069 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=99824.0, num_to_drop=1, layers_to_drop={2} 2023-04-29 07:25:06,072 INFO [optim.py:368] (1/8) 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 07:25:25,942 INFO [train.py:904] (1/8) Epoch 10, batch 8500, loss[loss=0.1894, simple_loss=0.2842, pruned_loss=0.04731, over 16773.00 frames. ], tot_loss[loss=0.1992, simple_loss=0.2864, pruned_loss=0.05594, over 3019647.83 frames. ], batch size: 102, lr: 6.73e-03, grad_scale: 8.0 2023-04-29 07:26:21,537 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 10, batch 8550, loss[loss=0.2115, simple_loss=0.2993, pruned_loss=0.06183, over 16909.00 frames. ], tot_loss[loss=0.1966, simple_loss=0.2838, pruned_loss=0.05467, over 3004865.45 frames. ], batch size: 109, lr: 6.73e-03, grad_scale: 8.0 2023-04-29 07:27:34,859 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=99924.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 07:27:34,899 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=99924.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 07:27:36,391 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.0681, 1.3813, 1.6708, 2.0542, 2.0608, 2.1042, 1.6129, 2.2204], device='cuda:1'), covar=tensor([0.0170, 0.0382, 0.0240, 0.0213, 0.0243, 0.0188, 0.0366, 0.0091], device='cuda:1'), in_proj_covar=tensor([0.0152, 0.0167, 0.0149, 0.0152, 0.0163, 0.0116, 0.0168, 0.0106], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001], device='cuda:1') 2023-04-29 07:27:50,547 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=99933.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 07:28:02,243 INFO [optim.py:368] (1/8) 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] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=99945.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 07:28:29,552 INFO [train.py:904] (1/8) Epoch 10, batch 8600, loss[loss=0.1901, simple_loss=0.2718, pruned_loss=0.05422, over 12702.00 frames. ], tot_loss[loss=0.1954, simple_loss=0.2844, pruned_loss=0.05319, over 3041414.93 frames. ], batch size: 247, lr: 6.73e-03, grad_scale: 8.0 2023-04-29 07:29:10,990 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=99972.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 07:29:25,998 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 10, batch 8650, loss[loss=0.1818, simple_loss=0.2756, pruned_loss=0.04402, over 16339.00 frames. ], tot_loss[loss=0.192, simple_loss=0.2816, pruned_loss=0.05123, over 3032980.38 frames. ], batch size: 146, lr: 6.73e-03, grad_scale: 8.0 2023-04-29 07:30:24,897 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=100007.0, num_to_drop=1, layers_to_drop={0} 2023-04-29 07:30:50,456 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.3017, 4.4156, 4.6169, 4.4658, 4.5158, 4.9933, 4.5906, 4.2931], device='cuda:1'), covar=tensor([0.1225, 0.1645, 0.1768, 0.1784, 0.2455, 0.0999, 0.1141, 0.2329], device='cuda:1'), in_proj_covar=tensor([0.0318, 0.0448, 0.0481, 0.0389, 0.0511, 0.0510, 0.0391, 0.0531], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-04-29 07:31:10,430 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=100028.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 07:31:27,926 INFO [zipformer.py:625] (1/8) 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] (1/8) 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] (1/8) Epoch 10, batch 8700, loss[loss=0.1838, simple_loss=0.2734, pruned_loss=0.04708, over 16968.00 frames. ], tot_loss[loss=0.1896, simple_loss=0.2791, pruned_loss=0.05002, over 3062214.10 frames. ], batch size: 109, lr: 6.73e-03, grad_scale: 8.0 2023-04-29 07:32:18,051 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.84 vs. limit=2.0 2023-04-29 07:32:26,371 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.7768, 3.7403, 3.9361, 3.7440, 3.7991, 4.2855, 3.9254, 3.6722], device='cuda:1'), covar=tensor([0.1787, 0.1935, 0.1795, 0.2158, 0.3031, 0.1528, 0.1571, 0.2467], device='cuda:1'), in_proj_covar=tensor([0.0316, 0.0445, 0.0477, 0.0384, 0.0507, 0.0506, 0.0390, 0.0526], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002], device='cuda:1') 2023-04-29 07:32:28,015 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=100068.0, num_to_drop=1, layers_to_drop={2} 2023-04-29 07:32:42,879 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=100076.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 07:33:35,357 INFO [train.py:904] (1/8) Epoch 10, batch 8750, loss[loss=0.2118, simple_loss=0.3076, pruned_loss=0.058, over 16708.00 frames. ], tot_loss[loss=0.189, simple_loss=0.279, pruned_loss=0.04951, over 3063721.03 frames. ], batch size: 134, lr: 6.72e-03, grad_scale: 8.0 2023-04-29 07:34:32,194 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=100124.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 07:34:32,268 INFO [zipformer.py:625] (1/8) 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] (1/8) 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 07:35:15,871 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.49 vs. limit=2.0 2023-04-29 07:35:29,749 INFO [train.py:904] (1/8) Epoch 10, batch 8800, loss[loss=0.1917, simple_loss=0.2781, pruned_loss=0.05267, over 12608.00 frames. ], tot_loss[loss=0.1873, simple_loss=0.2775, pruned_loss=0.04858, over 3068672.85 frames. ], batch size: 246, lr: 6.72e-03, grad_scale: 8.0 2023-04-29 07:35:44,624 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.1757, 3.2577, 1.7640, 3.4890, 2.2976, 3.4542, 2.0640, 2.6349], device='cuda:1'), covar=tensor([0.0247, 0.0336, 0.1742, 0.0162, 0.0949, 0.0552, 0.1496, 0.0719], device='cuda:1'), in_proj_covar=tensor([0.0141, 0.0153, 0.0180, 0.0115, 0.0159, 0.0192, 0.0189, 0.0164], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0004, 0.0004, 0.0003], device='cuda:1') 2023-04-29 07:36:10,203 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.47 vs. limit=2.0 2023-04-29 07:36:11,836 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=100172.0, num_to_drop=1, layers_to_drop={0} 2023-04-29 07:37:15,875 INFO [train.py:904] (1/8) Epoch 10, batch 8850, loss[loss=0.1869, simple_loss=0.2874, pruned_loss=0.04322, over 16651.00 frames. ], tot_loss[loss=0.1877, simple_loss=0.2793, pruned_loss=0.04802, over 3043482.22 frames. ], batch size: 134, lr: 6.72e-03, grad_scale: 8.0 2023-04-29 07:38:03,388 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=100224.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 07:38:34,758 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=100238.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 07:38:36,932 INFO [optim.py:368] (1/8) 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 07:38:48,408 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 10, batch 8900, loss[loss=0.1748, simple_loss=0.2675, pruned_loss=0.0411, over 12550.00 frames. ], tot_loss[loss=0.1878, simple_loss=0.2802, pruned_loss=0.04765, over 3051247.12 frames. ], batch size: 247, lr: 6.72e-03, grad_scale: 4.0 2023-04-29 07:39:41,466 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=4.85 vs. limit=5.0 2023-04-29 07:39:43,458 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=100272.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 07:40:45,779 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=100293.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 07:41:01,081 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=100299.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 07:41:05,910 INFO [train.py:904] (1/8) Epoch 10, batch 8950, loss[loss=0.1963, simple_loss=0.277, pruned_loss=0.05782, over 12703.00 frames. ], tot_loss[loss=0.1887, simple_loss=0.2807, pruned_loss=0.04834, over 3060687.11 frames. ], batch size: 246, lr: 6.72e-03, grad_scale: 4.0 2023-04-29 07:41:28,997 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.2765, 1.9047, 2.0515, 3.8617, 1.9586, 2.2978, 2.1031, 2.0541], device='cuda:1'), covar=tensor([0.0905, 0.3538, 0.2371, 0.0386, 0.4067, 0.2438, 0.3321, 0.3569], device='cuda:1'), in_proj_covar=tensor([0.0339, 0.0371, 0.0313, 0.0305, 0.0398, 0.0416, 0.0334, 0.0431], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-29 07:41:47,311 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([1.7962, 1.2886, 1.6205, 1.7584, 1.8288, 1.8906, 1.5566, 1.7722], device='cuda:1'), covar=tensor([0.0188, 0.0298, 0.0149, 0.0179, 0.0190, 0.0135, 0.0304, 0.0073], device='cuda:1'), in_proj_covar=tensor([0.0152, 0.0165, 0.0147, 0.0149, 0.0159, 0.0113, 0.0166, 0.0104], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001], device='cuda:1') 2023-04-29 07:42:21,909 INFO [zipformer.py:625] (1/8) 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] (1/8) 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 07:42:52,441 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.2793, 3.4191, 3.6735, 1.7570, 3.8678, 3.9436, 2.9481, 2.8467], device='cuda:1'), covar=tensor([0.0765, 0.0196, 0.0145, 0.1165, 0.0042, 0.0083, 0.0350, 0.0416], device='cuda:1'), in_proj_covar=tensor([0.0140, 0.0097, 0.0082, 0.0136, 0.0066, 0.0094, 0.0117, 0.0124], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0003, 0.0002, 0.0004, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:1') 2023-04-29 07:42:55,809 INFO [train.py:904] (1/8) Epoch 10, batch 9000, loss[loss=0.1623, simple_loss=0.2435, pruned_loss=0.04056, over 16584.00 frames. ], tot_loss[loss=0.1854, simple_loss=0.277, pruned_loss=0.0469, over 3057329.40 frames. ], batch size: 62, lr: 6.71e-03, grad_scale: 4.0 2023-04-29 07:42:55,809 INFO [train.py:929] (1/8) Computing validation loss 2023-04-29 07:43:03,866 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([5.6165, 5.6167, 5.9932, 5.9969, 5.9846, 5.7591, 5.6881, 5.4805], device='cuda:1'), covar=tensor([0.0205, 0.0328, 0.0234, 0.0261, 0.0286, 0.0216, 0.0553, 0.0301], device='cuda:1'), in_proj_covar=tensor([0.0298, 0.0306, 0.0309, 0.0297, 0.0355, 0.0328, 0.0418, 0.0268], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0002], device='cuda:1') 2023-04-29 07:43:05,469 INFO [train.py:938] (1/8) Epoch 10, validation: loss=0.1565, simple_loss=0.2604, pruned_loss=0.02634, over 944034.00 frames. 2023-04-29 07:43:05,470 INFO [train.py:939] (1/8) Maximum memory allocated so far is 17676MB 2023-04-29 07:43:29,847 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=100363.0, num_to_drop=1, layers_to_drop={0} 2023-04-29 07:44:12,993 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=100384.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 07:44:50,497 INFO [train.py:904] (1/8) Epoch 10, batch 9050, loss[loss=0.19, simple_loss=0.2839, pruned_loss=0.04807, over 16541.00 frames. ], tot_loss[loss=0.1867, simple_loss=0.2784, pruned_loss=0.04748, over 3063190.06 frames. ], batch size: 68, lr: 6.71e-03, grad_scale: 4.0 2023-04-29 07:44:51,440 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=100402.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 07:46:08,563 INFO [optim.py:368] (1/8) Clipping_scale=2.0, grad-norm quartiles 1.813e+02 2.533e+02 2.923e+02 3.957e+02 1.423e+03, threshold=5.846e+02, percent-clipped=5.0 2023-04-29 07:46:37,395 INFO [train.py:904] (1/8) Epoch 10, batch 9100, loss[loss=0.1926, simple_loss=0.2878, pruned_loss=0.04865, over 16895.00 frames. ], tot_loss[loss=0.1866, simple_loss=0.2777, pruned_loss=0.04772, over 3071389.38 frames. ], batch size: 116, lr: 6.71e-03, grad_scale: 4.0 2023-04-29 07:46:58,859 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=100463.0, num_to_drop=1, layers_to_drop={1} 2023-04-29 07:48:34,505 INFO [train.py:904] (1/8) Epoch 10, batch 9150, loss[loss=0.1821, simple_loss=0.2733, pruned_loss=0.04549, over 15508.00 frames. ], tot_loss[loss=0.1867, simple_loss=0.2786, pruned_loss=0.04736, over 3077398.70 frames. ], batch size: 191, lr: 6.71e-03, grad_scale: 4.0 2023-04-29 07:48:45,350 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=3.22 vs. limit=5.0 2023-04-29 07:49:08,878 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.38 vs. limit=2.0 2023-04-29 07:49:54,593 INFO [optim.py:368] (1/8) 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 07:50:13,379 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([1.9692, 2.2283, 2.3671, 2.9190, 2.1640, 3.2889, 1.6412, 2.8602], device='cuda:1'), covar=tensor([0.1173, 0.0574, 0.0915, 0.0101, 0.0095, 0.0355, 0.1300, 0.0622], device='cuda:1'), in_proj_covar=tensor([0.0155, 0.0155, 0.0178, 0.0132, 0.0188, 0.0202, 0.0179, 0.0179], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0003, 0.0004], device='cuda:1') 2023-04-29 07:50:13,981 INFO [train.py:904] (1/8) Epoch 10, batch 9200, loss[loss=0.1812, simple_loss=0.2657, pruned_loss=0.04835, over 12362.00 frames. ], tot_loss[loss=0.1832, simple_loss=0.274, pruned_loss=0.04613, over 3062056.11 frames. ], batch size: 248, lr: 6.71e-03, grad_scale: 8.0 2023-04-29 07:51:35,108 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=100594.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 07:51:50,957 INFO [train.py:904] (1/8) Epoch 10, batch 9250, loss[loss=0.1929, simple_loss=0.2838, pruned_loss=0.05097, over 15295.00 frames. ], tot_loss[loss=0.1829, simple_loss=0.2736, pruned_loss=0.0461, over 3059690.99 frames. ], batch size: 191, lr: 6.71e-03, grad_scale: 8.0 2023-04-29 07:53:14,008 INFO [optim.py:368] (1/8) 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 07:53:31,575 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.5121, 4.4710, 4.3433, 3.8711, 4.3536, 1.7896, 4.1814, 4.1877], device='cuda:1'), covar=tensor([0.0098, 0.0088, 0.0143, 0.0270, 0.0098, 0.2267, 0.0112, 0.0187], device='cuda:1'), in_proj_covar=tensor([0.0119, 0.0107, 0.0149, 0.0140, 0.0123, 0.0171, 0.0137, 0.0138], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0001, 0.0001, 0.0002, 0.0001, 0.0002], device='cuda:1') 2023-04-29 07:53:42,537 INFO [train.py:904] (1/8) Epoch 10, batch 9300, loss[loss=0.1624, simple_loss=0.2504, pruned_loss=0.03717, over 17019.00 frames. ], tot_loss[loss=0.181, simple_loss=0.2715, pruned_loss=0.04527, over 3051823.94 frames. ], batch size: 41, lr: 6.70e-03, grad_scale: 8.0 2023-04-29 07:54:09,131 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=100663.0, num_to_drop=1, layers_to_drop={2} 2023-04-29 07:55:29,731 INFO [train.py:904] (1/8) Epoch 10, batch 9350, loss[loss=0.1697, simple_loss=0.2568, pruned_loss=0.0413, over 17033.00 frames. ], tot_loss[loss=0.1805, simple_loss=0.271, pruned_loss=0.04496, over 3053763.94 frames. ], batch size: 53, lr: 6.70e-03, grad_scale: 8.0 2023-04-29 07:55:39,083 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.1774, 4.2777, 4.4322, 4.2920, 4.3396, 4.8235, 4.4328, 4.1608], device='cuda:1'), covar=tensor([0.1514, 0.1602, 0.1797, 0.1898, 0.2415, 0.0929, 0.1247, 0.2357], device='cuda:1'), in_proj_covar=tensor([0.0317, 0.0443, 0.0478, 0.0386, 0.0506, 0.0508, 0.0389, 0.0522], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002], device='cuda:1') 2023-04-29 07:55:50,101 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=100711.0, num_to_drop=1, layers_to_drop={0} 2023-04-29 07:56:28,468 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=100730.0, num_to_drop=1, layers_to_drop={0} 2023-04-29 07:56:42,410 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.38 vs. limit=2.0 2023-04-29 07:56:48,376 INFO [optim.py:368] (1/8) 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] (1/8) Epoch 10, batch 9400, loss[loss=0.1927, simple_loss=0.2935, pruned_loss=0.04591, over 16726.00 frames. ], tot_loss[loss=0.1791, simple_loss=0.2699, pruned_loss=0.04416, over 3049084.68 frames. ], batch size: 83, lr: 6.70e-03, grad_scale: 8.0 2023-04-29 07:57:25,167 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=100758.0, num_to_drop=1, layers_to_drop={3} 2023-04-29 07:57:40,884 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.4350, 3.0997, 3.1144, 1.8966, 2.7567, 2.2028, 2.9840, 3.1861], device='cuda:1'), covar=tensor([0.0298, 0.0600, 0.0480, 0.1697, 0.0709, 0.0893, 0.0705, 0.0683], device='cuda:1'), in_proj_covar=tensor([0.0137, 0.0132, 0.0152, 0.0139, 0.0132, 0.0122, 0.0132, 0.0141], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:1') 2023-04-29 07:58:33,104 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=100791.0, num_to_drop=1, layers_to_drop={2} 2023-04-29 07:58:52,903 INFO [train.py:904] (1/8) Epoch 10, batch 9450, loss[loss=0.1674, simple_loss=0.2602, pruned_loss=0.0373, over 16424.00 frames. ], tot_loss[loss=0.1806, simple_loss=0.2722, pruned_loss=0.04452, over 3067533.51 frames. ], batch size: 68, lr: 6.70e-03, grad_scale: 8.0 2023-04-29 07:58:57,644 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.2431, 4.2211, 4.6253, 2.3480, 4.8344, 4.8826, 3.5468, 3.8889], device='cuda:1'), covar=tensor([0.0531, 0.0148, 0.0130, 0.0985, 0.0025, 0.0055, 0.0259, 0.0281], device='cuda:1'), in_proj_covar=tensor([0.0137, 0.0095, 0.0081, 0.0133, 0.0064, 0.0092, 0.0114, 0.0122], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0003, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003, 0.0003], device='cuda:1') 2023-04-29 07:59:12,454 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.84 vs. limit=2.0 2023-04-29 08:00:10,245 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.0596, 4.0088, 3.9451, 3.5120, 3.9535, 1.7557, 3.7902, 3.6163], device='cuda:1'), covar=tensor([0.0067, 0.0068, 0.0107, 0.0177, 0.0066, 0.2258, 0.0089, 0.0146], device='cuda:1'), in_proj_covar=tensor([0.0119, 0.0106, 0.0148, 0.0139, 0.0123, 0.0171, 0.0136, 0.0138], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0001, 0.0001, 0.0002, 0.0001, 0.0002], device='cuda:1') 2023-04-29 08:00:10,871 INFO [optim.py:368] (1/8) 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 08:00:34,622 INFO [train.py:904] (1/8) Epoch 10, batch 9500, loss[loss=0.1712, simple_loss=0.2596, pruned_loss=0.04141, over 16684.00 frames. ], tot_loss[loss=0.1802, simple_loss=0.2718, pruned_loss=0.04434, over 3082159.10 frames. ], batch size: 62, lr: 6.70e-03, grad_scale: 8.0 2023-04-29 08:01:49,084 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.1550, 3.4477, 3.4637, 2.3853, 3.1129, 3.4306, 3.3551, 1.9013], device='cuda:1'), covar=tensor([0.0371, 0.0031, 0.0034, 0.0281, 0.0085, 0.0066, 0.0051, 0.0382], device='cuda:1'), in_proj_covar=tensor([0.0124, 0.0062, 0.0066, 0.0121, 0.0075, 0.0084, 0.0073, 0.0116], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:1') 2023-04-29 08:02:03,124 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=100894.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 08:02:20,731 INFO [train.py:904] (1/8) Epoch 10, batch 9550, loss[loss=0.1791, simple_loss=0.2783, pruned_loss=0.03998, over 16368.00 frames. ], tot_loss[loss=0.1809, simple_loss=0.2721, pruned_loss=0.04484, over 3086875.07 frames. ], batch size: 146, lr: 6.70e-03, grad_scale: 8.0 2023-04-29 08:03:40,249 INFO [optim.py:368] (1/8) 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 08:03:45,435 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=100942.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 08:04:02,627 INFO [train.py:904] (1/8) Epoch 10, batch 9600, loss[loss=0.1958, simple_loss=0.3016, pruned_loss=0.04494, over 16861.00 frames. ], tot_loss[loss=0.1829, simple_loss=0.2737, pruned_loss=0.04606, over 3072982.82 frames. ], batch size: 96, lr: 6.70e-03, grad_scale: 8.0 2023-04-29 08:05:52,879 INFO [train.py:904] (1/8) Epoch 10, batch 9650, loss[loss=0.201, simple_loss=0.296, pruned_loss=0.05299, over 16666.00 frames. ], tot_loss[loss=0.1842, simple_loss=0.2755, pruned_loss=0.04644, over 3068236.16 frames. ], batch size: 134, lr: 6.69e-03, grad_scale: 8.0 2023-04-29 08:06:43,411 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.5680, 3.5949, 3.9393, 1.9391, 4.1189, 4.2292, 3.0557, 3.2495], device='cuda:1'), covar=tensor([0.0686, 0.0201, 0.0130, 0.1048, 0.0048, 0.0072, 0.0341, 0.0328], device='cuda:1'), in_proj_covar=tensor([0.0136, 0.0095, 0.0080, 0.0133, 0.0064, 0.0092, 0.0114, 0.0121], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0003, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003, 0.0003], device='cuda:1') 2023-04-29 08:06:57,391 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.1473, 3.4133, 3.4512, 2.2491, 3.2290, 3.5127, 3.3487, 1.9428], device='cuda:1'), covar=tensor([0.0405, 0.0032, 0.0034, 0.0338, 0.0072, 0.0052, 0.0057, 0.0385], device='cuda:1'), in_proj_covar=tensor([0.0122, 0.0062, 0.0065, 0.0120, 0.0074, 0.0082, 0.0071, 0.0115], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:1') 2023-04-29 08:07:15,509 INFO [optim.py:368] (1/8) Clipping_scale=2.0, grad-norm quartiles 1.746e+02 2.568e+02 3.331e+02 4.260e+02 1.013e+03, threshold=6.663e+02, percent-clipped=7.0 2023-04-29 08:07:41,458 INFO [train.py:904] (1/8) Epoch 10, batch 9700, loss[loss=0.1719, simple_loss=0.2656, pruned_loss=0.03914, over 16239.00 frames. ], tot_loss[loss=0.1834, simple_loss=0.2745, pruned_loss=0.04608, over 3082340.14 frames. ], batch size: 165, lr: 6.69e-03, grad_scale: 8.0 2023-04-29 08:07:52,537 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=101058.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 08:08:53,872 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=101086.0, num_to_drop=1, layers_to_drop={2} 2023-04-29 08:09:25,194 INFO [train.py:904] (1/8) Epoch 10, batch 9750, loss[loss=0.1894, simple_loss=0.2859, pruned_loss=0.04647, over 16900.00 frames. ], tot_loss[loss=0.1829, simple_loss=0.2736, pruned_loss=0.04608, over 3071594.45 frames. ], batch size: 116, lr: 6.69e-03, grad_scale: 8.0 2023-04-29 08:09:32,959 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=101106.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 08:09:34,811 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=101107.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 08:10:45,034 INFO [optim.py:368] (1/8) 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 08:10:54,062 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.6341, 2.6489, 1.7852, 2.7966, 2.1234, 2.7546, 2.0330, 2.3732], device='cuda:1'), covar=tensor([0.0268, 0.0438, 0.1283, 0.0220, 0.0712, 0.0625, 0.1307, 0.0583], device='cuda:1'), in_proj_covar=tensor([0.0142, 0.0155, 0.0183, 0.0117, 0.0163, 0.0193, 0.0192, 0.0167], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0004, 0.0004, 0.0003], device='cuda:1') 2023-04-29 08:11:05,030 INFO [train.py:904] (1/8) Epoch 10, batch 9800, loss[loss=0.171, simple_loss=0.2729, pruned_loss=0.03456, over 15324.00 frames. ], tot_loss[loss=0.1818, simple_loss=0.2737, pruned_loss=0.04499, over 3088740.94 frames. ], batch size: 191, lr: 6.69e-03, grad_scale: 4.0 2023-04-29 08:11:36,071 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=101168.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 08:12:49,186 INFO [train.py:904] (1/8) Epoch 10, batch 9850, loss[loss=0.1849, simple_loss=0.2831, pruned_loss=0.04338, over 16852.00 frames. ], tot_loss[loss=0.1818, simple_loss=0.2747, pruned_loss=0.04444, over 3097604.09 frames. ], batch size: 102, lr: 6.69e-03, grad_scale: 4.0 2023-04-29 08:12:56,602 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.32 vs. limit=2.0 2023-04-29 08:13:04,787 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.5678, 3.6295, 3.4114, 3.2226, 3.2197, 3.5494, 3.2887, 3.3420], device='cuda:1'), covar=tensor([0.0543, 0.0489, 0.0239, 0.0217, 0.0536, 0.0378, 0.1126, 0.0514], device='cuda:1'), in_proj_covar=tensor([0.0221, 0.0273, 0.0257, 0.0237, 0.0276, 0.0269, 0.0178, 0.0301], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-29 08:13:11,741 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.6813, 4.2550, 2.9585, 2.3289, 2.7355, 2.2921, 4.4244, 3.5640], device='cuda:1'), covar=tensor([0.2657, 0.0549, 0.1519, 0.2333, 0.2546, 0.1905, 0.0385, 0.0913], device='cuda:1'), in_proj_covar=tensor([0.0293, 0.0246, 0.0274, 0.0266, 0.0256, 0.0212, 0.0254, 0.0277], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-29 08:13:54,841 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([1.6819, 2.6728, 2.2620, 3.8828, 2.6592, 3.8072, 1.4575, 2.7374], device='cuda:1'), covar=tensor([0.1212, 0.0579, 0.1117, 0.0113, 0.0109, 0.0404, 0.1377, 0.0738], device='cuda:1'), in_proj_covar=tensor([0.0151, 0.0151, 0.0175, 0.0128, 0.0179, 0.0197, 0.0175, 0.0174], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0003, 0.0004], device='cuda:1') 2023-04-29 08:14:17,926 INFO [optim.py:368] (1/8) 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] (1/8) Epoch 10, batch 9900, loss[loss=0.1781, simple_loss=0.2763, pruned_loss=0.03994, over 17157.00 frames. ], tot_loss[loss=0.182, simple_loss=0.275, pruned_loss=0.04454, over 3073556.86 frames. ], batch size: 48, lr: 6.69e-03, grad_scale: 4.0 2023-04-29 08:15:06,967 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.4021, 2.9581, 3.1551, 1.8851, 2.6972, 2.2238, 2.9771, 3.0419], device='cuda:1'), covar=tensor([0.0283, 0.0736, 0.0465, 0.1744, 0.0792, 0.0875, 0.0756, 0.0900], device='cuda:1'), in_proj_covar=tensor([0.0138, 0.0133, 0.0155, 0.0141, 0.0134, 0.0124, 0.0134, 0.0143], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:1') 2023-04-29 08:16:40,042 INFO [train.py:904] (1/8) Epoch 10, batch 9950, loss[loss=0.2023, simple_loss=0.3068, pruned_loss=0.04888, over 16941.00 frames. ], tot_loss[loss=0.1838, simple_loss=0.2775, pruned_loss=0.04507, over 3086333.21 frames. ], batch size: 109, lr: 6.68e-03, grad_scale: 4.0 2023-04-29 08:18:13,665 INFO [optim.py:368] (1/8) 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 08:18:42,216 INFO [train.py:904] (1/8) Epoch 10, batch 10000, loss[loss=0.1701, simple_loss=0.2691, pruned_loss=0.03557, over 16916.00 frames. ], tot_loss[loss=0.1828, simple_loss=0.2759, pruned_loss=0.04484, over 3086378.06 frames. ], batch size: 96, lr: 6.68e-03, grad_scale: 8.0 2023-04-29 08:18:56,329 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([5.1638, 4.9315, 5.1853, 5.3623, 5.5532, 4.9549, 5.5290, 5.4827], device='cuda:1'), covar=tensor([0.1288, 0.0881, 0.1351, 0.0545, 0.0380, 0.0576, 0.0370, 0.0479], device='cuda:1'), in_proj_covar=tensor([0.0467, 0.0586, 0.0703, 0.0600, 0.0455, 0.0453, 0.0469, 0.0534], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:1') 2023-04-29 08:19:15,608 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.7130, 2.6981, 1.7807, 2.8116, 2.2265, 2.8600, 2.0575, 2.4416], device='cuda:1'), covar=tensor([0.0223, 0.0289, 0.1239, 0.0222, 0.0661, 0.0419, 0.1147, 0.0536], device='cuda:1'), in_proj_covar=tensor([0.0141, 0.0153, 0.0181, 0.0116, 0.0161, 0.0190, 0.0190, 0.0164], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0004, 0.0004, 0.0003], device='cuda:1') 2023-04-29 08:19:54,021 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=101386.0, num_to_drop=1, layers_to_drop={0} 2023-04-29 08:20:23,597 INFO [train.py:904] (1/8) Epoch 10, batch 10050, loss[loss=0.198, simple_loss=0.2945, pruned_loss=0.05075, over 16264.00 frames. ], tot_loss[loss=0.183, simple_loss=0.2761, pruned_loss=0.04496, over 3069904.67 frames. ], batch size: 165, lr: 6.68e-03, grad_scale: 8.0 2023-04-29 08:20:32,245 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=2.67 vs. limit=5.0 2023-04-29 08:20:45,877 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.1590, 4.1723, 4.6367, 4.6527, 4.6094, 4.3474, 4.3214, 4.1398], device='cuda:1'), covar=tensor([0.0268, 0.0441, 0.0333, 0.0296, 0.0364, 0.0298, 0.0708, 0.0383], device='cuda:1'), in_proj_covar=tensor([0.0295, 0.0303, 0.0306, 0.0292, 0.0351, 0.0322, 0.0410, 0.0264], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0002], device='cuda:1') 2023-04-29 08:21:00,426 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.1549, 4.1717, 4.5705, 4.5719, 4.5507, 4.2895, 4.2305, 4.1397], device='cuda:1'), covar=tensor([0.0270, 0.0506, 0.0337, 0.0302, 0.0360, 0.0313, 0.0770, 0.0388], device='cuda:1'), in_proj_covar=tensor([0.0294, 0.0302, 0.0305, 0.0291, 0.0349, 0.0321, 0.0408, 0.0263], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0002], device='cuda:1') 2023-04-29 08:21:25,328 INFO [zipformer.py:625] (1/8) 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] (1/8) 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] (1/8) Epoch 10, batch 10100, loss[loss=0.1739, simple_loss=0.2563, pruned_loss=0.04575, over 12730.00 frames. ], tot_loss[loss=0.1831, simple_loss=0.2761, pruned_loss=0.04511, over 3073799.43 frames. ], batch size: 248, lr: 6.68e-03, grad_scale: 8.0 2023-04-29 08:22:16,008 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=101463.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 08:22:23,266 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.49 vs. limit=2.0 2023-04-29 08:22:52,212 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([5.1065, 5.1317, 5.5637, 5.5577, 5.5316, 5.2610, 5.1648, 4.8759], device='cuda:1'), covar=tensor([0.0218, 0.0450, 0.0317, 0.0316, 0.0337, 0.0286, 0.0768, 0.0349], device='cuda:1'), in_proj_covar=tensor([0.0290, 0.0299, 0.0302, 0.0288, 0.0344, 0.0319, 0.0404, 0.0261], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0002], device='cuda:1') 2023-04-29 08:23:38,425 INFO [train.py:904] (1/8) Epoch 11, batch 0, loss[loss=0.2933, simple_loss=0.3518, pruned_loss=0.1174, over 15364.00 frames. ], tot_loss[loss=0.2933, simple_loss=0.3518, pruned_loss=0.1174, over 15364.00 frames. ], batch size: 190, lr: 6.37e-03, grad_scale: 8.0 2023-04-29 08:23:38,425 INFO [train.py:929] (1/8) Computing validation loss 2023-04-29 08:23:45,826 INFO [train.py:938] (1/8) Epoch 11, validation: loss=0.1557, simple_loss=0.2595, pruned_loss=0.02591, over 944034.00 frames. 2023-04-29 08:23:45,826 INFO [train.py:939] (1/8) Maximum memory allocated so far is 17676MB 2023-04-29 08:24:43,126 INFO [optim.py:368] (1/8) 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 08:24:53,219 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.1979, 3.9963, 4.1863, 4.3830, 4.4857, 4.0469, 4.2591, 4.4418], device='cuda:1'), covar=tensor([0.1731, 0.1379, 0.1704, 0.0823, 0.0658, 0.1128, 0.1561, 0.1163], device='cuda:1'), in_proj_covar=tensor([0.0481, 0.0599, 0.0724, 0.0617, 0.0467, 0.0465, 0.0483, 0.0547], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-29 08:24:55,126 INFO [train.py:904] (1/8) Epoch 11, batch 50, loss[loss=0.2128, simple_loss=0.2973, pruned_loss=0.06418, over 16405.00 frames. ], tot_loss[loss=0.2079, simple_loss=0.2882, pruned_loss=0.06375, over 754196.12 frames. ], batch size: 68, lr: 6.37e-03, grad_scale: 1.0 2023-04-29 08:25:42,886 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=5.05 vs. limit=5.0 2023-04-29 08:26:05,598 INFO [train.py:904] (1/8) Epoch 11, batch 100, loss[loss=0.1973, simple_loss=0.2894, pruned_loss=0.05257, over 16759.00 frames. ], tot_loss[loss=0.2048, simple_loss=0.2854, pruned_loss=0.06211, over 1319566.82 frames. ], batch size: 57, lr: 6.37e-03, grad_scale: 1.0 2023-04-29 08:27:03,337 INFO [optim.py:368] (1/8) 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] (1/8) Epoch 11, batch 150, loss[loss=0.2026, simple_loss=0.2762, pruned_loss=0.06453, over 16517.00 frames. ], tot_loss[loss=0.2019, simple_loss=0.2831, pruned_loss=0.06033, over 1759792.52 frames. ], batch size: 146, lr: 6.37e-03, grad_scale: 1.0 2023-04-29 08:27:58,574 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=101684.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 08:28:02,255 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.8955, 4.1881, 3.9810, 4.0525, 3.6854, 3.7210, 3.8404, 4.1659], device='cuda:1'), covar=tensor([0.0974, 0.0917, 0.0992, 0.0670, 0.0727, 0.1706, 0.0801, 0.1009], device='cuda:1'), in_proj_covar=tensor([0.0514, 0.0639, 0.0532, 0.0445, 0.0405, 0.0425, 0.0543, 0.0489], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:1') 2023-04-29 08:28:03,548 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.0640, 3.9951, 3.9363, 3.0337, 3.9741, 1.4937, 3.6713, 3.5208], device='cuda:1'), covar=tensor([0.0178, 0.0139, 0.0208, 0.0499, 0.0137, 0.3165, 0.0178, 0.0362], device='cuda:1'), in_proj_covar=tensor([0.0125, 0.0111, 0.0156, 0.0143, 0.0128, 0.0177, 0.0142, 0.0143], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-29 08:28:05,243 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.49 vs. limit=2.0 2023-04-29 08:28:23,285 INFO [train.py:904] (1/8) Epoch 11, batch 200, loss[loss=0.181, simple_loss=0.2736, pruned_loss=0.04421, over 17112.00 frames. ], tot_loss[loss=0.2005, simple_loss=0.2818, pruned_loss=0.05958, over 2104132.88 frames. ], batch size: 49, lr: 6.37e-03, grad_scale: 1.0 2023-04-29 08:28:58,147 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.7406, 4.1541, 4.5120, 2.3148, 4.7179, 4.8035, 3.4174, 3.6374], device='cuda:1'), covar=tensor([0.0760, 0.0193, 0.0182, 0.1025, 0.0047, 0.0089, 0.0313, 0.0325], device='cuda:1'), in_proj_covar=tensor([0.0140, 0.0097, 0.0083, 0.0137, 0.0066, 0.0096, 0.0117, 0.0125], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0003, 0.0002, 0.0004, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:1') 2023-04-29 08:29:02,452 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.2197, 1.8862, 2.6342, 3.0110, 2.9407, 3.2673, 2.2770, 3.2521], device='cuda:1'), covar=tensor([0.0146, 0.0381, 0.0232, 0.0203, 0.0208, 0.0184, 0.0324, 0.0134], device='cuda:1'), in_proj_covar=tensor([0.0156, 0.0169, 0.0153, 0.0154, 0.0165, 0.0118, 0.0170, 0.0107], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001], device='cuda:1') 2023-04-29 08:29:20,174 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([5.1356, 5.1122, 4.9283, 4.4290, 4.9083, 2.0478, 4.7041, 4.8870], device='cuda:1'), covar=tensor([0.0072, 0.0065, 0.0133, 0.0297, 0.0082, 0.2160, 0.0112, 0.0164], device='cuda:1'), in_proj_covar=tensor([0.0125, 0.0111, 0.0156, 0.0144, 0.0128, 0.0176, 0.0142, 0.0144], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-29 08:29:21,756 INFO [optim.py:368] (1/8) Clipping_scale=2.0, grad-norm quartiles 1.605e+02 2.429e+02 2.923e+02 3.406e+02 5.474e+02, threshold=5.846e+02, percent-clipped=1.0 2023-04-29 08:29:23,320 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 11, batch 250, loss[loss=0.2202, simple_loss=0.2821, pruned_loss=0.07912, over 16828.00 frames. ], tot_loss[loss=0.1989, simple_loss=0.2793, pruned_loss=0.05928, over 2379340.86 frames. ], batch size: 102, lr: 6.37e-03, grad_scale: 1.0 2023-04-29 08:29:37,021 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.81 vs. limit=2.0 2023-04-29 08:29:39,239 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.5849, 3.4734, 3.8583, 2.7514, 3.5048, 3.8920, 3.6647, 2.2738], device='cuda:1'), covar=tensor([0.0354, 0.0197, 0.0038, 0.0262, 0.0081, 0.0074, 0.0064, 0.0354], device='cuda:1'), in_proj_covar=tensor([0.0127, 0.0069, 0.0070, 0.0126, 0.0078, 0.0086, 0.0076, 0.0120], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:1') 2023-04-29 08:29:46,226 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 11, batch 300, loss[loss=0.2153, simple_loss=0.2911, pruned_loss=0.06972, over 16864.00 frames. ], tot_loss[loss=0.1959, simple_loss=0.2762, pruned_loss=0.05776, over 2595672.77 frames. ], batch size: 102, lr: 6.37e-03, grad_scale: 1.0 2023-04-29 08:30:46,097 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([5.4584, 5.8358, 5.2723, 5.7000, 5.2609, 4.9404, 5.4061, 5.9356], device='cuda:1'), covar=tensor([0.1766, 0.1381, 0.2081, 0.1013, 0.1233, 0.1155, 0.1811, 0.1486], device='cuda:1'), in_proj_covar=tensor([0.0524, 0.0655, 0.0543, 0.0457, 0.0415, 0.0432, 0.0553, 0.0500], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:1') 2023-04-29 08:30:51,058 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=101811.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 08:30:59,323 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.8007, 4.7580, 4.6621, 4.1738, 4.6633, 1.8426, 4.4317, 4.5324], device='cuda:1'), covar=tensor([0.0086, 0.0073, 0.0141, 0.0272, 0.0084, 0.2207, 0.0121, 0.0173], device='cuda:1'), in_proj_covar=tensor([0.0125, 0.0111, 0.0157, 0.0145, 0.0128, 0.0176, 0.0143, 0.0144], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-29 08:31:35,544 INFO [optim.py:368] (1/8) 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] (1/8) Epoch 11, batch 350, loss[loss=0.191, simple_loss=0.2645, pruned_loss=0.05878, over 16738.00 frames. ], tot_loss[loss=0.1932, simple_loss=0.2735, pruned_loss=0.05648, over 2759904.32 frames. ], batch size: 134, lr: 6.36e-03, grad_scale: 1.0 2023-04-29 08:32:42,835 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.6133, 4.5818, 4.9966, 5.0367, 5.0949, 4.7027, 4.6803, 4.4566], device='cuda:1'), covar=tensor([0.0329, 0.0605, 0.0467, 0.0438, 0.0420, 0.0382, 0.0932, 0.0509], device='cuda:1'), in_proj_covar=tensor([0.0316, 0.0328, 0.0332, 0.0311, 0.0373, 0.0346, 0.0444, 0.0281], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003], device='cuda:1') 2023-04-29 08:32:56,331 INFO [train.py:904] (1/8) Epoch 11, batch 400, loss[loss=0.1536, simple_loss=0.2369, pruned_loss=0.03516, over 16787.00 frames. ], tot_loss[loss=0.191, simple_loss=0.2713, pruned_loss=0.05535, over 2887103.84 frames. ], batch size: 39, lr: 6.36e-03, grad_scale: 2.0 2023-04-29 08:32:56,807 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.2999, 2.2898, 1.6861, 2.1449, 2.6154, 2.4791, 2.7589, 2.7403], device='cuda:1'), covar=tensor([0.0149, 0.0231, 0.0365, 0.0294, 0.0158, 0.0235, 0.0159, 0.0187], device='cuda:1'), in_proj_covar=tensor([0.0140, 0.0201, 0.0199, 0.0197, 0.0200, 0.0202, 0.0201, 0.0187], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-29 08:33:22,046 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=101920.0, num_to_drop=1, layers_to_drop={0} 2023-04-29 08:33:54,606 INFO [optim.py:368] (1/8) Clipping_scale=2.0, grad-norm quartiles 1.501e+02 2.298e+02 2.716e+02 3.213e+02 6.833e+02, threshold=5.433e+02, percent-clipped=1.0 2023-04-29 08:34:06,160 INFO [train.py:904] (1/8) Epoch 11, batch 450, loss[loss=0.1985, simple_loss=0.27, pruned_loss=0.06354, over 16914.00 frames. ], tot_loss[loss=0.1895, simple_loss=0.2696, pruned_loss=0.05465, over 2977738.90 frames. ], batch size: 116, lr: 6.36e-03, grad_scale: 2.0 2023-04-29 08:34:46,987 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.6452, 3.7718, 4.1047, 2.0419, 4.2890, 4.3586, 3.1032, 3.2647], device='cuda:1'), covar=tensor([0.0742, 0.0211, 0.0236, 0.1147, 0.0063, 0.0130, 0.0400, 0.0387], device='cuda:1'), in_proj_covar=tensor([0.0143, 0.0100, 0.0086, 0.0140, 0.0069, 0.0101, 0.0120, 0.0128], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0003, 0.0002, 0.0004, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:1') 2023-04-29 08:34:47,003 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=101981.0, num_to_drop=1, layers_to_drop={2} 2023-04-29 08:34:54,589 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 11, batch 500, loss[loss=0.1732, simple_loss=0.2717, pruned_loss=0.03738, over 17019.00 frames. ], tot_loss[loss=0.1868, simple_loss=0.2678, pruned_loss=0.05291, over 3063119.53 frames. ], batch size: 50, lr: 6.36e-03, grad_scale: 2.0 2023-04-29 08:35:36,882 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.6494, 3.9877, 4.2104, 3.0354, 3.7440, 4.1424, 3.8968, 2.5039], device='cuda:1'), covar=tensor([0.0363, 0.0076, 0.0032, 0.0261, 0.0069, 0.0067, 0.0054, 0.0333], device='cuda:1'), in_proj_covar=tensor([0.0126, 0.0068, 0.0069, 0.0125, 0.0077, 0.0086, 0.0076, 0.0119], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:1') 2023-04-29 08:36:13,471 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=102040.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 08:36:19,093 INFO [optim.py:368] (1/8) 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 08:36:24,820 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=102048.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 08:36:29,129 INFO [train.py:904] (1/8) Epoch 11, batch 550, loss[loss=0.2025, simple_loss=0.2865, pruned_loss=0.05926, over 17056.00 frames. ], tot_loss[loss=0.1852, simple_loss=0.2664, pruned_loss=0.05197, over 3128159.21 frames. ], batch size: 53, lr: 6.36e-03, grad_scale: 2.0 2023-04-29 08:37:40,169 INFO [train.py:904] (1/8) Epoch 11, batch 600, loss[loss=0.18, simple_loss=0.256, pruned_loss=0.05203, over 15422.00 frames. ], tot_loss[loss=0.1849, simple_loss=0.2657, pruned_loss=0.05206, over 3178887.45 frames. ], batch size: 190, lr: 6.36e-03, grad_scale: 2.0 2023-04-29 08:37:46,055 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.8449, 1.7216, 2.2812, 2.6864, 2.7432, 2.4513, 1.6803, 2.7333], device='cuda:1'), covar=tensor([0.0111, 0.0304, 0.0197, 0.0176, 0.0150, 0.0172, 0.0320, 0.0107], device='cuda:1'), in_proj_covar=tensor([0.0159, 0.0170, 0.0154, 0.0157, 0.0168, 0.0121, 0.0171, 0.0111], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001], device='cuda:1') 2023-04-29 08:38:38,907 INFO [optim.py:368] (1/8) 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] (1/8) Epoch 11, batch 650, loss[loss=0.176, simple_loss=0.2677, pruned_loss=0.04216, over 17133.00 frames. ], tot_loss[loss=0.1832, simple_loss=0.264, pruned_loss=0.05118, over 3202256.56 frames. ], batch size: 47, lr: 6.35e-03, grad_scale: 2.0 2023-04-29 08:39:58,899 INFO [train.py:904] (1/8) Epoch 11, batch 700, loss[loss=0.179, simple_loss=0.2644, pruned_loss=0.04687, over 17201.00 frames. ], tot_loss[loss=0.1829, simple_loss=0.2637, pruned_loss=0.05101, over 3236854.30 frames. ], batch size: 44, lr: 6.35e-03, grad_scale: 2.0 2023-04-29 08:40:57,196 INFO [optim.py:368] (1/8) 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] (1/8) Epoch 11, batch 750, loss[loss=0.201, simple_loss=0.2753, pruned_loss=0.06337, over 16916.00 frames. ], tot_loss[loss=0.184, simple_loss=0.2647, pruned_loss=0.05166, over 3254504.07 frames. ], batch size: 116, lr: 6.35e-03, grad_scale: 2.0 2023-04-29 08:41:42,362 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=102276.0, num_to_drop=1, layers_to_drop={3} 2023-04-29 08:42:01,087 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.20 vs. limit=2.0 2023-04-29 08:42:10,610 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.89 vs. limit=2.0 2023-04-29 08:42:18,047 INFO [train.py:904] (1/8) Epoch 11, batch 800, loss[loss=0.1497, simple_loss=0.2301, pruned_loss=0.03462, over 16761.00 frames. ], tot_loss[loss=0.184, simple_loss=0.2647, pruned_loss=0.05166, over 3261940.88 frames. ], batch size: 39, lr: 6.35e-03, grad_scale: 4.0 2023-04-29 08:43:11,908 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=102340.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 08:43:14,272 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([5.3484, 5.2327, 5.1747, 4.7079, 4.6896, 5.1487, 5.2002, 4.8030], device='cuda:1'), covar=tensor([0.0533, 0.0423, 0.0258, 0.0292, 0.1128, 0.0417, 0.0251, 0.0731], device='cuda:1'), in_proj_covar=tensor([0.0254, 0.0315, 0.0294, 0.0275, 0.0318, 0.0313, 0.0202, 0.0345], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002], device='cuda:1') 2023-04-29 08:43:15,310 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=102343.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 08:43:16,113 INFO [optim.py:368] (1/8) 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 08:43:27,549 INFO [train.py:904] (1/8) Epoch 11, batch 850, loss[loss=0.1717, simple_loss=0.2592, pruned_loss=0.04212, over 16553.00 frames. ], tot_loss[loss=0.1832, simple_loss=0.2644, pruned_loss=0.05102, over 3271534.88 frames. ], batch size: 68, lr: 6.35e-03, grad_scale: 4.0 2023-04-29 08:44:16,958 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 11, batch 900, loss[loss=0.1872, simple_loss=0.2725, pruned_loss=0.05096, over 17226.00 frames. ], tot_loss[loss=0.1822, simple_loss=0.2635, pruned_loss=0.05046, over 3288573.62 frames. ], batch size: 45, lr: 6.35e-03, grad_scale: 4.0 2023-04-29 08:44:39,243 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.80 vs. limit=2.0 2023-04-29 08:45:35,217 INFO [optim.py:368] (1/8) 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] (1/8) Epoch 11, batch 950, loss[loss=0.1776, simple_loss=0.2729, pruned_loss=0.04114, over 17023.00 frames. ], tot_loss[loss=0.1819, simple_loss=0.2634, pruned_loss=0.05022, over 3305452.45 frames. ], batch size: 50, lr: 6.34e-03, grad_scale: 4.0 2023-04-29 08:46:54,282 INFO [train.py:904] (1/8) Epoch 11, batch 1000, loss[loss=0.1727, simple_loss=0.2574, pruned_loss=0.04402, over 17053.00 frames. ], tot_loss[loss=0.1809, simple_loss=0.2621, pruned_loss=0.04985, over 3317098.52 frames. ], batch size: 50, lr: 6.34e-03, grad_scale: 4.0 2023-04-29 08:47:52,028 INFO [optim.py:368] (1/8) 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 08:48:02,261 INFO [train.py:904] (1/8) Epoch 11, batch 1050, loss[loss=0.1885, simple_loss=0.2762, pruned_loss=0.0504, over 15575.00 frames. ], tot_loss[loss=0.1804, simple_loss=0.2617, pruned_loss=0.04958, over 3317786.99 frames. ], batch size: 191, lr: 6.34e-03, grad_scale: 4.0 2023-04-29 08:48:07,496 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.19 vs. limit=2.0 2023-04-29 08:48:35,798 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=102576.0, num_to_drop=1, layers_to_drop={0} 2023-04-29 08:49:12,617 INFO [train.py:904] (1/8) Epoch 11, batch 1100, loss[loss=0.167, simple_loss=0.2407, pruned_loss=0.04668, over 16839.00 frames. ], tot_loss[loss=0.1803, simple_loss=0.2612, pruned_loss=0.04971, over 3321124.03 frames. ], batch size: 96, lr: 6.34e-03, grad_scale: 4.0 2023-04-29 08:49:43,773 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=102624.0, num_to_drop=1, layers_to_drop={1} 2023-04-29 08:49:52,495 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.97 vs. limit=2.0 2023-04-29 08:50:07,620 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=102643.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 08:50:09,057 INFO [optim.py:368] (1/8) Clipping_scale=2.0, grad-norm quartiles 1.567e+02 2.199e+02 2.657e+02 3.494e+02 5.570e+02, threshold=5.313e+02, percent-clipped=0.0 2023-04-29 08:50:20,437 INFO [train.py:904] (1/8) Epoch 11, batch 1150, loss[loss=0.1758, simple_loss=0.256, pruned_loss=0.04782, over 16379.00 frames. ], tot_loss[loss=0.1795, simple_loss=0.2605, pruned_loss=0.04923, over 3316568.71 frames. ], batch size: 68, lr: 6.34e-03, grad_scale: 4.0 2023-04-29 08:51:13,699 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=4.43 vs. limit=5.0 2023-04-29 08:51:14,266 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=102691.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 08:51:27,908 INFO [train.py:904] (1/8) Epoch 11, batch 1200, loss[loss=0.185, simple_loss=0.2853, pruned_loss=0.04232, over 17129.00 frames. ], tot_loss[loss=0.1791, simple_loss=0.2601, pruned_loss=0.04901, over 3315726.08 frames. ], batch size: 49, lr: 6.34e-03, grad_scale: 8.0 2023-04-29 08:51:30,240 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.8159, 4.1998, 3.1790, 2.2538, 2.9123, 2.5551, 4.5708, 3.8050], device='cuda:1'), covar=tensor([0.2466, 0.0607, 0.1441, 0.2133, 0.2188, 0.1676, 0.0305, 0.0918], device='cuda:1'), in_proj_covar=tensor([0.0305, 0.0257, 0.0283, 0.0275, 0.0278, 0.0222, 0.0268, 0.0296], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-29 08:51:50,153 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.9390, 4.3673, 3.3251, 2.2941, 2.9021, 2.5740, 4.7086, 3.8018], device='cuda:1'), covar=tensor([0.2423, 0.0625, 0.1419, 0.2308, 0.2517, 0.1736, 0.0355, 0.1091], device='cuda:1'), in_proj_covar=tensor([0.0304, 0.0257, 0.0282, 0.0275, 0.0278, 0.0222, 0.0268, 0.0296], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-29 08:52:27,631 INFO [optim.py:368] (1/8) 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 08:52:39,083 INFO [train.py:904] (1/8) Epoch 11, batch 1250, loss[loss=0.1713, simple_loss=0.2491, pruned_loss=0.04676, over 15933.00 frames. ], tot_loss[loss=0.1799, simple_loss=0.2601, pruned_loss=0.04984, over 3302672.33 frames. ], batch size: 35, lr: 6.34e-03, grad_scale: 8.0 2023-04-29 08:52:41,163 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.7968, 3.7093, 3.8555, 3.9923, 4.0746, 3.6411, 3.8858, 4.0665], device='cuda:1'), covar=tensor([0.1387, 0.0898, 0.1231, 0.0570, 0.0501, 0.1753, 0.1435, 0.0611], device='cuda:1'), in_proj_covar=tensor([0.0539, 0.0670, 0.0828, 0.0693, 0.0521, 0.0521, 0.0535, 0.0611], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-29 08:53:49,412 INFO [train.py:904] (1/8) Epoch 11, batch 1300, loss[loss=0.1781, simple_loss=0.2654, pruned_loss=0.04539, over 16659.00 frames. ], tot_loss[loss=0.1802, simple_loss=0.2607, pruned_loss=0.04987, over 3312339.30 frames. ], batch size: 62, lr: 6.33e-03, grad_scale: 8.0 2023-04-29 08:53:54,541 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=4.64 vs. limit=5.0 2023-04-29 08:54:46,485 INFO [optim.py:368] (1/8) 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 08:54:51,094 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.5050, 2.5269, 2.1638, 2.3487, 2.8377, 2.6048, 3.3564, 3.1136], device='cuda:1'), covar=tensor([0.0076, 0.0280, 0.0370, 0.0327, 0.0184, 0.0277, 0.0170, 0.0183], device='cuda:1'), in_proj_covar=tensor([0.0149, 0.0206, 0.0202, 0.0202, 0.0206, 0.0205, 0.0211, 0.0196], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-29 08:54:57,221 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 11, batch 1350, loss[loss=0.1847, simple_loss=0.2603, pruned_loss=0.05454, over 16819.00 frames. ], tot_loss[loss=0.1798, simple_loss=0.2607, pruned_loss=0.04947, over 3319570.28 frames. ], batch size: 102, lr: 6.33e-03, grad_scale: 8.0 2023-04-29 08:55:45,165 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 11, batch 1400, loss[loss=0.1826, simple_loss=0.2609, pruned_loss=0.05221, over 16271.00 frames. ], tot_loss[loss=0.1801, simple_loss=0.2611, pruned_loss=0.04954, over 3321041.51 frames. ], batch size: 164, lr: 6.33e-03, grad_scale: 8.0 2023-04-29 08:56:20,064 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=102912.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 08:56:22,235 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=3.56 vs. limit=5.0 2023-04-29 08:56:39,700 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.18 vs. limit=2.0 2023-04-29 08:57:02,704 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.1815, 1.9431, 2.6707, 3.1783, 2.8927, 3.4971, 2.4363, 3.5718], device='cuda:1'), covar=tensor([0.0134, 0.0357, 0.0224, 0.0170, 0.0206, 0.0113, 0.0305, 0.0088], device='cuda:1'), in_proj_covar=tensor([0.0162, 0.0173, 0.0157, 0.0158, 0.0168, 0.0123, 0.0171, 0.0114], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001], device='cuda:1') 2023-04-29 08:57:05,104 INFO [optim.py:368] (1/8) 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 08:57:10,160 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 11, batch 1450, loss[loss=0.1816, simple_loss=0.2482, pruned_loss=0.0575, over 12379.00 frames. ], tot_loss[loss=0.179, simple_loss=0.2597, pruned_loss=0.04917, over 3313809.99 frames. ], batch size: 246, lr: 6.33e-03, grad_scale: 8.0 2023-04-29 08:58:25,000 INFO [train.py:904] (1/8) Epoch 11, batch 1500, loss[loss=0.1987, simple_loss=0.2682, pruned_loss=0.06464, over 16667.00 frames. ], tot_loss[loss=0.1792, simple_loss=0.2598, pruned_loss=0.0493, over 3311687.89 frames. ], batch size: 134, lr: 6.33e-03, grad_scale: 8.0 2023-04-29 08:58:53,997 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=4.43 vs. limit=5.0 2023-04-29 08:59:24,563 INFO [optim.py:368] (1/8) 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] (1/8) Epoch 11, batch 1550, loss[loss=0.2226, simple_loss=0.2771, pruned_loss=0.08406, over 16785.00 frames. ], tot_loss[loss=0.1802, simple_loss=0.2603, pruned_loss=0.05005, over 3317505.30 frames. ], batch size: 83, lr: 6.33e-03, grad_scale: 8.0 2023-04-29 09:00:39,918 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=103098.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 09:00:45,000 INFO [train.py:904] (1/8) Epoch 11, batch 1600, loss[loss=0.1708, simple_loss=0.2499, pruned_loss=0.0459, over 16995.00 frames. ], tot_loss[loss=0.1827, simple_loss=0.2627, pruned_loss=0.05131, over 3321919.95 frames. ], batch size: 41, lr: 6.32e-03, grad_scale: 8.0 2023-04-29 09:01:20,256 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.58 vs. limit=2.0 2023-04-29 09:01:26,034 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=103132.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 09:01:42,626 INFO [optim.py:368] (1/8) 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 09:01:53,521 INFO [train.py:904] (1/8) Epoch 11, batch 1650, loss[loss=0.202, simple_loss=0.2716, pruned_loss=0.06617, over 16833.00 frames. ], tot_loss[loss=0.1834, simple_loss=0.2638, pruned_loss=0.0515, over 3321959.97 frames. ], batch size: 96, lr: 6.32e-03, grad_scale: 8.0 2023-04-29 09:01:56,411 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.5964, 3.8940, 4.0148, 2.1781, 3.3712, 2.5523, 4.0799, 4.0453], device='cuda:1'), covar=tensor([0.0253, 0.0734, 0.0477, 0.1849, 0.0773, 0.0985, 0.0548, 0.1032], device='cuda:1'), in_proj_covar=tensor([0.0142, 0.0143, 0.0156, 0.0142, 0.0136, 0.0124, 0.0136, 0.0154], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:1') 2023-04-29 09:02:03,298 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=103159.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 09:02:50,104 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=103193.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 09:03:02,606 INFO [train.py:904] (1/8) Epoch 11, batch 1700, loss[loss=0.2305, simple_loss=0.3099, pruned_loss=0.07553, over 15562.00 frames. ], tot_loss[loss=0.1841, simple_loss=0.265, pruned_loss=0.05164, over 3321352.73 frames. ], batch size: 190, lr: 6.32e-03, grad_scale: 8.0 2023-04-29 09:03:10,657 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=103207.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 09:03:14,616 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.90 vs. limit=2.0 2023-04-29 09:03:50,237 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.4676, 3.5658, 3.2472, 3.0258, 3.1181, 3.4470, 3.2389, 3.2408], device='cuda:1'), covar=tensor([0.0561, 0.0455, 0.0274, 0.0261, 0.0547, 0.0334, 0.1314, 0.0489], device='cuda:1'), in_proj_covar=tensor([0.0259, 0.0323, 0.0302, 0.0281, 0.0324, 0.0320, 0.0206, 0.0355], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-04-29 09:04:01,646 INFO [zipformer.py:625] (1/8) 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] (1/8) 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 09:04:13,824 INFO [train.py:904] (1/8) Epoch 11, batch 1750, loss[loss=0.2095, simple_loss=0.2981, pruned_loss=0.06045, over 16833.00 frames. ], tot_loss[loss=0.1858, simple_loss=0.2672, pruned_loss=0.05221, over 3313922.19 frames. ], batch size: 57, lr: 6.32e-03, grad_scale: 4.0 2023-04-29 09:04:14,319 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.3621, 3.3927, 1.9359, 3.5513, 2.4833, 3.5148, 2.0146, 2.6472], device='cuda:1'), covar=tensor([0.0227, 0.0352, 0.1471, 0.0249, 0.0771, 0.0606, 0.1355, 0.0646], device='cuda:1'), in_proj_covar=tensor([0.0151, 0.0167, 0.0188, 0.0134, 0.0169, 0.0210, 0.0195, 0.0171], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-04-29 09:05:03,181 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.9487, 4.3993, 3.1855, 2.3283, 2.8575, 2.3780, 4.7326, 3.8247], device='cuda:1'), covar=tensor([0.2307, 0.0530, 0.1470, 0.2208, 0.2415, 0.1799, 0.0279, 0.0995], device='cuda:1'), in_proj_covar=tensor([0.0300, 0.0253, 0.0280, 0.0272, 0.0276, 0.0219, 0.0264, 0.0294], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-29 09:05:22,485 INFO [train.py:904] (1/8) Epoch 11, batch 1800, loss[loss=0.1537, simple_loss=0.2343, pruned_loss=0.03657, over 16945.00 frames. ], tot_loss[loss=0.1854, simple_loss=0.2677, pruned_loss=0.05157, over 3319833.64 frames. ], batch size: 41, lr: 6.32e-03, grad_scale: 4.0 2023-04-29 09:06:10,984 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.4690, 4.3200, 4.5041, 4.7252, 4.8373, 4.4014, 4.6315, 4.7618], device='cuda:1'), covar=tensor([0.1296, 0.0928, 0.1243, 0.0532, 0.0511, 0.0932, 0.1716, 0.0600], device='cuda:1'), in_proj_covar=tensor([0.0545, 0.0680, 0.0837, 0.0699, 0.0528, 0.0527, 0.0538, 0.0618], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-29 09:06:21,365 INFO [optim.py:368] (1/8) 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 09:06:26,667 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.88 vs. limit=2.0 2023-04-29 09:06:31,957 INFO [train.py:904] (1/8) Epoch 11, batch 1850, loss[loss=0.2209, simple_loss=0.2819, pruned_loss=0.07996, over 16835.00 frames. ], tot_loss[loss=0.1855, simple_loss=0.268, pruned_loss=0.0515, over 3326822.29 frames. ], batch size: 124, lr: 6.32e-03, grad_scale: 4.0 2023-04-29 09:06:50,771 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=103366.0, num_to_drop=1, layers_to_drop={0} 2023-04-29 09:07:39,137 INFO [train.py:904] (1/8) Epoch 11, batch 1900, loss[loss=0.178, simple_loss=0.2517, pruned_loss=0.05212, over 16853.00 frames. ], tot_loss[loss=0.1851, simple_loss=0.2678, pruned_loss=0.05118, over 3335841.93 frames. ], batch size: 116, lr: 6.32e-03, grad_scale: 4.0 2023-04-29 09:08:16,158 INFO [zipformer.py:625] (1/8) 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] (1/8) 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 09:08:51,895 INFO [train.py:904] (1/8) Epoch 11, batch 1950, loss[loss=0.1772, simple_loss=0.2669, pruned_loss=0.04373, over 17147.00 frames. ], tot_loss[loss=0.1849, simple_loss=0.2679, pruned_loss=0.05093, over 3327563.38 frames. ], batch size: 46, lr: 6.31e-03, grad_scale: 4.0 2023-04-29 09:08:55,093 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=103454.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 09:09:41,327 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 11, batch 2000, loss[loss=0.2006, simple_loss=0.2812, pruned_loss=0.06003, over 16319.00 frames. ], tot_loss[loss=0.1851, simple_loss=0.2679, pruned_loss=0.05114, over 3322222.31 frames. ], batch size: 165, lr: 6.31e-03, grad_scale: 8.0 2023-04-29 09:10:07,040 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=103507.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 09:10:58,891 INFO [zipformer.py:625] (1/8) 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] (1/8) 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 09:11:08,486 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.25 vs. limit=2.0 2023-04-29 09:11:11,322 INFO [train.py:904] (1/8) Epoch 11, batch 2050, loss[loss=0.2034, simple_loss=0.2741, pruned_loss=0.0664, over 16749.00 frames. ], tot_loss[loss=0.1849, simple_loss=0.2677, pruned_loss=0.05099, over 3321295.86 frames. ], batch size: 124, lr: 6.31e-03, grad_scale: 4.0 2023-04-29 09:11:15,122 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.16 vs. limit=2.0 2023-04-29 09:11:16,434 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=103555.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 09:11:44,284 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.15 vs. limit=2.0 2023-04-29 09:12:05,788 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=103591.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 09:12:21,596 INFO [train.py:904] (1/8) Epoch 11, batch 2100, loss[loss=0.2094, simple_loss=0.2878, pruned_loss=0.06546, over 16247.00 frames. ], tot_loss[loss=0.1867, simple_loss=0.269, pruned_loss=0.0522, over 3324340.36 frames. ], batch size: 164, lr: 6.31e-03, grad_scale: 4.0 2023-04-29 09:13:08,905 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([5.5707, 5.9652, 5.6563, 5.7883, 5.2905, 5.1232, 5.4060, 6.0723], device='cuda:1'), covar=tensor([0.1075, 0.0833, 0.0975, 0.0716, 0.0838, 0.0651, 0.1047, 0.0778], device='cuda:1'), in_proj_covar=tensor([0.0544, 0.0686, 0.0568, 0.0480, 0.0428, 0.0442, 0.0574, 0.0521], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-29 09:13:15,479 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.9513, 5.4397, 5.6260, 5.2780, 5.3655, 6.0075, 5.5708, 5.2897], device='cuda:1'), covar=tensor([0.0937, 0.1892, 0.2225, 0.1906, 0.2998, 0.0985, 0.1354, 0.2201], device='cuda:1'), in_proj_covar=tensor([0.0359, 0.0508, 0.0549, 0.0441, 0.0583, 0.0576, 0.0435, 0.0590], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-04-29 09:13:22,846 INFO [optim.py:368] (1/8) 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] (1/8) Epoch 11, batch 2150, loss[loss=0.2132, simple_loss=0.2833, pruned_loss=0.07153, over 16683.00 frames. ], tot_loss[loss=0.1875, simple_loss=0.27, pruned_loss=0.05247, over 3320030.49 frames. ], batch size: 134, lr: 6.31e-03, grad_scale: 4.0 2023-04-29 09:13:53,033 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.54 vs. limit=2.0 2023-04-29 09:14:42,107 INFO [train.py:904] (1/8) Epoch 11, batch 2200, loss[loss=0.2079, simple_loss=0.2818, pruned_loss=0.06699, over 16258.00 frames. ], tot_loss[loss=0.1876, simple_loss=0.2707, pruned_loss=0.05223, over 3325260.20 frames. ], batch size: 165, lr: 6.31e-03, grad_scale: 4.0 2023-04-29 09:15:10,360 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=103722.0, num_to_drop=1, layers_to_drop={1} 2023-04-29 09:15:35,197 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=103740.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 09:15:44,092 INFO [optim.py:368] (1/8) 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] (1/8) Epoch 11, batch 2250, loss[loss=0.2938, simple_loss=0.3523, pruned_loss=0.1177, over 12124.00 frames. ], tot_loss[loss=0.1889, simple_loss=0.2717, pruned_loss=0.05308, over 3326291.33 frames. ], batch size: 246, lr: 6.31e-03, grad_scale: 4.0 2023-04-29 09:15:54,340 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=103754.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 09:16:12,408 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=103765.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 09:16:44,194 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=103788.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 09:17:01,629 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 11, batch 2300, loss[loss=0.183, simple_loss=0.2646, pruned_loss=0.05065, over 16476.00 frames. ], tot_loss[loss=0.1894, simple_loss=0.2716, pruned_loss=0.05361, over 3333406.44 frames. ], batch size: 75, lr: 6.30e-03, grad_scale: 4.0 2023-04-29 09:17:02,656 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=103802.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 09:17:07,764 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([5.2439, 5.5320, 5.2775, 5.3444, 5.0229, 4.8176, 5.0093, 5.6764], device='cuda:1'), covar=tensor([0.1076, 0.0947, 0.1028, 0.0691, 0.0707, 0.0774, 0.1006, 0.0789], device='cuda:1'), in_proj_covar=tensor([0.0544, 0.0687, 0.0565, 0.0479, 0.0429, 0.0442, 0.0572, 0.0521], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-29 09:17:35,762 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=103826.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 09:17:38,890 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.36 vs. limit=2.0 2023-04-29 09:17:45,617 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.91 vs. limit=2.0 2023-04-29 09:17:48,571 INFO [zipformer.py:625] (1/8) 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] (1/8) 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 09:18:11,651 INFO [train.py:904] (1/8) Epoch 11, batch 2350, loss[loss=0.1833, simple_loss=0.2799, pruned_loss=0.04336, over 17036.00 frames. ], tot_loss[loss=0.1903, simple_loss=0.2723, pruned_loss=0.05409, over 3332764.39 frames. ], batch size: 50, lr: 6.30e-03, grad_scale: 4.0 2023-04-29 09:18:45,235 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=103876.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 09:19:19,656 INFO [train.py:904] (1/8) Epoch 11, batch 2400, loss[loss=0.202, simple_loss=0.2807, pruned_loss=0.06168, over 16494.00 frames. ], tot_loss[loss=0.1901, simple_loss=0.2728, pruned_loss=0.05371, over 3335201.48 frames. ], batch size: 75, lr: 6.30e-03, grad_scale: 8.0 2023-04-29 09:20:09,907 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=103937.0, num_to_drop=1, layers_to_drop={2} 2023-04-29 09:20:20,789 INFO [optim.py:368] (1/8) 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 09:20:28,627 INFO [train.py:904] (1/8) Epoch 11, batch 2450, loss[loss=0.1778, simple_loss=0.2546, pruned_loss=0.05049, over 16861.00 frames. ], tot_loss[loss=0.1911, simple_loss=0.2741, pruned_loss=0.05399, over 3333760.62 frames. ], batch size: 109, lr: 6.30e-03, grad_scale: 8.0 2023-04-29 09:20:33,710 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.6920, 4.8356, 4.9850, 4.8655, 4.8123, 5.4474, 5.0159, 4.7098], device='cuda:1'), covar=tensor([0.1345, 0.1938, 0.1946, 0.2067, 0.2793, 0.1022, 0.1403, 0.2485], device='cuda:1'), in_proj_covar=tensor([0.0358, 0.0501, 0.0545, 0.0439, 0.0576, 0.0571, 0.0428, 0.0588], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-04-29 09:21:21,637 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=103990.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 09:21:31,617 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=103997.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 09:21:42,813 INFO [train.py:904] (1/8) Epoch 11, batch 2500, loss[loss=0.2013, simple_loss=0.2705, pruned_loss=0.0661, over 16863.00 frames. ], tot_loss[loss=0.1904, simple_loss=0.2736, pruned_loss=0.05354, over 3330025.06 frames. ], batch size: 109, lr: 6.30e-03, grad_scale: 8.0 2023-04-29 09:22:11,531 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=104022.0, num_to_drop=1, layers_to_drop={1} 2023-04-29 09:22:39,880 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.47 vs. limit=2.0 2023-04-29 09:22:43,745 INFO [optim.py:368] (1/8) 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 09:22:47,544 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=104049.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 09:22:49,946 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 11, batch 2550, loss[loss=0.1923, simple_loss=0.2827, pruned_loss=0.05099, over 16634.00 frames. ], tot_loss[loss=0.19, simple_loss=0.2729, pruned_loss=0.05353, over 3333885.23 frames. ], batch size: 62, lr: 6.30e-03, grad_scale: 8.0 2023-04-29 09:22:59,848 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=104058.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 09:23:15,996 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=104070.0, num_to_drop=1, layers_to_drop={0} 2023-04-29 09:23:26,916 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=3.24 vs. limit=5.0 2023-04-29 09:23:51,203 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 11, batch 2600, loss[loss=0.2385, simple_loss=0.3008, pruned_loss=0.08811, over 16824.00 frames. ], tot_loss[loss=0.1901, simple_loss=0.2733, pruned_loss=0.05344, over 3335266.27 frames. ], batch size: 124, lr: 6.29e-03, grad_scale: 8.0 2023-04-29 09:24:11,992 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=104110.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 09:24:26,682 INFO [zipformer.py:625] (1/8) 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] (1/8) 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] (1/8) Epoch 11, batch 2650, loss[loss=0.1715, simple_loss=0.2631, pruned_loss=0.03988, over 17149.00 frames. ], tot_loss[loss=0.1898, simple_loss=0.2735, pruned_loss=0.05306, over 3338534.22 frames. ], batch size: 47, lr: 6.29e-03, grad_scale: 8.0 2023-04-29 09:26:18,840 INFO [train.py:904] (1/8) Epoch 11, batch 2700, loss[loss=0.1608, simple_loss=0.2492, pruned_loss=0.03625, over 17212.00 frames. ], tot_loss[loss=0.1887, simple_loss=0.273, pruned_loss=0.0522, over 3337858.10 frames. ], batch size: 45, lr: 6.29e-03, grad_scale: 8.0 2023-04-29 09:27:00,691 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=104232.0, num_to_drop=1, layers_to_drop={2} 2023-04-29 09:27:19,004 INFO [optim.py:368] (1/8) 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] (1/8) Epoch 11, batch 2750, loss[loss=0.1939, simple_loss=0.2651, pruned_loss=0.0613, over 16756.00 frames. ], tot_loss[loss=0.1888, simple_loss=0.2733, pruned_loss=0.0521, over 3332074.46 frames. ], batch size: 124, lr: 6.29e-03, grad_scale: 8.0 2023-04-29 09:28:01,739 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.63 vs. limit=2.0 2023-04-29 09:28:09,321 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.2538, 2.1139, 1.7098, 1.8986, 2.4521, 2.2919, 2.4888, 2.5762], device='cuda:1'), covar=tensor([0.0144, 0.0274, 0.0357, 0.0338, 0.0156, 0.0228, 0.0150, 0.0196], device='cuda:1'), in_proj_covar=tensor([0.0154, 0.0210, 0.0202, 0.0204, 0.0209, 0.0208, 0.0217, 0.0198], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-29 09:28:36,571 INFO [train.py:904] (1/8) Epoch 11, batch 2800, loss[loss=0.1845, simple_loss=0.2665, pruned_loss=0.05126, over 17245.00 frames. ], tot_loss[loss=0.1888, simple_loss=0.2735, pruned_loss=0.05211, over 3336754.40 frames. ], batch size: 45, lr: 6.29e-03, grad_scale: 8.0 2023-04-29 09:29:14,638 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.7206, 2.2027, 2.3621, 4.4795, 2.1012, 2.6635, 2.3031, 2.4404], device='cuda:1'), covar=tensor([0.0861, 0.3235, 0.2136, 0.0355, 0.3734, 0.2052, 0.2950, 0.3111], device='cuda:1'), in_proj_covar=tensor([0.0367, 0.0392, 0.0331, 0.0325, 0.0412, 0.0451, 0.0355, 0.0464], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-29 09:29:34,365 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.7956, 4.8507, 5.3437, 5.3155, 5.2914, 4.9239, 4.8997, 4.6182], device='cuda:1'), covar=tensor([0.0299, 0.0434, 0.0278, 0.0328, 0.0474, 0.0274, 0.0804, 0.0413], device='cuda:1'), in_proj_covar=tensor([0.0337, 0.0349, 0.0351, 0.0329, 0.0398, 0.0368, 0.0469, 0.0295], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:1') 2023-04-29 09:29:37,339 INFO [optim.py:368] (1/8) 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 09:29:37,603 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=104346.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 09:29:44,352 INFO [train.py:904] (1/8) Epoch 11, batch 2850, loss[loss=0.1777, simple_loss=0.2688, pruned_loss=0.04329, over 17155.00 frames. ], tot_loss[loss=0.1883, simple_loss=0.2727, pruned_loss=0.05189, over 3331803.44 frames. ], batch size: 46, lr: 6.29e-03, grad_scale: 8.0 2023-04-29 09:29:45,780 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=104353.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 09:30:09,935 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.7516, 2.2143, 2.2256, 4.5178, 2.0797, 2.6866, 2.3051, 2.3971], device='cuda:1'), covar=tensor([0.0837, 0.3300, 0.2300, 0.0343, 0.3802, 0.2306, 0.2829, 0.3551], device='cuda:1'), in_proj_covar=tensor([0.0367, 0.0392, 0.0331, 0.0324, 0.0412, 0.0450, 0.0355, 0.0464], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-29 09:30:44,207 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=104396.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 09:30:52,869 INFO [train.py:904] (1/8) Epoch 11, batch 2900, loss[loss=0.1721, simple_loss=0.2601, pruned_loss=0.04202, over 17094.00 frames. ], tot_loss[loss=0.1879, simple_loss=0.2717, pruned_loss=0.05202, over 3334800.52 frames. ], batch size: 53, lr: 6.29e-03, grad_scale: 8.0 2023-04-29 09:30:57,400 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=104405.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 09:31:19,942 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=104421.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 09:31:44,060 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.4450, 4.3991, 4.3871, 3.8700, 4.3561, 1.7597, 4.1516, 4.1046], device='cuda:1'), covar=tensor([0.0103, 0.0090, 0.0149, 0.0292, 0.0090, 0.2189, 0.0137, 0.0191], device='cuda:1'), in_proj_covar=tensor([0.0134, 0.0121, 0.0170, 0.0160, 0.0139, 0.0180, 0.0157, 0.0158], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-29 09:31:52,197 INFO [zipformer.py:625] (1/8) 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] (1/8) 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] (1/8) Epoch 11, batch 2950, loss[loss=0.2833, simple_loss=0.34, pruned_loss=0.1133, over 12025.00 frames. ], tot_loss[loss=0.1887, simple_loss=0.2717, pruned_loss=0.05283, over 3325310.86 frames. ], batch size: 246, lr: 6.28e-03, grad_scale: 8.0 2023-04-29 09:32:27,832 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=104469.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 09:33:03,401 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.7975, 1.7070, 2.2836, 2.6283, 2.6608, 2.6635, 1.8775, 2.9055], device='cuda:1'), covar=tensor([0.0136, 0.0333, 0.0245, 0.0174, 0.0194, 0.0180, 0.0329, 0.0084], device='cuda:1'), in_proj_covar=tensor([0.0166, 0.0175, 0.0160, 0.0161, 0.0171, 0.0126, 0.0172, 0.0118], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001], device='cuda:1') 2023-04-29 09:33:12,380 INFO [train.py:904] (1/8) Epoch 11, batch 3000, loss[loss=0.1962, simple_loss=0.2725, pruned_loss=0.05993, over 16255.00 frames. ], tot_loss[loss=0.1903, simple_loss=0.2724, pruned_loss=0.05409, over 3309501.79 frames. ], batch size: 165, lr: 6.28e-03, grad_scale: 8.0 2023-04-29 09:33:12,380 INFO [train.py:929] (1/8) Computing validation loss 2023-04-29 09:33:22,061 INFO [train.py:938] (1/8) Epoch 11, validation: loss=0.1413, simple_loss=0.2475, pruned_loss=0.01754, over 944034.00 frames. 2023-04-29 09:33:22,061 INFO [train.py:939] (1/8) Maximum memory allocated so far is 17676MB 2023-04-29 09:34:04,308 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=104532.0, num_to_drop=1, layers_to_drop={0} 2023-04-29 09:34:20,848 INFO [optim.py:368] (1/8) 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 09:34:30,327 INFO [train.py:904] (1/8) Epoch 11, batch 3050, loss[loss=0.1557, simple_loss=0.2468, pruned_loss=0.03228, over 17251.00 frames. ], tot_loss[loss=0.1899, simple_loss=0.2724, pruned_loss=0.05372, over 3316657.17 frames. ], batch size: 45, lr: 6.28e-03, grad_scale: 8.0 2023-04-29 09:34:37,502 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.8537, 1.7298, 2.2767, 2.6466, 2.6678, 2.7147, 1.7608, 2.9376], device='cuda:1'), covar=tensor([0.0124, 0.0329, 0.0244, 0.0175, 0.0182, 0.0178, 0.0345, 0.0078], device='cuda:1'), in_proj_covar=tensor([0.0167, 0.0175, 0.0160, 0.0162, 0.0172, 0.0127, 0.0173, 0.0118], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001], device='cuda:1') 2023-04-29 09:35:07,352 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=104580.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 09:35:36,546 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.3585, 3.8698, 3.4764, 1.8290, 2.7047, 2.1700, 3.7114, 4.0066], device='cuda:1'), covar=tensor([0.0293, 0.0700, 0.0664, 0.2167, 0.1065, 0.1310, 0.0669, 0.0945], device='cuda:1'), in_proj_covar=tensor([0.0146, 0.0148, 0.0159, 0.0143, 0.0136, 0.0126, 0.0138, 0.0159], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-04-29 09:35:37,111 INFO [train.py:904] (1/8) Epoch 11, batch 3100, loss[loss=0.1907, simple_loss=0.2788, pruned_loss=0.05128, over 17092.00 frames. ], tot_loss[loss=0.189, simple_loss=0.2718, pruned_loss=0.05312, over 3327716.24 frames. ], batch size: 49, lr: 6.28e-03, grad_scale: 8.0 2023-04-29 09:36:26,391 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.1878, 4.0503, 4.2450, 4.4177, 4.5291, 4.1037, 4.2338, 4.4901], device='cuda:1'), covar=tensor([0.1409, 0.0957, 0.1311, 0.0664, 0.0544, 0.1181, 0.1710, 0.0587], device='cuda:1'), in_proj_covar=tensor([0.0551, 0.0689, 0.0850, 0.0704, 0.0530, 0.0541, 0.0544, 0.0626], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-29 09:36:39,246 INFO [optim.py:368] (1/8) 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 09:36:39,599 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=104646.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 09:36:43,543 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.22 vs. limit=2.0 2023-04-29 09:36:47,518 INFO [train.py:904] (1/8) Epoch 11, batch 3150, loss[loss=0.18, simple_loss=0.2705, pruned_loss=0.04472, over 17115.00 frames. ], tot_loss[loss=0.1888, simple_loss=0.271, pruned_loss=0.05331, over 3312802.09 frames. ], batch size: 48, lr: 6.28e-03, grad_scale: 8.0 2023-04-29 09:36:49,119 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=104653.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 09:37:18,586 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.8289, 2.6670, 2.3749, 2.5937, 3.0417, 2.8601, 3.6344, 3.3084], device='cuda:1'), covar=tensor([0.0074, 0.0287, 0.0330, 0.0299, 0.0188, 0.0252, 0.0191, 0.0162], device='cuda:1'), in_proj_covar=tensor([0.0153, 0.0208, 0.0202, 0.0203, 0.0208, 0.0205, 0.0217, 0.0197], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-29 09:37:32,086 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.9682, 4.7436, 4.9511, 5.2210, 5.4174, 4.7625, 5.3939, 5.4137], device='cuda:1'), covar=tensor([0.1575, 0.1153, 0.1810, 0.0677, 0.0465, 0.0734, 0.0445, 0.0483], device='cuda:1'), in_proj_covar=tensor([0.0556, 0.0696, 0.0860, 0.0710, 0.0536, 0.0545, 0.0550, 0.0632], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-29 09:37:46,267 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=104694.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 09:37:56,591 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=104701.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 09:37:57,454 INFO [train.py:904] (1/8) Epoch 11, batch 3200, loss[loss=0.197, simple_loss=0.2884, pruned_loss=0.05281, over 17030.00 frames. ], tot_loss[loss=0.1872, simple_loss=0.27, pruned_loss=0.05221, over 3317814.14 frames. ], batch size: 53, lr: 6.28e-03, grad_scale: 8.0 2023-04-29 09:38:01,262 INFO [zipformer.py:625] (1/8) 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] (1/8) 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 09:39:06,600 INFO [train.py:904] (1/8) Epoch 11, batch 3250, loss[loss=0.1963, simple_loss=0.2875, pruned_loss=0.05259, over 16646.00 frames. ], tot_loss[loss=0.1874, simple_loss=0.2698, pruned_loss=0.05248, over 3321528.19 frames. ], batch size: 62, lr: 6.28e-03, grad_scale: 8.0 2023-04-29 09:39:08,057 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=104753.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 09:40:15,729 INFO [train.py:904] (1/8) Epoch 11, batch 3300, loss[loss=0.1773, simple_loss=0.2675, pruned_loss=0.04354, over 17139.00 frames. ], tot_loss[loss=0.1887, simple_loss=0.2714, pruned_loss=0.05304, over 3320911.99 frames. ], batch size: 48, lr: 6.27e-03, grad_scale: 8.0 2023-04-29 09:41:16,318 INFO [optim.py:368] (1/8) 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 09:41:16,683 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.8854, 4.9860, 5.4844, 5.4869, 5.4217, 5.0306, 5.0162, 4.7115], device='cuda:1'), covar=tensor([0.0351, 0.0435, 0.0347, 0.0344, 0.0465, 0.0339, 0.0860, 0.0437], device='cuda:1'), in_proj_covar=tensor([0.0342, 0.0357, 0.0357, 0.0335, 0.0404, 0.0373, 0.0478, 0.0301], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:1') 2023-04-29 09:41:24,649 INFO [train.py:904] (1/8) Epoch 11, batch 3350, loss[loss=0.234, simple_loss=0.3106, pruned_loss=0.07868, over 11923.00 frames. ], tot_loss[loss=0.1887, simple_loss=0.2717, pruned_loss=0.05284, over 3323293.19 frames. ], batch size: 247, lr: 6.27e-03, grad_scale: 8.0 2023-04-29 09:42:23,776 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.83 vs. limit=2.0 2023-04-29 09:42:33,960 INFO [train.py:904] (1/8) Epoch 11, batch 3400, loss[loss=0.1777, simple_loss=0.2613, pruned_loss=0.04704, over 17215.00 frames. ], tot_loss[loss=0.1878, simple_loss=0.2711, pruned_loss=0.05227, over 3329374.90 frames. ], batch size: 44, lr: 6.27e-03, grad_scale: 8.0 2023-04-29 09:43:18,718 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.6577, 4.7106, 4.9463, 4.7281, 4.7181, 5.3751, 4.9947, 4.6701], device='cuda:1'), covar=tensor([0.1414, 0.1889, 0.1816, 0.1874, 0.2697, 0.1007, 0.1345, 0.2316], device='cuda:1'), in_proj_covar=tensor([0.0358, 0.0506, 0.0546, 0.0436, 0.0581, 0.0576, 0.0432, 0.0593], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-04-29 09:43:33,847 INFO [optim.py:368] (1/8) 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] (1/8) Epoch 11, batch 3450, loss[loss=0.179, simple_loss=0.2719, pruned_loss=0.04312, over 17124.00 frames. ], tot_loss[loss=0.186, simple_loss=0.2694, pruned_loss=0.0513, over 3334487.90 frames. ], batch size: 48, lr: 6.27e-03, grad_scale: 8.0 2023-04-29 09:44:06,352 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.7672, 2.1648, 2.1731, 4.4446, 2.0796, 2.7266, 2.2989, 2.3762], device='cuda:1'), covar=tensor([0.0869, 0.3389, 0.2365, 0.0367, 0.3918, 0.2197, 0.2857, 0.3500], device='cuda:1'), in_proj_covar=tensor([0.0369, 0.0393, 0.0331, 0.0327, 0.0414, 0.0455, 0.0357, 0.0465], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-29 09:44:50,175 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.9438, 4.7452, 5.0193, 5.2472, 5.3953, 4.6880, 5.3833, 5.3630], device='cuda:1'), covar=tensor([0.1511, 0.0985, 0.1584, 0.0602, 0.0447, 0.0945, 0.0388, 0.0458], device='cuda:1'), in_proj_covar=tensor([0.0562, 0.0700, 0.0864, 0.0717, 0.0539, 0.0553, 0.0555, 0.0637], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-29 09:44:52,761 INFO [train.py:904] (1/8) Epoch 11, batch 3500, loss[loss=0.1791, simple_loss=0.2708, pruned_loss=0.04367, over 16678.00 frames. ], tot_loss[loss=0.1842, simple_loss=0.2676, pruned_loss=0.05043, over 3330070.57 frames. ], batch size: 57, lr: 6.27e-03, grad_scale: 8.0 2023-04-29 09:45:55,145 INFO [optim.py:368] (1/8) 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] (1/8) Epoch 11, batch 3550, loss[loss=0.1896, simple_loss=0.2661, pruned_loss=0.05655, over 15373.00 frames. ], tot_loss[loss=0.1841, simple_loss=0.267, pruned_loss=0.05056, over 3318539.57 frames. ], batch size: 190, lr: 6.27e-03, grad_scale: 4.0 2023-04-29 09:46:15,241 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([5.4011, 5.8461, 5.6137, 5.6435, 5.2067, 5.1628, 5.1904, 5.9733], device='cuda:1'), covar=tensor([0.1261, 0.0867, 0.0894, 0.0613, 0.0769, 0.0655, 0.0979, 0.0851], device='cuda:1'), in_proj_covar=tensor([0.0568, 0.0708, 0.0584, 0.0494, 0.0445, 0.0454, 0.0590, 0.0540], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-29 09:46:29,455 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([5.1947, 5.7023, 5.9126, 5.6365, 5.7791, 6.2601, 5.8686, 5.5600], device='cuda:1'), covar=tensor([0.0850, 0.1608, 0.1856, 0.1890, 0.2470, 0.1049, 0.1252, 0.2423], device='cuda:1'), in_proj_covar=tensor([0.0357, 0.0504, 0.0545, 0.0437, 0.0581, 0.0575, 0.0432, 0.0593], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-04-29 09:47:12,588 INFO [train.py:904] (1/8) Epoch 11, batch 3600, loss[loss=0.1808, simple_loss=0.2721, pruned_loss=0.04474, over 17069.00 frames. ], tot_loss[loss=0.1835, simple_loss=0.2662, pruned_loss=0.05034, over 3324406.83 frames. ], batch size: 55, lr: 6.26e-03, grad_scale: 8.0 2023-04-29 09:48:17,992 INFO [optim.py:368] (1/8) 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] (1/8) Epoch 11, batch 3650, loss[loss=0.1644, simple_loss=0.2447, pruned_loss=0.0421, over 15597.00 frames. ], tot_loss[loss=0.1832, simple_loss=0.2652, pruned_loss=0.05063, over 3302441.22 frames. ], batch size: 190, lr: 6.26e-03, grad_scale: 8.0 2023-04-29 09:48:48,505 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.5605, 3.9270, 4.2170, 2.9381, 3.7494, 4.2119, 3.8947, 2.5490], device='cuda:1'), covar=tensor([0.0400, 0.0176, 0.0038, 0.0275, 0.0068, 0.0058, 0.0050, 0.0328], device='cuda:1'), in_proj_covar=tensor([0.0127, 0.0071, 0.0071, 0.0126, 0.0079, 0.0090, 0.0077, 0.0118], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:1') 2023-04-29 09:49:37,393 INFO [train.py:904] (1/8) Epoch 11, batch 3700, loss[loss=0.1909, simple_loss=0.2678, pruned_loss=0.05698, over 16489.00 frames. ], tot_loss[loss=0.1837, simple_loss=0.2639, pruned_loss=0.05176, over 3268520.72 frames. ], batch size: 146, lr: 6.26e-03, grad_scale: 8.0 2023-04-29 09:49:49,802 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.54 vs. limit=2.0 2023-04-29 09:49:58,283 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([1.6570, 2.7450, 2.4823, 4.1528, 3.5337, 4.1096, 1.5068, 2.8243], device='cuda:1'), covar=tensor([0.1389, 0.0598, 0.1068, 0.0144, 0.0168, 0.0336, 0.1397, 0.0788], device='cuda:1'), in_proj_covar=tensor([0.0153, 0.0157, 0.0178, 0.0147, 0.0198, 0.0209, 0.0176, 0.0176], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0003, 0.0004], device='cuda:1') 2023-04-29 09:50:33,874 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.53 vs. limit=2.0 2023-04-29 09:50:41,009 INFO [optim.py:368] (1/8) 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] (1/8) Epoch 11, batch 3750, loss[loss=0.2021, simple_loss=0.2705, pruned_loss=0.06686, over 16307.00 frames. ], tot_loss[loss=0.1857, simple_loss=0.2643, pruned_loss=0.05353, over 3268218.36 frames. ], batch size: 165, lr: 6.26e-03, grad_scale: 8.0 2023-04-29 09:51:57,087 INFO [train.py:904] (1/8) Epoch 11, batch 3800, loss[loss=0.2098, simple_loss=0.2764, pruned_loss=0.07163, over 16852.00 frames. ], tot_loss[loss=0.1882, simple_loss=0.2663, pruned_loss=0.05508, over 3270605.23 frames. ], batch size: 116, lr: 6.26e-03, grad_scale: 8.0 2023-04-29 09:53:02,333 INFO [optim.py:368] (1/8) 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] (1/8) Epoch 11, batch 3850, loss[loss=0.1867, simple_loss=0.2538, pruned_loss=0.05982, over 16877.00 frames. ], tot_loss[loss=0.1883, simple_loss=0.2658, pruned_loss=0.05537, over 3276329.60 frames. ], batch size: 116, lr: 6.26e-03, grad_scale: 8.0 2023-04-29 09:53:59,956 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 11, batch 3900, loss[loss=0.1644, simple_loss=0.2421, pruned_loss=0.04332, over 16875.00 frames. ], tot_loss[loss=0.1883, simple_loss=0.2651, pruned_loss=0.05569, over 3272830.98 frames. ], batch size: 96, lr: 6.26e-03, grad_scale: 8.0 2023-04-29 09:55:25,284 INFO [optim.py:368] (1/8) 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 09:55:28,952 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 11, batch 3950, loss[loss=0.1961, simple_loss=0.2612, pruned_loss=0.06543, over 16467.00 frames. ], tot_loss[loss=0.1891, simple_loss=0.2649, pruned_loss=0.05665, over 3277776.97 frames. ], batch size: 146, lr: 6.25e-03, grad_scale: 8.0 2023-04-29 09:56:21,774 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 11, batch 4000, loss[loss=0.2204, simple_loss=0.2903, pruned_loss=0.07528, over 15508.00 frames. ], tot_loss[loss=0.1882, simple_loss=0.2639, pruned_loss=0.05621, over 3274779.24 frames. ], batch size: 190, lr: 6.25e-03, grad_scale: 8.0 2023-04-29 09:56:44,538 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=105502.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 09:57:48,088 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.73 vs. limit=2.0 2023-04-29 09:57:48,230 INFO [optim.py:368] (1/8) 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 09:57:48,819 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=105547.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 09:57:52,615 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.56 vs. limit=2.0 2023-04-29 09:57:55,346 INFO [train.py:904] (1/8) Epoch 11, batch 4050, loss[loss=0.1688, simple_loss=0.255, pruned_loss=0.04137, over 16676.00 frames. ], tot_loss[loss=0.1874, simple_loss=0.2642, pruned_loss=0.05531, over 3286174.49 frames. ], batch size: 83, lr: 6.25e-03, grad_scale: 8.0 2023-04-29 09:57:59,367 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.2228, 3.5132, 3.6433, 3.5964, 3.5880, 3.4109, 3.4527, 3.4509], device='cuda:1'), covar=tensor([0.0382, 0.0495, 0.0393, 0.0464, 0.0502, 0.0412, 0.0708, 0.0494], device='cuda:1'), in_proj_covar=tensor([0.0331, 0.0344, 0.0345, 0.0327, 0.0392, 0.0361, 0.0465, 0.0291], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:1') 2023-04-29 09:58:11,074 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=105563.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 09:58:47,094 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([5.7131, 6.0468, 5.7006, 5.8635, 5.3723, 5.1763, 5.4711, 6.1523], device='cuda:1'), covar=tensor([0.1011, 0.0606, 0.0813, 0.0639, 0.0672, 0.0627, 0.0799, 0.0647], device='cuda:1'), in_proj_covar=tensor([0.0556, 0.0687, 0.0569, 0.0480, 0.0436, 0.0445, 0.0575, 0.0526], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-29 09:59:08,730 INFO [train.py:904] (1/8) Epoch 11, batch 4100, loss[loss=0.2157, simple_loss=0.2907, pruned_loss=0.07039, over 15360.00 frames. ], tot_loss[loss=0.1873, simple_loss=0.2656, pruned_loss=0.05451, over 3286633.64 frames. ], batch size: 190, lr: 6.25e-03, grad_scale: 8.0 2023-04-29 10:00:18,622 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.70 vs. limit=2.0 2023-04-29 10:00:18,929 INFO [optim.py:368] (1/8) 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 10:00:26,731 INFO [train.py:904] (1/8) Epoch 11, batch 4150, loss[loss=0.2019, simple_loss=0.2938, pruned_loss=0.05503, over 16528.00 frames. ], tot_loss[loss=0.193, simple_loss=0.2725, pruned_loss=0.05671, over 3257640.88 frames. ], batch size: 75, lr: 6.25e-03, grad_scale: 8.0 2023-04-29 10:01:44,701 INFO [train.py:904] (1/8) Epoch 11, batch 4200, loss[loss=0.2218, simple_loss=0.3017, pruned_loss=0.07092, over 16578.00 frames. ], tot_loss[loss=0.1982, simple_loss=0.2797, pruned_loss=0.0583, over 3249707.64 frames. ], batch size: 68, lr: 6.25e-03, grad_scale: 8.0 2023-04-29 10:02:28,758 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.2570, 1.9889, 1.6363, 1.7742, 2.2528, 2.0105, 2.2096, 2.3991], device='cuda:1'), covar=tensor([0.0096, 0.0255, 0.0344, 0.0334, 0.0162, 0.0252, 0.0122, 0.0187], device='cuda:1'), in_proj_covar=tensor([0.0152, 0.0206, 0.0201, 0.0202, 0.0205, 0.0204, 0.0211, 0.0196], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-29 10:02:49,833 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=105744.0, num_to_drop=1, layers_to_drop={2} 2023-04-29 10:02:53,622 INFO [optim.py:368] (1/8) 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] (1/8) Epoch 11, batch 4250, loss[loss=0.2012, simple_loss=0.2885, pruned_loss=0.05695, over 16952.00 frames. ], tot_loss[loss=0.1995, simple_loss=0.2827, pruned_loss=0.05818, over 3219326.96 frames. ], batch size: 41, lr: 6.25e-03, grad_scale: 8.0 2023-04-29 10:03:13,465 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.5856, 3.7686, 2.8022, 2.1562, 2.6651, 2.2998, 3.8811, 3.3755], device='cuda:1'), covar=tensor([0.2689, 0.0621, 0.1675, 0.2414, 0.2448, 0.1800, 0.0527, 0.1046], device='cuda:1'), in_proj_covar=tensor([0.0299, 0.0253, 0.0281, 0.0275, 0.0285, 0.0220, 0.0265, 0.0295], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-29 10:03:56,949 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.6986, 3.7752, 2.8949, 2.2996, 2.8463, 2.4343, 4.0375, 3.5375], device='cuda:1'), covar=tensor([0.2406, 0.0733, 0.1538, 0.2025, 0.2012, 0.1702, 0.0447, 0.0991], device='cuda:1'), in_proj_covar=tensor([0.0300, 0.0254, 0.0282, 0.0276, 0.0286, 0.0221, 0.0267, 0.0296], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-29 10:03:59,241 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=105792.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 10:04:08,462 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=105799.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 10:04:12,691 INFO [train.py:904] (1/8) Epoch 11, batch 4300, loss[loss=0.1965, simple_loss=0.2889, pruned_loss=0.05204, over 16493.00 frames. ], tot_loss[loss=0.1989, simple_loss=0.2837, pruned_loss=0.05702, over 3229467.13 frames. ], batch size: 68, lr: 6.24e-03, grad_scale: 8.0 2023-04-29 10:04:27,981 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.5771, 4.6561, 4.6806, 2.9009, 3.8286, 4.5232, 4.0118, 2.5925], device='cuda:1'), covar=tensor([0.0427, 0.0014, 0.0018, 0.0298, 0.0063, 0.0044, 0.0050, 0.0317], device='cuda:1'), in_proj_covar=tensor([0.0127, 0.0069, 0.0070, 0.0124, 0.0078, 0.0089, 0.0076, 0.0118], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:1') 2023-04-29 10:05:11,562 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=105842.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 10:05:17,305 INFO [optim.py:368] (1/8) 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] (1/8) Epoch 11, batch 4350, loss[loss=0.2432, simple_loss=0.3116, pruned_loss=0.08739, over 11637.00 frames. ], tot_loss[loss=0.2026, simple_loss=0.2876, pruned_loss=0.05877, over 3205786.23 frames. ], batch size: 246, lr: 6.24e-03, grad_scale: 8.0 2023-04-29 10:05:27,929 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=105853.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 10:05:34,702 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=105858.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 10:05:37,381 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=105860.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 10:06:37,392 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.9840, 5.3915, 5.7039, 5.3752, 5.4526, 6.0240, 5.5597, 5.2137], device='cuda:1'), covar=tensor([0.0756, 0.1518, 0.1654, 0.1554, 0.2171, 0.0835, 0.1203, 0.2183], device='cuda:1'), in_proj_covar=tensor([0.0344, 0.0478, 0.0519, 0.0416, 0.0550, 0.0545, 0.0411, 0.0563], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-04-29 10:06:38,342 INFO [train.py:904] (1/8) Epoch 11, batch 4400, loss[loss=0.2246, simple_loss=0.3108, pruned_loss=0.06919, over 17239.00 frames. ], tot_loss[loss=0.2056, simple_loss=0.2901, pruned_loss=0.06051, over 3181248.10 frames. ], batch size: 52, lr: 6.24e-03, grad_scale: 8.0 2023-04-29 10:07:36,172 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([1.5566, 2.5698, 2.2463, 3.5925, 3.0407, 3.8377, 1.2151, 2.8786], device='cuda:1'), covar=tensor([0.1368, 0.0714, 0.1236, 0.0124, 0.0277, 0.0355, 0.1712, 0.0761], device='cuda:1'), in_proj_covar=tensor([0.0155, 0.0159, 0.0181, 0.0146, 0.0201, 0.0210, 0.0181, 0.0180], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-04-29 10:07:40,707 INFO [optim.py:368] (1/8) 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] (1/8) Epoch 11, batch 4450, loss[loss=0.1983, simple_loss=0.283, pruned_loss=0.05681, over 16389.00 frames. ], tot_loss[loss=0.2082, simple_loss=0.2934, pruned_loss=0.06151, over 3188094.76 frames. ], batch size: 68, lr: 6.24e-03, grad_scale: 8.0 2023-04-29 10:08:34,723 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([5.4421, 5.4955, 5.2813, 4.9434, 4.9140, 5.3934, 5.2302, 4.9846], device='cuda:1'), covar=tensor([0.0471, 0.0151, 0.0185, 0.0213, 0.0756, 0.0212, 0.0176, 0.0482], device='cuda:1'), in_proj_covar=tensor([0.0239, 0.0301, 0.0285, 0.0264, 0.0304, 0.0299, 0.0194, 0.0329], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-29 10:09:04,903 INFO [train.py:904] (1/8) Epoch 11, batch 4500, loss[loss=0.2012, simple_loss=0.287, pruned_loss=0.05769, over 16561.00 frames. ], tot_loss[loss=0.2085, simple_loss=0.2936, pruned_loss=0.06172, over 3198481.17 frames. ], batch size: 68, lr: 6.24e-03, grad_scale: 8.0 2023-04-29 10:10:04,944 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=106044.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 10:10:09,390 INFO [optim.py:368] (1/8) 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 10:10:17,384 INFO [train.py:904] (1/8) Epoch 11, batch 4550, loss[loss=0.2307, simple_loss=0.3092, pruned_loss=0.07605, over 16836.00 frames. ], tot_loss[loss=0.2097, simple_loss=0.2941, pruned_loss=0.06264, over 3184709.51 frames. ], batch size: 42, lr: 6.24e-03, grad_scale: 8.0 2023-04-29 10:10:53,614 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=106077.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 10:11:14,748 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 11, batch 4600, loss[loss=0.1905, simple_loss=0.284, pruned_loss=0.04852, over 16821.00 frames. ], tot_loss[loss=0.21, simple_loss=0.2947, pruned_loss=0.0627, over 3179154.50 frames. ], batch size: 102, lr: 6.24e-03, grad_scale: 8.0 2023-04-29 10:11:30,089 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=4.71 vs. limit=5.0 2023-04-29 10:12:22,254 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=106138.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 10:12:28,036 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=106142.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 10:12:35,552 INFO [optim.py:368] (1/8) 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] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=106148.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 10:12:41,946 INFO [train.py:904] (1/8) Epoch 11, batch 4650, loss[loss=0.1895, simple_loss=0.2753, pruned_loss=0.0519, over 16620.00 frames. ], tot_loss[loss=0.2096, simple_loss=0.2938, pruned_loss=0.06274, over 3188517.47 frames. ], batch size: 68, lr: 6.23e-03, grad_scale: 8.0 2023-04-29 10:12:47,341 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=106155.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 10:12:51,750 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=106158.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 10:13:37,558 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=106190.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 10:13:55,748 INFO [train.py:904] (1/8) Epoch 11, batch 4700, loss[loss=0.1911, simple_loss=0.2768, pruned_loss=0.05274, over 16918.00 frames. ], tot_loss[loss=0.2069, simple_loss=0.291, pruned_loss=0.06141, over 3204870.77 frames. ], batch size: 109, lr: 6.23e-03, grad_scale: 8.0 2023-04-29 10:14:01,834 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=106206.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 10:15:01,676 INFO [optim.py:368] (1/8) 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 10:15:09,056 INFO [train.py:904] (1/8) Epoch 11, batch 4750, loss[loss=0.2128, simple_loss=0.2837, pruned_loss=0.07092, over 11933.00 frames. ], tot_loss[loss=0.203, simple_loss=0.2869, pruned_loss=0.05951, over 3207940.04 frames. ], batch size: 247, lr: 6.23e-03, grad_scale: 8.0 2023-04-29 10:15:32,269 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=106268.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 10:15:51,648 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([5.1871, 5.1632, 5.1045, 4.3058, 5.0973, 1.6827, 4.8414, 5.0545], device='cuda:1'), covar=tensor([0.0074, 0.0060, 0.0098, 0.0438, 0.0064, 0.2364, 0.0097, 0.0144], device='cuda:1'), in_proj_covar=tensor([0.0128, 0.0116, 0.0164, 0.0158, 0.0134, 0.0177, 0.0151, 0.0151], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-29 10:16:22,052 INFO [train.py:904] (1/8) Epoch 11, batch 4800, loss[loss=0.2166, simple_loss=0.3024, pruned_loss=0.06543, over 16253.00 frames. ], tot_loss[loss=0.1988, simple_loss=0.2829, pruned_loss=0.05738, over 3214983.74 frames. ], batch size: 165, lr: 6.23e-03, grad_scale: 8.0 2023-04-29 10:16:48,314 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.97 vs. limit=2.0 2023-04-29 10:17:02,224 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=106329.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 10:17:12,263 INFO [zipformer.py:625] (1/8) 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] (1/8) 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] (1/8) Epoch 11, batch 4850, loss[loss=0.1853, simple_loss=0.2889, pruned_loss=0.04088, over 16863.00 frames. ], tot_loss[loss=0.1979, simple_loss=0.2836, pruned_loss=0.05606, over 3209094.58 frames. ], batch size: 96, lr: 6.23e-03, grad_scale: 8.0 2023-04-29 10:18:08,945 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.3341, 4.6656, 4.4520, 4.4960, 4.1312, 4.1511, 4.1835, 4.6834], device='cuda:1'), covar=tensor([0.1026, 0.0748, 0.0837, 0.0565, 0.0684, 0.1149, 0.0860, 0.0828], device='cuda:1'), in_proj_covar=tensor([0.0526, 0.0649, 0.0539, 0.0449, 0.0413, 0.0420, 0.0538, 0.0499], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:1') 2023-04-29 10:18:38,660 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 11, batch 4900, loss[loss=0.1883, simple_loss=0.2772, pruned_loss=0.04969, over 16410.00 frames. ], tot_loss[loss=0.1962, simple_loss=0.2828, pruned_loss=0.05479, over 3204417.49 frames. ], batch size: 146, lr: 6.23e-03, grad_scale: 8.0 2023-04-29 10:19:29,912 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=106433.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 10:19:50,195 INFO [optim.py:368] (1/8) 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 10:19:51,775 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 11, batch 4950, loss[loss=0.1893, simple_loss=0.284, pruned_loss=0.04733, over 16577.00 frames. ], tot_loss[loss=0.1956, simple_loss=0.2826, pruned_loss=0.05426, over 3199462.35 frames. ], batch size: 68, lr: 6.22e-03, grad_scale: 8.0 2023-04-29 10:20:01,906 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=106454.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 10:20:03,086 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=106455.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 10:20:56,812 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.1506, 1.8721, 2.0456, 3.6812, 1.6944, 2.3412, 1.9791, 2.0439], device='cuda:1'), covar=tensor([0.1203, 0.3768, 0.2450, 0.0576, 0.4707, 0.2627, 0.3361, 0.3670], device='cuda:1'), in_proj_covar=tensor([0.0363, 0.0392, 0.0327, 0.0319, 0.0411, 0.0451, 0.0355, 0.0460], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-29 10:21:00,650 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 11, batch 5000, loss[loss=0.1835, simple_loss=0.2763, pruned_loss=0.04532, over 16428.00 frames. ], tot_loss[loss=0.1964, simple_loss=0.2845, pruned_loss=0.05414, over 3202422.40 frames. ], batch size: 75, lr: 6.22e-03, grad_scale: 8.0 2023-04-29 10:21:10,910 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=106503.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 10:21:28,729 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=106515.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 10:21:54,390 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.2794, 3.3373, 1.4409, 3.4956, 2.2533, 3.5658, 1.7177, 2.4893], device='cuda:1'), covar=tensor([0.0182, 0.0257, 0.1676, 0.0095, 0.0785, 0.0343, 0.1453, 0.0662], device='cuda:1'), in_proj_covar=tensor([0.0148, 0.0164, 0.0188, 0.0127, 0.0168, 0.0206, 0.0193, 0.0169], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-04-29 10:22:04,189 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([1.8770, 3.2437, 2.5879, 4.8366, 3.8913, 4.1383, 1.6078, 2.9417], device='cuda:1'), covar=tensor([0.1119, 0.0529, 0.1040, 0.0087, 0.0248, 0.0378, 0.1339, 0.0797], device='cuda:1'), in_proj_covar=tensor([0.0153, 0.0158, 0.0181, 0.0145, 0.0199, 0.0208, 0.0180, 0.0180], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0003, 0.0004], device='cuda:1') 2023-04-29 10:22:14,238 INFO [optim.py:368] (1/8) 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] (1/8) Epoch 11, batch 5050, loss[loss=0.1979, simple_loss=0.2887, pruned_loss=0.05353, over 16571.00 frames. ], tot_loss[loss=0.1972, simple_loss=0.2854, pruned_loss=0.05451, over 3200904.60 frames. ], batch size: 75, lr: 6.22e-03, grad_scale: 8.0 2023-04-29 10:23:32,363 INFO [train.py:904] (1/8) Epoch 11, batch 5100, loss[loss=0.2251, simple_loss=0.2963, pruned_loss=0.07698, over 12518.00 frames. ], tot_loss[loss=0.1965, simple_loss=0.2841, pruned_loss=0.0545, over 3192083.18 frames. ], batch size: 246, lr: 6.22e-03, grad_scale: 8.0 2023-04-29 10:23:57,171 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.4541, 4.5024, 4.4552, 3.0225, 3.5879, 4.3174, 3.8745, 2.3410], device='cuda:1'), covar=tensor([0.0443, 0.0016, 0.0019, 0.0275, 0.0071, 0.0074, 0.0064, 0.0365], device='cuda:1'), in_proj_covar=tensor([0.0128, 0.0070, 0.0070, 0.0127, 0.0079, 0.0091, 0.0078, 0.0120], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:1') 2023-04-29 10:24:03,672 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=106624.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 10:24:05,065 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.73 vs. limit=2.0 2023-04-29 10:24:38,775 INFO [optim.py:368] (1/8) 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 10:24:44,357 INFO [train.py:904] (1/8) Epoch 11, batch 5150, loss[loss=0.2035, simple_loss=0.2973, pruned_loss=0.05479, over 16681.00 frames. ], tot_loss[loss=0.1958, simple_loss=0.2839, pruned_loss=0.05385, over 3195973.76 frames. ], batch size: 134, lr: 6.22e-03, grad_scale: 4.0 2023-04-29 10:24:45,071 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.16 vs. limit=2.0 2023-04-29 10:25:23,177 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.38 vs. limit=2.0 2023-04-29 10:25:43,744 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=106692.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 10:25:56,050 INFO [train.py:904] (1/8) Epoch 11, batch 5200, loss[loss=0.1863, simple_loss=0.2706, pruned_loss=0.05096, over 16587.00 frames. ], tot_loss[loss=0.1941, simple_loss=0.2821, pruned_loss=0.05309, over 3209160.85 frames. ], batch size: 76, lr: 6.22e-03, grad_scale: 8.0 2023-04-29 10:25:58,978 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.85 vs. limit=2.0 2023-04-29 10:26:27,296 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.2110, 2.0435, 2.1363, 3.9183, 2.0209, 2.5122, 2.1770, 2.2752], device='cuda:1'), covar=tensor([0.1047, 0.3181, 0.2196, 0.0361, 0.3517, 0.2117, 0.2973, 0.2592], device='cuda:1'), in_proj_covar=tensor([0.0361, 0.0389, 0.0325, 0.0319, 0.0408, 0.0448, 0.0352, 0.0457], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-29 10:26:34,324 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.8321, 2.4075, 2.4024, 4.7174, 2.2649, 2.8584, 2.4919, 2.7070], device='cuda:1'), covar=tensor([0.0867, 0.3070, 0.2147, 0.0276, 0.3469, 0.2144, 0.2813, 0.2517], device='cuda:1'), in_proj_covar=tensor([0.0361, 0.0389, 0.0324, 0.0318, 0.0407, 0.0448, 0.0352, 0.0456], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-29 10:26:42,255 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=106733.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 10:26:57,760 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.8851, 4.0091, 3.1082, 2.3953, 2.9408, 2.6355, 4.3833, 3.7296], device='cuda:1'), covar=tensor([0.2240, 0.0634, 0.1431, 0.2034, 0.1963, 0.1544, 0.0411, 0.0827], device='cuda:1'), in_proj_covar=tensor([0.0299, 0.0254, 0.0280, 0.0274, 0.0280, 0.0219, 0.0265, 0.0292], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-29 10:27:03,714 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.2111, 1.9862, 1.6580, 1.6869, 2.2302, 1.9889, 2.0674, 2.3750], device='cuda:1'), covar=tensor([0.0101, 0.0298, 0.0368, 0.0370, 0.0154, 0.0272, 0.0129, 0.0198], device='cuda:1'), in_proj_covar=tensor([0.0144, 0.0200, 0.0196, 0.0195, 0.0199, 0.0201, 0.0202, 0.0192], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-29 10:27:04,335 INFO [optim.py:368] (1/8) 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 10:27:11,356 INFO [train.py:904] (1/8) Epoch 11, batch 5250, loss[loss=0.1948, simple_loss=0.2753, pruned_loss=0.05711, over 16697.00 frames. ], tot_loss[loss=0.1927, simple_loss=0.2796, pruned_loss=0.05292, over 3217911.75 frames. ], batch size: 62, lr: 6.22e-03, grad_scale: 8.0 2023-04-29 10:27:28,151 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.9776, 5.3586, 5.6050, 5.3543, 5.3251, 5.9318, 5.4498, 5.1278], device='cuda:1'), covar=tensor([0.0747, 0.1417, 0.1199, 0.1411, 0.2046, 0.0704, 0.1048, 0.1970], device='cuda:1'), in_proj_covar=tensor([0.0343, 0.0468, 0.0508, 0.0410, 0.0542, 0.0537, 0.0405, 0.0561], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-04-29 10:27:52,545 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=106781.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 10:28:19,932 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.8696, 1.9419, 2.2689, 3.1947, 2.0953, 2.1879, 2.1733, 2.0782], device='cuda:1'), covar=tensor([0.1053, 0.3200, 0.1936, 0.0527, 0.3609, 0.2306, 0.2843, 0.2986], device='cuda:1'), in_proj_covar=tensor([0.0362, 0.0390, 0.0325, 0.0319, 0.0409, 0.0448, 0.0352, 0.0456], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-29 10:28:22,267 INFO [train.py:904] (1/8) Epoch 11, batch 5300, loss[loss=0.1639, simple_loss=0.2465, pruned_loss=0.04059, over 16606.00 frames. ], tot_loss[loss=0.1899, simple_loss=0.2764, pruned_loss=0.05169, over 3221558.16 frames. ], batch size: 57, lr: 6.21e-03, grad_scale: 8.0 2023-04-29 10:28:32,965 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=106810.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 10:29:24,515 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.5651, 3.5347, 3.5126, 2.9105, 3.4176, 1.9298, 3.2900, 2.9731], device='cuda:1'), covar=tensor([0.0119, 0.0108, 0.0130, 0.0294, 0.0086, 0.2247, 0.0126, 0.0209], device='cuda:1'), in_proj_covar=tensor([0.0127, 0.0116, 0.0164, 0.0159, 0.0135, 0.0178, 0.0151, 0.0153], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-29 10:29:27,200 INFO [optim.py:368] (1/8) 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 10:29:33,906 INFO [train.py:904] (1/8) Epoch 11, batch 5350, loss[loss=0.2065, simple_loss=0.2993, pruned_loss=0.05686, over 16408.00 frames. ], tot_loss[loss=0.1888, simple_loss=0.2753, pruned_loss=0.05116, over 3218400.00 frames. ], batch size: 146, lr: 6.21e-03, grad_scale: 8.0 2023-04-29 10:29:42,275 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.49 vs. limit=2.0 2023-04-29 10:30:06,280 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.6968, 3.7576, 4.0655, 4.0564, 4.0446, 3.8188, 3.7999, 3.7487], device='cuda:1'), covar=tensor([0.0308, 0.0605, 0.0394, 0.0405, 0.0453, 0.0364, 0.0823, 0.0484], device='cuda:1'), in_proj_covar=tensor([0.0326, 0.0339, 0.0339, 0.0326, 0.0387, 0.0360, 0.0465, 0.0289], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:1') 2023-04-29 10:30:45,865 INFO [train.py:904] (1/8) Epoch 11, batch 5400, loss[loss=0.2042, simple_loss=0.288, pruned_loss=0.06025, over 16495.00 frames. ], tot_loss[loss=0.1925, simple_loss=0.279, pruned_loss=0.05297, over 3198327.95 frames. ], batch size: 68, lr: 6.21e-03, grad_scale: 8.0 2023-04-29 10:31:18,207 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=106924.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 10:31:54,541 INFO [optim.py:368] (1/8) 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] (1/8) Epoch 11, batch 5450, loss[loss=0.2477, simple_loss=0.3248, pruned_loss=0.08536, over 11962.00 frames. ], tot_loss[loss=0.1964, simple_loss=0.2826, pruned_loss=0.05512, over 3183868.39 frames. ], batch size: 248, lr: 6.21e-03, grad_scale: 8.0 2023-04-29 10:32:04,633 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.9006, 3.3502, 3.2793, 1.9664, 2.9341, 2.3441, 3.3546, 3.5338], device='cuda:1'), covar=tensor([0.0255, 0.0596, 0.0525, 0.1667, 0.0707, 0.0839, 0.0602, 0.0785], device='cuda:1'), in_proj_covar=tensor([0.0142, 0.0144, 0.0157, 0.0142, 0.0135, 0.0124, 0.0136, 0.0154], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:1') 2023-04-29 10:32:34,435 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=106972.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 10:32:45,856 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.4149, 2.4074, 1.8197, 2.2536, 2.8323, 2.4356, 3.0768, 3.0729], device='cuda:1'), covar=tensor([0.0059, 0.0296, 0.0426, 0.0338, 0.0209, 0.0302, 0.0149, 0.0166], device='cuda:1'), in_proj_covar=tensor([0.0147, 0.0204, 0.0198, 0.0199, 0.0202, 0.0204, 0.0206, 0.0193], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-29 10:33:03,981 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=106992.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 10:33:19,271 INFO [train.py:904] (1/8) Epoch 11, batch 5500, loss[loss=0.2028, simple_loss=0.289, pruned_loss=0.05829, over 16684.00 frames. ], tot_loss[loss=0.2048, simple_loss=0.2901, pruned_loss=0.05972, over 3165448.40 frames. ], batch size: 76, lr: 6.21e-03, grad_scale: 8.0 2023-04-29 10:34:18,908 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=107040.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 10:34:31,621 INFO [optim.py:368] (1/8) 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] (1/8) Epoch 11, batch 5550, loss[loss=0.2427, simple_loss=0.3187, pruned_loss=0.08331, over 16386.00 frames. ], tot_loss[loss=0.2135, simple_loss=0.2974, pruned_loss=0.06477, over 3152359.38 frames. ], batch size: 146, lr: 6.21e-03, grad_scale: 8.0 2023-04-29 10:35:46,232 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.40 vs. limit=2.0 2023-04-29 10:35:57,902 INFO [train.py:904] (1/8) Epoch 11, batch 5600, loss[loss=0.237, simple_loss=0.3172, pruned_loss=0.07839, over 16698.00 frames. ], tot_loss[loss=0.2217, simple_loss=0.3033, pruned_loss=0.07001, over 3120575.09 frames. ], batch size: 134, lr: 6.21e-03, grad_scale: 8.0 2023-04-29 10:36:12,112 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=107110.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 10:36:12,255 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.6930, 3.5362, 3.9622, 1.8669, 4.1737, 4.1831, 3.0129, 3.1341], device='cuda:1'), covar=tensor([0.0659, 0.0211, 0.0195, 0.1157, 0.0061, 0.0097, 0.0399, 0.0390], device='cuda:1'), in_proj_covar=tensor([0.0142, 0.0100, 0.0089, 0.0139, 0.0070, 0.0102, 0.0120, 0.0128], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:1') 2023-04-29 10:37:02,642 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=107140.0, num_to_drop=1, layers_to_drop={0} 2023-04-29 10:37:17,301 INFO [optim.py:368] (1/8) 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] (1/8) Epoch 11, batch 5650, loss[loss=0.3049, simple_loss=0.3545, pruned_loss=0.1277, over 11541.00 frames. ], tot_loss[loss=0.2299, simple_loss=0.3093, pruned_loss=0.07529, over 3066755.93 frames. ], batch size: 248, lr: 6.20e-03, grad_scale: 4.0 2023-04-29 10:37:22,249 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=107152.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 10:37:32,229 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=107158.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 10:38:42,790 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 11, batch 5700, loss[loss=0.3398, simple_loss=0.3772, pruned_loss=0.1513, over 11293.00 frames. ], tot_loss[loss=0.2324, simple_loss=0.3111, pruned_loss=0.07685, over 3051021.83 frames. ], batch size: 248, lr: 6.20e-03, grad_scale: 4.0 2023-04-29 10:39:02,870 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=107213.0, num_to_drop=1, layers_to_drop={1} 2023-04-29 10:39:26,322 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=2.95 vs. limit=5.0 2023-04-29 10:39:59,317 INFO [optim.py:368] (1/8) 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] (1/8) Epoch 11, batch 5750, loss[loss=0.211, simple_loss=0.3016, pruned_loss=0.06013, over 16841.00 frames. ], tot_loss[loss=0.2362, simple_loss=0.3141, pruned_loss=0.07919, over 3005526.81 frames. ], batch size: 96, lr: 6.20e-03, grad_scale: 4.0 2023-04-29 10:40:35,693 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.64 vs. limit=2.0 2023-04-29 10:41:25,722 INFO [train.py:904] (1/8) Epoch 11, batch 5800, loss[loss=0.2068, simple_loss=0.2985, pruned_loss=0.05756, over 16767.00 frames. ], tot_loss[loss=0.2329, simple_loss=0.3127, pruned_loss=0.07654, over 3036610.67 frames. ], batch size: 124, lr: 6.20e-03, grad_scale: 4.0 2023-04-29 10:42:39,026 INFO [optim.py:368] (1/8) Clipping_scale=2.0, grad-norm quartiles 2.001e+02 2.956e+02 3.560e+02 4.660e+02 9.709e+02, threshold=7.120e+02, percent-clipped=1.0 2023-04-29 10:42:43,690 INFO [train.py:904] (1/8) Epoch 11, batch 5850, loss[loss=0.2111, simple_loss=0.2946, pruned_loss=0.06381, over 17140.00 frames. ], tot_loss[loss=0.2311, simple_loss=0.3114, pruned_loss=0.07538, over 3037096.55 frames. ], batch size: 47, lr: 6.20e-03, grad_scale: 4.0 2023-04-29 10:42:47,981 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.5240, 4.6370, 4.8800, 4.6887, 4.7603, 5.2920, 4.8193, 4.5729], device='cuda:1'), covar=tensor([0.1149, 0.1856, 0.1730, 0.1834, 0.2288, 0.0894, 0.1322, 0.2288], device='cuda:1'), in_proj_covar=tensor([0.0346, 0.0478, 0.0522, 0.0417, 0.0548, 0.0547, 0.0411, 0.0564], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-04-29 10:44:05,228 INFO [train.py:904] (1/8) Epoch 11, batch 5900, loss[loss=0.2048, simple_loss=0.2808, pruned_loss=0.06436, over 15463.00 frames. ], tot_loss[loss=0.2308, simple_loss=0.3109, pruned_loss=0.07536, over 3046914.73 frames. ], batch size: 191, lr: 6.20e-03, grad_scale: 4.0 2023-04-29 10:45:21,996 INFO [optim.py:368] (1/8) 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 10:45:26,029 INFO [train.py:904] (1/8) Epoch 11, batch 5950, loss[loss=0.2139, simple_loss=0.2988, pruned_loss=0.06446, over 15323.00 frames. ], tot_loss[loss=0.2287, simple_loss=0.3111, pruned_loss=0.07319, over 3065569.35 frames. ], batch size: 190, lr: 6.20e-03, grad_scale: 4.0 2023-04-29 10:46:40,597 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=107496.0, num_to_drop=1, layers_to_drop={3} 2023-04-29 10:46:48,955 INFO [train.py:904] (1/8) Epoch 11, batch 6000, loss[loss=0.2103, simple_loss=0.2942, pruned_loss=0.06323, over 16258.00 frames. ], tot_loss[loss=0.2263, simple_loss=0.3092, pruned_loss=0.07165, over 3099807.35 frames. ], batch size: 165, lr: 6.19e-03, grad_scale: 8.0 2023-04-29 10:46:48,955 INFO [train.py:929] (1/8) Computing validation loss 2023-04-29 10:46:56,230 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.7234, 4.7331, 5.0991, 5.0516, 5.0532, 4.8008, 4.7521, 4.5102], device='cuda:1'), covar=tensor([0.0233, 0.0325, 0.0277, 0.0291, 0.0302, 0.0234, 0.0710, 0.0346], device='cuda:1'), in_proj_covar=tensor([0.0319, 0.0333, 0.0337, 0.0318, 0.0379, 0.0354, 0.0452, 0.0289], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003], device='cuda:1') 2023-04-29 10:46:59,887 INFO [train.py:938] (1/8) Epoch 11, validation: loss=0.163, simple_loss=0.2761, pruned_loss=0.02492, over 944034.00 frames. 2023-04-29 10:46:59,888 INFO [train.py:939] (1/8) Maximum memory allocated so far is 17676MB 2023-04-29 10:47:09,650 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=107508.0, num_to_drop=1, layers_to_drop={3} 2023-04-29 10:47:13,606 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=107510.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 10:48:12,720 INFO [optim.py:368] (1/8) 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] (1/8) Epoch 11, batch 6050, loss[loss=0.189, simple_loss=0.2771, pruned_loss=0.05043, over 16334.00 frames. ], tot_loss[loss=0.2255, simple_loss=0.3078, pruned_loss=0.07162, over 3090787.92 frames. ], batch size: 35, lr: 6.19e-03, grad_scale: 8.0 2023-04-29 10:48:48,435 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=107571.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 10:49:12,888 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.9286, 4.1004, 4.6011, 1.9632, 4.7790, 4.7620, 3.2963, 3.4092], device='cuda:1'), covar=tensor([0.0680, 0.0173, 0.0126, 0.1174, 0.0042, 0.0105, 0.0305, 0.0389], device='cuda:1'), in_proj_covar=tensor([0.0141, 0.0097, 0.0087, 0.0137, 0.0068, 0.0100, 0.0118, 0.0126], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0003, 0.0002, 0.0004, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:1') 2023-04-29 10:49:13,242 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.82 vs. limit=2.0 2023-04-29 10:49:24,183 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.3281, 4.2080, 4.4158, 4.5806, 4.6985, 4.2875, 4.6519, 4.6934], device='cuda:1'), covar=tensor([0.1598, 0.1085, 0.1381, 0.0581, 0.0562, 0.0958, 0.0617, 0.0605], device='cuda:1'), in_proj_covar=tensor([0.0515, 0.0643, 0.0780, 0.0654, 0.0503, 0.0508, 0.0518, 0.0584], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-29 10:49:34,932 INFO [train.py:904] (1/8) Epoch 11, batch 6100, loss[loss=0.246, simple_loss=0.3317, pruned_loss=0.08018, over 16622.00 frames. ], tot_loss[loss=0.2233, simple_loss=0.3065, pruned_loss=0.07001, over 3108615.27 frames. ], batch size: 62, lr: 6.19e-03, grad_scale: 8.0 2023-04-29 10:50:51,420 INFO [optim.py:368] (1/8) 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 10:50:56,527 INFO [train.py:904] (1/8) Epoch 11, batch 6150, loss[loss=0.209, simple_loss=0.2936, pruned_loss=0.06218, over 16673.00 frames. ], tot_loss[loss=0.2222, simple_loss=0.3047, pruned_loss=0.06986, over 3113884.16 frames. ], batch size: 62, lr: 6.19e-03, grad_scale: 8.0 2023-04-29 10:51:07,651 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=107659.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 10:52:14,168 INFO [train.py:904] (1/8) Epoch 11, batch 6200, loss[loss=0.2386, simple_loss=0.2973, pruned_loss=0.08991, over 11664.00 frames. ], tot_loss[loss=0.2209, simple_loss=0.3029, pruned_loss=0.06945, over 3111949.09 frames. ], batch size: 248, lr: 6.19e-03, grad_scale: 4.0 2023-04-29 10:52:42,565 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=107720.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 10:53:07,978 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=107737.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 10:53:27,428 INFO [optim.py:368] (1/8) 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 10:53:27,970 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.1529, 3.9669, 4.1986, 4.3678, 4.5086, 4.0523, 4.4236, 4.4926], device='cuda:1'), covar=tensor([0.1538, 0.1149, 0.1542, 0.0691, 0.0566, 0.1221, 0.0756, 0.0643], device='cuda:1'), in_proj_covar=tensor([0.0517, 0.0646, 0.0783, 0.0654, 0.0506, 0.0509, 0.0522, 0.0586], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-29 10:53:29,999 INFO [train.py:904] (1/8) Epoch 11, batch 6250, loss[loss=0.2509, simple_loss=0.3422, pruned_loss=0.0798, over 16510.00 frames. ], tot_loss[loss=0.2209, simple_loss=0.3026, pruned_loss=0.06959, over 3099690.32 frames. ], batch size: 146, lr: 6.19e-03, grad_scale: 4.0 2023-04-29 10:54:19,152 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.7324, 3.1699, 3.0417, 1.7869, 2.6735, 2.2361, 3.2121, 3.3697], device='cuda:1'), covar=tensor([0.0273, 0.0645, 0.0583, 0.1818, 0.0790, 0.0863, 0.0640, 0.0738], device='cuda:1'), in_proj_covar=tensor([0.0140, 0.0141, 0.0156, 0.0142, 0.0134, 0.0124, 0.0135, 0.0153], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:1') 2023-04-29 10:54:36,274 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=107796.0, num_to_drop=1, layers_to_drop={1} 2023-04-29 10:54:36,659 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.90 vs. limit=2.0 2023-04-29 10:54:38,950 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=107798.0, num_to_drop=1, layers_to_drop={1} 2023-04-29 10:54:45,102 INFO [train.py:904] (1/8) Epoch 11, batch 6300, loss[loss=0.1975, simple_loss=0.2881, pruned_loss=0.05348, over 16821.00 frames. ], tot_loss[loss=0.2193, simple_loss=0.3018, pruned_loss=0.06837, over 3107766.22 frames. ], batch size: 102, lr: 6.19e-03, grad_scale: 4.0 2023-04-29 10:54:54,157 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=107808.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 10:55:52,342 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=107844.0, num_to_drop=1, layers_to_drop={1} 2023-04-29 10:56:00,513 INFO [optim.py:368] (1/8) 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 10:56:03,041 INFO [train.py:904] (1/8) Epoch 11, batch 6350, loss[loss=0.2277, simple_loss=0.3097, pruned_loss=0.07284, over 16659.00 frames. ], tot_loss[loss=0.2209, simple_loss=0.3028, pruned_loss=0.0695, over 3106768.93 frames. ], batch size: 89, lr: 6.18e-03, grad_scale: 4.0 2023-04-29 10:56:10,038 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=107856.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 10:56:25,830 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=107866.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 10:57:20,954 INFO [train.py:904] (1/8) Epoch 11, batch 6400, loss[loss=0.2157, simple_loss=0.29, pruned_loss=0.07069, over 17061.00 frames. ], tot_loss[loss=0.2231, simple_loss=0.3036, pruned_loss=0.07124, over 3088649.54 frames. ], batch size: 55, lr: 6.18e-03, grad_scale: 8.0 2023-04-29 10:58:08,865 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.7667, 4.0270, 4.2511, 2.2112, 3.4512, 2.8605, 4.1427, 4.2912], device='cuda:1'), covar=tensor([0.0221, 0.0589, 0.0473, 0.1745, 0.0663, 0.0799, 0.0531, 0.0660], device='cuda:1'), in_proj_covar=tensor([0.0141, 0.0142, 0.0156, 0.0142, 0.0135, 0.0124, 0.0136, 0.0154], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:1') 2023-04-29 10:58:35,864 INFO [optim.py:368] (1/8) 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 10:58:35,880 INFO [train.py:904] (1/8) Epoch 11, batch 6450, loss[loss=0.2283, simple_loss=0.3085, pruned_loss=0.07406, over 16657.00 frames. ], tot_loss[loss=0.2227, simple_loss=0.304, pruned_loss=0.0707, over 3089182.96 frames. ], batch size: 76, lr: 6.18e-03, grad_scale: 2.0 2023-04-29 10:59:54,943 INFO [train.py:904] (1/8) Epoch 11, batch 6500, loss[loss=0.2103, simple_loss=0.299, pruned_loss=0.0608, over 16290.00 frames. ], tot_loss[loss=0.2208, simple_loss=0.302, pruned_loss=0.0698, over 3102115.29 frames. ], batch size: 165, lr: 6.18e-03, grad_scale: 2.0 2023-04-29 11:00:11,394 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.81 vs. limit=2.0 2023-04-29 11:00:14,832 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=108015.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 11:01:12,919 INFO [optim.py:368] (1/8) 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 11:01:12,934 INFO [train.py:904] (1/8) Epoch 11, batch 6550, loss[loss=0.2344, simple_loss=0.3355, pruned_loss=0.0666, over 16870.00 frames. ], tot_loss[loss=0.222, simple_loss=0.3043, pruned_loss=0.06986, over 3125888.88 frames. ], batch size: 90, lr: 6.18e-03, grad_scale: 2.0 2023-04-29 11:02:01,341 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.4368, 4.4163, 4.8523, 4.8133, 4.8114, 4.4531, 4.4767, 4.2477], device='cuda:1'), covar=tensor([0.0292, 0.0456, 0.0404, 0.0419, 0.0574, 0.0373, 0.0905, 0.0489], device='cuda:1'), in_proj_covar=tensor([0.0327, 0.0343, 0.0345, 0.0324, 0.0391, 0.0362, 0.0463, 0.0293], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:1') 2023-04-29 11:02:13,590 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=108093.0, num_to_drop=1, layers_to_drop={0} 2023-04-29 11:02:25,852 INFO [train.py:904] (1/8) Epoch 11, batch 6600, loss[loss=0.2297, simple_loss=0.3113, pruned_loss=0.07405, over 16708.00 frames. ], tot_loss[loss=0.2233, simple_loss=0.3066, pruned_loss=0.07001, over 3124892.13 frames. ], batch size: 124, lr: 6.18e-03, grad_scale: 2.0 2023-04-29 11:03:41,607 INFO [optim.py:368] (1/8) 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] (1/8) Epoch 11, batch 6650, loss[loss=0.2114, simple_loss=0.2972, pruned_loss=0.06283, over 16889.00 frames. ], tot_loss[loss=0.2248, simple_loss=0.3069, pruned_loss=0.07135, over 3108780.40 frames. ], batch size: 96, lr: 6.18e-03, grad_scale: 2.0 2023-04-29 11:04:03,254 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=108166.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 11:04:50,274 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([1.7222, 2.6521, 2.3084, 3.7067, 2.6164, 3.8044, 1.3701, 2.8000], device='cuda:1'), covar=tensor([0.1238, 0.0640, 0.1174, 0.0145, 0.0237, 0.0389, 0.1523, 0.0760], device='cuda:1'), in_proj_covar=tensor([0.0154, 0.0158, 0.0182, 0.0144, 0.0200, 0.0209, 0.0181, 0.0181], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-04-29 11:04:56,954 INFO [train.py:904] (1/8) Epoch 11, batch 6700, loss[loss=0.2382, simple_loss=0.3126, pruned_loss=0.0819, over 16196.00 frames. ], tot_loss[loss=0.2244, simple_loss=0.3057, pruned_loss=0.07152, over 3097064.34 frames. ], batch size: 165, lr: 6.17e-03, grad_scale: 2.0 2023-04-29 11:05:14,875 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=108214.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 11:06:13,521 INFO [optim.py:368] (1/8) 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] (1/8) Epoch 11, batch 6750, loss[loss=0.2005, simple_loss=0.282, pruned_loss=0.05949, over 17147.00 frames. ], tot_loss[loss=0.2239, simple_loss=0.3045, pruned_loss=0.07164, over 3095256.38 frames. ], batch size: 48, lr: 6.17e-03, grad_scale: 2.0 2023-04-29 11:07:28,504 INFO [train.py:904] (1/8) Epoch 11, batch 6800, loss[loss=0.1834, simple_loss=0.2774, pruned_loss=0.04467, over 16829.00 frames. ], tot_loss[loss=0.2231, simple_loss=0.3043, pruned_loss=0.071, over 3100714.63 frames. ], batch size: 102, lr: 6.17e-03, grad_scale: 4.0 2023-04-29 11:07:48,577 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=108315.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 11:08:05,605 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=108325.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 11:08:27,142 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.4550, 2.5946, 1.9791, 2.3504, 2.9603, 2.6144, 3.2086, 3.2264], device='cuda:1'), covar=tensor([0.0073, 0.0299, 0.0434, 0.0364, 0.0178, 0.0283, 0.0182, 0.0158], device='cuda:1'), in_proj_covar=tensor([0.0144, 0.0199, 0.0196, 0.0196, 0.0200, 0.0201, 0.0204, 0.0189], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-29 11:08:45,531 INFO [optim.py:368] (1/8) 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 11:08:45,546 INFO [train.py:904] (1/8) Epoch 11, batch 6850, loss[loss=0.2149, simple_loss=0.312, pruned_loss=0.05888, over 16419.00 frames. ], tot_loss[loss=0.224, simple_loss=0.3054, pruned_loss=0.07132, over 3104880.25 frames. ], batch size: 146, lr: 6.17e-03, grad_scale: 4.0 2023-04-29 11:09:01,699 INFO [zipformer.py:625] (1/8) 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] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=108386.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 11:09:44,188 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([1.8538, 1.3540, 1.6407, 1.6427, 1.8096, 1.8321, 1.5155, 1.9054], device='cuda:1'), covar=tensor([0.0170, 0.0253, 0.0140, 0.0193, 0.0176, 0.0127, 0.0262, 0.0089], device='cuda:1'), in_proj_covar=tensor([0.0160, 0.0171, 0.0151, 0.0159, 0.0168, 0.0124, 0.0170, 0.0113], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001], device='cuda:1') 2023-04-29 11:09:46,552 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 11, batch 6900, loss[loss=0.2336, simple_loss=0.3119, pruned_loss=0.07768, over 16322.00 frames. ], tot_loss[loss=0.2248, simple_loss=0.3076, pruned_loss=0.07103, over 3111597.78 frames. ], batch size: 165, lr: 6.17e-03, grad_scale: 2.0 2023-04-29 11:10:45,912 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=108432.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 11:10:59,098 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=108441.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 11:11:15,792 INFO [train.py:904] (1/8) Epoch 11, batch 6950, loss[loss=0.2654, simple_loss=0.3284, pruned_loss=0.1011, over 11534.00 frames. ], tot_loss[loss=0.2286, simple_loss=0.31, pruned_loss=0.0736, over 3085083.32 frames. ], batch size: 247, lr: 6.17e-03, grad_scale: 2.0 2023-04-29 11:11:17,884 INFO [optim.py:368] (1/8) Clipping_scale=2.0, grad-norm quartiles 2.069e+02 2.933e+02 3.744e+02 4.621e+02 9.342e+02, threshold=7.489e+02, percent-clipped=9.0 2023-04-29 11:12:20,786 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=108493.0, num_to_drop=1, layers_to_drop={2} 2023-04-29 11:12:25,365 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.82 vs. limit=2.0 2023-04-29 11:12:33,547 INFO [train.py:904] (1/8) Epoch 11, batch 7000, loss[loss=0.212, simple_loss=0.3076, pruned_loss=0.05823, over 16432.00 frames. ], tot_loss[loss=0.2284, simple_loss=0.3106, pruned_loss=0.07312, over 3086494.55 frames. ], batch size: 165, lr: 6.17e-03, grad_scale: 2.0 2023-04-29 11:13:52,263 INFO [train.py:904] (1/8) Epoch 11, batch 7050, loss[loss=0.2252, simple_loss=0.3051, pruned_loss=0.07263, over 16897.00 frames. ], tot_loss[loss=0.2272, simple_loss=0.3103, pruned_loss=0.07206, over 3096269.08 frames. ], batch size: 109, lr: 6.16e-03, grad_scale: 2.0 2023-04-29 11:13:53,480 INFO [optim.py:368] (1/8) 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 11:14:32,897 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=108577.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 11:14:45,396 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.16 vs. limit=2.0 2023-04-29 11:15:11,199 INFO [train.py:904] (1/8) Epoch 11, batch 7100, loss[loss=0.2656, simple_loss=0.3213, pruned_loss=0.105, over 11451.00 frames. ], tot_loss[loss=0.2253, simple_loss=0.3082, pruned_loss=0.07124, over 3111813.44 frames. ], batch size: 248, lr: 6.16e-03, grad_scale: 2.0 2023-04-29 11:16:07,256 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=108638.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 11:16:23,233 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.51 vs. limit=2.0 2023-04-29 11:16:27,131 INFO [train.py:904] (1/8) Epoch 11, batch 7150, loss[loss=0.2564, simple_loss=0.3165, pruned_loss=0.09814, over 11462.00 frames. ], tot_loss[loss=0.2226, simple_loss=0.3051, pruned_loss=0.07001, over 3120784.34 frames. ], batch size: 246, lr: 6.16e-03, grad_scale: 2.0 2023-04-29 11:16:28,930 INFO [optim.py:368] (1/8) 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 11:16:42,167 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([1.8056, 1.7668, 1.5272, 1.5354, 1.8921, 1.5628, 1.7368, 1.9334], device='cuda:1'), covar=tensor([0.0108, 0.0175, 0.0268, 0.0224, 0.0125, 0.0185, 0.0124, 0.0125], device='cuda:1'), in_proj_covar=tensor([0.0143, 0.0199, 0.0196, 0.0196, 0.0200, 0.0200, 0.0203, 0.0188], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-29 11:17:10,531 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=108681.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 11:17:41,815 INFO [train.py:904] (1/8) Epoch 11, batch 7200, loss[loss=0.1863, simple_loss=0.2781, pruned_loss=0.04727, over 16795.00 frames. ], tot_loss[loss=0.2213, simple_loss=0.3038, pruned_loss=0.06937, over 3082586.94 frames. ], batch size: 116, lr: 6.16e-03, grad_scale: 4.0 2023-04-29 11:19:02,056 INFO [train.py:904] (1/8) Epoch 11, batch 7250, loss[loss=0.1831, simple_loss=0.2662, pruned_loss=0.04997, over 16896.00 frames. ], tot_loss[loss=0.2172, simple_loss=0.3002, pruned_loss=0.06714, over 3111349.11 frames. ], batch size: 42, lr: 6.16e-03, grad_scale: 4.0 2023-04-29 11:19:03,146 INFO [optim.py:368] (1/8) 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 11:19:55,672 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=108788.0, num_to_drop=1, layers_to_drop={2} 2023-04-29 11:20:00,944 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=108791.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 11:20:16,090 INFO [train.py:904] (1/8) Epoch 11, batch 7300, loss[loss=0.2257, simple_loss=0.318, pruned_loss=0.06671, over 16869.00 frames. ], tot_loss[loss=0.2167, simple_loss=0.2991, pruned_loss=0.06717, over 3090137.21 frames. ], batch size: 96, lr: 6.16e-03, grad_scale: 4.0 2023-04-29 11:21:34,169 INFO [train.py:904] (1/8) Epoch 11, batch 7350, loss[loss=0.2081, simple_loss=0.2938, pruned_loss=0.06122, over 16559.00 frames. ], tot_loss[loss=0.2183, simple_loss=0.3003, pruned_loss=0.06815, over 3073636.37 frames. ], batch size: 62, lr: 6.16e-03, grad_scale: 4.0 2023-04-29 11:21:34,701 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=108852.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 11:21:34,947 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.52 vs. limit=2.0 2023-04-29 11:21:35,287 INFO [optim.py:368] (1/8) 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 11:22:29,800 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=3.41 vs. limit=5.0 2023-04-29 11:22:54,323 INFO [train.py:904] (1/8) Epoch 11, batch 7400, loss[loss=0.2075, simple_loss=0.303, pruned_loss=0.05597, over 16881.00 frames. ], tot_loss[loss=0.2193, simple_loss=0.3012, pruned_loss=0.06873, over 3076975.86 frames. ], batch size: 102, lr: 6.15e-03, grad_scale: 4.0 2023-04-29 11:23:42,499 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 11, batch 7450, loss[loss=0.2136, simple_loss=0.2973, pruned_loss=0.06499, over 16458.00 frames. ], tot_loss[loss=0.2212, simple_loss=0.3028, pruned_loss=0.0698, over 3075870.89 frames. ], batch size: 75, lr: 6.15e-03, grad_scale: 4.0 2023-04-29 11:24:13,636 INFO [optim.py:368] (1/8) 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 11:24:28,558 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.32 vs. limit=2.0 2023-04-29 11:24:59,150 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=108981.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 11:25:31,627 INFO [train.py:904] (1/8) Epoch 11, batch 7500, loss[loss=0.2604, simple_loss=0.3242, pruned_loss=0.09825, over 11232.00 frames. ], tot_loss[loss=0.222, simple_loss=0.304, pruned_loss=0.06999, over 3065006.60 frames. ], batch size: 247, lr: 6.15e-03, grad_scale: 4.0 2023-04-29 11:26:16,356 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=109029.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 11:26:27,936 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=109037.0, num_to_drop=1, layers_to_drop={0} 2023-04-29 11:26:51,082 INFO [train.py:904] (1/8) Epoch 11, batch 7550, loss[loss=0.2264, simple_loss=0.3081, pruned_loss=0.07239, over 16765.00 frames. ], tot_loss[loss=0.2225, simple_loss=0.3036, pruned_loss=0.07072, over 3058941.73 frames. ], batch size: 124, lr: 6.15e-03, grad_scale: 4.0 2023-04-29 11:26:52,322 INFO [optim.py:368] (1/8) 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 11:27:46,081 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=109088.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 11:28:02,037 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=109098.0, num_to_drop=1, layers_to_drop={2} 2023-04-29 11:28:06,702 INFO [train.py:904] (1/8) Epoch 11, batch 7600, loss[loss=0.2657, simple_loss=0.3292, pruned_loss=0.101, over 11231.00 frames. ], tot_loss[loss=0.2218, simple_loss=0.3025, pruned_loss=0.0705, over 3074312.75 frames. ], batch size: 248, lr: 6.15e-03, grad_scale: 8.0 2023-04-29 11:28:42,488 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.3339, 3.0153, 2.7843, 1.9373, 2.5596, 2.1056, 2.9404, 3.1512], device='cuda:1'), covar=tensor([0.0329, 0.0641, 0.0743, 0.1890, 0.0994, 0.1000, 0.0763, 0.0754], device='cuda:1'), in_proj_covar=tensor([0.0141, 0.0143, 0.0157, 0.0144, 0.0137, 0.0126, 0.0137, 0.0155], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:1') 2023-04-29 11:28:54,843 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.7577, 4.0184, 4.3532, 1.9713, 4.5739, 4.5247, 3.2236, 3.1979], device='cuda:1'), covar=tensor([0.0673, 0.0145, 0.0132, 0.1104, 0.0034, 0.0081, 0.0327, 0.0378], device='cuda:1'), in_proj_covar=tensor([0.0141, 0.0098, 0.0087, 0.0137, 0.0068, 0.0100, 0.0118, 0.0126], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0003, 0.0002, 0.0004, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:1') 2023-04-29 11:28:56,881 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=109136.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 11:29:06,980 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.54 vs. limit=2.0 2023-04-29 11:29:12,289 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=109147.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 11:29:20,693 INFO [train.py:904] (1/8) Epoch 11, batch 7650, loss[loss=0.2572, simple_loss=0.346, pruned_loss=0.08425, over 15387.00 frames. ], tot_loss[loss=0.221, simple_loss=0.302, pruned_loss=0.07, over 3102684.04 frames. ], batch size: 191, lr: 6.15e-03, grad_scale: 4.0 2023-04-29 11:29:23,630 INFO [optim.py:368] (1/8) 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 11:29:26,878 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.1552, 4.1776, 4.0261, 3.8237, 3.6982, 4.1057, 3.8350, 3.8100], device='cuda:1'), covar=tensor([0.0574, 0.0426, 0.0277, 0.0278, 0.0870, 0.0423, 0.0734, 0.0675], device='cuda:1'), in_proj_covar=tensor([0.0235, 0.0306, 0.0279, 0.0256, 0.0298, 0.0297, 0.0194, 0.0326], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-29 11:29:52,394 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([1.6604, 2.8110, 2.3857, 4.2900, 3.1569, 4.0570, 1.4469, 2.9227], device='cuda:1'), covar=tensor([0.1313, 0.0680, 0.1238, 0.0160, 0.0252, 0.0388, 0.1561, 0.0814], device='cuda:1'), in_proj_covar=tensor([0.0156, 0.0160, 0.0183, 0.0145, 0.0200, 0.0210, 0.0182, 0.0182], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-04-29 11:30:25,785 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=109194.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 11:30:36,525 INFO [train.py:904] (1/8) Epoch 11, batch 7700, loss[loss=0.293, simple_loss=0.3397, pruned_loss=0.1231, over 11681.00 frames. ], tot_loss[loss=0.222, simple_loss=0.3026, pruned_loss=0.07071, over 3095370.96 frames. ], batch size: 247, lr: 6.15e-03, grad_scale: 4.0 2023-04-29 11:31:25,259 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=109233.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 11:31:53,769 INFO [train.py:904] (1/8) Epoch 11, batch 7750, loss[loss=0.2131, simple_loss=0.303, pruned_loss=0.06157, over 16720.00 frames. ], tot_loss[loss=0.2219, simple_loss=0.3031, pruned_loss=0.0703, over 3110757.74 frames. ], batch size: 83, lr: 6.14e-03, grad_scale: 4.0 2023-04-29 11:31:56,716 INFO [optim.py:368] (1/8) 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 11:31:59,161 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=109255.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 11:32:39,019 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=109281.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 11:32:48,546 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.0368, 3.0502, 1.7339, 3.2509, 2.3014, 3.2967, 1.9292, 2.5969], device='cuda:1'), covar=tensor([0.0252, 0.0383, 0.1548, 0.0151, 0.0765, 0.0554, 0.1447, 0.0658], device='cuda:1'), in_proj_covar=tensor([0.0148, 0.0163, 0.0188, 0.0125, 0.0167, 0.0205, 0.0195, 0.0169], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0003, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-04-29 11:32:58,003 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=109294.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 11:33:10,717 INFO [train.py:904] (1/8) Epoch 11, batch 7800, loss[loss=0.2222, simple_loss=0.3149, pruned_loss=0.06476, over 16359.00 frames. ], tot_loss[loss=0.2238, simple_loss=0.3045, pruned_loss=0.07153, over 3089183.83 frames. ], batch size: 146, lr: 6.14e-03, grad_scale: 4.0 2023-04-29 11:33:31,249 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.44 vs. limit=2.0 2023-04-29 11:33:35,159 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 11, batch 7850, loss[loss=0.229, simple_loss=0.3156, pruned_loss=0.0712, over 16843.00 frames. ], tot_loss[loss=0.224, simple_loss=0.3053, pruned_loss=0.07133, over 3093366.68 frames. ], batch size: 124, lr: 6.14e-03, grad_scale: 2.0 2023-04-29 11:34:30,495 INFO [optim.py:368] (1/8) 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 11:34:30,920 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=109355.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 11:34:37,011 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.9270, 1.7386, 2.3967, 2.9171, 2.7898, 3.2547, 2.0781, 3.2924], device='cuda:1'), covar=tensor([0.0145, 0.0364, 0.0228, 0.0189, 0.0182, 0.0118, 0.0334, 0.0088], device='cuda:1'), in_proj_covar=tensor([0.0158, 0.0169, 0.0148, 0.0156, 0.0165, 0.0123, 0.0168, 0.0112], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001], device='cuda:1') 2023-04-29 11:35:07,145 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=109379.0, num_to_drop=1, layers_to_drop={3} 2023-04-29 11:35:27,273 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=109393.0, num_to_drop=1, layers_to_drop={0} 2023-04-29 11:35:41,205 INFO [train.py:904] (1/8) Epoch 11, batch 7900, loss[loss=0.2117, simple_loss=0.2977, pruned_loss=0.06281, over 16769.00 frames. ], tot_loss[loss=0.2217, simple_loss=0.3037, pruned_loss=0.0699, over 3107270.66 frames. ], batch size: 83, lr: 6.14e-03, grad_scale: 2.0 2023-04-29 11:36:02,565 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.6701, 3.7703, 4.1232, 4.0622, 4.0585, 3.8106, 3.8186, 3.8291], device='cuda:1'), covar=tensor([0.0321, 0.0531, 0.0330, 0.0410, 0.0432, 0.0386, 0.0879, 0.0471], device='cuda:1'), in_proj_covar=tensor([0.0322, 0.0336, 0.0339, 0.0318, 0.0387, 0.0355, 0.0461, 0.0288], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003], device='cuda:1') 2023-04-29 11:36:05,592 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.23 vs. limit=2.0 2023-04-29 11:36:52,244 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 11, batch 7950, loss[loss=0.2299, simple_loss=0.3167, pruned_loss=0.07153, over 16292.00 frames. ], tot_loss[loss=0.2229, simple_loss=0.3044, pruned_loss=0.07068, over 3100952.64 frames. ], batch size: 35, lr: 6.14e-03, grad_scale: 2.0 2023-04-29 11:37:04,709 INFO [optim.py:368] (1/8) Clipping_scale=2.0, grad-norm quartiles 2.170e+02 3.029e+02 3.447e+02 4.404e+02 6.785e+02, threshold=6.894e+02, percent-clipped=0.0 2023-04-29 11:37:24,451 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.7146, 4.5058, 4.7426, 4.9131, 5.0813, 4.5985, 5.0441, 5.0357], device='cuda:1'), covar=tensor([0.1499, 0.1059, 0.1377, 0.0616, 0.0493, 0.0736, 0.0507, 0.0582], device='cuda:1'), in_proj_covar=tensor([0.0516, 0.0642, 0.0770, 0.0651, 0.0504, 0.0499, 0.0515, 0.0584], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-29 11:38:04,154 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=109495.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 11:38:14,152 INFO [train.py:904] (1/8) Epoch 11, batch 8000, loss[loss=0.1987, simple_loss=0.2859, pruned_loss=0.05574, over 16820.00 frames. ], tot_loss[loss=0.2232, simple_loss=0.3046, pruned_loss=0.07085, over 3104260.55 frames. ], batch size: 116, lr: 6.14e-03, grad_scale: 4.0 2023-04-29 11:38:49,426 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.1948, 3.9959, 4.2941, 4.4024, 4.5603, 4.1674, 4.4916, 4.5368], device='cuda:1'), covar=tensor([0.1465, 0.1222, 0.1335, 0.0686, 0.0535, 0.1046, 0.0653, 0.0655], device='cuda:1'), in_proj_covar=tensor([0.0513, 0.0637, 0.0768, 0.0648, 0.0502, 0.0495, 0.0511, 0.0581], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-29 11:39:24,929 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=109550.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 11:39:27,473 INFO [train.py:904] (1/8) Epoch 11, batch 8050, loss[loss=0.2399, simple_loss=0.318, pruned_loss=0.08094, over 16408.00 frames. ], tot_loss[loss=0.2234, simple_loss=0.3048, pruned_loss=0.07097, over 3098389.81 frames. ], batch size: 146, lr: 6.14e-03, grad_scale: 4.0 2023-04-29 11:39:31,003 INFO [optim.py:368] (1/8) 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 11:39:37,728 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.1467, 1.9745, 2.1024, 3.7097, 1.9674, 2.3646, 2.1205, 2.1370], device='cuda:1'), covar=tensor([0.1038, 0.3478, 0.2493, 0.0458, 0.3967, 0.2376, 0.3184, 0.3296], device='cuda:1'), in_proj_covar=tensor([0.0360, 0.0392, 0.0327, 0.0319, 0.0415, 0.0447, 0.0354, 0.0459], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-29 11:40:06,541 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=109579.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 11:40:40,896 INFO [train.py:904] (1/8) Epoch 11, batch 8100, loss[loss=0.217, simple_loss=0.3009, pruned_loss=0.06653, over 16937.00 frames. ], tot_loss[loss=0.2218, simple_loss=0.3039, pruned_loss=0.06987, over 3118312.21 frames. ], batch size: 109, lr: 6.13e-03, grad_scale: 4.0 2023-04-29 11:41:00,926 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=2.33 vs. limit=5.0 2023-04-29 11:41:39,237 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=109640.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 11:41:54,478 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=109650.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 11:41:56,612 INFO [train.py:904] (1/8) Epoch 11, batch 8150, loss[loss=0.2056, simple_loss=0.2863, pruned_loss=0.0625, over 16252.00 frames. ], tot_loss[loss=0.218, simple_loss=0.3, pruned_loss=0.06796, over 3131969.47 frames. ], batch size: 165, lr: 6.13e-03, grad_scale: 4.0 2023-04-29 11:42:01,362 INFO [optim.py:368] (1/8) 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 11:42:08,188 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=109659.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 11:42:30,923 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=109674.0, num_to_drop=1, layers_to_drop={3} 2023-04-29 11:42:59,331 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=109693.0, num_to_drop=1, layers_to_drop={2} 2023-04-29 11:43:12,283 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=5.06 vs. limit=5.0 2023-04-29 11:43:12,476 INFO [train.py:904] (1/8) Epoch 11, batch 8200, loss[loss=0.2748, simple_loss=0.3272, pruned_loss=0.1112, over 11618.00 frames. ], tot_loss[loss=0.2175, simple_loss=0.2984, pruned_loss=0.06823, over 3116754.49 frames. ], batch size: 248, lr: 6.13e-03, grad_scale: 4.0 2023-04-29 11:43:41,773 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=109720.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 11:44:16,311 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=109741.0, num_to_drop=1, layers_to_drop={0} 2023-04-29 11:44:33,541 INFO [train.py:904] (1/8) Epoch 11, batch 8250, loss[loss=0.2067, simple_loss=0.298, pruned_loss=0.05771, over 16635.00 frames. ], tot_loss[loss=0.2151, simple_loss=0.2976, pruned_loss=0.06627, over 3108369.80 frames. ], batch size: 62, lr: 6.13e-03, grad_scale: 4.0 2023-04-29 11:44:38,003 INFO [optim.py:368] (1/8) 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 11:45:52,507 INFO [train.py:904] (1/8) Epoch 11, batch 8300, loss[loss=0.1892, simple_loss=0.2684, pruned_loss=0.05497, over 11999.00 frames. ], tot_loss[loss=0.211, simple_loss=0.2953, pruned_loss=0.0633, over 3098386.53 frames. ], batch size: 247, lr: 6.13e-03, grad_scale: 4.0 2023-04-29 11:47:09,116 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 11, batch 8350, loss[loss=0.2111, simple_loss=0.3031, pruned_loss=0.05956, over 15483.00 frames. ], tot_loss[loss=0.2088, simple_loss=0.2941, pruned_loss=0.06169, over 3062566.64 frames. ], batch size: 191, lr: 6.13e-03, grad_scale: 4.0 2023-04-29 11:47:16,940 INFO [optim.py:368] (1/8) 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 11:48:15,168 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.4194, 3.3362, 3.4704, 3.5792, 3.6131, 3.2771, 3.5376, 3.6403], device='cuda:1'), covar=tensor([0.1094, 0.0911, 0.1059, 0.0583, 0.0574, 0.2440, 0.0906, 0.0685], device='cuda:1'), in_proj_covar=tensor([0.0508, 0.0638, 0.0763, 0.0645, 0.0499, 0.0495, 0.0509, 0.0579], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-29 11:48:22,935 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.0209, 3.1280, 1.8626, 3.2748, 2.3311, 3.3324, 2.0247, 2.6892], device='cuda:1'), covar=tensor([0.0225, 0.0287, 0.1457, 0.0168, 0.0765, 0.0343, 0.1504, 0.0608], device='cuda:1'), in_proj_covar=tensor([0.0145, 0.0158, 0.0183, 0.0122, 0.0164, 0.0197, 0.0191, 0.0166], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0003, 0.0004, 0.0004, 0.0003], device='cuda:1') 2023-04-29 11:48:24,628 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 11, batch 8400, loss[loss=0.1886, simple_loss=0.2807, pruned_loss=0.04827, over 16280.00 frames. ], tot_loss[loss=0.2045, simple_loss=0.291, pruned_loss=0.059, over 3068017.87 frames. ], batch size: 165, lr: 6.13e-03, grad_scale: 8.0 2023-04-29 11:48:32,327 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.4333, 3.0228, 3.1174, 1.8272, 2.7560, 2.1513, 3.0429, 3.1366], device='cuda:1'), covar=tensor([0.0296, 0.0730, 0.0539, 0.2006, 0.0837, 0.1037, 0.0668, 0.0791], device='cuda:1'), in_proj_covar=tensor([0.0139, 0.0140, 0.0154, 0.0141, 0.0134, 0.0123, 0.0133, 0.0149], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:1') 2023-04-29 11:48:35,722 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.5304, 1.5312, 2.0563, 2.4417, 2.5162, 2.8000, 1.6786, 2.7389], device='cuda:1'), covar=tensor([0.0134, 0.0433, 0.0245, 0.0222, 0.0200, 0.0141, 0.0383, 0.0104], device='cuda:1'), in_proj_covar=tensor([0.0156, 0.0166, 0.0148, 0.0154, 0.0163, 0.0120, 0.0167, 0.0111], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001], device='cuda:1') 2023-04-29 11:49:14,127 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.3081, 3.4421, 1.9811, 3.6373, 2.5249, 3.6491, 2.1846, 2.9180], device='cuda:1'), covar=tensor([0.0210, 0.0255, 0.1397, 0.0158, 0.0725, 0.0394, 0.1306, 0.0528], device='cuda:1'), in_proj_covar=tensor([0.0145, 0.0158, 0.0182, 0.0122, 0.0164, 0.0197, 0.0191, 0.0165], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0003, 0.0004, 0.0004, 0.0003], device='cuda:1') 2023-04-29 11:49:21,772 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=109935.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 11:49:45,365 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 11, batch 8450, loss[loss=0.1744, simple_loss=0.2721, pruned_loss=0.03831, over 16236.00 frames. ], tot_loss[loss=0.2015, simple_loss=0.2891, pruned_loss=0.05695, over 3073181.20 frames. ], batch size: 165, lr: 6.13e-03, grad_scale: 8.0 2023-04-29 11:49:52,354 INFO [optim.py:368] (1/8) 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 11:50:01,166 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=2.43 vs. limit=5.0 2023-04-29 11:50:10,729 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.57 vs. limit=2.0 2023-04-29 11:50:23,530 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=109974.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 11:51:00,423 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=109998.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 11:51:09,331 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.53 vs. limit=2.0 2023-04-29 11:51:09,681 INFO [train.py:904] (1/8) Epoch 11, batch 8500, loss[loss=0.1615, simple_loss=0.2422, pruned_loss=0.04042, over 11747.00 frames. ], tot_loss[loss=0.1968, simple_loss=0.2852, pruned_loss=0.05421, over 3074037.36 frames. ], batch size: 248, lr: 6.12e-03, grad_scale: 8.0 2023-04-29 11:51:31,233 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=110015.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 11:51:32,599 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.9584, 4.9991, 4.8345, 4.5257, 4.4636, 4.9175, 4.8219, 4.5928], device='cuda:1'), covar=tensor([0.0576, 0.0472, 0.0245, 0.0296, 0.0912, 0.0408, 0.0301, 0.0674], device='cuda:1'), in_proj_covar=tensor([0.0232, 0.0304, 0.0274, 0.0255, 0.0291, 0.0293, 0.0191, 0.0320], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-29 11:51:33,952 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=110017.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 11:51:37,157 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([1.6856, 2.7047, 2.3996, 3.8928, 2.7923, 4.0849, 1.4057, 3.0649], device='cuda:1'), covar=tensor([0.1376, 0.0663, 0.1141, 0.0134, 0.0149, 0.0334, 0.1649, 0.0630], device='cuda:1'), in_proj_covar=tensor([0.0153, 0.0157, 0.0178, 0.0140, 0.0196, 0.0206, 0.0179, 0.0178], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0003, 0.0004], device='cuda:1') 2023-04-29 11:51:41,404 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=110022.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 11:52:17,064 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 11, batch 8550, loss[loss=0.187, simple_loss=0.2673, pruned_loss=0.05333, over 11841.00 frames. ], tot_loss[loss=0.1937, simple_loss=0.282, pruned_loss=0.05274, over 3040076.80 frames. ], batch size: 247, lr: 6.12e-03, grad_scale: 8.0 2023-04-29 11:52:37,011 INFO [optim.py:368] (1/8) Clipping_scale=2.0, grad-norm quartiles 1.639e+02 2.340e+02 2.855e+02 3.457e+02 5.505e+02, threshold=5.709e+02, percent-clipped=2.0 2023-04-29 11:53:19,579 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=110078.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 11:53:42,241 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.5850, 2.6677, 1.6643, 2.7991, 2.0630, 2.8145, 1.8567, 2.4110], device='cuda:1'), covar=tensor([0.0220, 0.0325, 0.1320, 0.0230, 0.0641, 0.0537, 0.1338, 0.0562], device='cuda:1'), in_proj_covar=tensor([0.0144, 0.0157, 0.0182, 0.0122, 0.0162, 0.0195, 0.0190, 0.0164], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0003, 0.0004, 0.0004, 0.0003], device='cuda:1') 2023-04-29 11:54:07,122 INFO [train.py:904] (1/8) Epoch 11, batch 8600, loss[loss=0.1685, simple_loss=0.2688, pruned_loss=0.0341, over 16738.00 frames. ], tot_loss[loss=0.1934, simple_loss=0.2825, pruned_loss=0.0522, over 3025297.89 frames. ], batch size: 76, lr: 6.12e-03, grad_scale: 8.0 2023-04-29 11:54:14,058 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=110105.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 11:54:22,973 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=110110.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 11:54:34,984 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=110115.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 11:54:41,188 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.3599, 4.6591, 4.4682, 4.4289, 4.1076, 4.1243, 4.1577, 4.6961], device='cuda:1'), covar=tensor([0.0896, 0.0841, 0.0906, 0.0621, 0.0774, 0.1406, 0.1013, 0.0721], device='cuda:1'), in_proj_covar=tensor([0.0519, 0.0638, 0.0531, 0.0445, 0.0400, 0.0423, 0.0535, 0.0490], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:1') 2023-04-29 11:55:05,384 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.90 vs. limit=2.0 2023-04-29 11:55:18,599 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.6277, 2.7194, 1.7768, 2.8480, 2.1178, 2.8613, 2.0739, 2.4684], device='cuda:1'), covar=tensor([0.0230, 0.0327, 0.1201, 0.0206, 0.0664, 0.0425, 0.1192, 0.0559], device='cuda:1'), in_proj_covar=tensor([0.0144, 0.0156, 0.0181, 0.0121, 0.0162, 0.0195, 0.0190, 0.0164], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0003, 0.0004, 0.0004, 0.0003], device='cuda:1') 2023-04-29 11:55:43,444 INFO [train.py:904] (1/8) Epoch 11, batch 8650, loss[loss=0.1689, simple_loss=0.2672, pruned_loss=0.03529, over 16376.00 frames. ], tot_loss[loss=0.1913, simple_loss=0.2808, pruned_loss=0.05092, over 3014896.60 frames. ], batch size: 146, lr: 6.12e-03, grad_scale: 4.0 2023-04-29 11:55:53,865 INFO [optim.py:368] (1/8) 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 11:56:28,217 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=110171.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 11:56:39,712 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=110176.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 11:57:30,808 INFO [train.py:904] (1/8) Epoch 11, batch 8700, loss[loss=0.1745, simple_loss=0.28, pruned_loss=0.03451, over 15414.00 frames. ], tot_loss[loss=0.1881, simple_loss=0.2776, pruned_loss=0.04924, over 3031095.85 frames. ], batch size: 191, lr: 6.12e-03, grad_scale: 4.0 2023-04-29 11:58:34,095 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=110235.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 11:59:05,967 INFO [train.py:904] (1/8) Epoch 11, batch 8750, loss[loss=0.1899, simple_loss=0.2906, pruned_loss=0.04456, over 16432.00 frames. ], tot_loss[loss=0.1874, simple_loss=0.2773, pruned_loss=0.04874, over 3026300.55 frames. ], batch size: 68, lr: 6.12e-03, grad_scale: 4.0 2023-04-29 11:59:15,713 INFO [optim.py:368] (1/8) 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 12:00:22,506 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=110283.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 12:00:22,696 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 11, batch 8800, loss[loss=0.1706, simple_loss=0.2679, pruned_loss=0.03661, over 16749.00 frames. ], tot_loss[loss=0.1852, simple_loss=0.2758, pruned_loss=0.04733, over 3047910.18 frames. ], batch size: 83, lr: 6.12e-03, grad_scale: 8.0 2023-04-29 12:01:03,297 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=110303.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 12:01:27,539 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=110315.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 12:01:34,190 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=110318.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 12:01:40,157 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.0910, 2.6762, 2.6943, 1.8144, 2.8561, 2.9105, 2.4685, 2.4266], device='cuda:1'), covar=tensor([0.0659, 0.0206, 0.0187, 0.0969, 0.0086, 0.0163, 0.0418, 0.0416], device='cuda:1'), in_proj_covar=tensor([0.0139, 0.0098, 0.0085, 0.0137, 0.0067, 0.0099, 0.0118, 0.0125], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0003, 0.0002, 0.0004, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:1') 2023-04-29 12:02:13,843 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.8525, 4.8223, 4.5678, 4.1252, 4.6533, 1.7038, 4.5056, 4.4598], device='cuda:1'), covar=tensor([0.0048, 0.0049, 0.0132, 0.0209, 0.0064, 0.2352, 0.0082, 0.0152], device='cuda:1'), in_proj_covar=tensor([0.0125, 0.0111, 0.0159, 0.0147, 0.0130, 0.0178, 0.0146, 0.0145], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-29 12:02:29,299 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 11, batch 8850, loss[loss=0.1938, simple_loss=0.3018, pruned_loss=0.04284, over 16663.00 frames. ], tot_loss[loss=0.186, simple_loss=0.2788, pruned_loss=0.04662, over 3063849.84 frames. ], batch size: 76, lr: 6.11e-03, grad_scale: 8.0 2023-04-29 12:02:52,480 INFO [optim.py:368] (1/8) 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] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=110363.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 12:03:11,371 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=110364.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 12:03:17,118 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.3343, 3.2702, 2.6744, 2.0642, 2.1360, 2.1227, 3.3611, 2.9109], device='cuda:1'), covar=tensor([0.2871, 0.0809, 0.1627, 0.2394, 0.2186, 0.2058, 0.0525, 0.1187], device='cuda:1'), in_proj_covar=tensor([0.0297, 0.0249, 0.0275, 0.0268, 0.0265, 0.0216, 0.0258, 0.0282], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-29 12:03:29,776 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=110373.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 12:03:42,462 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=110379.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 12:04:18,870 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.7211, 2.1323, 2.2692, 4.3789, 2.0779, 2.5255, 2.2835, 2.3747], device='cuda:1'), covar=tensor([0.0830, 0.3550, 0.2304, 0.0339, 0.4250, 0.2442, 0.3216, 0.3483], device='cuda:1'), in_proj_covar=tensor([0.0350, 0.0382, 0.0323, 0.0310, 0.0404, 0.0433, 0.0344, 0.0446], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-29 12:04:25,795 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([1.9766, 1.7990, 1.5727, 1.4709, 1.9558, 1.5895, 1.6680, 1.9840], device='cuda:1'), covar=tensor([0.0118, 0.0232, 0.0329, 0.0349, 0.0168, 0.0256, 0.0134, 0.0165], device='cuda:1'), in_proj_covar=tensor([0.0141, 0.0200, 0.0196, 0.0195, 0.0198, 0.0199, 0.0197, 0.0185], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-29 12:04:27,935 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 11, batch 8900, loss[loss=0.1644, simple_loss=0.2615, pruned_loss=0.03371, over 16822.00 frames. ], tot_loss[loss=0.1861, simple_loss=0.28, pruned_loss=0.04617, over 3088789.91 frames. ], batch size: 76, lr: 6.11e-03, grad_scale: 8.0 2023-04-29 12:04:47,154 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.8117, 5.1371, 4.9350, 4.9120, 4.5722, 4.5824, 4.5655, 5.2063], device='cuda:1'), covar=tensor([0.0987, 0.0737, 0.0819, 0.0612, 0.0743, 0.0883, 0.0894, 0.0787], device='cuda:1'), in_proj_covar=tensor([0.0523, 0.0641, 0.0533, 0.0446, 0.0404, 0.0426, 0.0540, 0.0493], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:1') 2023-04-29 12:05:01,878 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.94 vs. limit=2.0 2023-04-29 12:06:36,590 INFO [train.py:904] (1/8) Epoch 11, batch 8950, loss[loss=0.1806, simple_loss=0.2769, pruned_loss=0.04216, over 16722.00 frames. ], tot_loss[loss=0.1863, simple_loss=0.2794, pruned_loss=0.04658, over 3094343.58 frames. ], batch size: 76, lr: 6.11e-03, grad_scale: 8.0 2023-04-29 12:06:45,507 INFO [optim.py:368] (1/8) 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 12:06:51,191 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=4.83 vs. limit=5.0 2023-04-29 12:07:07,584 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=110466.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 12:07:17,794 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=110471.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 12:08:26,524 INFO [train.py:904] (1/8) Epoch 11, batch 9000, loss[loss=0.1755, simple_loss=0.2633, pruned_loss=0.04383, over 16912.00 frames. ], tot_loss[loss=0.1824, simple_loss=0.2754, pruned_loss=0.04473, over 3107794.89 frames. ], batch size: 116, lr: 6.11e-03, grad_scale: 8.0 2023-04-29 12:08:26,525 INFO [train.py:929] (1/8) Computing validation loss 2023-04-29 12:08:36,934 INFO [train.py:938] (1/8) Epoch 11, validation: loss=0.1545, simple_loss=0.2586, pruned_loss=0.02523, over 944034.00 frames. 2023-04-29 12:08:36,935 INFO [train.py:939] (1/8) Maximum memory allocated so far is 17845MB 2023-04-29 12:09:22,952 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.2844, 1.9433, 2.0053, 3.9240, 1.9513, 2.3665, 2.0838, 2.1475], device='cuda:1'), covar=tensor([0.0920, 0.3425, 0.2509, 0.0396, 0.3846, 0.2354, 0.3296, 0.3255], device='cuda:1'), in_proj_covar=tensor([0.0349, 0.0380, 0.0323, 0.0309, 0.0402, 0.0432, 0.0344, 0.0443], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-29 12:09:54,404 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.50 vs. limit=2.0 2023-04-29 12:09:54,498 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.44 vs. limit=2.0 2023-04-29 12:10:21,906 INFO [train.py:904] (1/8) Epoch 11, batch 9050, loss[loss=0.1653, simple_loss=0.2562, pruned_loss=0.03724, over 16917.00 frames. ], tot_loss[loss=0.1829, simple_loss=0.2755, pruned_loss=0.04516, over 3092447.62 frames. ], batch size: 96, lr: 6.11e-03, grad_scale: 8.0 2023-04-29 12:10:28,906 INFO [optim.py:368] (1/8) 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 12:11:11,545 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.2939, 3.7126, 3.8624, 2.1319, 3.1889, 2.6271, 3.7874, 3.8088], device='cuda:1'), covar=tensor([0.0198, 0.0587, 0.0401, 0.1630, 0.0665, 0.0788, 0.0516, 0.0781], device='cuda:1'), in_proj_covar=tensor([0.0140, 0.0137, 0.0155, 0.0141, 0.0135, 0.0123, 0.0134, 0.0148], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:1') 2023-04-29 12:12:06,397 INFO [train.py:904] (1/8) Epoch 11, batch 9100, loss[loss=0.1612, simple_loss=0.2636, pruned_loss=0.02935, over 16765.00 frames. ], tot_loss[loss=0.1834, simple_loss=0.2751, pruned_loss=0.04579, over 3072409.27 frames. ], batch size: 83, lr: 6.11e-03, grad_scale: 8.0 2023-04-29 12:12:25,077 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=110611.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 12:13:16,391 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.28 vs. limit=2.0 2023-04-29 12:13:28,027 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.4245, 2.9479, 2.7427, 2.2412, 2.1841, 2.1608, 2.9945, 2.8504], device='cuda:1'), covar=tensor([0.2146, 0.0904, 0.1338, 0.2070, 0.2052, 0.1804, 0.0519, 0.1148], device='cuda:1'), in_proj_covar=tensor([0.0296, 0.0250, 0.0275, 0.0269, 0.0262, 0.0217, 0.0259, 0.0281], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-29 12:13:34,264 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.8469, 3.7674, 3.8862, 3.9952, 4.0759, 3.6208, 4.0531, 4.1286], device='cuda:1'), covar=tensor([0.1152, 0.0796, 0.1072, 0.0560, 0.0476, 0.1755, 0.0501, 0.0526], device='cuda:1'), in_proj_covar=tensor([0.0495, 0.0618, 0.0734, 0.0629, 0.0481, 0.0486, 0.0495, 0.0564], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-29 12:13:36,692 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 11, batch 9150, loss[loss=0.1926, simple_loss=0.2728, pruned_loss=0.05616, over 11816.00 frames. ], tot_loss[loss=0.1834, simple_loss=0.2754, pruned_loss=0.04568, over 3060879.09 frames. ], batch size: 250, lr: 6.11e-03, grad_scale: 8.0 2023-04-29 12:14:15,980 INFO [optim.py:368] (1/8) 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 12:14:23,373 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=110659.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 12:14:52,241 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=110672.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 12:14:53,801 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=110673.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 12:14:56,078 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=110674.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 12:15:46,180 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 11, batch 9200, loss[loss=0.1739, simple_loss=0.2596, pruned_loss=0.04411, over 16634.00 frames. ], tot_loss[loss=0.1798, simple_loss=0.2707, pruned_loss=0.04447, over 3070385.67 frames. ], batch size: 62, lr: 6.10e-03, grad_scale: 8.0 2023-04-29 12:16:24,194 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=110721.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 12:16:53,015 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=110736.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 12:17:07,737 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.23 vs. limit=2.0 2023-04-29 12:17:18,046 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 11, batch 9250, loss[loss=0.1762, simple_loss=0.2628, pruned_loss=0.04482, over 16288.00 frames. ], tot_loss[loss=0.1799, simple_loss=0.2706, pruned_loss=0.04454, over 3074699.02 frames. ], batch size: 35, lr: 6.10e-03, grad_scale: 8.0 2023-04-29 12:17:32,955 INFO [optim.py:368] (1/8) 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 12:17:53,021 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=110766.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 12:18:05,065 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=110771.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 12:19:06,504 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=110797.0, num_to_drop=1, layers_to_drop={1} 2023-04-29 12:19:16,917 INFO [train.py:904] (1/8) Epoch 11, batch 9300, loss[loss=0.1646, simple_loss=0.2504, pruned_loss=0.0394, over 16434.00 frames. ], tot_loss[loss=0.1787, simple_loss=0.2695, pruned_loss=0.04396, over 3092038.27 frames. ], batch size: 68, lr: 6.10e-03, grad_scale: 8.0 2023-04-29 12:19:17,592 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.3552, 3.3318, 3.4268, 3.5156, 3.5554, 3.2299, 3.4926, 3.6003], device='cuda:1'), covar=tensor([0.1266, 0.0912, 0.1188, 0.0701, 0.0657, 0.2784, 0.1115, 0.0687], device='cuda:1'), in_proj_covar=tensor([0.0490, 0.0610, 0.0729, 0.0619, 0.0474, 0.0479, 0.0490, 0.0555], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:1') 2023-04-29 12:19:45,570 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=110814.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 12:19:59,650 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=110819.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 12:21:01,157 INFO [train.py:904] (1/8) Epoch 11, batch 9350, loss[loss=0.1936, simple_loss=0.2801, pruned_loss=0.05353, over 16234.00 frames. ], tot_loss[loss=0.1783, simple_loss=0.2693, pruned_loss=0.04369, over 3086520.42 frames. ], batch size: 165, lr: 6.10e-03, grad_scale: 8.0 2023-04-29 12:21:10,062 INFO [optim.py:368] (1/8) 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 12:21:35,932 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=110868.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 12:22:41,697 INFO [train.py:904] (1/8) Epoch 11, batch 9400, loss[loss=0.1877, simple_loss=0.2856, pruned_loss=0.04491, over 15384.00 frames. ], tot_loss[loss=0.1785, simple_loss=0.2696, pruned_loss=0.0437, over 3078078.12 frames. ], batch size: 192, lr: 6.10e-03, grad_scale: 8.0 2023-04-29 12:23:36,564 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=110929.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 12:23:57,277 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=110939.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 12:24:21,276 INFO [train.py:904] (1/8) Epoch 11, batch 9450, loss[loss=0.1837, simple_loss=0.2755, pruned_loss=0.04602, over 16356.00 frames. ], tot_loss[loss=0.1797, simple_loss=0.2713, pruned_loss=0.04402, over 3077491.08 frames. ], batch size: 146, lr: 6.10e-03, grad_scale: 8.0 2023-04-29 12:24:27,348 INFO [optim.py:368] (1/8) 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 12:24:33,039 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=110958.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 12:24:34,606 INFO [zipformer.py:625] (1/8) 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] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=110967.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 12:25:06,689 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=110974.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 12:25:33,498 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=110987.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 12:26:00,932 INFO [train.py:904] (1/8) Epoch 11, batch 9500, loss[loss=0.1778, simple_loss=0.2683, pruned_loss=0.04361, over 16768.00 frames. ], tot_loss[loss=0.1788, simple_loss=0.2705, pruned_loss=0.04359, over 3079134.22 frames. ], batch size: 83, lr: 6.10e-03, grad_scale: 8.0 2023-04-29 12:26:15,363 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=111007.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 12:26:37,618 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=111019.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 12:26:43,501 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=111022.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 12:26:47,211 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([5.4190, 5.7469, 5.5421, 5.5757, 5.1807, 5.0694, 5.2216, 5.8500], device='cuda:1'), covar=tensor([0.1065, 0.0838, 0.0884, 0.0601, 0.0754, 0.0659, 0.0881, 0.0784], device='cuda:1'), in_proj_covar=tensor([0.0514, 0.0637, 0.0521, 0.0441, 0.0399, 0.0417, 0.0530, 0.0483], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:1') 2023-04-29 12:27:46,818 INFO [train.py:904] (1/8) Epoch 11, batch 9550, loss[loss=0.1961, simple_loss=0.2942, pruned_loss=0.04899, over 16898.00 frames. ], tot_loss[loss=0.1784, simple_loss=0.2696, pruned_loss=0.0436, over 3078048.61 frames. ], batch size: 116, lr: 6.09e-03, grad_scale: 8.0 2023-04-29 12:27:55,298 INFO [optim.py:368] (1/8) 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 12:29:10,485 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=111092.0, num_to_drop=1, layers_to_drop={1} 2023-04-29 12:29:26,769 INFO [train.py:904] (1/8) Epoch 11, batch 9600, loss[loss=0.2155, simple_loss=0.3102, pruned_loss=0.06038, over 16801.00 frames. ], tot_loss[loss=0.1797, simple_loss=0.2708, pruned_loss=0.04433, over 3044408.62 frames. ], batch size: 116, lr: 6.09e-03, grad_scale: 8.0 2023-04-29 12:29:43,162 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.1757, 1.4700, 1.6686, 2.1686, 2.1322, 2.2711, 1.5313, 2.3067], device='cuda:1'), covar=tensor([0.0186, 0.0387, 0.0253, 0.0241, 0.0261, 0.0158, 0.0378, 0.0084], device='cuda:1'), in_proj_covar=tensor([0.0154, 0.0166, 0.0147, 0.0151, 0.0161, 0.0118, 0.0166, 0.0108], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001], device='cuda:1') 2023-04-29 12:30:14,564 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.3884, 2.0577, 2.0958, 3.9011, 1.9558, 2.4631, 2.1737, 2.1500], device='cuda:1'), covar=tensor([0.0876, 0.3289, 0.2368, 0.0381, 0.3956, 0.2236, 0.3060, 0.3318], device='cuda:1'), in_proj_covar=tensor([0.0347, 0.0374, 0.0320, 0.0304, 0.0399, 0.0423, 0.0339, 0.0436], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-29 12:31:14,963 INFO [train.py:904] (1/8) Epoch 11, batch 9650, loss[loss=0.1976, simple_loss=0.2906, pruned_loss=0.05233, over 16852.00 frames. ], tot_loss[loss=0.1811, simple_loss=0.2731, pruned_loss=0.04459, over 3066444.17 frames. ], batch size: 124, lr: 6.09e-03, grad_scale: 8.0 2023-04-29 12:31:24,135 INFO [optim.py:368] (1/8) Clipping_scale=2.0, grad-norm quartiles 1.791e+02 2.367e+02 2.756e+02 3.328e+02 5.495e+02, threshold=5.512e+02, percent-clipped=0.0 2023-04-29 12:31:51,899 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.40 vs. limit=2.0 2023-04-29 12:33:03,902 INFO [train.py:904] (1/8) Epoch 11, batch 9700, loss[loss=0.1663, simple_loss=0.2616, pruned_loss=0.03545, over 16634.00 frames. ], tot_loss[loss=0.1809, simple_loss=0.2728, pruned_loss=0.04447, over 3073752.38 frames. ], batch size: 62, lr: 6.09e-03, grad_scale: 8.0 2023-04-29 12:33:49,082 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=111224.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 12:33:51,782 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=111225.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 12:34:46,298 INFO [train.py:904] (1/8) Epoch 11, batch 9750, loss[loss=0.1873, simple_loss=0.284, pruned_loss=0.0453, over 16750.00 frames. ], tot_loss[loss=0.1807, simple_loss=0.2718, pruned_loss=0.04481, over 3060353.35 frames. ], batch size: 124, lr: 6.09e-03, grad_scale: 8.0 2023-04-29 12:34:51,500 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.1229, 4.1059, 4.5446, 4.5110, 4.5375, 4.2765, 4.2774, 4.1245], device='cuda:1'), covar=tensor([0.0301, 0.0560, 0.0435, 0.0537, 0.0515, 0.0360, 0.0802, 0.0457], device='cuda:1'), in_proj_covar=tensor([0.0303, 0.0314, 0.0319, 0.0298, 0.0358, 0.0335, 0.0424, 0.0271], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0002], device='cuda:1') 2023-04-29 12:34:53,729 INFO [optim.py:368] (1/8) 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 12:35:17,246 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=111267.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 12:35:57,982 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=111286.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 12:36:25,538 INFO [train.py:904] (1/8) Epoch 11, batch 9800, loss[loss=0.1667, simple_loss=0.2682, pruned_loss=0.03261, over 16436.00 frames. ], tot_loss[loss=0.1799, simple_loss=0.2721, pruned_loss=0.04392, over 3065044.42 frames. ], batch size: 68, lr: 6.09e-03, grad_scale: 8.0 2023-04-29 12:36:49,124 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=111314.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 12:36:50,549 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=111315.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 12:37:34,963 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.3514, 4.3554, 4.1255, 3.7173, 4.2051, 1.6099, 3.9926, 3.9573], device='cuda:1'), covar=tensor([0.0071, 0.0062, 0.0151, 0.0209, 0.0076, 0.2262, 0.0097, 0.0165], device='cuda:1'), in_proj_covar=tensor([0.0124, 0.0109, 0.0156, 0.0142, 0.0128, 0.0176, 0.0144, 0.0142], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-29 12:38:11,446 INFO [train.py:904] (1/8) Epoch 11, batch 9850, loss[loss=0.1698, simple_loss=0.2577, pruned_loss=0.04091, over 12346.00 frames. ], tot_loss[loss=0.1803, simple_loss=0.2734, pruned_loss=0.04365, over 3068164.49 frames. ], batch size: 248, lr: 6.09e-03, grad_scale: 8.0 2023-04-29 12:38:20,190 INFO [optim.py:368] (1/8) 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 12:38:23,118 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.5339, 3.7341, 2.1628, 4.1934, 2.8000, 4.0727, 2.3087, 2.8768], device='cuda:1'), covar=tensor([0.0248, 0.0324, 0.1660, 0.0123, 0.0828, 0.0443, 0.1684, 0.0748], device='cuda:1'), in_proj_covar=tensor([0.0146, 0.0157, 0.0185, 0.0121, 0.0164, 0.0196, 0.0195, 0.0167], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0003, 0.0004, 0.0004, 0.0003], device='cuda:1') 2023-04-29 12:39:42,428 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=111392.0, num_to_drop=1, layers_to_drop={0} 2023-04-29 12:40:02,804 INFO [train.py:904] (1/8) Epoch 11, batch 9900, loss[loss=0.1869, simple_loss=0.287, pruned_loss=0.04346, over 16709.00 frames. ], tot_loss[loss=0.1805, simple_loss=0.2738, pruned_loss=0.04359, over 3077822.09 frames. ], batch size: 134, lr: 6.09e-03, grad_scale: 8.0 2023-04-29 12:41:34,259 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 11, batch 9950, loss[loss=0.1799, simple_loss=0.2757, pruned_loss=0.04201, over 16684.00 frames. ], tot_loss[loss=0.1816, simple_loss=0.2752, pruned_loss=0.04403, over 3056112.52 frames. ], batch size: 62, lr: 6.08e-03, grad_scale: 8.0 2023-04-29 12:42:11,464 INFO [optim.py:368] (1/8) 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 12:42:40,503 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=111468.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 12:43:28,186 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.0292, 3.9203, 4.3755, 4.3756, 4.3902, 4.1283, 4.1187, 4.0643], device='cuda:1'), covar=tensor([0.0265, 0.0646, 0.0421, 0.0437, 0.0385, 0.0366, 0.0814, 0.0364], device='cuda:1'), in_proj_covar=tensor([0.0297, 0.0309, 0.0313, 0.0294, 0.0354, 0.0330, 0.0419, 0.0267], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0002], device='cuda:1') 2023-04-29 12:44:02,030 INFO [train.py:904] (1/8) Epoch 11, batch 10000, loss[loss=0.2148, simple_loss=0.2899, pruned_loss=0.06988, over 12606.00 frames. ], tot_loss[loss=0.1802, simple_loss=0.2736, pruned_loss=0.04345, over 3070934.22 frames. ], batch size: 250, lr: 6.08e-03, grad_scale: 8.0 2023-04-29 12:44:16,674 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.8803, 3.2087, 3.4045, 1.8653, 3.0082, 2.1902, 3.4122, 3.3130], device='cuda:1'), covar=tensor([0.0275, 0.0708, 0.0480, 0.1792, 0.0624, 0.0927, 0.0642, 0.0908], device='cuda:1'), in_proj_covar=tensor([0.0137, 0.0132, 0.0152, 0.0139, 0.0132, 0.0121, 0.0131, 0.0144], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:1') 2023-04-29 12:44:45,213 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=111524.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 12:44:55,140 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=111529.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 12:45:10,153 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.3142, 4.6378, 4.4753, 4.4829, 4.1453, 4.1046, 4.1661, 4.6610], device='cuda:1'), covar=tensor([0.1061, 0.0862, 0.0962, 0.0626, 0.0797, 0.1316, 0.0967, 0.0940], device='cuda:1'), in_proj_covar=tensor([0.0505, 0.0626, 0.0508, 0.0434, 0.0392, 0.0411, 0.0520, 0.0478], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:1') 2023-04-29 12:45:42,134 INFO [train.py:904] (1/8) Epoch 11, batch 10050, loss[loss=0.212, simple_loss=0.2944, pruned_loss=0.06479, over 12393.00 frames. ], tot_loss[loss=0.1805, simple_loss=0.2738, pruned_loss=0.04357, over 3059445.77 frames. ], batch size: 248, lr: 6.08e-03, grad_scale: 8.0 2023-04-29 12:45:50,236 INFO [optim.py:368] (1/8) Clipping_scale=2.0, grad-norm quartiles 1.518e+02 2.160e+02 2.693e+02 3.599e+02 6.171e+02, threshold=5.385e+02, percent-clipped=1.0 2023-04-29 12:46:06,609 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=4.75 vs. limit=5.0 2023-04-29 12:46:22,295 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=111572.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 12:46:38,804 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=111581.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 12:47:14,825 INFO [train.py:904] (1/8) Epoch 11, batch 10100, loss[loss=0.1778, simple_loss=0.2734, pruned_loss=0.04113, over 16486.00 frames. ], tot_loss[loss=0.1815, simple_loss=0.2748, pruned_loss=0.04415, over 3041598.04 frames. ], batch size: 68, lr: 6.08e-03, grad_scale: 8.0 2023-04-29 12:47:35,953 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=111614.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 12:48:58,023 INFO [train.py:904] (1/8) Epoch 12, batch 0, loss[loss=0.1939, simple_loss=0.2768, pruned_loss=0.05551, over 17017.00 frames. ], tot_loss[loss=0.1939, simple_loss=0.2768, pruned_loss=0.05551, over 17017.00 frames. ], batch size: 41, lr: 5.82e-03, grad_scale: 8.0 2023-04-29 12:48:58,023 INFO [train.py:929] (1/8) Computing validation loss 2023-04-29 12:49:05,313 INFO [train.py:938] (1/8) Epoch 12, validation: loss=0.1543, simple_loss=0.2577, pruned_loss=0.0254, over 944034.00 frames. 2023-04-29 12:49:05,314 INFO [train.py:939] (1/8) Maximum memory allocated so far is 17845MB 2023-04-29 12:49:12,536 INFO [optim.py:368] (1/8) 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] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=111662.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 12:49:48,498 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.2842, 3.3385, 3.5440, 1.8462, 3.6557, 3.6273, 2.8691, 2.6700], device='cuda:1'), covar=tensor([0.0816, 0.0158, 0.0105, 0.1077, 0.0064, 0.0140, 0.0367, 0.0409], device='cuda:1'), in_proj_covar=tensor([0.0138, 0.0096, 0.0082, 0.0136, 0.0066, 0.0097, 0.0116, 0.0123], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0003, 0.0002, 0.0004, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:1') 2023-04-29 12:50:15,966 INFO [train.py:904] (1/8) Epoch 12, batch 50, loss[loss=0.1701, simple_loss=0.2513, pruned_loss=0.04451, over 16859.00 frames. ], tot_loss[loss=0.2077, simple_loss=0.2865, pruned_loss=0.0645, over 746633.43 frames. ], batch size: 42, lr: 5.82e-03, grad_scale: 2.0 2023-04-29 12:50:44,657 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.1272, 4.2335, 4.6249, 2.2213, 4.7689, 4.7679, 3.5212, 3.7635], device='cuda:1'), covar=tensor([0.0656, 0.0163, 0.0144, 0.1011, 0.0048, 0.0096, 0.0314, 0.0306], device='cuda:1'), in_proj_covar=tensor([0.0140, 0.0097, 0.0084, 0.0138, 0.0067, 0.0099, 0.0118, 0.0124], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0003, 0.0002, 0.0004, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:1') 2023-04-29 12:50:58,435 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.5418, 4.4163, 4.3930, 4.1851, 4.1101, 4.4250, 4.3265, 4.1910], device='cuda:1'), covar=tensor([0.0591, 0.0540, 0.0293, 0.0262, 0.0855, 0.0478, 0.0462, 0.0607], device='cuda:1'), in_proj_covar=tensor([0.0233, 0.0301, 0.0275, 0.0254, 0.0294, 0.0296, 0.0192, 0.0318], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-29 12:51:25,697 INFO [train.py:904] (1/8) Epoch 12, batch 100, loss[loss=0.2532, simple_loss=0.3383, pruned_loss=0.08406, over 12614.00 frames. ], tot_loss[loss=0.2002, simple_loss=0.2808, pruned_loss=0.05981, over 1314804.84 frames. ], batch size: 246, lr: 5.82e-03, grad_scale: 2.0 2023-04-29 12:51:34,341 INFO [optim.py:368] (1/8) 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 12:52:07,147 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.9334, 1.9004, 2.2877, 2.9034, 2.6884, 2.9864, 2.0676, 3.1211], device='cuda:1'), covar=tensor([0.0156, 0.0363, 0.0252, 0.0193, 0.0201, 0.0168, 0.0365, 0.0106], device='cuda:1'), in_proj_covar=tensor([0.0159, 0.0171, 0.0152, 0.0156, 0.0166, 0.0122, 0.0170, 0.0112], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001], device='cuda:1') 2023-04-29 12:52:18,934 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 12, batch 150, loss[loss=0.184, simple_loss=0.2576, pruned_loss=0.05521, over 16792.00 frames. ], tot_loss[loss=0.1959, simple_loss=0.2775, pruned_loss=0.05718, over 1758555.75 frames. ], batch size: 83, lr: 5.82e-03, grad_scale: 2.0 2023-04-29 12:53:02,368 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=111824.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 12:53:30,982 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.9374, 1.8437, 2.3373, 2.9093, 2.6789, 2.9135, 2.1711, 3.1125], device='cuda:1'), covar=tensor([0.0160, 0.0375, 0.0263, 0.0218, 0.0230, 0.0181, 0.0339, 0.0123], device='cuda:1'), in_proj_covar=tensor([0.0158, 0.0170, 0.0151, 0.0157, 0.0166, 0.0122, 0.0169, 0.0112], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001], device='cuda:1') 2023-04-29 12:53:40,720 INFO [train.py:904] (1/8) Epoch 12, batch 200, loss[loss=0.202, simple_loss=0.2832, pruned_loss=0.06038, over 15882.00 frames. ], tot_loss[loss=0.1962, simple_loss=0.2781, pruned_loss=0.05719, over 2106519.07 frames. ], batch size: 35, lr: 5.82e-03, grad_scale: 2.0 2023-04-29 12:53:41,289 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=111852.0, num_to_drop=1, layers_to_drop={1} 2023-04-29 12:53:50,232 INFO [optim.py:368] (1/8) 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 12:54:21,409 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 12, batch 250, loss[loss=0.1965, simple_loss=0.2835, pruned_loss=0.05477, over 16699.00 frames. ], tot_loss[loss=0.1959, simple_loss=0.2767, pruned_loss=0.05754, over 2374898.30 frames. ], batch size: 57, lr: 5.82e-03, grad_scale: 2.0 2023-04-29 12:55:26,515 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=111929.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 12:55:32,026 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([1.7808, 1.3339, 1.5473, 1.7055, 1.8074, 1.9642, 1.5674, 1.7451], device='cuda:1'), covar=tensor([0.0190, 0.0272, 0.0141, 0.0214, 0.0178, 0.0127, 0.0276, 0.0078], device='cuda:1'), in_proj_covar=tensor([0.0160, 0.0171, 0.0153, 0.0159, 0.0168, 0.0123, 0.0170, 0.0113], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001], device='cuda:1') 2023-04-29 12:55:57,751 INFO [train.py:904] (1/8) Epoch 12, batch 300, loss[loss=0.1857, simple_loss=0.2652, pruned_loss=0.05311, over 16429.00 frames. ], tot_loss[loss=0.1924, simple_loss=0.2732, pruned_loss=0.05581, over 2576343.74 frames. ], batch size: 146, lr: 5.82e-03, grad_scale: 1.0 2023-04-29 12:56:09,472 INFO [optim.py:368] (1/8) 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 12:57:10,668 INFO [train.py:904] (1/8) Epoch 12, batch 350, loss[loss=0.1767, simple_loss=0.2475, pruned_loss=0.05296, over 16844.00 frames. ], tot_loss[loss=0.1885, simple_loss=0.2693, pruned_loss=0.05379, over 2731674.02 frames. ], batch size: 96, lr: 5.81e-03, grad_scale: 1.0 2023-04-29 12:58:17,741 INFO [train.py:904] (1/8) Epoch 12, batch 400, loss[loss=0.1693, simple_loss=0.2519, pruned_loss=0.04333, over 16514.00 frames. ], tot_loss[loss=0.1868, simple_loss=0.2683, pruned_loss=0.05263, over 2870200.80 frames. ], batch size: 68, lr: 5.81e-03, grad_scale: 2.0 2023-04-29 12:58:27,674 INFO [optim.py:368] (1/8) 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 12:58:36,941 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=112066.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 12:58:55,838 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=4.96 vs. limit=5.0 2023-04-29 12:59:26,055 INFO [train.py:904] (1/8) Epoch 12, batch 450, loss[loss=0.1689, simple_loss=0.2668, pruned_loss=0.03549, over 17027.00 frames. ], tot_loss[loss=0.1839, simple_loss=0.2653, pruned_loss=0.05126, over 2974648.88 frames. ], batch size: 50, lr: 5.81e-03, grad_scale: 2.0 2023-04-29 12:59:55,910 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=112124.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 13:00:00,099 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=112127.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 13:00:27,068 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=112147.0, num_to_drop=1, layers_to_drop={3} 2023-04-29 13:00:33,800 INFO [train.py:904] (1/8) Epoch 12, batch 500, loss[loss=0.1988, simple_loss=0.2695, pruned_loss=0.06403, over 16806.00 frames. ], tot_loss[loss=0.1829, simple_loss=0.2647, pruned_loss=0.0506, over 3061915.88 frames. ], batch size: 102, lr: 5.81e-03, grad_scale: 2.0 2023-04-29 13:00:45,214 INFO [optim.py:368] (1/8) 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 13:01:02,758 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=112172.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 13:01:41,406 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.0032, 4.4794, 4.5471, 3.1249, 3.8313, 4.4808, 4.0474, 2.8206], device='cuda:1'), covar=tensor([0.0351, 0.0042, 0.0029, 0.0285, 0.0072, 0.0070, 0.0057, 0.0302], device='cuda:1'), in_proj_covar=tensor([0.0128, 0.0070, 0.0069, 0.0125, 0.0078, 0.0088, 0.0077, 0.0120], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:1') 2023-04-29 13:01:44,736 INFO [train.py:904] (1/8) Epoch 12, batch 550, loss[loss=0.1778, simple_loss=0.2649, pruned_loss=0.04534, over 17174.00 frames. ], tot_loss[loss=0.1824, simple_loss=0.2641, pruned_loss=0.05039, over 3123794.96 frames. ], batch size: 46, lr: 5.81e-03, grad_scale: 2.0 2023-04-29 13:02:16,349 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.96 vs. limit=2.0 2023-04-29 13:02:34,163 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([5.6975, 6.1174, 5.7660, 5.8318, 5.3462, 5.3069, 5.4943, 6.1662], device='cuda:1'), covar=tensor([0.1289, 0.0883, 0.1193, 0.0735, 0.0938, 0.0671, 0.1072, 0.1039], device='cuda:1'), in_proj_covar=tensor([0.0560, 0.0695, 0.0569, 0.0484, 0.0438, 0.0447, 0.0581, 0.0533], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:1') 2023-04-29 13:02:55,015 INFO [train.py:904] (1/8) Epoch 12, batch 600, loss[loss=0.1799, simple_loss=0.2553, pruned_loss=0.05222, over 16380.00 frames. ], tot_loss[loss=0.1824, simple_loss=0.2642, pruned_loss=0.05031, over 3169765.20 frames. ], batch size: 165, lr: 5.81e-03, grad_scale: 2.0 2023-04-29 13:02:57,007 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.28 vs. limit=2.0 2023-04-29 13:03:06,853 INFO [optim.py:368] (1/8) Clipping_scale=2.0, grad-norm quartiles 1.658e+02 2.365e+02 2.773e+02 3.420e+02 1.272e+03, threshold=5.547e+02, percent-clipped=1.0 2023-04-29 13:04:04,841 INFO [train.py:904] (1/8) Epoch 12, batch 650, loss[loss=0.1745, simple_loss=0.2668, pruned_loss=0.04113, over 17041.00 frames. ], tot_loss[loss=0.1814, simple_loss=0.2635, pruned_loss=0.0496, over 3208614.60 frames. ], batch size: 53, lr: 5.81e-03, grad_scale: 2.0 2023-04-29 13:04:26,251 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.9644, 3.9003, 4.4032, 1.8732, 4.6739, 4.6366, 3.2104, 3.4780], device='cuda:1'), covar=tensor([0.0658, 0.0203, 0.0158, 0.1151, 0.0045, 0.0115, 0.0397, 0.0349], device='cuda:1'), in_proj_covar=tensor([0.0141, 0.0099, 0.0086, 0.0139, 0.0069, 0.0104, 0.0120, 0.0126], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0003, 0.0002, 0.0004, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:1') 2023-04-29 13:04:50,482 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=4.88 vs. limit=5.0 2023-04-29 13:05:14,225 INFO [train.py:904] (1/8) Epoch 12, batch 700, loss[loss=0.1637, simple_loss=0.2427, pruned_loss=0.04241, over 15970.00 frames. ], tot_loss[loss=0.1815, simple_loss=0.2638, pruned_loss=0.04958, over 3227228.84 frames. ], batch size: 35, lr: 5.81e-03, grad_scale: 2.0 2023-04-29 13:05:14,615 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([5.2582, 5.7373, 5.9297, 5.6480, 5.6984, 6.2648, 5.8553, 5.5591], device='cuda:1'), covar=tensor([0.0803, 0.1881, 0.1898, 0.2005, 0.2905, 0.0970, 0.1257, 0.2455], device='cuda:1'), in_proj_covar=tensor([0.0354, 0.0507, 0.0555, 0.0439, 0.0586, 0.0584, 0.0435, 0.0587], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-04-29 13:05:26,018 INFO [optim.py:368] (1/8) 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 13:06:06,999 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 12, batch 750, loss[loss=0.2223, simple_loss=0.2883, pruned_loss=0.07817, over 16763.00 frames. ], tot_loss[loss=0.1823, simple_loss=0.264, pruned_loss=0.05029, over 3241365.56 frames. ], batch size: 124, lr: 5.80e-03, grad_scale: 2.0 2023-04-29 13:06:51,814 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=112422.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 13:07:19,585 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.6443, 2.2554, 2.4435, 4.4150, 2.2921, 2.7023, 2.3980, 2.4590], device='cuda:1'), covar=tensor([0.0957, 0.3293, 0.2289, 0.0393, 0.3623, 0.2198, 0.3016, 0.3084], device='cuda:1'), in_proj_covar=tensor([0.0366, 0.0394, 0.0333, 0.0322, 0.0415, 0.0449, 0.0357, 0.0461], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-29 13:07:29,124 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=112447.0, num_to_drop=1, layers_to_drop={0} 2023-04-29 13:07:32,728 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 12, batch 800, loss[loss=0.1804, simple_loss=0.2652, pruned_loss=0.04783, over 17109.00 frames. ], tot_loss[loss=0.1808, simple_loss=0.2632, pruned_loss=0.0492, over 3264467.64 frames. ], batch size: 53, lr: 5.80e-03, grad_scale: 4.0 2023-04-29 13:07:45,051 INFO [optim.py:368] (1/8) 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] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=112495.0, num_to_drop=1, layers_to_drop={0} 2023-04-29 13:08:42,805 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.89 vs. limit=2.0 2023-04-29 13:08:42,884 INFO [train.py:904] (1/8) Epoch 12, batch 850, loss[loss=0.1568, simple_loss=0.2517, pruned_loss=0.03099, over 17120.00 frames. ], tot_loss[loss=0.1806, simple_loss=0.2632, pruned_loss=0.049, over 3271681.08 frames. ], batch size: 47, lr: 5.80e-03, grad_scale: 4.0 2023-04-29 13:08:58,634 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([5.2977, 5.0288, 5.2550, 5.5233, 5.6702, 4.9241, 5.5997, 5.5938], device='cuda:1'), covar=tensor([0.1415, 0.1012, 0.1445, 0.0536, 0.0400, 0.0722, 0.0479, 0.0527], device='cuda:1'), in_proj_covar=tensor([0.0560, 0.0693, 0.0841, 0.0703, 0.0534, 0.0540, 0.0553, 0.0635], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-29 13:09:26,268 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=3.74 vs. limit=5.0 2023-04-29 13:09:51,999 INFO [train.py:904] (1/8) Epoch 12, batch 900, loss[loss=0.1851, simple_loss=0.2724, pruned_loss=0.04892, over 17108.00 frames. ], tot_loss[loss=0.1792, simple_loss=0.2621, pruned_loss=0.04809, over 3290052.02 frames. ], batch size: 53, lr: 5.80e-03, grad_scale: 4.0 2023-04-29 13:10:02,361 INFO [optim.py:368] (1/8) 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 13:10:05,175 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=112562.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 13:10:26,085 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=112577.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 13:10:35,118 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.9119, 4.5828, 3.4050, 2.3673, 2.9148, 2.6700, 4.7914, 3.8889], device='cuda:1'), covar=tensor([0.2608, 0.0507, 0.1452, 0.2330, 0.2579, 0.1695, 0.0309, 0.1061], device='cuda:1'), in_proj_covar=tensor([0.0303, 0.0259, 0.0284, 0.0277, 0.0275, 0.0223, 0.0266, 0.0298], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-29 13:10:59,465 INFO [train.py:904] (1/8) Epoch 12, batch 950, loss[loss=0.187, simple_loss=0.2536, pruned_loss=0.06025, over 16552.00 frames. ], tot_loss[loss=0.18, simple_loss=0.2621, pruned_loss=0.04888, over 3289671.58 frames. ], batch size: 146, lr: 5.80e-03, grad_scale: 4.0 2023-04-29 13:11:21,282 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=112617.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 13:11:28,362 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=112623.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 13:11:50,145 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=112638.0, num_to_drop=1, layers_to_drop={1} 2023-04-29 13:12:07,106 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.0791, 3.0919, 1.8038, 3.2269, 2.3803, 3.2739, 2.1049, 2.6338], device='cuda:1'), covar=tensor([0.0266, 0.0368, 0.1600, 0.0291, 0.0742, 0.0597, 0.1286, 0.0649], device='cuda:1'), in_proj_covar=tensor([0.0153, 0.0166, 0.0189, 0.0134, 0.0167, 0.0208, 0.0198, 0.0171], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0003, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-04-29 13:12:07,853 INFO [train.py:904] (1/8) Epoch 12, batch 1000, loss[loss=0.1872, simple_loss=0.2606, pruned_loss=0.05693, over 15518.00 frames. ], tot_loss[loss=0.1795, simple_loss=0.2612, pruned_loss=0.04888, over 3295824.28 frames. ], batch size: 190, lr: 5.80e-03, grad_scale: 4.0 2023-04-29 13:12:18,367 INFO [optim.py:368] (1/8) 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 13:12:43,254 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=112678.0, num_to_drop=1, layers_to_drop={3} 2023-04-29 13:12:49,716 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.0136, 4.3077, 3.1640, 2.3828, 2.8958, 2.5020, 4.6051, 3.8186], device='cuda:1'), covar=tensor([0.2268, 0.0565, 0.1465, 0.2130, 0.2389, 0.1720, 0.0382, 0.0988], device='cuda:1'), in_proj_covar=tensor([0.0302, 0.0258, 0.0283, 0.0276, 0.0275, 0.0223, 0.0266, 0.0298], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-29 13:13:15,633 INFO [train.py:904] (1/8) Epoch 12, batch 1050, loss[loss=0.1771, simple_loss=0.2598, pruned_loss=0.04719, over 17065.00 frames. ], tot_loss[loss=0.1794, simple_loss=0.2607, pruned_loss=0.04902, over 3301998.09 frames. ], batch size: 50, lr: 5.80e-03, grad_scale: 4.0 2023-04-29 13:13:32,582 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.3680, 3.4638, 3.8113, 2.7396, 3.4428, 3.8017, 3.5618, 2.2241], device='cuda:1'), covar=tensor([0.0415, 0.0173, 0.0041, 0.0285, 0.0079, 0.0079, 0.0073, 0.0377], device='cuda:1'), in_proj_covar=tensor([0.0128, 0.0070, 0.0069, 0.0125, 0.0079, 0.0089, 0.0078, 0.0121], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:1') 2023-04-29 13:13:35,450 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.42 vs. limit=2.0 2023-04-29 13:13:36,046 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.1473, 3.9537, 4.4111, 1.9513, 4.6255, 4.6687, 3.2555, 3.7080], device='cuda:1'), covar=tensor([0.0614, 0.0206, 0.0198, 0.1183, 0.0059, 0.0108, 0.0375, 0.0288], device='cuda:1'), in_proj_covar=tensor([0.0144, 0.0101, 0.0089, 0.0141, 0.0070, 0.0106, 0.0121, 0.0127], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:1') 2023-04-29 13:13:42,912 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=112722.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 13:14:14,517 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=112745.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 13:14:22,679 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.7311, 4.0899, 4.2806, 3.2185, 3.6674, 4.1672, 3.8128, 2.5079], device='cuda:1'), covar=tensor([0.0376, 0.0058, 0.0034, 0.0242, 0.0086, 0.0077, 0.0065, 0.0353], device='cuda:1'), in_proj_covar=tensor([0.0128, 0.0070, 0.0069, 0.0125, 0.0079, 0.0089, 0.0078, 0.0121], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:1') 2023-04-29 13:14:23,335 INFO [train.py:904] (1/8) Epoch 12, batch 1100, loss[loss=0.1845, simple_loss=0.2704, pruned_loss=0.04934, over 16644.00 frames. ], tot_loss[loss=0.1789, simple_loss=0.2601, pruned_loss=0.0488, over 3307226.83 frames. ], batch size: 57, lr: 5.79e-03, grad_scale: 4.0 2023-04-29 13:14:34,069 INFO [optim.py:368] (1/8) 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] (1/8) 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] (1/8) Epoch 12, batch 1150, loss[loss=0.1808, simple_loss=0.2693, pruned_loss=0.04612, over 16554.00 frames. ], tot_loss[loss=0.1776, simple_loss=0.26, pruned_loss=0.04759, over 3322902.41 frames. ], batch size: 62, lr: 5.79e-03, grad_scale: 4.0 2023-04-29 13:15:34,016 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.3422, 3.3133, 3.5992, 2.6563, 3.3001, 3.6582, 3.3237, 2.1622], device='cuda:1'), covar=tensor([0.0377, 0.0106, 0.0040, 0.0257, 0.0076, 0.0072, 0.0072, 0.0342], device='cuda:1'), in_proj_covar=tensor([0.0129, 0.0071, 0.0070, 0.0127, 0.0080, 0.0089, 0.0078, 0.0122], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:1') 2023-04-29 13:16:42,798 INFO [train.py:904] (1/8) Epoch 12, batch 1200, loss[loss=0.2051, simple_loss=0.2778, pruned_loss=0.06624, over 16763.00 frames. ], tot_loss[loss=0.1772, simple_loss=0.2592, pruned_loss=0.0476, over 3312962.05 frames. ], batch size: 134, lr: 5.79e-03, grad_scale: 8.0 2023-04-29 13:16:52,675 INFO [optim.py:368] (1/8) 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] (1/8) Epoch 12, batch 1250, loss[loss=0.188, simple_loss=0.2583, pruned_loss=0.05886, over 16589.00 frames. ], tot_loss[loss=0.1772, simple_loss=0.259, pruned_loss=0.04774, over 3316940.32 frames. ], batch size: 146, lr: 5.79e-03, grad_scale: 8.0 2023-04-29 13:18:12,314 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=112918.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 13:18:16,576 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=112921.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 13:18:32,350 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=112933.0, num_to_drop=1, layers_to_drop={0} 2023-04-29 13:18:41,584 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.99 vs. limit=2.0 2023-04-29 13:18:57,920 INFO [train.py:904] (1/8) Epoch 12, batch 1300, loss[loss=0.1631, simple_loss=0.2592, pruned_loss=0.03348, over 17050.00 frames. ], tot_loss[loss=0.1767, simple_loss=0.259, pruned_loss=0.04721, over 3318295.66 frames. ], batch size: 50, lr: 5.79e-03, grad_scale: 8.0 2023-04-29 13:19:09,587 INFO [optim.py:368] (1/8) 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 13:19:11,455 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.7710, 2.2485, 2.2959, 4.6593, 2.1742, 2.7178, 2.4102, 2.5302], device='cuda:1'), covar=tensor([0.0928, 0.3596, 0.2532, 0.0315, 0.3972, 0.2416, 0.3144, 0.3634], device='cuda:1'), in_proj_covar=tensor([0.0369, 0.0397, 0.0336, 0.0324, 0.0416, 0.0455, 0.0361, 0.0465], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-29 13:19:27,113 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=112973.0, num_to_drop=1, layers_to_drop={0} 2023-04-29 13:19:41,015 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=112982.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 13:20:08,547 INFO [train.py:904] (1/8) Epoch 12, batch 1350, loss[loss=0.1799, simple_loss=0.2552, pruned_loss=0.05233, over 16680.00 frames. ], tot_loss[loss=0.1763, simple_loss=0.2589, pruned_loss=0.04686, over 3318660.49 frames. ], batch size: 134, lr: 5.79e-03, grad_scale: 8.0 2023-04-29 13:20:44,726 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.7568, 2.7931, 2.3611, 2.8170, 3.0656, 2.8843, 3.5339, 3.2422], device='cuda:1'), covar=tensor([0.0085, 0.0270, 0.0355, 0.0267, 0.0185, 0.0257, 0.0164, 0.0213], device='cuda:1'), in_proj_covar=tensor([0.0160, 0.0213, 0.0205, 0.0204, 0.0212, 0.0212, 0.0217, 0.0202], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-29 13:21:07,348 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=113045.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 13:21:16,371 INFO [train.py:904] (1/8) Epoch 12, batch 1400, loss[loss=0.1687, simple_loss=0.2648, pruned_loss=0.0363, over 17123.00 frames. ], tot_loss[loss=0.1767, simple_loss=0.259, pruned_loss=0.04722, over 3324754.00 frames. ], batch size: 49, lr: 5.79e-03, grad_scale: 8.0 2023-04-29 13:21:26,436 INFO [optim.py:368] (1/8) Clipping_scale=2.0, grad-norm quartiles 1.826e+02 2.440e+02 3.018e+02 3.818e+02 8.239e+02, threshold=6.035e+02, percent-clipped=8.0 2023-04-29 13:22:12,583 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=113093.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 13:22:24,189 INFO [train.py:904] (1/8) Epoch 12, batch 1450, loss[loss=0.1624, simple_loss=0.2497, pruned_loss=0.03749, over 17176.00 frames. ], tot_loss[loss=0.1766, simple_loss=0.2586, pruned_loss=0.04731, over 3326087.72 frames. ], batch size: 46, lr: 5.79e-03, grad_scale: 8.0 2023-04-29 13:23:05,054 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.6118, 4.7158, 5.2079, 5.1394, 5.1695, 4.8196, 4.7958, 4.5984], device='cuda:1'), covar=tensor([0.0323, 0.0517, 0.0345, 0.0447, 0.0400, 0.0337, 0.0749, 0.0419], device='cuda:1'), in_proj_covar=tensor([0.0346, 0.0360, 0.0362, 0.0340, 0.0404, 0.0382, 0.0481, 0.0309], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:1') 2023-04-29 13:23:35,102 INFO [train.py:904] (1/8) Epoch 12, batch 1500, loss[loss=0.2098, simple_loss=0.2814, pruned_loss=0.06907, over 16555.00 frames. ], tot_loss[loss=0.1768, simple_loss=0.2584, pruned_loss=0.04764, over 3327105.38 frames. ], batch size: 75, lr: 5.78e-03, grad_scale: 8.0 2023-04-29 13:23:45,775 INFO [optim.py:368] (1/8) 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 13:24:18,298 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.37 vs. limit=2.0 2023-04-29 13:24:30,711 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.41 vs. limit=2.0 2023-04-29 13:24:43,433 INFO [train.py:904] (1/8) Epoch 12, batch 1550, loss[loss=0.2111, simple_loss=0.2786, pruned_loss=0.07182, over 16729.00 frames. ], tot_loss[loss=0.1802, simple_loss=0.2604, pruned_loss=0.05, over 3322329.38 frames. ], batch size: 76, lr: 5.78e-03, grad_scale: 8.0 2023-04-29 13:25:06,538 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=113218.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 13:25:27,356 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=113233.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 13:25:28,518 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=113234.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 13:25:36,412 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.33 vs. limit=2.0 2023-04-29 13:25:54,130 INFO [train.py:904] (1/8) Epoch 12, batch 1600, loss[loss=0.1496, simple_loss=0.2415, pruned_loss=0.02882, over 16856.00 frames. ], tot_loss[loss=0.1818, simple_loss=0.2624, pruned_loss=0.0506, over 3322801.40 frames. ], batch size: 39, lr: 5.78e-03, grad_scale: 8.0 2023-04-29 13:26:04,703 INFO [optim.py:368] (1/8) 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 13:26:12,639 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=113266.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 13:26:14,128 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([1.6480, 2.6281, 2.4439, 4.8054, 3.6279, 4.3401, 1.5728, 2.7559], device='cuda:1'), covar=tensor([0.1503, 0.0868, 0.1423, 0.0221, 0.0343, 0.0444, 0.1638, 0.1073], device='cuda:1'), in_proj_covar=tensor([0.0155, 0.0159, 0.0180, 0.0148, 0.0194, 0.0211, 0.0181, 0.0181], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0003, 0.0004], device='cuda:1') 2023-04-29 13:26:22,763 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=113273.0, num_to_drop=1, layers_to_drop={1} 2023-04-29 13:26:27,893 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=113277.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 13:26:33,950 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=113281.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 13:26:53,660 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=113295.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 13:27:02,727 INFO [train.py:904] (1/8) Epoch 12, batch 1650, loss[loss=0.1744, simple_loss=0.2708, pruned_loss=0.03899, over 17079.00 frames. ], tot_loss[loss=0.1816, simple_loss=0.2631, pruned_loss=0.05003, over 3329780.42 frames. ], batch size: 48, lr: 5.78e-03, grad_scale: 8.0 2023-04-29 13:27:29,756 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=113321.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 13:27:58,005 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.2899, 2.0912, 2.1894, 3.9276, 2.0464, 2.5403, 2.1380, 2.3320], device='cuda:1'), covar=tensor([0.1010, 0.3381, 0.2326, 0.0441, 0.3487, 0.2177, 0.3288, 0.2758], device='cuda:1'), in_proj_covar=tensor([0.0367, 0.0395, 0.0335, 0.0323, 0.0412, 0.0453, 0.0358, 0.0463], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-29 13:28:12,390 INFO [train.py:904] (1/8) Epoch 12, batch 1700, loss[loss=0.2179, simple_loss=0.2934, pruned_loss=0.07117, over 15462.00 frames. ], tot_loss[loss=0.1839, simple_loss=0.2659, pruned_loss=0.05092, over 3313503.63 frames. ], batch size: 190, lr: 5.78e-03, grad_scale: 8.0 2023-04-29 13:28:23,617 INFO [optim.py:368] (1/8) 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 13:29:21,805 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.60 vs. limit=2.0 2023-04-29 13:29:22,272 INFO [train.py:904] (1/8) Epoch 12, batch 1750, loss[loss=0.1765, simple_loss=0.2671, pruned_loss=0.04297, over 17196.00 frames. ], tot_loss[loss=0.1849, simple_loss=0.2672, pruned_loss=0.05133, over 3313081.08 frames. ], batch size: 44, lr: 5.78e-03, grad_scale: 8.0 2023-04-29 13:29:27,347 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.92 vs. limit=2.0 2023-04-29 13:30:32,315 INFO [train.py:904] (1/8) Epoch 12, batch 1800, loss[loss=0.204, simple_loss=0.2891, pruned_loss=0.0594, over 17056.00 frames. ], tot_loss[loss=0.1861, simple_loss=0.2684, pruned_loss=0.05184, over 3307003.44 frames. ], batch size: 55, lr: 5.78e-03, grad_scale: 8.0 2023-04-29 13:30:43,442 INFO [optim.py:368] (1/8) 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] (1/8) Epoch 12, batch 1850, loss[loss=0.1654, simple_loss=0.2461, pruned_loss=0.04234, over 17219.00 frames. ], tot_loss[loss=0.1855, simple_loss=0.2686, pruned_loss=0.05124, over 3312448.97 frames. ], batch size: 44, lr: 5.78e-03, grad_scale: 8.0 2023-04-29 13:31:43,584 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=3.39 vs. limit=5.0 2023-04-29 13:32:04,638 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.3229, 2.0844, 2.2959, 4.0712, 2.0944, 2.5252, 2.1748, 2.2906], device='cuda:1'), covar=tensor([0.1045, 0.3417, 0.2278, 0.0419, 0.3403, 0.2281, 0.3416, 0.2889], device='cuda:1'), in_proj_covar=tensor([0.0371, 0.0397, 0.0336, 0.0325, 0.0415, 0.0457, 0.0362, 0.0465], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-29 13:32:28,930 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.5342, 3.4698, 3.9849, 2.9683, 3.5743, 3.9494, 3.6799, 2.3895], device='cuda:1'), covar=tensor([0.0402, 0.0224, 0.0035, 0.0250, 0.0074, 0.0081, 0.0068, 0.0345], device='cuda:1'), in_proj_covar=tensor([0.0130, 0.0073, 0.0071, 0.0127, 0.0081, 0.0091, 0.0080, 0.0122], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:1') 2023-04-29 13:32:53,583 INFO [train.py:904] (1/8) Epoch 12, batch 1900, loss[loss=0.1629, simple_loss=0.255, pruned_loss=0.03535, over 17146.00 frames. ], tot_loss[loss=0.1842, simple_loss=0.2677, pruned_loss=0.05032, over 3318565.66 frames. ], batch size: 47, lr: 5.77e-03, grad_scale: 8.0 2023-04-29 13:33:04,793 INFO [optim.py:368] (1/8) 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 13:33:29,745 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=113577.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 13:33:48,964 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=113590.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 13:34:05,627 INFO [train.py:904] (1/8) Epoch 12, batch 1950, loss[loss=0.1896, simple_loss=0.2654, pruned_loss=0.0569, over 16842.00 frames. ], tot_loss[loss=0.1833, simple_loss=0.2674, pruned_loss=0.04962, over 3316185.81 frames. ], batch size: 102, lr: 5.77e-03, grad_scale: 8.0 2023-04-29 13:34:06,064 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.2516, 4.1089, 4.5285, 2.1147, 4.8019, 4.7942, 3.3169, 3.7079], device='cuda:1'), covar=tensor([0.0593, 0.0185, 0.0159, 0.1105, 0.0042, 0.0086, 0.0384, 0.0330], device='cuda:1'), in_proj_covar=tensor([0.0144, 0.0101, 0.0089, 0.0141, 0.0071, 0.0108, 0.0122, 0.0127], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:1') 2023-04-29 13:34:27,370 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=113617.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 13:34:38,486 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=113625.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 13:35:08,083 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 12, batch 2000, loss[loss=0.203, simple_loss=0.2915, pruned_loss=0.05727, over 17033.00 frames. ], tot_loss[loss=0.1827, simple_loss=0.267, pruned_loss=0.04921, over 3310165.17 frames. ], batch size: 55, lr: 5.77e-03, grad_scale: 8.0 2023-04-29 13:35:27,900 INFO [optim.py:368] (1/8) 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 13:35:52,954 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=113678.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 13:36:25,542 INFO [train.py:904] (1/8) Epoch 12, batch 2050, loss[loss=0.2201, simple_loss=0.2999, pruned_loss=0.07018, over 11807.00 frames. ], tot_loss[loss=0.1841, simple_loss=0.2676, pruned_loss=0.0503, over 3308296.22 frames. ], batch size: 246, lr: 5.77e-03, grad_scale: 8.0 2023-04-29 13:36:32,422 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=113707.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 13:36:54,193 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.7213, 3.9441, 2.9843, 2.2116, 2.7661, 2.3127, 4.1920, 3.5281], device='cuda:1'), covar=tensor([0.2388, 0.0641, 0.1473, 0.2293, 0.2293, 0.1868, 0.0386, 0.1196], device='cuda:1'), in_proj_covar=tensor([0.0302, 0.0260, 0.0286, 0.0282, 0.0282, 0.0225, 0.0269, 0.0303], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-29 13:37:33,719 INFO [train.py:904] (1/8) Epoch 12, batch 2100, loss[loss=0.1889, simple_loss=0.2814, pruned_loss=0.04816, over 17119.00 frames. ], tot_loss[loss=0.186, simple_loss=0.2693, pruned_loss=0.05134, over 3305695.12 frames. ], batch size: 48, lr: 5.77e-03, grad_scale: 8.0 2023-04-29 13:37:45,424 INFO [optim.py:368] (1/8) 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 13:38:44,790 INFO [train.py:904] (1/8) Epoch 12, batch 2150, loss[loss=0.1573, simple_loss=0.2427, pruned_loss=0.03596, over 17015.00 frames. ], tot_loss[loss=0.1872, simple_loss=0.2703, pruned_loss=0.05203, over 3310219.83 frames. ], batch size: 41, lr: 5.77e-03, grad_scale: 8.0 2023-04-29 13:38:57,524 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.4743, 2.1428, 2.3617, 4.2097, 2.1442, 2.6025, 2.2098, 2.3783], device='cuda:1'), covar=tensor([0.1049, 0.3447, 0.2234, 0.0487, 0.3699, 0.2205, 0.3261, 0.2918], device='cuda:1'), in_proj_covar=tensor([0.0372, 0.0397, 0.0335, 0.0324, 0.0414, 0.0458, 0.0362, 0.0467], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-29 13:39:54,119 INFO [train.py:904] (1/8) Epoch 12, batch 2200, loss[loss=0.196, simple_loss=0.2861, pruned_loss=0.05301, over 16714.00 frames. ], tot_loss[loss=0.1879, simple_loss=0.271, pruned_loss=0.05234, over 3315101.63 frames. ], batch size: 57, lr: 5.77e-03, grad_scale: 8.0 2023-04-29 13:40:05,120 INFO [optim.py:368] (1/8) 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 13:40:05,562 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.5431, 4.4778, 4.4380, 4.2475, 4.1882, 4.5019, 4.2966, 4.2579], device='cuda:1'), covar=tensor([0.0556, 0.0547, 0.0268, 0.0254, 0.0759, 0.0438, 0.0557, 0.0641], device='cuda:1'), in_proj_covar=tensor([0.0268, 0.0348, 0.0317, 0.0295, 0.0337, 0.0341, 0.0215, 0.0370], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-04-29 13:40:06,775 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([5.0300, 4.9831, 4.8868, 4.5652, 4.5427, 4.9434, 4.8945, 4.5792], device='cuda:1'), covar=tensor([0.0558, 0.0463, 0.0261, 0.0276, 0.0898, 0.0417, 0.0355, 0.0766], device='cuda:1'), in_proj_covar=tensor([0.0268, 0.0349, 0.0317, 0.0295, 0.0337, 0.0341, 0.0215, 0.0370], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-04-29 13:40:27,147 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.2486, 4.1703, 4.5369, 2.1625, 4.7656, 4.7077, 3.3771, 3.7615], device='cuda:1'), covar=tensor([0.0591, 0.0178, 0.0187, 0.1084, 0.0051, 0.0124, 0.0334, 0.0314], device='cuda:1'), in_proj_covar=tensor([0.0145, 0.0101, 0.0090, 0.0140, 0.0071, 0.0110, 0.0121, 0.0128], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:1') 2023-04-29 13:40:36,975 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([5.0247, 4.9491, 4.8542, 4.5159, 4.4739, 4.9205, 4.9176, 4.5209], device='cuda:1'), covar=tensor([0.0590, 0.0485, 0.0305, 0.0283, 0.1048, 0.0376, 0.0364, 0.0734], device='cuda:1'), in_proj_covar=tensor([0.0268, 0.0348, 0.0316, 0.0295, 0.0337, 0.0340, 0.0215, 0.0369], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-04-29 13:40:48,889 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=113890.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 13:41:03,695 INFO [train.py:904] (1/8) Epoch 12, batch 2250, loss[loss=0.1457, simple_loss=0.2339, pruned_loss=0.02873, over 17199.00 frames. ], tot_loss[loss=0.1874, simple_loss=0.2707, pruned_loss=0.0521, over 3312340.20 frames. ], batch size: 44, lr: 5.77e-03, grad_scale: 8.0 2023-04-29 13:41:52,997 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.82 vs. limit=2.0 2023-04-29 13:41:54,643 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=113938.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 13:42:14,192 INFO [train.py:904] (1/8) Epoch 12, batch 2300, loss[loss=0.1776, simple_loss=0.2803, pruned_loss=0.0374, over 17053.00 frames. ], tot_loss[loss=0.1878, simple_loss=0.2714, pruned_loss=0.05207, over 3305492.42 frames. ], batch size: 55, lr: 5.76e-03, grad_scale: 16.0 2023-04-29 13:42:24,215 INFO [optim.py:368] (1/8) Clipping_scale=2.0, grad-norm quartiles 1.424e+02 2.413e+02 2.819e+02 3.398e+02 7.599e+02, threshold=5.637e+02, percent-clipped=2.0 2023-04-29 13:42:43,121 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 12, batch 2350, loss[loss=0.1877, simple_loss=0.2806, pruned_loss=0.04736, over 17226.00 frames. ], tot_loss[loss=0.1879, simple_loss=0.271, pruned_loss=0.05236, over 3308788.62 frames. ], batch size: 45, lr: 5.76e-03, grad_scale: 16.0 2023-04-29 13:43:25,834 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=114002.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 13:43:54,057 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.86 vs. limit=2.0 2023-04-29 13:44:35,987 INFO [train.py:904] (1/8) Epoch 12, batch 2400, loss[loss=0.1928, simple_loss=0.2698, pruned_loss=0.05791, over 16756.00 frames. ], tot_loss[loss=0.1892, simple_loss=0.2723, pruned_loss=0.05304, over 3302791.88 frames. ], batch size: 83, lr: 5.76e-03, grad_scale: 16.0 2023-04-29 13:44:48,486 INFO [optim.py:368] (1/8) 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 13:45:27,349 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.46 vs. limit=2.0 2023-04-29 13:45:49,044 INFO [train.py:904] (1/8) Epoch 12, batch 2450, loss[loss=0.1506, simple_loss=0.2359, pruned_loss=0.03265, over 16954.00 frames. ], tot_loss[loss=0.188, simple_loss=0.2719, pruned_loss=0.05208, over 3302264.76 frames. ], batch size: 41, lr: 5.76e-03, grad_scale: 16.0 2023-04-29 13:46:23,295 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=114127.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 13:46:54,794 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([5.7280, 6.1296, 5.8268, 5.9779, 5.5038, 5.4403, 5.7123, 6.2216], device='cuda:1'), covar=tensor([0.1187, 0.0960, 0.1145, 0.0733, 0.0937, 0.0746, 0.0932, 0.0981], device='cuda:1'), in_proj_covar=tensor([0.0575, 0.0715, 0.0581, 0.0500, 0.0448, 0.0455, 0.0592, 0.0548], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:1') 2023-04-29 13:46:57,384 INFO [train.py:904] (1/8) Epoch 12, batch 2500, loss[loss=0.1968, simple_loss=0.2943, pruned_loss=0.04969, over 17122.00 frames. ], tot_loss[loss=0.1864, simple_loss=0.2708, pruned_loss=0.051, over 3315536.35 frames. ], batch size: 49, lr: 5.76e-03, grad_scale: 16.0 2023-04-29 13:47:09,683 INFO [optim.py:368] (1/8) 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 13:47:48,626 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 12, batch 2550, loss[loss=0.1572, simple_loss=0.2496, pruned_loss=0.03239, over 17096.00 frames. ], tot_loss[loss=0.1863, simple_loss=0.2707, pruned_loss=0.05091, over 3319069.49 frames. ], batch size: 47, lr: 5.76e-03, grad_scale: 16.0 2023-04-29 13:49:15,389 INFO [train.py:904] (1/8) Epoch 12, batch 2600, loss[loss=0.1949, simple_loss=0.2868, pruned_loss=0.05148, over 16666.00 frames. ], tot_loss[loss=0.1858, simple_loss=0.2704, pruned_loss=0.05057, over 3317968.75 frames. ], batch size: 62, lr: 5.76e-03, grad_scale: 16.0 2023-04-29 13:49:25,933 INFO [optim.py:368] (1/8) 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 13:49:45,774 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=114273.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 13:49:47,700 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([1.7836, 3.0975, 3.0338, 4.9883, 4.1611, 4.4502, 1.6235, 3.2902], device='cuda:1'), covar=tensor([0.1299, 0.0664, 0.0979, 0.0206, 0.0215, 0.0389, 0.1450, 0.0705], device='cuda:1'), in_proj_covar=tensor([0.0154, 0.0160, 0.0180, 0.0152, 0.0196, 0.0212, 0.0181, 0.0180], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0003, 0.0004], device='cuda:1') 2023-04-29 13:49:52,206 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([5.0793, 4.8081, 5.1067, 5.3506, 5.4891, 4.8044, 5.4491, 5.4747], device='cuda:1'), covar=tensor([0.1477, 0.1160, 0.1642, 0.0629, 0.0498, 0.0807, 0.0577, 0.0518], device='cuda:1'), in_proj_covar=tensor([0.0572, 0.0714, 0.0864, 0.0728, 0.0547, 0.0562, 0.0565, 0.0652], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-29 13:50:24,315 INFO [train.py:904] (1/8) Epoch 12, batch 2650, loss[loss=0.1786, simple_loss=0.2701, pruned_loss=0.04355, over 17194.00 frames. ], tot_loss[loss=0.1851, simple_loss=0.2703, pruned_loss=0.04993, over 3322095.06 frames. ], batch size: 46, lr: 5.76e-03, grad_scale: 16.0 2023-04-29 13:50:24,680 INFO [zipformer.py:625] (1/8) 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] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=114321.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 13:51:12,708 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.79 vs. limit=2.0 2023-04-29 13:51:32,414 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=114350.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 13:51:35,146 INFO [train.py:904] (1/8) Epoch 12, batch 2700, loss[loss=0.152, simple_loss=0.2395, pruned_loss=0.03226, over 16813.00 frames. ], tot_loss[loss=0.184, simple_loss=0.2697, pruned_loss=0.04912, over 3325229.48 frames. ], batch size: 39, lr: 5.75e-03, grad_scale: 16.0 2023-04-29 13:51:45,488 INFO [optim.py:368] (1/8) 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 13:52:29,134 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.39 vs. limit=2.0 2023-04-29 13:52:44,855 INFO [train.py:904] (1/8) Epoch 12, batch 2750, loss[loss=0.1818, simple_loss=0.2703, pruned_loss=0.04663, over 16558.00 frames. ], tot_loss[loss=0.1843, simple_loss=0.2701, pruned_loss=0.04923, over 3327514.92 frames. ], batch size: 75, lr: 5.75e-03, grad_scale: 16.0 2023-04-29 13:53:05,742 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([5.3703, 5.2225, 5.1714, 4.8209, 4.7723, 5.2169, 5.2680, 4.8004], device='cuda:1'), covar=tensor([0.0524, 0.0412, 0.0312, 0.0280, 0.1057, 0.0362, 0.0251, 0.0784], device='cuda:1'), in_proj_covar=tensor([0.0268, 0.0349, 0.0318, 0.0294, 0.0336, 0.0341, 0.0216, 0.0370], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-04-29 13:53:54,633 INFO [train.py:904] (1/8) Epoch 12, batch 2800, loss[loss=0.1932, simple_loss=0.2646, pruned_loss=0.06088, over 16483.00 frames. ], tot_loss[loss=0.1845, simple_loss=0.27, pruned_loss=0.04952, over 3326944.27 frames. ], batch size: 146, lr: 5.75e-03, grad_scale: 16.0 2023-04-29 13:54:06,072 INFO [optim.py:368] (1/8) 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 13:54:38,608 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=114483.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 13:55:04,045 INFO [train.py:904] (1/8) Epoch 12, batch 2850, loss[loss=0.179, simple_loss=0.2601, pruned_loss=0.04891, over 16632.00 frames. ], tot_loss[loss=0.1837, simple_loss=0.2689, pruned_loss=0.04926, over 3326628.24 frames. ], batch size: 89, lr: 5.75e-03, grad_scale: 16.0 2023-04-29 13:55:38,358 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.8004, 2.4957, 1.9390, 2.3757, 2.9036, 2.7521, 2.9710, 3.0057], device='cuda:1'), covar=tensor([0.0150, 0.0290, 0.0380, 0.0315, 0.0174, 0.0236, 0.0220, 0.0203], device='cuda:1'), in_proj_covar=tensor([0.0163, 0.0214, 0.0203, 0.0206, 0.0213, 0.0212, 0.0222, 0.0204], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-29 13:55:49,493 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.6767, 2.2377, 2.2743, 4.5303, 2.1596, 2.7314, 2.3362, 2.4207], device='cuda:1'), covar=tensor([0.0957, 0.3321, 0.2437, 0.0363, 0.3800, 0.2389, 0.3120, 0.3342], device='cuda:1'), in_proj_covar=tensor([0.0372, 0.0397, 0.0336, 0.0326, 0.0414, 0.0460, 0.0361, 0.0469], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-29 13:56:13,202 INFO [train.py:904] (1/8) Epoch 12, batch 2900, loss[loss=0.2763, simple_loss=0.3221, pruned_loss=0.1153, over 11521.00 frames. ], tot_loss[loss=0.1844, simple_loss=0.2686, pruned_loss=0.05008, over 3319486.62 frames. ], batch size: 247, lr: 5.75e-03, grad_scale: 8.0 2023-04-29 13:56:24,537 INFO [optim.py:368] (1/8) 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 13:57:20,462 INFO [train.py:904] (1/8) Epoch 12, batch 2950, loss[loss=0.1926, simple_loss=0.2694, pruned_loss=0.05788, over 16745.00 frames. ], tot_loss[loss=0.1843, simple_loss=0.268, pruned_loss=0.0503, over 3323805.08 frames. ], batch size: 134, lr: 5.75e-03, grad_scale: 8.0 2023-04-29 13:57:42,462 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.3244, 3.7109, 3.7799, 2.1625, 3.0856, 2.4800, 3.8678, 3.8166], device='cuda:1'), covar=tensor([0.0297, 0.0740, 0.0503, 0.1683, 0.0750, 0.0898, 0.0624, 0.1003], device='cuda:1'), in_proj_covar=tensor([0.0145, 0.0147, 0.0158, 0.0144, 0.0136, 0.0124, 0.0138, 0.0160], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:1') 2023-04-29 13:58:28,574 INFO [train.py:904] (1/8) Epoch 12, batch 3000, loss[loss=0.1573, simple_loss=0.2454, pruned_loss=0.03456, over 16887.00 frames. ], tot_loss[loss=0.1843, simple_loss=0.2681, pruned_loss=0.05024, over 3332556.94 frames. ], batch size: 42, lr: 5.75e-03, grad_scale: 8.0 2023-04-29 13:58:28,574 INFO [train.py:929] (1/8) Computing validation loss 2023-04-29 13:58:38,473 INFO [train.py:938] (1/8) Epoch 12, validation: loss=0.14, simple_loss=0.2459, pruned_loss=0.01708, over 944034.00 frames. 2023-04-29 13:58:38,473 INFO [train.py:939] (1/8) Maximum memory allocated so far is 17845MB 2023-04-29 13:58:39,180 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.86 vs. limit=2.0 2023-04-29 13:58:50,228 INFO [optim.py:368] (1/8) 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] (1/8) Epoch 12, batch 3050, loss[loss=0.1902, simple_loss=0.2824, pruned_loss=0.04901, over 17066.00 frames. ], tot_loss[loss=0.1845, simple_loss=0.2682, pruned_loss=0.05043, over 3332030.35 frames. ], batch size: 53, lr: 5.75e-03, grad_scale: 8.0 2023-04-29 14:00:08,877 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=114716.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 14:00:56,625 INFO [train.py:904] (1/8) Epoch 12, batch 3100, loss[loss=0.1715, simple_loss=0.2638, pruned_loss=0.03962, over 17112.00 frames. ], tot_loss[loss=0.1837, simple_loss=0.2677, pruned_loss=0.04985, over 3332975.77 frames. ], batch size: 49, lr: 5.74e-03, grad_scale: 8.0 2023-04-29 14:01:04,340 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.2734, 1.9259, 2.5692, 3.1477, 2.7643, 3.5781, 1.8253, 3.5310], device='cuda:1'), covar=tensor([0.0121, 0.0420, 0.0228, 0.0181, 0.0238, 0.0113, 0.0501, 0.0117], device='cuda:1'), in_proj_covar=tensor([0.0170, 0.0177, 0.0159, 0.0165, 0.0175, 0.0131, 0.0175, 0.0121], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001], device='cuda:1') 2023-04-29 14:01:10,357 INFO [optim.py:368] (1/8) 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 14:01:31,848 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=114777.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 14:01:40,449 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 12, batch 3150, loss[loss=0.1886, simple_loss=0.2767, pruned_loss=0.05022, over 16705.00 frames. ], tot_loss[loss=0.1841, simple_loss=0.2676, pruned_loss=0.05025, over 3327265.11 frames. ], batch size: 57, lr: 5.74e-03, grad_scale: 8.0 2023-04-29 14:02:45,190 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=114831.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 14:02:50,753 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.31 vs. limit=2.0 2023-04-29 14:03:00,677 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.0523, 3.1668, 3.1906, 2.1292, 2.8163, 2.2017, 3.5791, 3.4342], device='cuda:1'), covar=tensor([0.0214, 0.0835, 0.0571, 0.1633, 0.0799, 0.0975, 0.0509, 0.0889], device='cuda:1'), in_proj_covar=tensor([0.0145, 0.0148, 0.0158, 0.0144, 0.0136, 0.0124, 0.0137, 0.0160], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:1') 2023-04-29 14:03:14,011 INFO [train.py:904] (1/8) Epoch 12, batch 3200, loss[loss=0.1745, simple_loss=0.2543, pruned_loss=0.04734, over 16747.00 frames. ], tot_loss[loss=0.1831, simple_loss=0.2666, pruned_loss=0.04985, over 3326469.50 frames. ], batch size: 124, lr: 5.74e-03, grad_scale: 8.0 2023-04-29 14:03:26,056 INFO [optim.py:368] (1/8) 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 14:03:40,822 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=114871.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 14:04:22,405 INFO [train.py:904] (1/8) Epoch 12, batch 3250, loss[loss=0.1565, simple_loss=0.2438, pruned_loss=0.03454, over 16828.00 frames. ], tot_loss[loss=0.1834, simple_loss=0.2667, pruned_loss=0.05008, over 3322641.05 frames. ], batch size: 42, lr: 5.74e-03, grad_scale: 4.0 2023-04-29 14:05:05,283 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 12, batch 3300, loss[loss=0.1751, simple_loss=0.2743, pruned_loss=0.03791, over 17123.00 frames. ], tot_loss[loss=0.1843, simple_loss=0.2676, pruned_loss=0.05046, over 3321333.55 frames. ], batch size: 49, lr: 5.74e-03, grad_scale: 4.0 2023-04-29 14:05:45,365 INFO [optim.py:368] (1/8) 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 14:06:01,174 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.4009, 4.2219, 4.6348, 2.2331, 4.7678, 4.7341, 3.3827, 3.9374], device='cuda:1'), covar=tensor([0.0532, 0.0170, 0.0165, 0.1004, 0.0079, 0.0128, 0.0348, 0.0260], device='cuda:1'), in_proj_covar=tensor([0.0143, 0.0101, 0.0089, 0.0138, 0.0070, 0.0109, 0.0122, 0.0127], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:1') 2023-04-29 14:06:10,313 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=5.01 vs. limit=5.0 2023-04-29 14:06:22,022 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.2770, 3.7896, 3.7652, 1.9744, 3.0233, 2.5477, 3.7900, 3.7618], device='cuda:1'), covar=tensor([0.0267, 0.0691, 0.0515, 0.1792, 0.0747, 0.0855, 0.0630, 0.1041], device='cuda:1'), in_proj_covar=tensor([0.0146, 0.0149, 0.0159, 0.0145, 0.0137, 0.0125, 0.0138, 0.0162], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:1') 2023-04-29 14:06:41,208 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.7796, 4.8996, 5.1119, 4.8598, 4.8887, 5.5460, 5.0614, 4.7547], device='cuda:1'), covar=tensor([0.1185, 0.1773, 0.1864, 0.2085, 0.2716, 0.0905, 0.1421, 0.2293], device='cuda:1'), in_proj_covar=tensor([0.0361, 0.0516, 0.0558, 0.0441, 0.0599, 0.0583, 0.0443, 0.0593], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-04-29 14:06:42,021 INFO [train.py:904] (1/8) Epoch 12, batch 3350, loss[loss=0.2018, simple_loss=0.2793, pruned_loss=0.06222, over 16851.00 frames. ], tot_loss[loss=0.1851, simple_loss=0.2686, pruned_loss=0.05082, over 3318612.70 frames. ], batch size: 96, lr: 5.74e-03, grad_scale: 4.0 2023-04-29 14:07:34,972 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.5064, 2.4165, 2.0377, 2.2824, 2.8248, 2.5783, 3.1993, 3.1231], device='cuda:1'), covar=tensor([0.0095, 0.0358, 0.0418, 0.0355, 0.0224, 0.0313, 0.0205, 0.0204], device='cuda:1'), in_proj_covar=tensor([0.0165, 0.0214, 0.0205, 0.0206, 0.0215, 0.0214, 0.0223, 0.0205], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-29 14:07:50,787 INFO [train.py:904] (1/8) Epoch 12, batch 3400, loss[loss=0.1897, simple_loss=0.2698, pruned_loss=0.05485, over 16687.00 frames. ], tot_loss[loss=0.1836, simple_loss=0.2676, pruned_loss=0.04982, over 3327900.25 frames. ], batch size: 83, lr: 5.74e-03, grad_scale: 4.0 2023-04-29 14:08:04,044 INFO [optim.py:368] (1/8) 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 14:08:08,707 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.7563, 3.9881, 3.1186, 2.3621, 2.7283, 2.4068, 3.9198, 3.6039], device='cuda:1'), covar=tensor([0.2433, 0.0573, 0.1409, 0.2357, 0.2300, 0.1715, 0.0521, 0.1112], device='cuda:1'), in_proj_covar=tensor([0.0303, 0.0261, 0.0285, 0.0282, 0.0285, 0.0225, 0.0271, 0.0305], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:1') 2023-04-29 14:08:18,295 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=115072.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 14:08:37,797 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=115086.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 14:08:56,402 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.7418, 1.8098, 2.2253, 2.6420, 2.6819, 2.5966, 1.7877, 2.8409], device='cuda:1'), covar=tensor([0.0126, 0.0380, 0.0236, 0.0224, 0.0202, 0.0188, 0.0390, 0.0116], device='cuda:1'), in_proj_covar=tensor([0.0171, 0.0178, 0.0160, 0.0165, 0.0177, 0.0132, 0.0176, 0.0122], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-29 14:09:00,023 INFO [train.py:904] (1/8) Epoch 12, batch 3450, loss[loss=0.1986, simple_loss=0.2685, pruned_loss=0.06434, over 16477.00 frames. ], tot_loss[loss=0.1821, simple_loss=0.2657, pruned_loss=0.04928, over 3330744.01 frames. ], batch size: 146, lr: 5.74e-03, grad_scale: 4.0 2023-04-29 14:10:01,695 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=115147.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 14:10:08,337 INFO [train.py:904] (1/8) Epoch 12, batch 3500, loss[loss=0.1846, simple_loss=0.2632, pruned_loss=0.05302, over 16682.00 frames. ], tot_loss[loss=0.1804, simple_loss=0.2642, pruned_loss=0.04831, over 3329181.92 frames. ], batch size: 83, lr: 5.73e-03, grad_scale: 4.0 2023-04-29 14:10:23,310 INFO [optim.py:368] (1/8) 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] (1/8) Epoch 12, batch 3550, loss[loss=0.1793, simple_loss=0.2693, pruned_loss=0.04465, over 17045.00 frames. ], tot_loss[loss=0.1795, simple_loss=0.2635, pruned_loss=0.04776, over 3324446.23 frames. ], batch size: 53, lr: 5.73e-03, grad_scale: 4.0 2023-04-29 14:11:53,669 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=115227.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 14:12:25,181 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.21 vs. limit=2.0 2023-04-29 14:12:28,659 INFO [train.py:904] (1/8) Epoch 12, batch 3600, loss[loss=0.1518, simple_loss=0.2426, pruned_loss=0.03045, over 17135.00 frames. ], tot_loss[loss=0.1781, simple_loss=0.2616, pruned_loss=0.04736, over 3326201.73 frames. ], batch size: 48, lr: 5.73e-03, grad_scale: 8.0 2023-04-29 14:12:43,848 INFO [optim.py:368] (1/8) 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 14:12:50,585 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([5.0110, 5.4589, 5.6777, 5.3578, 5.4144, 6.0390, 5.6379, 5.3411], device='cuda:1'), covar=tensor([0.0873, 0.1875, 0.1952, 0.2089, 0.2571, 0.0834, 0.1364, 0.2275], device='cuda:1'), in_proj_covar=tensor([0.0362, 0.0516, 0.0559, 0.0438, 0.0595, 0.0581, 0.0442, 0.0594], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-04-29 14:13:17,033 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([5.1404, 5.7048, 5.8640, 5.6074, 5.6011, 6.2024, 5.8474, 5.5554], device='cuda:1'), covar=tensor([0.0740, 0.1825, 0.1695, 0.1773, 0.2647, 0.0874, 0.1143, 0.2093], device='cuda:1'), in_proj_covar=tensor([0.0362, 0.0515, 0.0559, 0.0437, 0.0594, 0.0580, 0.0441, 0.0593], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-04-29 14:13:40,328 INFO [train.py:904] (1/8) Epoch 12, batch 3650, loss[loss=0.1732, simple_loss=0.2434, pruned_loss=0.05145, over 16736.00 frames. ], tot_loss[loss=0.1783, simple_loss=0.2609, pruned_loss=0.04787, over 3313127.19 frames. ], batch size: 89, lr: 5.73e-03, grad_scale: 8.0 2023-04-29 14:14:55,142 INFO [train.py:904] (1/8) Epoch 12, batch 3700, loss[loss=0.2054, simple_loss=0.2716, pruned_loss=0.06957, over 16764.00 frames. ], tot_loss[loss=0.1798, simple_loss=0.2605, pruned_loss=0.04952, over 3299680.16 frames. ], batch size: 124, lr: 5.73e-03, grad_scale: 8.0 2023-04-29 14:14:55,787 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.7619, 3.8210, 2.9722, 2.2582, 2.6104, 2.3904, 3.8507, 3.4957], device='cuda:1'), covar=tensor([0.2349, 0.0597, 0.1471, 0.2630, 0.2352, 0.1808, 0.0498, 0.1260], device='cuda:1'), in_proj_covar=tensor([0.0301, 0.0258, 0.0284, 0.0281, 0.0284, 0.0224, 0.0269, 0.0303], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-29 14:15:09,331 INFO [optim.py:368] (1/8) 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 14:15:22,918 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=115370.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 14:15:24,706 INFO [zipformer.py:625] (1/8) 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] (1/8) 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] (1/8) Epoch 12, batch 3750, loss[loss=0.1838, simple_loss=0.26, pruned_loss=0.05378, over 16403.00 frames. ], tot_loss[loss=0.1822, simple_loss=0.2614, pruned_loss=0.05154, over 3299183.27 frames. ], batch size: 146, lr: 5.73e-03, grad_scale: 8.0 2023-04-29 14:16:36,698 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=115420.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 14:16:54,200 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=115431.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 14:16:55,315 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=115432.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 14:17:09,416 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=115442.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 14:17:23,444 INFO [train.py:904] (1/8) Epoch 12, batch 3800, loss[loss=0.1723, simple_loss=0.2472, pruned_loss=0.04873, over 16851.00 frames. ], tot_loss[loss=0.1849, simple_loss=0.2631, pruned_loss=0.05331, over 3293836.62 frames. ], batch size: 102, lr: 5.73e-03, grad_scale: 8.0 2023-04-29 14:17:38,964 INFO [optim.py:368] (1/8) Clipping_scale=2.0, grad-norm quartiles 1.548e+02 2.466e+02 2.817e+02 3.424e+02 6.065e+02, threshold=5.633e+02, percent-clipped=2.0 2023-04-29 14:18:37,587 INFO [train.py:904] (1/8) Epoch 12, batch 3850, loss[loss=0.2281, simple_loss=0.2822, pruned_loss=0.08698, over 16799.00 frames. ], tot_loss[loss=0.1854, simple_loss=0.2629, pruned_loss=0.05398, over 3289770.69 frames. ], batch size: 124, lr: 5.73e-03, grad_scale: 8.0 2023-04-29 14:18:54,298 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.3793, 4.4032, 4.7774, 4.7765, 4.7742, 4.4662, 4.4946, 4.2959], device='cuda:1'), covar=tensor([0.0312, 0.0599, 0.0347, 0.0364, 0.0465, 0.0343, 0.0814, 0.0572], device='cuda:1'), in_proj_covar=tensor([0.0356, 0.0371, 0.0371, 0.0345, 0.0416, 0.0389, 0.0497, 0.0313], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:1') 2023-04-29 14:19:16,894 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=115527.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 14:19:52,697 INFO [train.py:904] (1/8) Epoch 12, batch 3900, loss[loss=0.1814, simple_loss=0.2535, pruned_loss=0.0547, over 16749.00 frames. ], tot_loss[loss=0.1855, simple_loss=0.2626, pruned_loss=0.05423, over 3284244.92 frames. ], batch size: 124, lr: 5.72e-03, grad_scale: 8.0 2023-04-29 14:19:59,830 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([1.6547, 2.7256, 2.4458, 4.0731, 3.3507, 4.0171, 1.4124, 2.9706], device='cuda:1'), covar=tensor([0.1325, 0.0630, 0.1081, 0.0163, 0.0179, 0.0368, 0.1499, 0.0728], device='cuda:1'), in_proj_covar=tensor([0.0155, 0.0162, 0.0183, 0.0155, 0.0202, 0.0214, 0.0183, 0.0183], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-04-29 14:20:07,961 INFO [optim.py:368] (1/8) 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 14:20:19,431 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=3.45 vs. limit=5.0 2023-04-29 14:20:29,519 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=115575.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 14:20:40,215 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.32 vs. limit=2.0 2023-04-29 14:21:01,285 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([1.7364, 2.6975, 2.5527, 1.8458, 2.6604, 2.8127, 2.6216, 1.6165], device='cuda:1'), covar=tensor([0.0393, 0.0076, 0.0057, 0.0350, 0.0088, 0.0096, 0.0095, 0.0434], device='cuda:1'), in_proj_covar=tensor([0.0127, 0.0071, 0.0071, 0.0125, 0.0080, 0.0090, 0.0080, 0.0120], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:1') 2023-04-29 14:21:08,905 INFO [train.py:904] (1/8) Epoch 12, batch 3950, loss[loss=0.1833, simple_loss=0.2494, pruned_loss=0.05863, over 16958.00 frames. ], tot_loss[loss=0.1856, simple_loss=0.2619, pruned_loss=0.0546, over 3277305.72 frames. ], batch size: 90, lr: 5.72e-03, grad_scale: 8.0 2023-04-29 14:22:08,289 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.25 vs. limit=2.0 2023-04-29 14:22:17,297 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=3.04 vs. limit=5.0 2023-04-29 14:22:21,470 INFO [train.py:904] (1/8) Epoch 12, batch 4000, loss[loss=0.1831, simple_loss=0.2631, pruned_loss=0.05156, over 16528.00 frames. ], tot_loss[loss=0.1854, simple_loss=0.2615, pruned_loss=0.05464, over 3280216.67 frames. ], batch size: 68, lr: 5.72e-03, grad_scale: 8.0 2023-04-29 14:22:34,737 INFO [optim.py:368] (1/8) 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 14:22:51,222 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.3838, 4.2812, 4.3642, 2.9174, 3.8012, 4.2817, 3.8734, 2.3538], device='cuda:1'), covar=tensor([0.0440, 0.0026, 0.0024, 0.0290, 0.0056, 0.0063, 0.0056, 0.0342], device='cuda:1'), in_proj_covar=tensor([0.0127, 0.0071, 0.0071, 0.0125, 0.0080, 0.0090, 0.0080, 0.0120], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:1') 2023-04-29 14:22:55,555 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([1.6277, 2.7359, 2.4304, 3.7481, 2.9660, 4.0060, 1.6275, 3.0312], device='cuda:1'), covar=tensor([0.1469, 0.0758, 0.1245, 0.0179, 0.0348, 0.0369, 0.1574, 0.0798], device='cuda:1'), in_proj_covar=tensor([0.0154, 0.0161, 0.0182, 0.0155, 0.0201, 0.0213, 0.0182, 0.0182], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-04-29 14:23:30,739 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.3121, 2.5403, 2.0183, 2.3105, 2.9101, 2.5505, 3.1060, 3.2034], device='cuda:1'), covar=tensor([0.0075, 0.0273, 0.0411, 0.0321, 0.0155, 0.0276, 0.0135, 0.0147], device='cuda:1'), in_proj_covar=tensor([0.0160, 0.0208, 0.0202, 0.0203, 0.0209, 0.0208, 0.0216, 0.0200], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-29 14:23:35,774 INFO [train.py:904] (1/8) Epoch 12, batch 4050, loss[loss=0.1769, simple_loss=0.2611, pruned_loss=0.04639, over 16678.00 frames. ], tot_loss[loss=0.1838, simple_loss=0.2612, pruned_loss=0.05319, over 3277896.87 frames. ], batch size: 76, lr: 5.72e-03, grad_scale: 8.0 2023-04-29 14:24:12,117 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=115726.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 14:24:13,899 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=115727.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 14:24:27,814 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.1856, 1.9618, 2.1991, 3.8337, 1.8392, 2.2958, 2.0896, 2.1071], device='cuda:1'), covar=tensor([0.1254, 0.3990, 0.2539, 0.0532, 0.4770, 0.2793, 0.3527, 0.3814], device='cuda:1'), in_proj_covar=tensor([0.0376, 0.0407, 0.0339, 0.0327, 0.0416, 0.0468, 0.0369, 0.0475], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-29 14:24:35,713 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 12, batch 4100, loss[loss=0.1945, simple_loss=0.2823, pruned_loss=0.05338, over 16521.00 frames. ], tot_loss[loss=0.1838, simple_loss=0.2624, pruned_loss=0.0526, over 3275108.83 frames. ], batch size: 75, lr: 5.72e-03, grad_scale: 8.0 2023-04-29 14:25:05,531 INFO [optim.py:368] (1/8) 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 14:25:28,279 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([1.7730, 2.5928, 2.3051, 3.3911, 2.5674, 3.5494, 1.5270, 2.7541], device='cuda:1'), covar=tensor([0.1202, 0.0623, 0.1164, 0.0136, 0.0197, 0.0400, 0.1453, 0.0750], device='cuda:1'), in_proj_covar=tensor([0.0155, 0.0162, 0.0183, 0.0156, 0.0203, 0.0214, 0.0184, 0.0184], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-04-29 14:25:48,281 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=115790.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 14:26:05,812 INFO [train.py:904] (1/8) Epoch 12, batch 4150, loss[loss=0.2015, simple_loss=0.2925, pruned_loss=0.05532, over 16772.00 frames. ], tot_loss[loss=0.1905, simple_loss=0.2702, pruned_loss=0.05539, over 3247968.69 frames. ], batch size: 83, lr: 5.72e-03, grad_scale: 8.0 2023-04-29 14:27:02,009 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.6462, 3.7372, 2.1623, 4.3012, 2.7798, 4.2349, 2.4381, 2.9362], device='cuda:1'), covar=tensor([0.0220, 0.0294, 0.1554, 0.0105, 0.0716, 0.0350, 0.1346, 0.0682], device='cuda:1'), in_proj_covar=tensor([0.0154, 0.0166, 0.0188, 0.0136, 0.0168, 0.0210, 0.0195, 0.0170], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-04-29 14:27:09,911 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.17 vs. limit=2.0 2023-04-29 14:27:23,037 INFO [train.py:904] (1/8) Epoch 12, batch 4200, loss[loss=0.2039, simple_loss=0.293, pruned_loss=0.05744, over 16659.00 frames. ], tot_loss[loss=0.1961, simple_loss=0.2777, pruned_loss=0.05719, over 3224939.54 frames. ], batch size: 57, lr: 5.72e-03, grad_scale: 8.0 2023-04-29 14:27:37,143 INFO [optim.py:368] (1/8) Clipping_scale=2.0, grad-norm quartiles 1.566e+02 2.546e+02 2.889e+02 3.538e+02 7.743e+02, threshold=5.778e+02, percent-clipped=11.0 2023-04-29 14:27:40,773 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.5004, 3.5834, 3.2525, 3.0523, 3.1792, 3.4745, 3.2974, 3.2144], device='cuda:1'), covar=tensor([0.0558, 0.0520, 0.0263, 0.0270, 0.0640, 0.0424, 0.1127, 0.0480], device='cuda:1'), in_proj_covar=tensor([0.0252, 0.0328, 0.0300, 0.0277, 0.0320, 0.0322, 0.0202, 0.0348], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002], device='cuda:1') 2023-04-29 14:28:03,182 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.7693, 4.6765, 4.8148, 5.0036, 5.1712, 4.5194, 5.1345, 5.1736], device='cuda:1'), covar=tensor([0.1430, 0.0988, 0.1529, 0.0596, 0.0428, 0.0864, 0.0530, 0.0475], device='cuda:1'), in_proj_covar=tensor([0.0539, 0.0673, 0.0804, 0.0683, 0.0514, 0.0532, 0.0533, 0.0617], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-29 14:28:22,466 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=115892.0, num_to_drop=1, layers_to_drop={1} 2023-04-29 14:28:25,123 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.5820, 3.8805, 2.9538, 2.2029, 2.5886, 2.3086, 4.0898, 3.4478], device='cuda:1'), covar=tensor([0.2591, 0.0617, 0.1483, 0.2414, 0.2589, 0.1872, 0.0454, 0.1017], device='cuda:1'), in_proj_covar=tensor([0.0297, 0.0256, 0.0283, 0.0278, 0.0284, 0.0222, 0.0267, 0.0298], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-29 14:28:36,618 INFO [train.py:904] (1/8) Epoch 12, batch 4250, loss[loss=0.1973, simple_loss=0.2887, pruned_loss=0.05296, over 16427.00 frames. ], tot_loss[loss=0.1975, simple_loss=0.281, pruned_loss=0.05696, over 3203857.28 frames. ], batch size: 68, lr: 5.72e-03, grad_scale: 8.0 2023-04-29 14:29:25,798 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.19 vs. limit=2.0 2023-04-29 14:29:49,145 INFO [train.py:904] (1/8) Epoch 12, batch 4300, loss[loss=0.2178, simple_loss=0.2996, pruned_loss=0.06806, over 11896.00 frames. ], tot_loss[loss=0.1969, simple_loss=0.2817, pruned_loss=0.05603, over 3195973.58 frames. ], batch size: 247, lr: 5.71e-03, grad_scale: 8.0 2023-04-29 14:29:50,910 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=115953.0, num_to_drop=1, layers_to_drop={3} 2023-04-29 14:29:54,245 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.1066, 3.3401, 3.4432, 1.9627, 2.8857, 2.3916, 3.3733, 3.5913], device='cuda:1'), covar=tensor([0.0258, 0.0745, 0.0574, 0.1798, 0.0791, 0.0841, 0.0687, 0.0906], device='cuda:1'), in_proj_covar=tensor([0.0145, 0.0149, 0.0158, 0.0144, 0.0136, 0.0124, 0.0138, 0.0159], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:1') 2023-04-29 14:30:04,742 INFO [optim.py:368] (1/8) 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] (1/8) Epoch 12, batch 4350, loss[loss=0.2211, simple_loss=0.2951, pruned_loss=0.07361, over 16393.00 frames. ], tot_loss[loss=0.2002, simple_loss=0.2856, pruned_loss=0.0574, over 3176887.96 frames. ], batch size: 35, lr: 5.71e-03, grad_scale: 8.0 2023-04-29 14:31:12,764 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.7928, 2.4107, 2.5338, 4.6675, 2.3103, 2.8670, 2.4281, 2.6610], device='cuda:1'), covar=tensor([0.0871, 0.2974, 0.2023, 0.0305, 0.3322, 0.1921, 0.2645, 0.2886], device='cuda:1'), in_proj_covar=tensor([0.0373, 0.0404, 0.0335, 0.0324, 0.0417, 0.0464, 0.0366, 0.0471], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-29 14:31:27,458 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=116014.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 14:31:46,026 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=116026.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 14:31:47,880 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 12, batch 4400, loss[loss=0.2185, simple_loss=0.3077, pruned_loss=0.0647, over 16740.00 frames. ], tot_loss[loss=0.2028, simple_loss=0.288, pruned_loss=0.05882, over 3183347.25 frames. ], batch size: 124, lr: 5.71e-03, grad_scale: 8.0 2023-04-29 14:32:37,554 INFO [optim.py:368] (1/8) 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 14:32:54,756 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=116074.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 14:32:56,778 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=116075.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 14:32:56,931 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=116075.0, num_to_drop=1, layers_to_drop={3} 2023-04-29 14:33:35,426 INFO [train.py:904] (1/8) Epoch 12, batch 4450, loss[loss=0.218, simple_loss=0.3081, pruned_loss=0.06395, over 16485.00 frames. ], tot_loss[loss=0.2065, simple_loss=0.2926, pruned_loss=0.06027, over 3198422.32 frames. ], batch size: 68, lr: 5.71e-03, grad_scale: 8.0 2023-04-29 14:33:40,088 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=116105.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 14:33:56,997 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 12, batch 4500, loss[loss=0.2076, simple_loss=0.2949, pruned_loss=0.06012, over 16481.00 frames. ], tot_loss[loss=0.2067, simple_loss=0.2925, pruned_loss=0.06047, over 3201922.97 frames. ], batch size: 68, lr: 5.71e-03, grad_scale: 8.0 2023-04-29 14:35:03,486 INFO [optim.py:368] (1/8) 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 14:35:10,282 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=116166.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 14:35:25,600 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=116177.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 14:36:02,108 INFO [train.py:904] (1/8) Epoch 12, batch 4550, loss[loss=0.2316, simple_loss=0.3163, pruned_loss=0.07338, over 16794.00 frames. ], tot_loss[loss=0.208, simple_loss=0.2932, pruned_loss=0.06136, over 3209956.27 frames. ], batch size: 83, lr: 5.71e-03, grad_scale: 8.0 2023-04-29 14:36:06,809 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.2757, 3.5907, 3.7573, 1.9963, 3.0119, 2.3683, 3.6283, 3.7246], device='cuda:1'), covar=tensor([0.0211, 0.0589, 0.0490, 0.1883, 0.0788, 0.0924, 0.0583, 0.0781], device='cuda:1'), in_proj_covar=tensor([0.0145, 0.0149, 0.0158, 0.0144, 0.0136, 0.0124, 0.0137, 0.0159], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:1') 2023-04-29 14:36:26,522 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.8698, 3.5269, 3.1842, 1.9417, 2.6371, 2.3299, 3.1956, 3.6315], device='cuda:1'), covar=tensor([0.0389, 0.0622, 0.0698, 0.1915, 0.0994, 0.0930, 0.0906, 0.0830], device='cuda:1'), in_proj_covar=tensor([0.0145, 0.0149, 0.0158, 0.0144, 0.0136, 0.0124, 0.0137, 0.0159], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:1') 2023-04-29 14:37:08,830 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=116248.0, num_to_drop=1, layers_to_drop={1} 2023-04-29 14:37:14,080 INFO [train.py:904] (1/8) Epoch 12, batch 4600, loss[loss=0.2002, simple_loss=0.2871, pruned_loss=0.05664, over 16605.00 frames. ], tot_loss[loss=0.2076, simple_loss=0.2934, pruned_loss=0.06083, over 3231816.75 frames. ], batch size: 68, lr: 5.71e-03, grad_scale: 8.0 2023-04-29 14:37:29,433 INFO [optim.py:368] (1/8) Clipping_scale=2.0, grad-norm quartiles 1.395e+02 1.985e+02 2.258e+02 2.658e+02 3.690e+02, threshold=4.517e+02, percent-clipped=0.0 2023-04-29 14:37:50,934 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.38 vs. limit=2.0 2023-04-29 14:38:26,067 INFO [train.py:904] (1/8) Epoch 12, batch 4650, loss[loss=0.1973, simple_loss=0.2808, pruned_loss=0.05689, over 16780.00 frames. ], tot_loss[loss=0.2063, simple_loss=0.2917, pruned_loss=0.06043, over 3247106.47 frames. ], batch size: 83, lr: 5.71e-03, grad_scale: 8.0 2023-04-29 14:38:55,754 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=116323.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 14:39:04,417 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=2.00 vs. limit=2.0 2023-04-29 14:39:24,091 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([1.8614, 2.6690, 2.5131, 1.8641, 2.4695, 2.7139, 2.5528, 1.8718], device='cuda:1'), covar=tensor([0.0314, 0.0052, 0.0044, 0.0294, 0.0082, 0.0091, 0.0077, 0.0279], device='cuda:1'), in_proj_covar=tensor([0.0127, 0.0070, 0.0071, 0.0126, 0.0081, 0.0090, 0.0079, 0.0119], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:1') 2023-04-29 14:39:38,311 INFO [train.py:904] (1/8) Epoch 12, batch 4700, loss[loss=0.1879, simple_loss=0.2693, pruned_loss=0.05325, over 16644.00 frames. ], tot_loss[loss=0.2044, simple_loss=0.2894, pruned_loss=0.05966, over 3224350.39 frames. ], batch size: 62, lr: 5.70e-03, grad_scale: 8.0 2023-04-29 14:39:53,945 INFO [optim.py:368] (1/8) 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 14:40:05,831 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=116370.0, num_to_drop=1, layers_to_drop={0} 2023-04-29 14:40:27,457 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=116384.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 14:40:36,702 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.0969, 3.1827, 1.6114, 3.4762, 2.3721, 3.4731, 1.9224, 2.5425], device='cuda:1'), covar=tensor([0.0283, 0.0379, 0.1883, 0.0121, 0.0839, 0.0509, 0.1622, 0.0754], device='cuda:1'), in_proj_covar=tensor([0.0153, 0.0165, 0.0189, 0.0133, 0.0165, 0.0207, 0.0196, 0.0170], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0003, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-04-29 14:40:53,370 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.7774, 4.5586, 4.8128, 5.0210, 5.1615, 4.6085, 5.1691, 5.1633], device='cuda:1'), covar=tensor([0.1398, 0.1119, 0.1559, 0.0638, 0.0521, 0.0732, 0.0507, 0.0473], device='cuda:1'), in_proj_covar=tensor([0.0527, 0.0660, 0.0790, 0.0676, 0.0502, 0.0521, 0.0523, 0.0607], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-29 14:40:54,105 INFO [train.py:904] (1/8) Epoch 12, batch 4750, loss[loss=0.1912, simple_loss=0.272, pruned_loss=0.05523, over 16722.00 frames. ], tot_loss[loss=0.2, simple_loss=0.2853, pruned_loss=0.05738, over 3231132.55 frames. ], batch size: 57, lr: 5.70e-03, grad_scale: 8.0 2023-04-29 14:41:29,762 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.25 vs. limit=2.0 2023-04-29 14:41:53,541 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=4.78 vs. limit=5.0 2023-04-29 14:41:58,944 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=116446.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 14:42:05,747 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=4.75 vs. limit=5.0 2023-04-29 14:42:07,121 INFO [train.py:904] (1/8) Epoch 12, batch 4800, loss[loss=0.1906, simple_loss=0.2848, pruned_loss=0.0482, over 15408.00 frames. ], tot_loss[loss=0.1953, simple_loss=0.2808, pruned_loss=0.05494, over 3246298.77 frames. ], batch size: 190, lr: 5.70e-03, grad_scale: 8.0 2023-04-29 14:42:22,123 INFO [zipformer.py:625] (1/8) 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] (1/8) 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] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=116472.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 14:42:48,715 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.89 vs. limit=2.0 2023-04-29 14:43:23,413 INFO [train.py:904] (1/8) Epoch 12, batch 4850, loss[loss=0.2001, simple_loss=0.2892, pruned_loss=0.05555, over 16692.00 frames. ], tot_loss[loss=0.195, simple_loss=0.2815, pruned_loss=0.05426, over 3232841.54 frames. ], batch size: 134, lr: 5.70e-03, grad_scale: 8.0 2023-04-29 14:43:28,299 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=116505.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 14:43:30,746 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=116507.0, num_to_drop=1, layers_to_drop={1} 2023-04-29 14:43:50,300 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.0073, 3.9729, 3.9655, 3.2792, 3.9200, 1.7837, 3.7597, 3.5637], device='cuda:1'), covar=tensor([0.0101, 0.0092, 0.0110, 0.0331, 0.0084, 0.2290, 0.0122, 0.0202], device='cuda:1'), in_proj_covar=tensor([0.0131, 0.0119, 0.0164, 0.0155, 0.0136, 0.0178, 0.0154, 0.0153], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-29 14:44:07,645 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.54 vs. limit=2.0 2023-04-29 14:44:31,792 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=116548.0, num_to_drop=1, layers_to_drop={2} 2023-04-29 14:44:38,143 INFO [train.py:904] (1/8) Epoch 12, batch 4900, loss[loss=0.1853, simple_loss=0.2777, pruned_loss=0.04648, over 16437.00 frames. ], tot_loss[loss=0.1934, simple_loss=0.2807, pruned_loss=0.053, over 3223352.98 frames. ], batch size: 146, lr: 5.70e-03, grad_scale: 8.0 2023-04-29 14:44:52,629 INFO [optim.py:368] (1/8) 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 14:44:58,595 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=116566.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 14:45:42,263 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 12, batch 4950, loss[loss=0.1876, simple_loss=0.2803, pruned_loss=0.04738, over 16718.00 frames. ], tot_loss[loss=0.1929, simple_loss=0.2805, pruned_loss=0.05267, over 3222636.09 frames. ], batch size: 83, lr: 5.70e-03, grad_scale: 8.0 2023-04-29 14:46:16,317 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=116618.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 14:46:18,620 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([5.1911, 5.1436, 4.9816, 4.7045, 4.5546, 5.0630, 5.0665, 4.7459], device='cuda:1'), covar=tensor([0.0557, 0.0484, 0.0298, 0.0260, 0.1156, 0.0411, 0.0235, 0.0660], device='cuda:1'), in_proj_covar=tensor([0.0242, 0.0320, 0.0291, 0.0270, 0.0311, 0.0311, 0.0198, 0.0338], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-29 14:47:04,272 INFO [train.py:904] (1/8) Epoch 12, batch 5000, loss[loss=0.1873, simple_loss=0.2713, pruned_loss=0.0517, over 17207.00 frames. ], tot_loss[loss=0.1942, simple_loss=0.2827, pruned_loss=0.05291, over 3225287.11 frames. ], batch size: 44, lr: 5.70e-03, grad_scale: 8.0 2023-04-29 14:47:17,030 INFO [optim.py:368] (1/8) 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 14:47:30,331 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=116670.0, num_to_drop=1, layers_to_drop={1} 2023-04-29 14:47:42,665 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=116679.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 14:47:42,797 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=116679.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 14:48:15,688 INFO [train.py:904] (1/8) Epoch 12, batch 5050, loss[loss=0.1786, simple_loss=0.2606, pruned_loss=0.04829, over 17142.00 frames. ], tot_loss[loss=0.1937, simple_loss=0.2825, pruned_loss=0.05248, over 3236600.59 frames. ], batch size: 48, lr: 5.70e-03, grad_scale: 8.0 2023-04-29 14:48:18,345 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=116704.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 14:48:38,167 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=116718.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 14:49:24,609 INFO [train.py:904] (1/8) Epoch 12, batch 5100, loss[loss=0.1761, simple_loss=0.2566, pruned_loss=0.04779, over 17039.00 frames. ], tot_loss[loss=0.1922, simple_loss=0.2804, pruned_loss=0.05199, over 3233260.78 frames. ], batch size: 53, lr: 5.69e-03, grad_scale: 8.0 2023-04-29 14:49:36,748 INFO [zipformer.py:625] (1/8) 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] (1/8) 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 14:49:43,660 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=116765.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 14:49:52,546 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=116772.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 14:49:58,663 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.3075, 4.3594, 4.7207, 4.7242, 4.7011, 4.4091, 4.3807, 4.2400], device='cuda:1'), covar=tensor([0.0298, 0.0523, 0.0338, 0.0345, 0.0434, 0.0311, 0.0895, 0.0434], device='cuda:1'), in_proj_covar=tensor([0.0340, 0.0356, 0.0356, 0.0335, 0.0406, 0.0375, 0.0482, 0.0301], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:1') 2023-04-29 14:50:35,772 INFO [train.py:904] (1/8) Epoch 12, batch 5150, loss[loss=0.2218, simple_loss=0.3158, pruned_loss=0.06392, over 16392.00 frames. ], tot_loss[loss=0.1916, simple_loss=0.2809, pruned_loss=0.05113, over 3233899.55 frames. ], batch size: 146, lr: 5.69e-03, grad_scale: 8.0 2023-04-29 14:50:36,702 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=116802.0, num_to_drop=1, layers_to_drop={2} 2023-04-29 14:50:36,782 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=116802.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 14:50:47,786 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=116809.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 14:51:03,102 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=116820.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 14:51:11,792 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.38 vs. limit=2.0 2023-04-29 14:51:15,170 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.0939, 1.9910, 2.0929, 3.6965, 1.9534, 2.3364, 2.1106, 2.1673], device='cuda:1'), covar=tensor([0.1160, 0.3476, 0.2380, 0.0429, 0.3823, 0.2366, 0.3270, 0.2903], device='cuda:1'), in_proj_covar=tensor([0.0366, 0.0397, 0.0331, 0.0318, 0.0412, 0.0455, 0.0361, 0.0460], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-29 14:51:47,915 INFO [train.py:904] (1/8) Epoch 12, batch 5200, loss[loss=0.1984, simple_loss=0.2966, pruned_loss=0.0501, over 16736.00 frames. ], tot_loss[loss=0.1906, simple_loss=0.2796, pruned_loss=0.0508, over 3247678.90 frames. ], batch size: 134, lr: 5.69e-03, grad_scale: 8.0 2023-04-29 14:52:00,670 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=116861.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 14:52:01,643 INFO [optim.py:368] (1/8) 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 14:52:03,384 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=116863.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 14:52:58,582 INFO [train.py:904] (1/8) Epoch 12, batch 5250, loss[loss=0.1753, simple_loss=0.267, pruned_loss=0.04182, over 16741.00 frames. ], tot_loss[loss=0.1894, simple_loss=0.2771, pruned_loss=0.0508, over 3250656.53 frames. ], batch size: 89, lr: 5.69e-03, grad_scale: 16.0 2023-04-29 14:54:01,838 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.22 vs. limit=2.0 2023-04-29 14:54:11,418 INFO [train.py:904] (1/8) Epoch 12, batch 5300, loss[loss=0.173, simple_loss=0.2519, pruned_loss=0.04704, over 16199.00 frames. ], tot_loss[loss=0.1861, simple_loss=0.2734, pruned_loss=0.0494, over 3247413.71 frames. ], batch size: 35, lr: 5.69e-03, grad_scale: 8.0 2023-04-29 14:54:27,261 INFO [optim.py:368] (1/8) Clipping_scale=2.0, grad-norm quartiles 1.433e+02 2.208e+02 2.588e+02 3.045e+02 5.450e+02, threshold=5.175e+02, percent-clipped=1.0 2023-04-29 14:54:42,027 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=116974.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 14:54:48,031 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.6021, 4.6037, 4.4323, 3.7872, 4.4474, 1.5010, 4.1978, 4.1853], device='cuda:1'), covar=tensor([0.0078, 0.0065, 0.0123, 0.0375, 0.0085, 0.2545, 0.0114, 0.0184], device='cuda:1'), in_proj_covar=tensor([0.0131, 0.0120, 0.0164, 0.0157, 0.0137, 0.0179, 0.0154, 0.0154], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-29 14:54:49,814 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=116979.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 14:55:21,993 INFO [train.py:904] (1/8) Epoch 12, batch 5350, loss[loss=0.1986, simple_loss=0.2905, pruned_loss=0.05333, over 15388.00 frames. ], tot_loss[loss=0.1849, simple_loss=0.2718, pruned_loss=0.049, over 3241168.88 frames. ], batch size: 190, lr: 5.69e-03, grad_scale: 4.0 2023-04-29 14:55:58,442 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 12, batch 5400, loss[loss=0.2152, simple_loss=0.2983, pruned_loss=0.06602, over 12064.00 frames. ], tot_loss[loss=0.1867, simple_loss=0.2743, pruned_loss=0.04949, over 3234061.35 frames. ], batch size: 248, lr: 5.69e-03, grad_scale: 4.0 2023-04-29 14:56:43,918 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=117060.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 14:56:49,065 INFO [optim.py:368] (1/8) 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] (1/8) Epoch 12, batch 5450, loss[loss=0.226, simple_loss=0.3114, pruned_loss=0.07031, over 16135.00 frames. ], tot_loss[loss=0.1909, simple_loss=0.2785, pruned_loss=0.05161, over 3218038.22 frames. ], batch size: 165, lr: 5.69e-03, grad_scale: 4.0 2023-04-29 14:57:46,742 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=117102.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 14:57:54,835 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.9196, 4.0237, 3.8126, 3.6288, 3.5247, 3.9395, 3.6124, 3.6823], device='cuda:1'), covar=tensor([0.0637, 0.0627, 0.0348, 0.0283, 0.0895, 0.0498, 0.1249, 0.0642], device='cuda:1'), in_proj_covar=tensor([0.0250, 0.0330, 0.0300, 0.0278, 0.0319, 0.0324, 0.0203, 0.0348], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002], device='cuda:1') 2023-04-29 14:58:57,078 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=117150.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 14:59:00,350 INFO [train.py:904] (1/8) Epoch 12, batch 5500, loss[loss=0.2179, simple_loss=0.2998, pruned_loss=0.06798, over 16324.00 frames. ], tot_loss[loss=0.2, simple_loss=0.2864, pruned_loss=0.05677, over 3187392.04 frames. ], batch size: 146, lr: 5.69e-03, grad_scale: 4.0 2023-04-29 14:59:09,312 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=117158.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 14:59:13,807 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=117161.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 14:59:18,171 INFO [optim.py:368] (1/8) 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 14:59:47,075 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.4007, 3.3425, 3.3866, 3.5155, 3.5202, 3.2333, 3.4895, 3.5659], device='cuda:1'), covar=tensor([0.1052, 0.0875, 0.1107, 0.0547, 0.0631, 0.2436, 0.0873, 0.0733], device='cuda:1'), in_proj_covar=tensor([0.0527, 0.0663, 0.0796, 0.0671, 0.0503, 0.0520, 0.0523, 0.0610], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-29 15:00:18,009 INFO [train.py:904] (1/8) Epoch 12, batch 5550, loss[loss=0.2061, simple_loss=0.2869, pruned_loss=0.06262, over 16329.00 frames. ], tot_loss[loss=0.2096, simple_loss=0.294, pruned_loss=0.06257, over 3161189.03 frames. ], batch size: 35, lr: 5.68e-03, grad_scale: 4.0 2023-04-29 15:00:30,330 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=117209.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 15:00:43,456 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.66 vs. limit=2.0 2023-04-29 15:00:49,811 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=117221.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 15:01:39,143 INFO [train.py:904] (1/8) Epoch 12, batch 5600, loss[loss=0.3139, simple_loss=0.3629, pruned_loss=0.1324, over 11092.00 frames. ], tot_loss[loss=0.2172, simple_loss=0.2997, pruned_loss=0.06738, over 3112570.11 frames. ], batch size: 246, lr: 5.68e-03, grad_scale: 8.0 2023-04-29 15:01:56,335 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.2793, 2.0244, 1.7163, 1.7702, 2.2184, 1.9278, 2.1038, 2.3918], device='cuda:1'), covar=tensor([0.0122, 0.0246, 0.0350, 0.0336, 0.0168, 0.0249, 0.0155, 0.0162], device='cuda:1'), in_proj_covar=tensor([0.0155, 0.0205, 0.0200, 0.0199, 0.0205, 0.0203, 0.0210, 0.0198], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-29 15:01:58,901 INFO [optim.py:368] (1/8) 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 15:02:16,218 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=117274.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 15:02:30,331 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=117282.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 15:03:02,068 INFO [train.py:904] (1/8) Epoch 12, batch 5650, loss[loss=0.2116, simple_loss=0.2843, pruned_loss=0.06947, over 16295.00 frames. ], tot_loss[loss=0.2235, simple_loss=0.3048, pruned_loss=0.07115, over 3086072.98 frames. ], batch size: 35, lr: 5.68e-03, grad_scale: 4.0 2023-04-29 15:03:10,845 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.6904, 2.3186, 1.9538, 2.1102, 2.6426, 2.3557, 2.6885, 2.8435], device='cuda:1'), covar=tensor([0.0118, 0.0301, 0.0373, 0.0365, 0.0169, 0.0272, 0.0155, 0.0186], device='cuda:1'), in_proj_covar=tensor([0.0155, 0.0206, 0.0201, 0.0200, 0.0206, 0.0204, 0.0211, 0.0199], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-29 15:03:32,807 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([1.6736, 1.7656, 1.6095, 1.5794, 1.8455, 1.5781, 1.6085, 1.8934], device='cuda:1'), covar=tensor([0.0131, 0.0179, 0.0265, 0.0235, 0.0134, 0.0173, 0.0138, 0.0140], device='cuda:1'), in_proj_covar=tensor([0.0155, 0.0206, 0.0201, 0.0200, 0.0205, 0.0203, 0.0210, 0.0198], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-29 15:03:33,930 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=117322.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 15:03:45,002 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=117329.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 15:03:53,232 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.1423, 3.1897, 1.8742, 3.4879, 2.3544, 3.4979, 1.9410, 2.5997], device='cuda:1'), covar=tensor([0.0251, 0.0394, 0.1610, 0.0185, 0.0860, 0.0484, 0.1562, 0.0758], device='cuda:1'), in_proj_covar=tensor([0.0152, 0.0166, 0.0189, 0.0131, 0.0167, 0.0206, 0.0197, 0.0171], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0003, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-04-29 15:04:08,964 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.3478, 4.3492, 4.2076, 3.5975, 4.2611, 1.7977, 4.0787, 3.9618], device='cuda:1'), covar=tensor([0.0080, 0.0063, 0.0123, 0.0281, 0.0069, 0.2345, 0.0096, 0.0158], device='cuda:1'), in_proj_covar=tensor([0.0129, 0.0117, 0.0161, 0.0155, 0.0134, 0.0176, 0.0150, 0.0150], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-29 15:04:14,565 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.0599, 3.9065, 4.0847, 4.2512, 4.3330, 3.9212, 4.2367, 4.3324], device='cuda:1'), covar=tensor([0.1416, 0.1107, 0.1342, 0.0569, 0.0559, 0.1381, 0.0711, 0.0567], device='cuda:1'), in_proj_covar=tensor([0.0519, 0.0655, 0.0784, 0.0663, 0.0498, 0.0516, 0.0519, 0.0602], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-29 15:04:18,926 INFO [train.py:904] (1/8) Epoch 12, batch 5700, loss[loss=0.2897, simple_loss=0.3427, pruned_loss=0.1184, over 11238.00 frames. ], tot_loss[loss=0.2273, simple_loss=0.3075, pruned_loss=0.07355, over 3069243.52 frames. ], batch size: 248, lr: 5.68e-03, grad_scale: 2.0 2023-04-29 15:04:32,704 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=117360.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 15:04:41,587 INFO [optim.py:368] (1/8) 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 15:05:21,405 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=117390.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 15:05:39,179 INFO [train.py:904] (1/8) Epoch 12, batch 5750, loss[loss=0.2319, simple_loss=0.3173, pruned_loss=0.07323, over 15362.00 frames. ], tot_loss[loss=0.2304, simple_loss=0.3102, pruned_loss=0.07525, over 3053393.42 frames. ], batch size: 190, lr: 5.68e-03, grad_scale: 2.0 2023-04-29 15:05:48,655 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=117408.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 15:06:53,755 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.5121, 3.5853, 2.0838, 3.8874, 2.6555, 3.9269, 2.2463, 2.8944], device='cuda:1'), covar=tensor([0.0203, 0.0335, 0.1471, 0.0235, 0.0701, 0.0434, 0.1420, 0.0614], device='cuda:1'), in_proj_covar=tensor([0.0151, 0.0165, 0.0189, 0.0131, 0.0166, 0.0205, 0.0196, 0.0170], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0003, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-04-29 15:07:00,135 INFO [train.py:904] (1/8) Epoch 12, batch 5800, loss[loss=0.1942, simple_loss=0.2913, pruned_loss=0.04856, over 16736.00 frames. ], tot_loss[loss=0.2304, simple_loss=0.3108, pruned_loss=0.07499, over 3030575.70 frames. ], batch size: 83, lr: 5.68e-03, grad_scale: 2.0 2023-04-29 15:07:09,799 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=117458.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 15:07:21,329 INFO [optim.py:368] (1/8) 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] (1/8) Epoch 12, batch 5850, loss[loss=0.2215, simple_loss=0.3118, pruned_loss=0.06557, over 16436.00 frames. ], tot_loss[loss=0.227, simple_loss=0.3085, pruned_loss=0.07273, over 3048182.55 frames. ], batch size: 146, lr: 5.68e-03, grad_scale: 2.0 2023-04-29 15:08:24,194 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=117506.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 15:09:37,157 INFO [train.py:904] (1/8) Epoch 12, batch 5900, loss[loss=0.2186, simple_loss=0.3117, pruned_loss=0.0627, over 16578.00 frames. ], tot_loss[loss=0.2258, simple_loss=0.3077, pruned_loss=0.07196, over 3077656.23 frames. ], batch size: 75, lr: 5.68e-03, grad_scale: 2.0 2023-04-29 15:10:01,561 INFO [optim.py:368] (1/8) 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 15:10:18,439 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 12, batch 5950, loss[loss=0.2156, simple_loss=0.3008, pruned_loss=0.06521, over 11533.00 frames. ], tot_loss[loss=0.2246, simple_loss=0.3083, pruned_loss=0.07045, over 3084453.97 frames. ], batch size: 246, lr: 5.67e-03, grad_scale: 2.0 2023-04-29 15:12:14,103 INFO [train.py:904] (1/8) Epoch 12, batch 6000, loss[loss=0.2065, simple_loss=0.2903, pruned_loss=0.06132, over 16189.00 frames. ], tot_loss[loss=0.2228, simple_loss=0.3066, pruned_loss=0.06948, over 3097050.97 frames. ], batch size: 165, lr: 5.67e-03, grad_scale: 4.0 2023-04-29 15:12:14,104 INFO [train.py:929] (1/8) Computing validation loss 2023-04-29 15:12:25,320 INFO [train.py:938] (1/8) Epoch 12, validation: loss=0.161, simple_loss=0.2739, pruned_loss=0.02405, over 944034.00 frames. 2023-04-29 15:12:25,320 INFO [train.py:939] (1/8) Maximum memory allocated so far is 17845MB 2023-04-29 15:12:46,505 INFO [optim.py:368] (1/8) 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 15:13:18,109 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 12, batch 6050, loss[loss=0.2219, simple_loss=0.3102, pruned_loss=0.06675, over 16882.00 frames. ], tot_loss[loss=0.2212, simple_loss=0.3052, pruned_loss=0.06862, over 3106673.14 frames. ], batch size: 116, lr: 5.67e-03, grad_scale: 4.0 2023-04-29 15:15:02,188 INFO [train.py:904] (1/8) Epoch 12, batch 6100, loss[loss=0.2235, simple_loss=0.312, pruned_loss=0.0675, over 16892.00 frames. ], tot_loss[loss=0.22, simple_loss=0.3045, pruned_loss=0.06774, over 3111813.79 frames. ], batch size: 109, lr: 5.67e-03, grad_scale: 4.0 2023-04-29 15:15:03,176 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.2268, 2.0053, 1.7362, 1.7525, 2.2411, 1.8793, 2.1281, 2.4069], device='cuda:1'), covar=tensor([0.0126, 0.0269, 0.0354, 0.0338, 0.0163, 0.0242, 0.0148, 0.0157], device='cuda:1'), in_proj_covar=tensor([0.0155, 0.0204, 0.0200, 0.0199, 0.0204, 0.0202, 0.0210, 0.0196], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-29 15:15:15,859 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.6129, 3.8342, 4.3459, 2.0838, 4.3896, 4.4133, 3.2963, 3.2372], device='cuda:1'), covar=tensor([0.0780, 0.0188, 0.0124, 0.1147, 0.0059, 0.0108, 0.0310, 0.0428], device='cuda:1'), in_proj_covar=tensor([0.0145, 0.0101, 0.0089, 0.0139, 0.0071, 0.0106, 0.0122, 0.0129], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:1') 2023-04-29 15:15:24,819 INFO [optim.py:368] (1/8) Clipping_scale=2.0, grad-norm quartiles 1.907e+02 2.974e+02 3.801e+02 4.781e+02 1.516e+03, threshold=7.602e+02, percent-clipped=11.0 2023-04-29 15:16:19,691 INFO [train.py:904] (1/8) Epoch 12, batch 6150, loss[loss=0.1945, simple_loss=0.2876, pruned_loss=0.05065, over 16645.00 frames. ], tot_loss[loss=0.2194, simple_loss=0.3028, pruned_loss=0.06803, over 3088779.44 frames. ], batch size: 76, lr: 5.67e-03, grad_scale: 4.0 2023-04-29 15:17:21,814 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.0728, 3.6668, 3.6136, 2.3044, 3.4004, 3.6481, 3.4731, 2.1700], device='cuda:1'), covar=tensor([0.0454, 0.0031, 0.0039, 0.0351, 0.0066, 0.0083, 0.0060, 0.0340], device='cuda:1'), in_proj_covar=tensor([0.0127, 0.0068, 0.0070, 0.0124, 0.0080, 0.0090, 0.0079, 0.0117], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:1') 2023-04-29 15:17:38,950 INFO [train.py:904] (1/8) Epoch 12, batch 6200, loss[loss=0.2414, simple_loss=0.306, pruned_loss=0.08843, over 11402.00 frames. ], tot_loss[loss=0.2189, simple_loss=0.3014, pruned_loss=0.06825, over 3059197.38 frames. ], batch size: 248, lr: 5.67e-03, grad_scale: 4.0 2023-04-29 15:18:00,666 INFO [optim.py:368] (1/8) 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 15:18:10,274 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.8924, 4.1788, 3.9655, 4.0034, 3.6756, 3.7581, 3.8364, 4.1386], device='cuda:1'), covar=tensor([0.1027, 0.0847, 0.0960, 0.0709, 0.0737, 0.1461, 0.0888, 0.0989], device='cuda:1'), in_proj_covar=tensor([0.0554, 0.0690, 0.0569, 0.0487, 0.0438, 0.0449, 0.0571, 0.0532], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:1') 2023-04-29 15:18:18,016 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=117877.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 15:18:31,393 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.3599, 4.2251, 4.4216, 4.6310, 4.7382, 4.3150, 4.6874, 4.7694], device='cuda:1'), covar=tensor([0.1663, 0.1212, 0.1711, 0.0660, 0.0597, 0.0993, 0.0700, 0.0556], device='cuda:1'), in_proj_covar=tensor([0.0528, 0.0660, 0.0792, 0.0673, 0.0507, 0.0518, 0.0532, 0.0609], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-29 15:18:52,884 INFO [train.py:904] (1/8) Epoch 12, batch 6250, loss[loss=0.2061, simple_loss=0.3047, pruned_loss=0.05377, over 16537.00 frames. ], tot_loss[loss=0.2181, simple_loss=0.3008, pruned_loss=0.06773, over 3066190.94 frames. ], batch size: 68, lr: 5.67e-03, grad_scale: 4.0 2023-04-29 15:19:27,991 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=117925.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 15:20:06,469 INFO [train.py:904] (1/8) Epoch 12, batch 6300, loss[loss=0.2052, simple_loss=0.2945, pruned_loss=0.05797, over 16799.00 frames. ], tot_loss[loss=0.2178, simple_loss=0.3011, pruned_loss=0.06725, over 3072069.26 frames. ], batch size: 83, lr: 5.67e-03, grad_scale: 4.0 2023-04-29 15:20:22,909 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.2958, 2.2243, 2.2435, 4.2068, 2.0639, 2.6431, 2.2536, 2.3585], device='cuda:1'), covar=tensor([0.1068, 0.3295, 0.2433, 0.0378, 0.3903, 0.2185, 0.3199, 0.2919], device='cuda:1'), in_proj_covar=tensor([0.0368, 0.0395, 0.0332, 0.0318, 0.0414, 0.0457, 0.0362, 0.0463], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-29 15:20:28,835 INFO [optim.py:368] (1/8) 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 15:20:44,573 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([5.0392, 5.0310, 4.8517, 4.5961, 4.4922, 4.9198, 4.9109, 4.6347], device='cuda:1'), covar=tensor([0.0508, 0.0379, 0.0271, 0.0259, 0.0987, 0.0411, 0.0280, 0.0616], device='cuda:1'), in_proj_covar=tensor([0.0242, 0.0323, 0.0289, 0.0269, 0.0308, 0.0311, 0.0198, 0.0338], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-29 15:20:57,532 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=117985.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 15:21:25,211 INFO [train.py:904] (1/8) Epoch 12, batch 6350, loss[loss=0.2745, simple_loss=0.329, pruned_loss=0.11, over 11531.00 frames. ], tot_loss[loss=0.2192, simple_loss=0.3015, pruned_loss=0.0684, over 3079392.22 frames. ], batch size: 248, lr: 5.66e-03, grad_scale: 4.0 2023-04-29 15:22:11,807 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=118033.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 15:22:21,783 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=118040.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 15:22:37,559 INFO [train.py:904] (1/8) Epoch 12, batch 6400, loss[loss=0.2043, simple_loss=0.2815, pruned_loss=0.06355, over 16901.00 frames. ], tot_loss[loss=0.2199, simple_loss=0.3016, pruned_loss=0.0691, over 3073224.03 frames. ], batch size: 96, lr: 5.66e-03, grad_scale: 8.0 2023-04-29 15:22:57,664 INFO [optim.py:368] (1/8) 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 15:23:49,769 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 12, batch 6450, loss[loss=0.2146, simple_loss=0.3047, pruned_loss=0.06221, over 16883.00 frames. ], tot_loss[loss=0.2183, simple_loss=0.3009, pruned_loss=0.0679, over 3067832.53 frames. ], batch size: 90, lr: 5.66e-03, grad_scale: 8.0 2023-04-29 15:25:08,003 INFO [train.py:904] (1/8) Epoch 12, batch 6500, loss[loss=0.2471, simple_loss=0.3047, pruned_loss=0.09474, over 11646.00 frames. ], tot_loss[loss=0.2166, simple_loss=0.2986, pruned_loss=0.06727, over 3073228.44 frames. ], batch size: 250, lr: 5.66e-03, grad_scale: 8.0 2023-04-29 15:25:29,348 INFO [optim.py:368] (1/8) 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 15:25:32,701 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=118168.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 15:26:28,541 INFO [train.py:904] (1/8) Epoch 12, batch 6550, loss[loss=0.2365, simple_loss=0.3293, pruned_loss=0.07181, over 16352.00 frames. ], tot_loss[loss=0.2186, simple_loss=0.3013, pruned_loss=0.068, over 3076823.18 frames. ], batch size: 146, lr: 5.66e-03, grad_scale: 8.0 2023-04-29 15:26:39,524 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([5.0873, 5.6956, 5.8395, 5.4871, 5.6119, 6.1710, 5.6493, 5.4040], device='cuda:1'), covar=tensor([0.0782, 0.1489, 0.2125, 0.1805, 0.2175, 0.0873, 0.1411, 0.2070], device='cuda:1'), in_proj_covar=tensor([0.0357, 0.0495, 0.0545, 0.0429, 0.0580, 0.0568, 0.0432, 0.0584], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-04-29 15:27:10,794 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=118229.0, num_to_drop=1, layers_to_drop={0} 2023-04-29 15:27:44,389 INFO [train.py:904] (1/8) Epoch 12, batch 6600, loss[loss=0.2089, simple_loss=0.3001, pruned_loss=0.05879, over 16846.00 frames. ], tot_loss[loss=0.2211, simple_loss=0.3039, pruned_loss=0.06911, over 3069963.46 frames. ], batch size: 102, lr: 5.66e-03, grad_scale: 8.0 2023-04-29 15:27:57,263 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.44 vs. limit=2.0 2023-04-29 15:28:05,473 INFO [optim.py:368] (1/8) 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 15:28:10,199 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=118269.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 15:28:41,331 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=118289.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 15:29:01,041 INFO [train.py:904] (1/8) Epoch 12, batch 6650, loss[loss=0.1855, simple_loss=0.2723, pruned_loss=0.04934, over 16657.00 frames. ], tot_loss[loss=0.2196, simple_loss=0.3025, pruned_loss=0.06837, over 3093687.58 frames. ], batch size: 62, lr: 5.66e-03, grad_scale: 8.0 2023-04-29 15:29:43,611 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=118329.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 15:29:45,503 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=118330.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 15:29:51,136 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.4102, 2.1499, 2.2817, 4.1111, 2.1100, 2.6400, 2.3042, 2.3455], device='cuda:1'), covar=tensor([0.0973, 0.3047, 0.2294, 0.0391, 0.3716, 0.2066, 0.2865, 0.3008], device='cuda:1'), in_proj_covar=tensor([0.0363, 0.0392, 0.0329, 0.0315, 0.0410, 0.0452, 0.0359, 0.0458], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-29 15:30:16,331 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=118350.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 15:30:18,844 INFO [train.py:904] (1/8) Epoch 12, batch 6700, loss[loss=0.2109, simple_loss=0.2918, pruned_loss=0.06501, over 17205.00 frames. ], tot_loss[loss=0.2192, simple_loss=0.3017, pruned_loss=0.06839, over 3094743.37 frames. ], batch size: 44, lr: 5.66e-03, grad_scale: 8.0 2023-04-29 15:30:22,029 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=3.12 vs. limit=5.0 2023-04-29 15:30:39,918 INFO [optim.py:368] (1/8) Clipping_scale=2.0, grad-norm quartiles 1.931e+02 3.120e+02 3.560e+02 4.149e+02 6.786e+02, threshold=7.119e+02, percent-clipped=0.0 2023-04-29 15:31:17,487 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=118390.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 15:31:26,557 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 12, batch 6750, loss[loss=0.2414, simple_loss=0.307, pruned_loss=0.08791, over 12149.00 frames. ], tot_loss[loss=0.2185, simple_loss=0.3006, pruned_loss=0.06818, over 3097839.60 frames. ], batch size: 246, lr: 5.66e-03, grad_scale: 8.0 2023-04-29 15:32:10,738 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.16 vs. limit=2.0 2023-04-29 15:32:49,915 INFO [train.py:904] (1/8) Epoch 12, batch 6800, loss[loss=0.2142, simple_loss=0.2953, pruned_loss=0.06651, over 16508.00 frames. ], tot_loss[loss=0.2182, simple_loss=0.3005, pruned_loss=0.06798, over 3100683.04 frames. ], batch size: 68, lr: 5.65e-03, grad_scale: 8.0 2023-04-29 15:33:11,648 INFO [optim.py:368] (1/8) Clipping_scale=2.0, grad-norm quartiles 1.917e+02 3.071e+02 3.777e+02 4.759e+02 7.416e+02, threshold=7.554e+02, percent-clipped=1.0 2023-04-29 15:33:14,503 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([5.2582, 5.5878, 5.3085, 5.2750, 4.9781, 4.8493, 4.9922, 5.6227], device='cuda:1'), covar=tensor([0.0961, 0.0629, 0.0843, 0.0660, 0.0697, 0.0776, 0.1026, 0.0818], device='cuda:1'), in_proj_covar=tensor([0.0558, 0.0692, 0.0571, 0.0490, 0.0440, 0.0455, 0.0574, 0.0541], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:1') 2023-04-29 15:33:26,394 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=2.88 vs. limit=5.0 2023-04-29 15:34:02,537 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.64 vs. limit=2.0 2023-04-29 15:34:04,806 INFO [train.py:904] (1/8) Epoch 12, batch 6850, loss[loss=0.228, simple_loss=0.3306, pruned_loss=0.06265, over 16473.00 frames. ], tot_loss[loss=0.2209, simple_loss=0.3031, pruned_loss=0.06934, over 3097756.23 frames. ], batch size: 146, lr: 5.65e-03, grad_scale: 8.0 2023-04-29 15:34:36,245 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 12, batch 6900, loss[loss=0.2153, simple_loss=0.3051, pruned_loss=0.06276, over 16710.00 frames. ], tot_loss[loss=0.2207, simple_loss=0.3048, pruned_loss=0.06832, over 3116652.06 frames. ], batch size: 89, lr: 5.65e-03, grad_scale: 8.0 2023-04-29 15:35:36,846 INFO [optim.py:368] (1/8) 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 15:36:16,808 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.52 vs. limit=2.0 2023-04-29 15:36:30,543 INFO [train.py:904] (1/8) Epoch 12, batch 6950, loss[loss=0.2353, simple_loss=0.3156, pruned_loss=0.07752, over 16729.00 frames. ], tot_loss[loss=0.2246, simple_loss=0.3077, pruned_loss=0.07073, over 3103170.08 frames. ], batch size: 134, lr: 5.65e-03, grad_scale: 8.0 2023-04-29 15:36:43,503 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([5.1681, 5.4946, 5.2037, 5.1841, 4.9520, 4.8340, 4.8965, 5.5605], device='cuda:1'), covar=tensor([0.1008, 0.0718, 0.0962, 0.0725, 0.0737, 0.0736, 0.1076, 0.0768], device='cuda:1'), in_proj_covar=tensor([0.0555, 0.0691, 0.0570, 0.0488, 0.0439, 0.0453, 0.0574, 0.0535], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:1') 2023-04-29 15:37:04,562 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=118625.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 15:37:14,717 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=118631.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 15:37:34,928 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=118645.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 15:37:44,792 INFO [train.py:904] (1/8) Epoch 12, batch 7000, loss[loss=0.242, simple_loss=0.3111, pruned_loss=0.08649, over 11549.00 frames. ], tot_loss[loss=0.2242, simple_loss=0.3077, pruned_loss=0.0704, over 3077862.65 frames. ], batch size: 247, lr: 5.65e-03, grad_scale: 8.0 2023-04-29 15:38:05,443 INFO [optim.py:368] (1/8) 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 15:38:33,777 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=118685.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 15:38:44,049 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=118692.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 15:38:50,369 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=118696.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 15:38:59,534 INFO [train.py:904] (1/8) Epoch 12, batch 7050, loss[loss=0.2236, simple_loss=0.3062, pruned_loss=0.07049, over 16638.00 frames. ], tot_loss[loss=0.2237, simple_loss=0.3079, pruned_loss=0.06977, over 3110630.35 frames. ], batch size: 62, lr: 5.65e-03, grad_scale: 8.0 2023-04-29 15:40:01,768 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 12, batch 7100, loss[loss=0.2183, simple_loss=0.3066, pruned_loss=0.06495, over 16833.00 frames. ], tot_loss[loss=0.223, simple_loss=0.3065, pruned_loss=0.06976, over 3087694.91 frames. ], batch size: 102, lr: 5.65e-03, grad_scale: 8.0 2023-04-29 15:40:36,859 INFO [optim.py:368] (1/8) 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 15:41:24,016 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.63 vs. limit=2.0 2023-04-29 15:41:29,296 INFO [train.py:904] (1/8) Epoch 12, batch 7150, loss[loss=0.2173, simple_loss=0.2962, pruned_loss=0.06913, over 16966.00 frames. ], tot_loss[loss=0.2215, simple_loss=0.3043, pruned_loss=0.06939, over 3099198.31 frames. ], batch size: 41, lr: 5.65e-03, grad_scale: 8.0 2023-04-29 15:41:34,753 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.77 vs. limit=2.0 2023-04-29 15:42:01,925 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=118824.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 15:42:41,519 INFO [train.py:904] (1/8) Epoch 12, batch 7200, loss[loss=0.2062, simple_loss=0.2902, pruned_loss=0.0611, over 16768.00 frames. ], tot_loss[loss=0.2183, simple_loss=0.3019, pruned_loss=0.06733, over 3097221.25 frames. ], batch size: 39, lr: 5.64e-03, grad_scale: 8.0 2023-04-29 15:43:03,207 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.9981, 4.0436, 4.4193, 4.3735, 4.3716, 4.0837, 4.0960, 4.0101], device='cuda:1'), covar=tensor([0.0303, 0.0497, 0.0309, 0.0368, 0.0434, 0.0363, 0.0820, 0.0480], device='cuda:1'), in_proj_covar=tensor([0.0338, 0.0353, 0.0353, 0.0334, 0.0403, 0.0375, 0.0474, 0.0301], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:1') 2023-04-29 15:43:03,913 INFO [optim.py:368] (1/8) 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] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=118872.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 15:43:28,300 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=4.41 vs. limit=5.0 2023-04-29 15:44:00,074 INFO [train.py:904] (1/8) Epoch 12, batch 7250, loss[loss=0.215, simple_loss=0.2907, pruned_loss=0.06961, over 16672.00 frames. ], tot_loss[loss=0.2157, simple_loss=0.2995, pruned_loss=0.06595, over 3082433.08 frames. ], batch size: 134, lr: 5.64e-03, grad_scale: 4.0 2023-04-29 15:44:35,521 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=118925.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 15:44:54,292 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([1.6839, 2.6458, 2.8374, 4.5734, 3.7929, 4.2023, 1.3683, 3.3515], device='cuda:1'), covar=tensor([0.1404, 0.0761, 0.1081, 0.0137, 0.0367, 0.0382, 0.1667, 0.0714], device='cuda:1'), in_proj_covar=tensor([0.0156, 0.0163, 0.0184, 0.0152, 0.0202, 0.0210, 0.0185, 0.0184], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-04-29 15:45:04,492 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=118945.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 15:45:08,788 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.2708, 4.3241, 4.0844, 3.8916, 3.8388, 4.2403, 3.9757, 3.9145], device='cuda:1'), covar=tensor([0.0529, 0.0375, 0.0283, 0.0261, 0.0842, 0.0388, 0.0620, 0.0616], device='cuda:1'), in_proj_covar=tensor([0.0238, 0.0320, 0.0285, 0.0265, 0.0305, 0.0306, 0.0197, 0.0333], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-29 15:45:15,120 INFO [train.py:904] (1/8) Epoch 12, batch 7300, loss[loss=0.2294, simple_loss=0.3143, pruned_loss=0.07226, over 15358.00 frames. ], tot_loss[loss=0.2149, simple_loss=0.2987, pruned_loss=0.06554, over 3085184.57 frames. ], batch size: 191, lr: 5.64e-03, grad_scale: 4.0 2023-04-29 15:45:18,567 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.75 vs. limit=2.0 2023-04-29 15:45:36,389 INFO [optim.py:368] (1/8) Clipping_scale=2.0, grad-norm quartiles 1.980e+02 3.034e+02 3.599e+02 4.380e+02 7.583e+02, threshold=7.199e+02, percent-clipped=5.0 2023-04-29 15:45:45,782 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=118973.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 15:46:03,494 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=118985.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 15:46:05,801 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=118987.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 15:46:15,622 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=118993.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 15:46:28,523 INFO [train.py:904] (1/8) Epoch 12, batch 7350, loss[loss=0.2423, simple_loss=0.3226, pruned_loss=0.08095, over 16911.00 frames. ], tot_loss[loss=0.2168, simple_loss=0.2998, pruned_loss=0.06697, over 3053199.93 frames. ], batch size: 109, lr: 5.64e-03, grad_scale: 4.0 2023-04-29 15:47:14,406 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=119033.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 15:47:42,936 INFO [train.py:904] (1/8) Epoch 12, batch 7400, loss[loss=0.1949, simple_loss=0.2901, pruned_loss=0.04986, over 16866.00 frames. ], tot_loss[loss=0.2175, simple_loss=0.3003, pruned_loss=0.06734, over 3057053.02 frames. ], batch size: 102, lr: 5.64e-03, grad_scale: 4.0 2023-04-29 15:48:06,306 INFO [optim.py:368] (1/8) 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 15:48:19,173 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=119075.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 15:48:47,062 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 12, batch 7450, loss[loss=0.1992, simple_loss=0.2792, pruned_loss=0.05958, over 17027.00 frames. ], tot_loss[loss=0.2194, simple_loss=0.3014, pruned_loss=0.0687, over 3061133.98 frames. ], batch size: 50, lr: 5.64e-03, grad_scale: 4.0 2023-04-29 15:49:55,914 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 12, batch 7500, loss[loss=0.221, simple_loss=0.3016, pruned_loss=0.07019, over 11640.00 frames. ], tot_loss[loss=0.2182, simple_loss=0.3015, pruned_loss=0.06743, over 3081291.60 frames. ], batch size: 248, lr: 5.64e-03, grad_scale: 4.0 2023-04-29 15:50:23,480 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=119154.0, num_to_drop=1, layers_to_drop={2} 2023-04-29 15:50:42,266 INFO [optim.py:368] (1/8) Clipping_scale=2.0, grad-norm quartiles 1.862e+02 2.952e+02 3.435e+02 4.438e+02 7.679e+02, threshold=6.870e+02, percent-clipped=0.0 2023-04-29 15:51:13,639 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.0650, 3.8988, 4.1271, 4.2701, 4.3670, 3.9468, 4.2985, 4.3441], device='cuda:1'), covar=tensor([0.1588, 0.1130, 0.1362, 0.0602, 0.0532, 0.1375, 0.0761, 0.0608], device='cuda:1'), in_proj_covar=tensor([0.0521, 0.0653, 0.0785, 0.0664, 0.0502, 0.0516, 0.0533, 0.0603], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-29 15:51:35,617 INFO [train.py:904] (1/8) Epoch 12, batch 7550, loss[loss=0.1794, simple_loss=0.2652, pruned_loss=0.04683, over 16832.00 frames. ], tot_loss[loss=0.2178, simple_loss=0.3004, pruned_loss=0.06765, over 3085196.53 frames. ], batch size: 83, lr: 5.64e-03, grad_scale: 4.0 2023-04-29 15:52:50,120 INFO [train.py:904] (1/8) Epoch 12, batch 7600, loss[loss=0.2177, simple_loss=0.2946, pruned_loss=0.07043, over 16924.00 frames. ], tot_loss[loss=0.218, simple_loss=0.2999, pruned_loss=0.06806, over 3076205.39 frames. ], batch size: 109, lr: 5.64e-03, grad_scale: 8.0 2023-04-29 15:53:12,404 INFO [optim.py:368] (1/8) 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 15:53:29,574 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.23 vs. limit=2.0 2023-04-29 15:53:43,257 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 12, batch 7650, loss[loss=0.186, simple_loss=0.2734, pruned_loss=0.04926, over 16844.00 frames. ], tot_loss[loss=0.2178, simple_loss=0.2999, pruned_loss=0.0679, over 3084999.51 frames. ], batch size: 83, lr: 5.63e-03, grad_scale: 8.0 2023-04-29 15:54:12,361 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.8586, 4.1253, 3.9121, 3.9540, 3.6505, 3.7941, 3.8063, 4.1058], device='cuda:1'), covar=tensor([0.1109, 0.1050, 0.1106, 0.0798, 0.0824, 0.1413, 0.0949, 0.1088], device='cuda:1'), in_proj_covar=tensor([0.0542, 0.0674, 0.0559, 0.0477, 0.0429, 0.0445, 0.0562, 0.0524], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:1') 2023-04-29 15:54:50,501 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=2.42 vs. limit=5.0 2023-04-29 15:54:55,612 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 12, batch 7700, loss[loss=0.2406, simple_loss=0.3305, pruned_loss=0.07535, over 16230.00 frames. ], tot_loss[loss=0.2193, simple_loss=0.3008, pruned_loss=0.06891, over 3068199.07 frames. ], batch size: 165, lr: 5.63e-03, grad_scale: 8.0 2023-04-29 15:55:42,612 INFO [optim.py:368] (1/8) 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 15:55:44,193 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.17 vs. limit=2.0 2023-04-29 15:56:36,113 INFO [train.py:904] (1/8) Epoch 12, batch 7750, loss[loss=0.2681, simple_loss=0.3273, pruned_loss=0.1044, over 11635.00 frames. ], tot_loss[loss=0.2196, simple_loss=0.3012, pruned_loss=0.06895, over 3059108.39 frames. ], batch size: 247, lr: 5.63e-03, grad_scale: 8.0 2023-04-29 15:56:53,027 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.5159, 3.6324, 2.0495, 4.1131, 2.6760, 4.0486, 2.1711, 2.7286], device='cuda:1'), covar=tensor([0.0246, 0.0321, 0.1667, 0.0133, 0.0739, 0.0518, 0.1470, 0.0771], device='cuda:1'), in_proj_covar=tensor([0.0152, 0.0164, 0.0190, 0.0130, 0.0168, 0.0204, 0.0196, 0.0171], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0003, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-04-29 15:57:18,792 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=119431.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 15:57:30,841 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.64 vs. limit=2.0 2023-04-29 15:57:45,519 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=119449.0, num_to_drop=1, layers_to_drop={1} 2023-04-29 15:57:48,706 INFO [train.py:904] (1/8) Epoch 12, batch 7800, loss[loss=0.1926, simple_loss=0.2863, pruned_loss=0.04948, over 16534.00 frames. ], tot_loss[loss=0.2219, simple_loss=0.303, pruned_loss=0.07042, over 3059379.02 frames. ], batch size: 75, lr: 5.63e-03, grad_scale: 8.0 2023-04-29 15:57:50,405 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([1.9422, 2.6396, 2.5991, 1.9735, 2.4814, 2.6283, 2.5902, 1.8810], device='cuda:1'), covar=tensor([0.0330, 0.0058, 0.0058, 0.0282, 0.0098, 0.0113, 0.0087, 0.0328], device='cuda:1'), in_proj_covar=tensor([0.0128, 0.0068, 0.0071, 0.0125, 0.0079, 0.0091, 0.0080, 0.0118], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:1') 2023-04-29 15:57:59,419 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.1583, 3.2592, 1.8696, 3.4920, 2.4430, 3.4563, 1.9956, 2.5504], device='cuda:1'), covar=tensor([0.0242, 0.0348, 0.1573, 0.0190, 0.0733, 0.0698, 0.1461, 0.0712], device='cuda:1'), in_proj_covar=tensor([0.0151, 0.0163, 0.0189, 0.0130, 0.0167, 0.0204, 0.0197, 0.0171], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0003, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-04-29 15:58:11,185 INFO [optim.py:368] (1/8) 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 15:58:56,527 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([1.7044, 2.9118, 2.6613, 4.6361, 3.5763, 4.1299, 1.7813, 3.1513], device='cuda:1'), covar=tensor([0.1278, 0.0654, 0.1053, 0.0119, 0.0280, 0.0349, 0.1392, 0.0730], device='cuda:1'), in_proj_covar=tensor([0.0156, 0.0163, 0.0183, 0.0151, 0.0200, 0.0208, 0.0184, 0.0182], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-04-29 15:59:04,896 INFO [train.py:904] (1/8) Epoch 12, batch 7850, loss[loss=0.1955, simple_loss=0.2859, pruned_loss=0.05257, over 16822.00 frames. ], tot_loss[loss=0.2221, simple_loss=0.3034, pruned_loss=0.07046, over 3049865.23 frames. ], batch size: 102, lr: 5.63e-03, grad_scale: 4.0 2023-04-29 15:59:23,787 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=119514.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 15:59:26,676 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.8734, 2.2951, 1.8075, 2.0347, 2.6164, 2.3198, 2.7845, 2.9207], device='cuda:1'), covar=tensor([0.0120, 0.0314, 0.0423, 0.0387, 0.0185, 0.0304, 0.0152, 0.0162], device='cuda:1'), in_proj_covar=tensor([0.0154, 0.0207, 0.0203, 0.0202, 0.0208, 0.0205, 0.0211, 0.0197], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-29 16:00:05,453 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.6223, 4.7589, 5.0060, 4.7741, 4.8559, 5.3826, 4.9032, 4.5758], device='cuda:1'), covar=tensor([0.1128, 0.1709, 0.1763, 0.1909, 0.2225, 0.0903, 0.1475, 0.2577], device='cuda:1'), in_proj_covar=tensor([0.0362, 0.0504, 0.0551, 0.0435, 0.0586, 0.0573, 0.0434, 0.0593], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-04-29 16:00:18,265 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.17 vs. limit=2.0 2023-04-29 16:00:21,535 INFO [train.py:904] (1/8) Epoch 12, batch 7900, loss[loss=0.2197, simple_loss=0.3098, pruned_loss=0.06477, over 16245.00 frames. ], tot_loss[loss=0.2209, simple_loss=0.3022, pruned_loss=0.06981, over 3056563.57 frames. ], batch size: 165, lr: 5.63e-03, grad_scale: 4.0 2023-04-29 16:00:45,732 INFO [optim.py:368] (1/8) 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 16:00:56,279 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=119575.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 16:01:38,577 INFO [train.py:904] (1/8) Epoch 12, batch 7950, loss[loss=0.2153, simple_loss=0.2962, pruned_loss=0.06719, over 16883.00 frames. ], tot_loss[loss=0.2211, simple_loss=0.3025, pruned_loss=0.06988, over 3066926.87 frames. ], batch size: 109, lr: 5.63e-03, grad_scale: 4.0 2023-04-29 16:02:53,353 INFO [train.py:904] (1/8) Epoch 12, batch 8000, loss[loss=0.2208, simple_loss=0.3161, pruned_loss=0.0627, over 16180.00 frames. ], tot_loss[loss=0.2216, simple_loss=0.3031, pruned_loss=0.07003, over 3074730.16 frames. ], batch size: 165, lr: 5.63e-03, grad_scale: 8.0 2023-04-29 16:03:12,830 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=119665.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 16:03:16,941 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.28 vs. limit=2.0 2023-04-29 16:03:17,120 INFO [optim.py:368] (1/8) 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] (1/8) Epoch 12, batch 8050, loss[loss=0.2242, simple_loss=0.3002, pruned_loss=0.07413, over 16486.00 frames. ], tot_loss[loss=0.2208, simple_loss=0.3023, pruned_loss=0.06966, over 3069618.83 frames. ], batch size: 68, lr: 5.62e-03, grad_scale: 4.0 2023-04-29 16:04:42,941 INFO [zipformer.py:625] (1/8) 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] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=119731.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 16:04:50,038 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.52 vs. limit=2.0 2023-04-29 16:05:17,372 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=119749.0, num_to_drop=1, layers_to_drop={2} 2023-04-29 16:05:21,185 INFO [train.py:904] (1/8) Epoch 12, batch 8100, loss[loss=0.2511, simple_loss=0.3102, pruned_loss=0.09604, over 11351.00 frames. ], tot_loss[loss=0.2204, simple_loss=0.3021, pruned_loss=0.06931, over 3068547.07 frames. ], batch size: 248, lr: 5.62e-03, grad_scale: 4.0 2023-04-29 16:05:42,986 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.8730, 3.3345, 3.2687, 1.9314, 2.8393, 2.1279, 3.4105, 3.4828], device='cuda:1'), covar=tensor([0.0290, 0.0721, 0.0624, 0.2008, 0.0847, 0.1044, 0.0683, 0.0863], device='cuda:1'), in_proj_covar=tensor([0.0141, 0.0144, 0.0158, 0.0143, 0.0136, 0.0124, 0.0137, 0.0155], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:1') 2023-04-29 16:05:43,334 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.53 vs. limit=2.0 2023-04-29 16:05:47,749 INFO [optim.py:368] (1/8) 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] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=119779.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 16:06:28,056 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=119797.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 16:06:34,772 INFO [train.py:904] (1/8) Epoch 12, batch 8150, loss[loss=0.2005, simple_loss=0.2793, pruned_loss=0.06083, over 16865.00 frames. ], tot_loss[loss=0.2177, simple_loss=0.2994, pruned_loss=0.06805, over 3078911.54 frames. ], batch size: 116, lr: 5.62e-03, grad_scale: 2.0 2023-04-29 16:07:50,608 INFO [train.py:904] (1/8) Epoch 12, batch 8200, loss[loss=0.2091, simple_loss=0.2988, pruned_loss=0.05965, over 15323.00 frames. ], tot_loss[loss=0.2162, simple_loss=0.2971, pruned_loss=0.0677, over 3071597.61 frames. ], batch size: 190, lr: 5.62e-03, grad_scale: 2.0 2023-04-29 16:08:11,505 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.8303, 4.1644, 3.2417, 2.2336, 2.7906, 2.4832, 4.4233, 3.7612], device='cuda:1'), covar=tensor([0.2613, 0.0496, 0.1401, 0.2564, 0.2513, 0.1841, 0.0322, 0.0899], device='cuda:1'), in_proj_covar=tensor([0.0308, 0.0258, 0.0288, 0.0282, 0.0282, 0.0223, 0.0269, 0.0298], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-29 16:08:18,226 INFO [optim.py:368] (1/8) 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] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=119870.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 16:09:09,079 INFO [train.py:904] (1/8) Epoch 12, batch 8250, loss[loss=0.1824, simple_loss=0.2738, pruned_loss=0.04548, over 16413.00 frames. ], tot_loss[loss=0.2125, simple_loss=0.2957, pruned_loss=0.06459, over 3066733.77 frames. ], batch size: 75, lr: 5.62e-03, grad_scale: 2.0 2023-04-29 16:10:28,023 INFO [train.py:904] (1/8) Epoch 12, batch 8300, loss[loss=0.1922, simple_loss=0.2877, pruned_loss=0.04833, over 16376.00 frames. ], tot_loss[loss=0.2082, simple_loss=0.2934, pruned_loss=0.06156, over 3056736.10 frames. ], batch size: 146, lr: 5.62e-03, grad_scale: 2.0 2023-04-29 16:10:57,564 INFO [optim.py:368] (1/8) Clipping_scale=2.0, grad-norm quartiles 1.672e+02 2.352e+02 2.901e+02 3.420e+02 5.863e+02, threshold=5.801e+02, percent-clipped=0.0 2023-04-29 16:11:17,583 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.43 vs. limit=2.0 2023-04-29 16:11:52,965 INFO [train.py:904] (1/8) Epoch 12, batch 8350, loss[loss=0.2204, simple_loss=0.2926, pruned_loss=0.07406, over 11942.00 frames. ], tot_loss[loss=0.2051, simple_loss=0.292, pruned_loss=0.05906, over 3048652.37 frames. ], batch size: 248, lr: 5.62e-03, grad_scale: 2.0 2023-04-29 16:12:24,540 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=120021.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 16:12:37,952 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.0738, 4.0354, 4.4333, 4.4185, 4.3917, 4.1594, 4.1000, 4.1082], device='cuda:1'), covar=tensor([0.0282, 0.0581, 0.0336, 0.0349, 0.0429, 0.0338, 0.0880, 0.0427], device='cuda:1'), in_proj_covar=tensor([0.0335, 0.0353, 0.0349, 0.0332, 0.0402, 0.0374, 0.0470, 0.0301], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:1') 2023-04-29 16:13:14,450 INFO [train.py:904] (1/8) Epoch 12, batch 8400, loss[loss=0.1858, simple_loss=0.2756, pruned_loss=0.04807, over 16516.00 frames. ], tot_loss[loss=0.201, simple_loss=0.2887, pruned_loss=0.05663, over 3041946.75 frames. ], batch size: 62, lr: 5.62e-03, grad_scale: 4.0 2023-04-29 16:13:42,956 INFO [optim.py:368] (1/8) Clipping_scale=2.0, grad-norm quartiles 1.720e+02 2.479e+02 2.965e+02 3.428e+02 6.852e+02, threshold=5.930e+02, percent-clipped=2.0 2023-04-29 16:14:31,502 INFO [train.py:904] (1/8) Epoch 12, batch 8450, loss[loss=0.1648, simple_loss=0.2636, pruned_loss=0.03302, over 16756.00 frames. ], tot_loss[loss=0.1978, simple_loss=0.2864, pruned_loss=0.05467, over 3044892.09 frames. ], batch size: 83, lr: 5.62e-03, grad_scale: 4.0 2023-04-29 16:15:47,447 INFO [train.py:904] (1/8) Epoch 12, batch 8500, loss[loss=0.1648, simple_loss=0.2475, pruned_loss=0.04102, over 11982.00 frames. ], tot_loss[loss=0.1936, simple_loss=0.2828, pruned_loss=0.05214, over 3061085.68 frames. ], batch size: 247, lr: 5.61e-03, grad_scale: 4.0 2023-04-29 16:16:15,431 INFO [optim.py:368] (1/8) 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 16:16:15,940 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=120170.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 16:17:07,725 INFO [train.py:904] (1/8) Epoch 12, batch 8550, loss[loss=0.223, simple_loss=0.3074, pruned_loss=0.06933, over 16754.00 frames. ], tot_loss[loss=0.1918, simple_loss=0.281, pruned_loss=0.05134, over 3059844.64 frames. ], batch size: 124, lr: 5.61e-03, grad_scale: 4.0 2023-04-29 16:17:36,930 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=120218.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 16:18:35,164 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.0203, 1.8249, 1.6406, 1.5429, 1.9542, 1.6072, 1.7062, 1.9968], device='cuda:1'), covar=tensor([0.0131, 0.0252, 0.0317, 0.0294, 0.0180, 0.0221, 0.0181, 0.0177], device='cuda:1'), in_proj_covar=tensor([0.0149, 0.0201, 0.0196, 0.0195, 0.0201, 0.0198, 0.0201, 0.0189], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-29 16:18:41,310 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.4552, 3.0489, 3.1329, 1.8530, 2.7303, 2.2075, 3.0182, 3.2211], device='cuda:1'), covar=tensor([0.0279, 0.0680, 0.0509, 0.1865, 0.0767, 0.0976, 0.0702, 0.0759], device='cuda:1'), in_proj_covar=tensor([0.0139, 0.0141, 0.0155, 0.0141, 0.0133, 0.0123, 0.0134, 0.0151], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:1') 2023-04-29 16:18:41,889 INFO [train.py:904] (1/8) Epoch 12, batch 8600, loss[loss=0.1976, simple_loss=0.2879, pruned_loss=0.05369, over 15395.00 frames. ], tot_loss[loss=0.1905, simple_loss=0.2806, pruned_loss=0.05022, over 3054849.84 frames. ], batch size: 190, lr: 5.61e-03, grad_scale: 4.0 2023-04-29 16:19:19,442 INFO [optim.py:368] (1/8) 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 16:19:28,686 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.6110, 3.5674, 2.7451, 2.0914, 2.3077, 2.1707, 3.7496, 3.2069], device='cuda:1'), covar=tensor([0.2568, 0.0671, 0.1540, 0.2652, 0.2554, 0.1976, 0.0412, 0.1221], device='cuda:1'), in_proj_covar=tensor([0.0298, 0.0249, 0.0278, 0.0273, 0.0269, 0.0217, 0.0260, 0.0287], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-29 16:20:19,201 INFO [train.py:904] (1/8) Epoch 12, batch 8650, loss[loss=0.1599, simple_loss=0.258, pruned_loss=0.03088, over 16658.00 frames. ], tot_loss[loss=0.1879, simple_loss=0.2783, pruned_loss=0.04876, over 3032417.92 frames. ], batch size: 134, lr: 5.61e-03, grad_scale: 4.0 2023-04-29 16:20:51,598 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=2.95 vs. limit=5.0 2023-04-29 16:21:01,178 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=120321.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 16:22:02,145 INFO [train.py:904] (1/8) Epoch 12, batch 8700, loss[loss=0.1779, simple_loss=0.2665, pruned_loss=0.04465, over 16366.00 frames. ], tot_loss[loss=0.1853, simple_loss=0.2758, pruned_loss=0.04745, over 3038075.62 frames. ], batch size: 146, lr: 5.61e-03, grad_scale: 4.0 2023-04-29 16:22:21,598 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.2987, 3.3607, 2.0844, 3.6497, 2.4724, 3.5963, 2.2491, 2.7226], device='cuda:1'), covar=tensor([0.0234, 0.0317, 0.1445, 0.0143, 0.0756, 0.0464, 0.1417, 0.0688], device='cuda:1'), in_proj_covar=tensor([0.0149, 0.0157, 0.0183, 0.0125, 0.0163, 0.0197, 0.0193, 0.0165], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0003, 0.0004, 0.0004, 0.0003], device='cuda:1') 2023-04-29 16:22:33,095 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=120369.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 16:22:34,421 INFO [optim.py:368] (1/8) 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 16:23:36,531 INFO [train.py:904] (1/8) Epoch 12, batch 8750, loss[loss=0.162, simple_loss=0.2503, pruned_loss=0.0369, over 12096.00 frames. ], tot_loss[loss=0.1858, simple_loss=0.2766, pruned_loss=0.04751, over 3051332.43 frames. ], batch size: 248, lr: 5.61e-03, grad_scale: 4.0 2023-04-29 16:23:40,009 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([5.0033, 5.3787, 5.1350, 5.0894, 4.8594, 4.7473, 4.7489, 5.4416], device='cuda:1'), covar=tensor([0.1176, 0.0755, 0.0915, 0.0632, 0.0676, 0.0877, 0.1067, 0.0766], device='cuda:1'), in_proj_covar=tensor([0.0529, 0.0651, 0.0540, 0.0464, 0.0416, 0.0434, 0.0545, 0.0507], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:1') 2023-04-29 16:25:30,723 INFO [train.py:904] (1/8) Epoch 12, batch 8800, loss[loss=0.166, simple_loss=0.2671, pruned_loss=0.03248, over 16613.00 frames. ], tot_loss[loss=0.1834, simple_loss=0.2744, pruned_loss=0.04625, over 3037131.49 frames. ], batch size: 62, lr: 5.61e-03, grad_scale: 8.0 2023-04-29 16:26:08,429 INFO [optim.py:368] (1/8) 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 16:26:28,739 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=3.24 vs. limit=5.0 2023-04-29 16:27:09,158 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.7883, 3.7960, 4.0824, 4.0725, 4.1280, 3.8745, 3.9323, 3.8244], device='cuda:1'), covar=tensor([0.0276, 0.0523, 0.0466, 0.0428, 0.0381, 0.0378, 0.0723, 0.0415], device='cuda:1'), in_proj_covar=tensor([0.0325, 0.0341, 0.0337, 0.0322, 0.0386, 0.0362, 0.0450, 0.0291], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003], device='cuda:1') 2023-04-29 16:27:15,508 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.2975, 3.3983, 3.6237, 3.6292, 3.6598, 3.4560, 3.5225, 3.4871], device='cuda:1'), covar=tensor([0.0341, 0.0604, 0.0492, 0.0446, 0.0410, 0.0479, 0.0683, 0.0438], device='cuda:1'), in_proj_covar=tensor([0.0325, 0.0341, 0.0337, 0.0322, 0.0386, 0.0362, 0.0450, 0.0291], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003], device='cuda:1') 2023-04-29 16:27:17,129 INFO [train.py:904] (1/8) Epoch 12, batch 8850, loss[loss=0.1807, simple_loss=0.2663, pruned_loss=0.04757, over 12455.00 frames. ], tot_loss[loss=0.1835, simple_loss=0.2764, pruned_loss=0.04537, over 3041509.37 frames. ], batch size: 248, lr: 5.61e-03, grad_scale: 8.0 2023-04-29 16:27:35,945 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.4017, 3.3502, 3.4272, 3.5283, 3.5882, 3.2656, 3.5764, 3.6199], device='cuda:1'), covar=tensor([0.0963, 0.0852, 0.1062, 0.0592, 0.0512, 0.2393, 0.0726, 0.0589], device='cuda:1'), in_proj_covar=tensor([0.0499, 0.0629, 0.0751, 0.0638, 0.0482, 0.0495, 0.0509, 0.0581], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:1') 2023-04-29 16:29:04,467 INFO [train.py:904] (1/8) Epoch 12, batch 8900, loss[loss=0.1806, simple_loss=0.2796, pruned_loss=0.04077, over 16955.00 frames. ], tot_loss[loss=0.1831, simple_loss=0.2766, pruned_loss=0.04475, over 3044282.01 frames. ], batch size: 96, lr: 5.60e-03, grad_scale: 8.0 2023-04-29 16:29:39,305 INFO [optim.py:368] (1/8) 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 16:31:10,991 INFO [train.py:904] (1/8) Epoch 12, batch 8950, loss[loss=0.1678, simple_loss=0.2652, pruned_loss=0.03521, over 12652.00 frames. ], tot_loss[loss=0.1838, simple_loss=0.2768, pruned_loss=0.04537, over 3050041.79 frames. ], batch size: 248, lr: 5.60e-03, grad_scale: 8.0 2023-04-29 16:33:00,365 INFO [train.py:904] (1/8) Epoch 12, batch 9000, loss[loss=0.1767, simple_loss=0.2589, pruned_loss=0.04728, over 16876.00 frames. ], tot_loss[loss=0.1805, simple_loss=0.2733, pruned_loss=0.04389, over 3051834.51 frames. ], batch size: 116, lr: 5.60e-03, grad_scale: 8.0 2023-04-29 16:33:00,365 INFO [train.py:929] (1/8) Computing validation loss 2023-04-29 16:33:10,348 INFO [train.py:938] (1/8) Epoch 12, validation: loss=0.1532, simple_loss=0.2571, pruned_loss=0.02465, over 944034.00 frames. 2023-04-29 16:33:10,348 INFO [train.py:939] (1/8) Maximum memory allocated so far is 17845MB 2023-04-29 16:33:28,005 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.49 vs. limit=2.0 2023-04-29 16:33:49,107 INFO [optim.py:368] (1/8) 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 16:34:01,167 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=120676.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 16:34:54,266 INFO [train.py:904] (1/8) Epoch 12, batch 9050, loss[loss=0.1616, simple_loss=0.2516, pruned_loss=0.03579, over 17012.00 frames. ], tot_loss[loss=0.181, simple_loss=0.2737, pruned_loss=0.04419, over 3049776.30 frames. ], batch size: 53, lr: 5.60e-03, grad_scale: 8.0 2023-04-29 16:36:04,414 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=120737.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 16:36:39,407 INFO [train.py:904] (1/8) Epoch 12, batch 9100, loss[loss=0.1864, simple_loss=0.2837, pruned_loss=0.04456, over 16765.00 frames. ], tot_loss[loss=0.1814, simple_loss=0.2735, pruned_loss=0.04468, over 3060003.42 frames. ], batch size: 124, lr: 5.60e-03, grad_scale: 8.0 2023-04-29 16:37:03,456 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.81 vs. limit=2.0 2023-04-29 16:37:15,445 INFO [optim.py:368] (1/8) 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 16:38:36,941 INFO [train.py:904] (1/8) Epoch 12, batch 9150, loss[loss=0.1621, simple_loss=0.2595, pruned_loss=0.03236, over 16810.00 frames. ], tot_loss[loss=0.1821, simple_loss=0.2749, pruned_loss=0.04463, over 3070459.02 frames. ], batch size: 90, lr: 5.60e-03, grad_scale: 8.0 2023-04-29 16:40:21,513 INFO [train.py:904] (1/8) Epoch 12, batch 9200, loss[loss=0.1782, simple_loss=0.2713, pruned_loss=0.04255, over 16928.00 frames. ], tot_loss[loss=0.1782, simple_loss=0.2699, pruned_loss=0.04323, over 3073551.01 frames. ], batch size: 116, lr: 5.60e-03, grad_scale: 8.0 2023-04-29 16:40:26,654 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.3095, 1.5439, 1.7888, 2.2779, 2.3002, 2.5064, 1.6058, 2.3579], device='cuda:1'), covar=tensor([0.0161, 0.0370, 0.0251, 0.0226, 0.0230, 0.0148, 0.0378, 0.0123], device='cuda:1'), in_proj_covar=tensor([0.0159, 0.0170, 0.0156, 0.0157, 0.0167, 0.0121, 0.0169, 0.0115], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001], device='cuda:1') 2023-04-29 16:40:55,537 INFO [optim.py:368] (1/8) 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 16:42:00,504 INFO [train.py:904] (1/8) Epoch 12, batch 9250, loss[loss=0.1822, simple_loss=0.2732, pruned_loss=0.04557, over 16471.00 frames. ], tot_loss[loss=0.1777, simple_loss=0.2696, pruned_loss=0.04294, over 3095442.43 frames. ], batch size: 146, lr: 5.60e-03, grad_scale: 8.0 2023-04-29 16:42:05,908 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.2083, 3.3691, 3.6331, 3.6234, 3.6188, 3.4207, 3.4438, 3.5048], device='cuda:1'), covar=tensor([0.0446, 0.1020, 0.0474, 0.0475, 0.0560, 0.0618, 0.0908, 0.0502], device='cuda:1'), in_proj_covar=tensor([0.0322, 0.0336, 0.0336, 0.0319, 0.0382, 0.0360, 0.0446, 0.0288], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0002], device='cuda:1') 2023-04-29 16:43:42,009 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.35 vs. limit=2.0 2023-04-29 16:43:49,117 INFO [train.py:904] (1/8) Epoch 12, batch 9300, loss[loss=0.171, simple_loss=0.2607, pruned_loss=0.04061, over 15382.00 frames. ], tot_loss[loss=0.1764, simple_loss=0.2679, pruned_loss=0.0424, over 3080057.40 frames. ], batch size: 192, lr: 5.60e-03, grad_scale: 4.0 2023-04-29 16:44:31,802 INFO [optim.py:368] (1/8) Clipping_scale=2.0, grad-norm quartiles 1.540e+02 2.549e+02 3.003e+02 3.626e+02 5.994e+02, threshold=6.005e+02, percent-clipped=1.0 2023-04-29 16:45:32,147 INFO [train.py:904] (1/8) Epoch 12, batch 9350, loss[loss=0.185, simple_loss=0.268, pruned_loss=0.05103, over 12179.00 frames. ], tot_loss[loss=0.177, simple_loss=0.2681, pruned_loss=0.04292, over 3072846.50 frames. ], batch size: 250, lr: 5.59e-03, grad_scale: 4.0 2023-04-29 16:45:34,988 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=121003.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 16:46:33,083 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=121032.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 16:47:13,261 INFO [train.py:904] (1/8) Epoch 12, batch 9400, loss[loss=0.1794, simple_loss=0.285, pruned_loss=0.03688, over 16701.00 frames. ], tot_loss[loss=0.1767, simple_loss=0.2679, pruned_loss=0.04273, over 3058819.76 frames. ], batch size: 83, lr: 5.59e-03, grad_scale: 4.0 2023-04-29 16:47:39,126 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=121064.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 16:47:50,524 INFO [optim.py:368] (1/8) 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 16:48:34,552 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.22 vs. limit=2.0 2023-04-29 16:48:55,034 INFO [train.py:904] (1/8) Epoch 12, batch 9450, loss[loss=0.1726, simple_loss=0.2637, pruned_loss=0.04073, over 16679.00 frames. ], tot_loss[loss=0.1776, simple_loss=0.2694, pruned_loss=0.04284, over 3049122.03 frames. ], batch size: 134, lr: 5.59e-03, grad_scale: 4.0 2023-04-29 16:49:13,411 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=5.09 vs. limit=5.0 2023-04-29 16:50:34,751 INFO [train.py:904] (1/8) Epoch 12, batch 9500, loss[loss=0.1618, simple_loss=0.2619, pruned_loss=0.03081, over 16639.00 frames. ], tot_loss[loss=0.1766, simple_loss=0.2687, pruned_loss=0.0422, over 3062592.75 frames. ], batch size: 89, lr: 5.59e-03, grad_scale: 4.0 2023-04-29 16:51:01,755 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=121164.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 16:51:13,619 INFO [optim.py:368] (1/8) 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 16:52:20,001 INFO [train.py:904] (1/8) Epoch 12, batch 9550, loss[loss=0.1923, simple_loss=0.288, pruned_loss=0.04828, over 15175.00 frames. ], tot_loss[loss=0.1765, simple_loss=0.2681, pruned_loss=0.04245, over 3049384.36 frames. ], batch size: 190, lr: 5.59e-03, grad_scale: 4.0 2023-04-29 16:52:49,374 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.0475, 1.4728, 1.8262, 2.1078, 2.1543, 2.2048, 1.6934, 2.2421], device='cuda:1'), covar=tensor([0.0207, 0.0400, 0.0227, 0.0232, 0.0248, 0.0166, 0.0376, 0.0101], device='cuda:1'), in_proj_covar=tensor([0.0158, 0.0170, 0.0155, 0.0156, 0.0167, 0.0120, 0.0168, 0.0113], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001], device='cuda:1') 2023-04-29 16:53:08,747 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=121225.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 16:53:13,024 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.2981, 1.9981, 2.0875, 3.8934, 2.0237, 2.3656, 2.1752, 2.1971], device='cuda:1'), covar=tensor([0.0946, 0.3447, 0.2484, 0.0400, 0.3897, 0.2424, 0.3082, 0.3457], device='cuda:1'), in_proj_covar=tensor([0.0353, 0.0381, 0.0325, 0.0306, 0.0404, 0.0434, 0.0351, 0.0445], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-29 16:53:51,791 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.0344, 4.0362, 3.9377, 3.4904, 3.9425, 1.7568, 3.7503, 3.6012], device='cuda:1'), covar=tensor([0.0084, 0.0080, 0.0140, 0.0203, 0.0084, 0.2322, 0.0113, 0.0196], device='cuda:1'), in_proj_covar=tensor([0.0126, 0.0114, 0.0156, 0.0144, 0.0131, 0.0177, 0.0146, 0.0143], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-29 16:53:59,926 INFO [train.py:904] (1/8) Epoch 12, batch 9600, loss[loss=0.2155, simple_loss=0.3162, pruned_loss=0.05741, over 15337.00 frames. ], tot_loss[loss=0.1791, simple_loss=0.2705, pruned_loss=0.04383, over 3048888.75 frames. ], batch size: 191, lr: 5.59e-03, grad_scale: 8.0 2023-04-29 16:54:35,186 INFO [optim.py:368] (1/8) 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 16:54:46,377 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.2520, 3.3835, 3.6349, 3.6209, 3.6038, 3.4237, 3.4395, 3.4607], device='cuda:1'), covar=tensor([0.0374, 0.0698, 0.0443, 0.0459, 0.0600, 0.0463, 0.0896, 0.0482], device='cuda:1'), in_proj_covar=tensor([0.0317, 0.0332, 0.0331, 0.0317, 0.0379, 0.0357, 0.0441, 0.0285], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0002], device='cuda:1') 2023-04-29 16:55:32,764 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([1.8938, 3.3556, 3.0718, 5.0746, 4.0209, 4.6037, 1.7105, 3.3811], device='cuda:1'), covar=tensor([0.1368, 0.0642, 0.0994, 0.0143, 0.0192, 0.0312, 0.1565, 0.0647], device='cuda:1'), in_proj_covar=tensor([0.0156, 0.0159, 0.0180, 0.0144, 0.0187, 0.0204, 0.0182, 0.0179], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0003, 0.0004], device='cuda:1') 2023-04-29 16:55:45,648 INFO [train.py:904] (1/8) Epoch 12, batch 9650, loss[loss=0.1882, simple_loss=0.2824, pruned_loss=0.04697, over 16898.00 frames. ], tot_loss[loss=0.1802, simple_loss=0.2724, pruned_loss=0.04402, over 3047745.60 frames. ], batch size: 116, lr: 5.59e-03, grad_scale: 8.0 2023-04-29 16:55:56,708 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.4425, 4.0154, 3.9825, 2.0918, 3.4540, 2.7093, 4.0232, 3.9860], device='cuda:1'), covar=tensor([0.0186, 0.0601, 0.0424, 0.1876, 0.0615, 0.0831, 0.0569, 0.0818], device='cuda:1'), in_proj_covar=tensor([0.0139, 0.0137, 0.0154, 0.0140, 0.0133, 0.0122, 0.0132, 0.0148], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:1') 2023-04-29 16:56:09,914 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=121311.0, num_to_drop=1, layers_to_drop={1} 2023-04-29 16:56:52,030 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 12, batch 9700, loss[loss=0.183, simple_loss=0.2754, pruned_loss=0.04528, over 16879.00 frames. ], tot_loss[loss=0.1796, simple_loss=0.2715, pruned_loss=0.04388, over 3036011.63 frames. ], batch size: 116, lr: 5.59e-03, grad_scale: 8.0 2023-04-29 16:57:44,383 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=121359.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 16:57:50,853 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([1.8991, 2.2638, 2.2196, 3.0007, 1.8078, 3.2161, 1.7216, 2.7505], device='cuda:1'), covar=tensor([0.1325, 0.0699, 0.1055, 0.0156, 0.0083, 0.0368, 0.1545, 0.0687], device='cuda:1'), in_proj_covar=tensor([0.0154, 0.0157, 0.0179, 0.0143, 0.0185, 0.0202, 0.0181, 0.0178], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0003, 0.0004], device='cuda:1') 2023-04-29 16:58:05,418 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.2846, 1.6163, 1.8933, 2.2565, 2.3287, 2.3929, 1.7477, 2.3917], device='cuda:1'), covar=tensor([0.0162, 0.0344, 0.0220, 0.0214, 0.0210, 0.0146, 0.0337, 0.0108], device='cuda:1'), in_proj_covar=tensor([0.0158, 0.0168, 0.0154, 0.0156, 0.0165, 0.0120, 0.0168, 0.0113], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001], device='cuda:1') 2023-04-29 16:58:07,289 INFO [optim.py:368] (1/8) 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 16:58:10,411 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=121372.0, num_to_drop=1, layers_to_drop={3} 2023-04-29 16:58:31,090 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=121380.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 16:59:14,332 INFO [train.py:904] (1/8) Epoch 12, batch 9750, loss[loss=0.18, simple_loss=0.2756, pruned_loss=0.04221, over 16700.00 frames. ], tot_loss[loss=0.1781, simple_loss=0.2697, pruned_loss=0.04329, over 3066200.52 frames. ], batch size: 134, lr: 5.59e-03, grad_scale: 8.0 2023-04-29 16:59:23,465 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 12, batch 9800, loss[loss=0.1822, simple_loss=0.2847, pruned_loss=0.03985, over 16729.00 frames. ], tot_loss[loss=0.1776, simple_loss=0.2702, pruned_loss=0.04247, over 3077207.27 frames. ], batch size: 134, lr: 5.58e-03, grad_scale: 8.0 2023-04-29 17:01:18,653 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([1.6157, 2.3962, 2.2873, 3.9269, 2.5439, 3.8465, 1.3790, 2.8698], device='cuda:1'), covar=tensor([0.1395, 0.0831, 0.1216, 0.0154, 0.0128, 0.0358, 0.1645, 0.0705], device='cuda:1'), in_proj_covar=tensor([0.0154, 0.0157, 0.0179, 0.0143, 0.0184, 0.0202, 0.0180, 0.0177], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0003, 0.0004], device='cuda:1') 2023-04-29 17:01:24,643 INFO [zipformer.py:625] (1/8) 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] (1/8) 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 17:02:38,486 INFO [train.py:904] (1/8) Epoch 12, batch 9850, loss[loss=0.1685, simple_loss=0.2659, pruned_loss=0.03557, over 16905.00 frames. ], tot_loss[loss=0.1776, simple_loss=0.2713, pruned_loss=0.04196, over 3104651.45 frames. ], batch size: 102, lr: 5.58e-03, grad_scale: 8.0 2023-04-29 17:03:17,192 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=121520.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 17:03:17,395 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.9592, 2.2658, 1.8457, 2.0453, 2.5983, 2.3261, 2.7927, 2.8015], device='cuda:1'), covar=tensor([0.0083, 0.0326, 0.0416, 0.0406, 0.0205, 0.0325, 0.0133, 0.0199], device='cuda:1'), in_proj_covar=tensor([0.0148, 0.0203, 0.0197, 0.0197, 0.0203, 0.0201, 0.0198, 0.0188], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-29 17:04:03,238 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.1579, 1.3920, 1.8528, 2.0478, 2.1161, 2.2006, 1.7048, 2.2411], device='cuda:1'), covar=tensor([0.0213, 0.0372, 0.0235, 0.0236, 0.0255, 0.0185, 0.0350, 0.0124], device='cuda:1'), in_proj_covar=tensor([0.0160, 0.0170, 0.0155, 0.0157, 0.0167, 0.0122, 0.0170, 0.0114], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001], device='cuda:1') 2023-04-29 17:04:28,370 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.23 vs. limit=2.0 2023-04-29 17:04:30,193 INFO [train.py:904] (1/8) Epoch 12, batch 9900, loss[loss=0.194, simple_loss=0.2902, pruned_loss=0.04887, over 16911.00 frames. ], tot_loss[loss=0.1779, simple_loss=0.272, pruned_loss=0.04194, over 3103436.87 frames. ], batch size: 116, lr: 5.58e-03, grad_scale: 8.0 2023-04-29 17:04:50,392 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.70 vs. limit=2.0 2023-04-29 17:05:12,985 INFO [optim.py:368] (1/8) 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 17:06:26,640 INFO [train.py:904] (1/8) Epoch 12, batch 9950, loss[loss=0.1715, simple_loss=0.2828, pruned_loss=0.03014, over 16900.00 frames. ], tot_loss[loss=0.1798, simple_loss=0.2743, pruned_loss=0.04268, over 3096233.02 frames. ], batch size: 102, lr: 5.58e-03, grad_scale: 8.0 2023-04-29 17:07:14,257 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([1.8827, 1.3223, 1.5844, 1.7428, 1.8547, 1.8747, 1.6254, 1.8944], device='cuda:1'), covar=tensor([0.0173, 0.0302, 0.0163, 0.0215, 0.0220, 0.0142, 0.0305, 0.0101], device='cuda:1'), in_proj_covar=tensor([0.0161, 0.0171, 0.0157, 0.0159, 0.0169, 0.0123, 0.0171, 0.0115], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001], device='cuda:1') 2023-04-29 17:08:27,932 INFO [train.py:904] (1/8) Epoch 12, batch 10000, loss[loss=0.1919, simple_loss=0.3026, pruned_loss=0.04062, over 15247.00 frames. ], tot_loss[loss=0.1786, simple_loss=0.2727, pruned_loss=0.04219, over 3089717.79 frames. ], batch size: 190, lr: 5.58e-03, grad_scale: 8.0 2023-04-29 17:08:44,216 INFO [zipformer.py:625] (1/8) 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] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=121667.0, num_to_drop=1, layers_to_drop={3} 2023-04-29 17:09:06,353 INFO [optim.py:368] (1/8) Clipping_scale=2.0, grad-norm quartiles 1.215e+02 2.322e+02 2.719e+02 3.378e+02 7.215e+02, threshold=5.438e+02, percent-clipped=2.0 2023-04-29 17:10:10,827 INFO [train.py:904] (1/8) Epoch 12, batch 10050, loss[loss=0.178, simple_loss=0.2753, pruned_loss=0.04035, over 16775.00 frames. ], tot_loss[loss=0.1781, simple_loss=0.2721, pruned_loss=0.04202, over 3073502.78 frames. ], batch size: 124, lr: 5.58e-03, grad_scale: 8.0 2023-04-29 17:10:21,425 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=121707.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 17:11:29,234 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.9864, 4.2540, 4.0949, 4.0843, 3.7458, 3.8260, 3.8651, 4.2297], device='cuda:1'), covar=tensor([0.0991, 0.0839, 0.0865, 0.0663, 0.0750, 0.1551, 0.0838, 0.0961], device='cuda:1'), in_proj_covar=tensor([0.0516, 0.0650, 0.0529, 0.0457, 0.0414, 0.0426, 0.0538, 0.0498], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:1') 2023-04-29 17:11:42,168 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.9752, 2.8079, 2.8283, 2.0911, 2.7249, 2.1670, 2.6756, 2.9599], device='cuda:1'), covar=tensor([0.0293, 0.0744, 0.0449, 0.1590, 0.0700, 0.0958, 0.0626, 0.0718], device='cuda:1'), in_proj_covar=tensor([0.0139, 0.0137, 0.0155, 0.0141, 0.0134, 0.0122, 0.0131, 0.0147], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:1') 2023-04-29 17:11:46,088 INFO [train.py:904] (1/8) Epoch 12, batch 10100, loss[loss=0.1608, simple_loss=0.2515, pruned_loss=0.0351, over 12469.00 frames. ], tot_loss[loss=0.1789, simple_loss=0.2727, pruned_loss=0.04255, over 3071784.09 frames. ], batch size: 246, lr: 5.58e-03, grad_scale: 8.0 2023-04-29 17:12:05,572 INFO [zipformer.py:625] (1/8) 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] (1/8) 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 17:13:30,551 INFO [train.py:904] (1/8) Epoch 13, batch 0, loss[loss=0.2344, simple_loss=0.2996, pruned_loss=0.08461, over 16803.00 frames. ], tot_loss[loss=0.2344, simple_loss=0.2996, pruned_loss=0.08461, over 16803.00 frames. ], batch size: 102, lr: 5.36e-03, grad_scale: 8.0 2023-04-29 17:13:30,551 INFO [train.py:929] (1/8) Computing validation loss 2023-04-29 17:13:38,109 INFO [train.py:938] (1/8) Epoch 13, validation: loss=0.1523, simple_loss=0.2559, pruned_loss=0.02431, over 944034.00 frames. 2023-04-29 17:13:38,110 INFO [train.py:939] (1/8) Maximum memory allocated so far is 17845MB 2023-04-29 17:14:04,097 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=121820.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 17:14:23,073 INFO [scaling.py:679] (1/8) 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] (1/8) Epoch 13, batch 50, loss[loss=0.191, simple_loss=0.271, pruned_loss=0.05552, over 16902.00 frames. ], tot_loss[loss=0.2008, simple_loss=0.2812, pruned_loss=0.06016, over 756487.64 frames. ], batch size: 96, lr: 5.36e-03, grad_scale: 2.0 2023-04-29 17:15:11,434 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=121868.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 17:15:18,817 INFO [optim.py:368] (1/8) 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 17:15:31,301 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.54 vs. limit=2.0 2023-04-29 17:15:58,273 INFO [train.py:904] (1/8) Epoch 13, batch 100, loss[loss=0.2373, simple_loss=0.3093, pruned_loss=0.08265, over 11883.00 frames. ], tot_loss[loss=0.1949, simple_loss=0.2768, pruned_loss=0.05651, over 1326584.93 frames. ], batch size: 248, lr: 5.36e-03, grad_scale: 2.0 2023-04-29 17:16:31,077 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.14 vs. limit=2.0 2023-04-29 17:17:04,532 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=121950.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 17:17:07,073 INFO [train.py:904] (1/8) Epoch 13, batch 150, loss[loss=0.159, simple_loss=0.2444, pruned_loss=0.03686, over 17234.00 frames. ], tot_loss[loss=0.1929, simple_loss=0.2751, pruned_loss=0.05539, over 1768916.15 frames. ], batch size: 45, lr: 5.35e-03, grad_scale: 2.0 2023-04-29 17:17:27,831 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=121967.0, num_to_drop=1, layers_to_drop={2} 2023-04-29 17:17:35,630 INFO [optim.py:368] (1/8) 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 17:18:01,376 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.7348, 2.2766, 1.6871, 2.0794, 2.7226, 2.5365, 2.9866, 2.7958], device='cuda:1'), covar=tensor([0.0223, 0.0404, 0.0575, 0.0477, 0.0242, 0.0316, 0.0210, 0.0244], device='cuda:1'), in_proj_covar=tensor([0.0159, 0.0214, 0.0206, 0.0205, 0.0212, 0.0210, 0.0212, 0.0198], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-29 17:18:18,195 INFO [train.py:904] (1/8) Epoch 13, batch 200, loss[loss=0.1841, simple_loss=0.2792, pruned_loss=0.04447, over 17220.00 frames. ], tot_loss[loss=0.194, simple_loss=0.2756, pruned_loss=0.05622, over 2113390.10 frames. ], batch size: 52, lr: 5.35e-03, grad_scale: 2.0 2023-04-29 17:18:30,536 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=122011.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 17:18:31,823 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.1870, 3.5747, 3.7714, 2.1894, 2.9900, 2.4703, 3.5054, 3.6988], device='cuda:1'), covar=tensor([0.0288, 0.0720, 0.0500, 0.1758, 0.0837, 0.0936, 0.0641, 0.0978], device='cuda:1'), in_proj_covar=tensor([0.0141, 0.0140, 0.0157, 0.0143, 0.0137, 0.0124, 0.0133, 0.0152], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:1') 2023-04-29 17:18:35,253 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=122015.0, num_to_drop=1, layers_to_drop={1} 2023-04-29 17:19:26,241 INFO [train.py:904] (1/8) Epoch 13, batch 250, loss[loss=0.1784, simple_loss=0.26, pruned_loss=0.04833, over 16797.00 frames. ], tot_loss[loss=0.1927, simple_loss=0.2742, pruned_loss=0.05559, over 2374633.00 frames. ], batch size: 102, lr: 5.35e-03, grad_scale: 2.0 2023-04-29 17:19:30,093 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.7017, 3.7408, 4.0936, 2.0818, 4.2241, 4.1674, 3.2761, 3.2260], device='cuda:1'), covar=tensor([0.0737, 0.0195, 0.0127, 0.1115, 0.0058, 0.0150, 0.0344, 0.0392], device='cuda:1'), in_proj_covar=tensor([0.0142, 0.0099, 0.0086, 0.0137, 0.0068, 0.0105, 0.0120, 0.0126], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0003, 0.0002, 0.0004, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:1') 2023-04-29 17:19:41,281 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=122063.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 17:19:54,204 INFO [optim.py:368] (1/8) Clipping_scale=2.0, grad-norm quartiles 1.510e+02 2.354e+02 2.915e+02 3.517e+02 1.210e+03, threshold=5.831e+02, percent-clipped=2.0 2023-04-29 17:20:34,740 INFO [train.py:904] (1/8) Epoch 13, batch 300, loss[loss=0.2173, simple_loss=0.281, pruned_loss=0.07678, over 16683.00 frames. ], tot_loss[loss=0.1897, simple_loss=0.2712, pruned_loss=0.05412, over 2585576.53 frames. ], batch size: 89, lr: 5.35e-03, grad_scale: 2.0 2023-04-29 17:20:47,419 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=122111.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 17:20:53,690 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.7816, 2.1396, 2.3232, 4.6489, 2.1751, 2.6509, 2.3430, 2.3927], device='cuda:1'), covar=tensor([0.0915, 0.3538, 0.2533, 0.0357, 0.4034, 0.2559, 0.3113, 0.3519], device='cuda:1'), in_proj_covar=tensor([0.0363, 0.0393, 0.0334, 0.0317, 0.0412, 0.0449, 0.0360, 0.0458], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-29 17:20:55,794 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 13, batch 350, loss[loss=0.177, simple_loss=0.2522, pruned_loss=0.05088, over 16725.00 frames. ], tot_loss[loss=0.1873, simple_loss=0.2687, pruned_loss=0.05298, over 2744217.29 frames. ], batch size: 124, lr: 5.35e-03, grad_scale: 2.0 2023-04-29 17:22:13,955 INFO [optim.py:368] (1/8) 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 17:22:20,584 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=122178.0, num_to_drop=1, layers_to_drop={3} 2023-04-29 17:22:43,792 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 13, batch 400, loss[loss=0.1817, simple_loss=0.2679, pruned_loss=0.04772, over 16719.00 frames. ], tot_loss[loss=0.1855, simple_loss=0.2671, pruned_loss=0.05199, over 2876121.64 frames. ], batch size: 57, lr: 5.35e-03, grad_scale: 4.0 2023-04-29 17:24:03,760 INFO [train.py:904] (1/8) Epoch 13, batch 450, loss[loss=0.1662, simple_loss=0.2453, pruned_loss=0.04357, over 16841.00 frames. ], tot_loss[loss=0.1835, simple_loss=0.2656, pruned_loss=0.05066, over 2978506.58 frames. ], batch size: 96, lr: 5.35e-03, grad_scale: 4.0 2023-04-29 17:24:06,527 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([5.2838, 5.1651, 5.0302, 4.5428, 4.6100, 4.9898, 5.0247, 4.5828], device='cuda:1'), covar=tensor([0.0511, 0.0455, 0.0272, 0.0301, 0.1034, 0.0448, 0.0356, 0.0798], device='cuda:1'), in_proj_covar=tensor([0.0250, 0.0335, 0.0298, 0.0277, 0.0317, 0.0317, 0.0203, 0.0347], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-29 17:24:09,692 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=122256.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 17:24:18,714 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=3.40 vs. limit=5.0 2023-04-29 17:24:32,943 INFO [optim.py:368] (1/8) 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 17:24:44,890 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.5760, 4.4343, 4.4240, 4.1037, 4.1100, 4.4740, 4.2567, 4.1577], device='cuda:1'), covar=tensor([0.0748, 0.0930, 0.0366, 0.0301, 0.0908, 0.0513, 0.0622, 0.0716], device='cuda:1'), in_proj_covar=tensor([0.0251, 0.0337, 0.0299, 0.0278, 0.0318, 0.0318, 0.0204, 0.0348], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-29 17:25:13,886 INFO [train.py:904] (1/8) Epoch 13, batch 500, loss[loss=0.1613, simple_loss=0.2522, pruned_loss=0.03521, over 17050.00 frames. ], tot_loss[loss=0.1818, simple_loss=0.2646, pruned_loss=0.04954, over 3056826.73 frames. ], batch size: 50, lr: 5.35e-03, grad_scale: 4.0 2023-04-29 17:25:18,836 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=122306.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 17:25:24,467 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.9207, 3.9607, 4.4240, 1.9416, 4.6096, 4.6979, 3.1313, 3.5937], device='cuda:1'), covar=tensor([0.0669, 0.0209, 0.0155, 0.1145, 0.0057, 0.0104, 0.0400, 0.0362], device='cuda:1'), in_proj_covar=tensor([0.0144, 0.0101, 0.0088, 0.0139, 0.0069, 0.0108, 0.0122, 0.0128], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:1') 2023-04-29 17:25:26,588 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=4.88 vs. limit=5.0 2023-04-29 17:25:36,909 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=122319.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 17:26:02,951 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=3.71 vs. limit=5.0 2023-04-29 17:26:21,438 INFO [train.py:904] (1/8) Epoch 13, batch 550, loss[loss=0.2091, simple_loss=0.2746, pruned_loss=0.0718, over 16431.00 frames. ], tot_loss[loss=0.1807, simple_loss=0.2633, pruned_loss=0.04906, over 3111774.10 frames. ], batch size: 146, lr: 5.35e-03, grad_scale: 4.0 2023-04-29 17:26:50,261 INFO [optim.py:368] (1/8) 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 17:27:00,817 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=122380.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 17:27:30,044 INFO [train.py:904] (1/8) Epoch 13, batch 600, loss[loss=0.1765, simple_loss=0.2644, pruned_loss=0.04434, over 16695.00 frames. ], tot_loss[loss=0.1809, simple_loss=0.2634, pruned_loss=0.04923, over 3163591.39 frames. ], batch size: 62, lr: 5.35e-03, grad_scale: 4.0 2023-04-29 17:27:39,419 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.75 vs. limit=2.0 2023-04-29 17:28:17,524 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.67 vs. limit=2.0 2023-04-29 17:28:37,794 INFO [train.py:904] (1/8) Epoch 13, batch 650, loss[loss=0.1475, simple_loss=0.2317, pruned_loss=0.03166, over 15987.00 frames. ], tot_loss[loss=0.1798, simple_loss=0.2618, pruned_loss=0.04891, over 3194265.52 frames. ], batch size: 35, lr: 5.34e-03, grad_scale: 4.0 2023-04-29 17:29:01,174 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=122468.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 17:29:07,722 INFO [optim.py:368] (1/8) 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 17:29:08,000 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=122473.0, num_to_drop=1, layers_to_drop={1} 2023-04-29 17:29:47,551 INFO [train.py:904] (1/8) Epoch 13, batch 700, loss[loss=0.1628, simple_loss=0.2462, pruned_loss=0.03968, over 15931.00 frames. ], tot_loss[loss=0.1788, simple_loss=0.2614, pruned_loss=0.04809, over 3233074.27 frames. ], batch size: 35, lr: 5.34e-03, grad_scale: 4.0 2023-04-29 17:30:25,433 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=122529.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 17:30:56,259 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=122551.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 17:30:57,589 INFO [train.py:904] (1/8) Epoch 13, batch 750, loss[loss=0.1817, simple_loss=0.2706, pruned_loss=0.04639, over 16672.00 frames. ], tot_loss[loss=0.1792, simple_loss=0.2619, pruned_loss=0.0483, over 3251689.82 frames. ], batch size: 57, lr: 5.34e-03, grad_scale: 4.0 2023-04-29 17:31:27,840 INFO [optim.py:368] (1/8) 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 17:32:02,985 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.0370, 4.0337, 3.9060, 3.6343, 3.6621, 4.0135, 3.6569, 3.8039], device='cuda:1'), covar=tensor([0.0634, 0.0635, 0.0300, 0.0264, 0.0654, 0.0427, 0.0821, 0.0556], device='cuda:1'), in_proj_covar=tensor([0.0257, 0.0345, 0.0307, 0.0287, 0.0327, 0.0328, 0.0208, 0.0358], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-29 17:32:09,553 INFO [train.py:904] (1/8) Epoch 13, batch 800, loss[loss=0.1685, simple_loss=0.2491, pruned_loss=0.04395, over 11956.00 frames. ], tot_loss[loss=0.1796, simple_loss=0.2617, pruned_loss=0.04873, over 3249518.73 frames. ], batch size: 246, lr: 5.34e-03, grad_scale: 8.0 2023-04-29 17:32:15,182 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=122606.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 17:33:15,730 INFO [train.py:904] (1/8) Epoch 13, batch 850, loss[loss=0.1788, simple_loss=0.2541, pruned_loss=0.05178, over 16281.00 frames. ], tot_loss[loss=0.1794, simple_loss=0.2613, pruned_loss=0.04874, over 3268696.62 frames. ], batch size: 165, lr: 5.34e-03, grad_scale: 8.0 2023-04-29 17:33:18,234 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=122654.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 17:33:23,682 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.8204, 4.0556, 2.4954, 4.5554, 3.0573, 4.5148, 2.3891, 3.0907], device='cuda:1'), covar=tensor([0.0251, 0.0313, 0.1343, 0.0205, 0.0721, 0.0458, 0.1432, 0.0707], device='cuda:1'), in_proj_covar=tensor([0.0155, 0.0164, 0.0188, 0.0136, 0.0168, 0.0206, 0.0196, 0.0172], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-04-29 17:33:44,225 INFO [optim.py:368] (1/8) 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 17:33:46,987 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=122675.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 17:34:23,501 INFO [train.py:904] (1/8) Epoch 13, batch 900, loss[loss=0.1517, simple_loss=0.2481, pruned_loss=0.02762, over 17143.00 frames. ], tot_loss[loss=0.1783, simple_loss=0.2603, pruned_loss=0.04818, over 3285054.86 frames. ], batch size: 48, lr: 5.34e-03, grad_scale: 8.0 2023-04-29 17:34:35,538 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([5.0525, 5.0700, 4.8052, 4.2856, 4.8809, 1.8892, 4.6601, 4.7483], device='cuda:1'), covar=tensor([0.0073, 0.0064, 0.0161, 0.0346, 0.0095, 0.2434, 0.0117, 0.0171], device='cuda:1'), in_proj_covar=tensor([0.0137, 0.0125, 0.0170, 0.0158, 0.0143, 0.0187, 0.0158, 0.0157], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-29 17:35:33,155 INFO [train.py:904] (1/8) Epoch 13, batch 950, loss[loss=0.1779, simple_loss=0.2713, pruned_loss=0.04227, over 16689.00 frames. ], tot_loss[loss=0.1777, simple_loss=0.2599, pruned_loss=0.04781, over 3293296.15 frames. ], batch size: 57, lr: 5.34e-03, grad_scale: 8.0 2023-04-29 17:36:02,667 INFO [optim.py:368] (1/8) 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 17:36:03,706 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=122773.0, num_to_drop=1, layers_to_drop={2} 2023-04-29 17:36:40,525 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.9147, 5.0307, 5.4551, 5.4244, 5.4362, 5.1006, 5.0386, 4.8680], device='cuda:1'), covar=tensor([0.0295, 0.0420, 0.0329, 0.0405, 0.0390, 0.0339, 0.0824, 0.0372], device='cuda:1'), in_proj_covar=tensor([0.0355, 0.0374, 0.0369, 0.0354, 0.0421, 0.0398, 0.0494, 0.0321], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:1') 2023-04-29 17:36:43,313 INFO [train.py:904] (1/8) Epoch 13, batch 1000, loss[loss=0.1643, simple_loss=0.2368, pruned_loss=0.04587, over 16729.00 frames. ], tot_loss[loss=0.1773, simple_loss=0.2587, pruned_loss=0.04795, over 3305056.90 frames. ], batch size: 76, lr: 5.34e-03, grad_scale: 8.0 2023-04-29 17:37:09,710 INFO [zipformer.py:625] (1/8) 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] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=122824.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 17:37:46,209 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.50 vs. limit=2.0 2023-04-29 17:37:51,967 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=122851.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 17:37:52,814 INFO [train.py:904] (1/8) Epoch 13, batch 1050, loss[loss=0.1577, simple_loss=0.2447, pruned_loss=0.03537, over 17186.00 frames. ], tot_loss[loss=0.1769, simple_loss=0.2584, pruned_loss=0.0477, over 3314201.83 frames. ], batch size: 44, lr: 5.34e-03, grad_scale: 8.0 2023-04-29 17:38:20,524 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=122872.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 17:38:21,177 INFO [optim.py:368] (1/8) Clipping_scale=2.0, grad-norm quartiles 1.558e+02 2.183e+02 2.662e+02 3.059e+02 5.432e+02, threshold=5.323e+02, percent-clipped=2.0 2023-04-29 17:38:32,667 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.2078, 4.5222, 4.6177, 3.5825, 3.9751, 4.5330, 4.1582, 2.8402], device='cuda:1'), covar=tensor([0.0359, 0.0042, 0.0029, 0.0228, 0.0080, 0.0057, 0.0052, 0.0361], device='cuda:1'), in_proj_covar=tensor([0.0133, 0.0073, 0.0072, 0.0129, 0.0084, 0.0092, 0.0082, 0.0124], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:1') 2023-04-29 17:38:58,265 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 13, batch 1100, loss[loss=0.208, simple_loss=0.2695, pruned_loss=0.07321, over 16912.00 frames. ], tot_loss[loss=0.1765, simple_loss=0.2578, pruned_loss=0.04756, over 3312505.70 frames. ], batch size: 109, lr: 5.33e-03, grad_scale: 8.0 2023-04-29 17:39:45,805 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=122933.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 17:40:11,765 INFO [train.py:904] (1/8) Epoch 13, batch 1150, loss[loss=0.1688, simple_loss=0.2435, pruned_loss=0.04702, over 16811.00 frames. ], tot_loss[loss=0.1754, simple_loss=0.2571, pruned_loss=0.0468, over 3309623.88 frames. ], batch size: 83, lr: 5.33e-03, grad_scale: 8.0 2023-04-29 17:40:39,510 INFO [optim.py:368] (1/8) 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 17:40:43,346 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=122975.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 17:41:20,392 INFO [train.py:904] (1/8) Epoch 13, batch 1200, loss[loss=0.1435, simple_loss=0.2263, pruned_loss=0.03031, over 16781.00 frames. ], tot_loss[loss=0.1736, simple_loss=0.2562, pruned_loss=0.04554, over 3316436.23 frames. ], batch size: 39, lr: 5.33e-03, grad_scale: 8.0 2023-04-29 17:41:50,191 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=123023.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 17:42:30,139 INFO [train.py:904] (1/8) Epoch 13, batch 1250, loss[loss=0.1467, simple_loss=0.2383, pruned_loss=0.02754, over 17133.00 frames. ], tot_loss[loss=0.1745, simple_loss=0.2566, pruned_loss=0.0462, over 3313564.98 frames. ], batch size: 47, lr: 5.33e-03, grad_scale: 8.0 2023-04-29 17:42:59,814 INFO [optim.py:368] (1/8) 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 17:43:32,162 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.8068, 2.1054, 2.3636, 4.3935, 2.0487, 2.5079, 2.2568, 2.3051], device='cuda:1'), covar=tensor([0.1002, 0.4108, 0.2610, 0.0482, 0.4815, 0.2881, 0.3548, 0.4117], device='cuda:1'), in_proj_covar=tensor([0.0371, 0.0403, 0.0339, 0.0327, 0.0419, 0.0463, 0.0369, 0.0472], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-29 17:43:40,470 INFO [train.py:904] (1/8) Epoch 13, batch 1300, loss[loss=0.1832, simple_loss=0.2725, pruned_loss=0.04697, over 16803.00 frames. ], tot_loss[loss=0.1743, simple_loss=0.257, pruned_loss=0.04576, over 3320832.37 frames. ], batch size: 62, lr: 5.33e-03, grad_scale: 8.0 2023-04-29 17:44:06,154 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.9712, 4.2116, 2.4155, 4.6492, 3.0991, 4.6820, 2.4387, 3.2310], device='cuda:1'), covar=tensor([0.0262, 0.0294, 0.1542, 0.0195, 0.0760, 0.0376, 0.1556, 0.0708], device='cuda:1'), in_proj_covar=tensor([0.0158, 0.0168, 0.0191, 0.0141, 0.0171, 0.0211, 0.0200, 0.0176], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-04-29 17:44:12,319 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=123124.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 17:44:49,681 INFO [train.py:904] (1/8) Epoch 13, batch 1350, loss[loss=0.203, simple_loss=0.2611, pruned_loss=0.07249, over 16868.00 frames. ], tot_loss[loss=0.1739, simple_loss=0.2569, pruned_loss=0.04539, over 3328905.24 frames. ], batch size: 90, lr: 5.33e-03, grad_scale: 8.0 2023-04-29 17:45:17,070 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=123172.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 17:45:17,975 INFO [optim.py:368] (1/8) 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 17:45:23,868 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=4.70 vs. limit=5.0 2023-04-29 17:45:58,577 INFO [train.py:904] (1/8) Epoch 13, batch 1400, loss[loss=0.1656, simple_loss=0.2594, pruned_loss=0.0359, over 17065.00 frames. ], tot_loss[loss=0.1742, simple_loss=0.2572, pruned_loss=0.04565, over 3318544.33 frames. ], batch size: 50, lr: 5.33e-03, grad_scale: 8.0 2023-04-29 17:46:35,857 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=123228.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 17:47:09,269 INFO [train.py:904] (1/8) Epoch 13, batch 1450, loss[loss=0.1758, simple_loss=0.2479, pruned_loss=0.05181, over 12064.00 frames. ], tot_loss[loss=0.1741, simple_loss=0.2564, pruned_loss=0.04594, over 3309802.88 frames. ], batch size: 247, lr: 5.33e-03, grad_scale: 8.0 2023-04-29 17:47:38,907 INFO [optim.py:368] (1/8) 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 17:48:19,759 INFO [train.py:904] (1/8) Epoch 13, batch 1500, loss[loss=0.1617, simple_loss=0.2572, pruned_loss=0.03306, over 17282.00 frames. ], tot_loss[loss=0.1742, simple_loss=0.2564, pruned_loss=0.04594, over 3314111.53 frames. ], batch size: 52, lr: 5.33e-03, grad_scale: 4.0 2023-04-29 17:48:37,116 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=123314.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 17:49:19,866 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.8985, 2.5538, 1.9483, 2.2620, 2.9127, 2.7223, 3.0435, 2.9863], device='cuda:1'), covar=tensor([0.0130, 0.0250, 0.0403, 0.0334, 0.0178, 0.0230, 0.0173, 0.0189], device='cuda:1'), in_proj_covar=tensor([0.0169, 0.0217, 0.0208, 0.0208, 0.0217, 0.0215, 0.0222, 0.0206], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-29 17:49:30,720 INFO [train.py:904] (1/8) Epoch 13, batch 1550, loss[loss=0.1993, simple_loss=0.2839, pruned_loss=0.05734, over 16660.00 frames. ], tot_loss[loss=0.177, simple_loss=0.2583, pruned_loss=0.04781, over 3313368.14 frames. ], batch size: 62, lr: 5.32e-03, grad_scale: 4.0 2023-04-29 17:49:31,158 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([1.6999, 2.7609, 2.4902, 4.1125, 3.3380, 4.0830, 1.6513, 2.8412], device='cuda:1'), covar=tensor([0.1352, 0.0571, 0.1065, 0.0159, 0.0134, 0.0358, 0.1344, 0.0737], device='cuda:1'), in_proj_covar=tensor([0.0154, 0.0160, 0.0182, 0.0153, 0.0194, 0.0209, 0.0182, 0.0180], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0003, 0.0004], device='cuda:1') 2023-04-29 17:50:00,255 INFO [optim.py:368] (1/8) 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 17:50:01,843 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=123375.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 17:50:11,954 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.0024, 2.9958, 3.1897, 2.1769, 2.9537, 3.2182, 3.0328, 1.8135], device='cuda:1'), covar=tensor([0.0432, 0.0108, 0.0046, 0.0318, 0.0103, 0.0084, 0.0082, 0.0415], device='cuda:1'), in_proj_covar=tensor([0.0133, 0.0072, 0.0072, 0.0129, 0.0084, 0.0093, 0.0083, 0.0124], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:1') 2023-04-29 17:50:27,764 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.8158, 3.6716, 3.8319, 3.6474, 3.7572, 4.1762, 3.8245, 3.5391], device='cuda:1'), covar=tensor([0.1852, 0.2087, 0.1995, 0.2512, 0.3115, 0.1950, 0.1605, 0.2879], device='cuda:1'), in_proj_covar=tensor([0.0370, 0.0526, 0.0574, 0.0455, 0.0616, 0.0597, 0.0454, 0.0606], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-04-29 17:50:39,392 INFO [train.py:904] (1/8) Epoch 13, batch 1600, loss[loss=0.2168, simple_loss=0.2999, pruned_loss=0.06687, over 15555.00 frames. ], tot_loss[loss=0.1787, simple_loss=0.2599, pruned_loss=0.04874, over 3313730.58 frames. ], batch size: 191, lr: 5.32e-03, grad_scale: 8.0 2023-04-29 17:50:52,350 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.8523, 3.7958, 4.2781, 2.0423, 4.4518, 4.4913, 3.1318, 3.5601], device='cuda:1'), covar=tensor([0.0692, 0.0216, 0.0185, 0.1127, 0.0069, 0.0124, 0.0406, 0.0338], device='cuda:1'), in_proj_covar=tensor([0.0143, 0.0101, 0.0090, 0.0138, 0.0069, 0.0110, 0.0121, 0.0127], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:1') 2023-04-29 17:50:59,399 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=4.69 vs. limit=5.0 2023-04-29 17:51:14,478 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=123428.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 17:51:47,303 INFO [train.py:904] (1/8) Epoch 13, batch 1650, loss[loss=0.1801, simple_loss=0.2513, pruned_loss=0.05448, over 16798.00 frames. ], tot_loss[loss=0.1797, simple_loss=0.2617, pruned_loss=0.04885, over 3320885.97 frames. ], batch size: 124, lr: 5.32e-03, grad_scale: 8.0 2023-04-29 17:52:10,692 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.0885, 2.5870, 1.9710, 2.2506, 2.9419, 2.7232, 3.1249, 3.0185], device='cuda:1'), covar=tensor([0.0175, 0.0304, 0.0425, 0.0346, 0.0189, 0.0264, 0.0193, 0.0209], device='cuda:1'), in_proj_covar=tensor([0.0170, 0.0217, 0.0209, 0.0208, 0.0218, 0.0216, 0.0223, 0.0207], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-29 17:52:18,026 INFO [optim.py:368] (1/8) 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 17:52:38,142 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 13, batch 1700, loss[loss=0.2363, simple_loss=0.3118, pruned_loss=0.08045, over 12382.00 frames. ], tot_loss[loss=0.1816, simple_loss=0.264, pruned_loss=0.04963, over 3311128.87 frames. ], batch size: 246, lr: 5.32e-03, grad_scale: 8.0 2023-04-29 17:53:12,587 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.9162, 3.2242, 2.9561, 1.9270, 2.5738, 2.0675, 3.3534, 3.4511], device='cuda:1'), covar=tensor([0.0242, 0.0730, 0.0673, 0.2016, 0.1031, 0.1115, 0.0568, 0.0799], device='cuda:1'), in_proj_covar=tensor([0.0145, 0.0149, 0.0159, 0.0144, 0.0137, 0.0125, 0.0137, 0.0159], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:1') 2023-04-29 17:53:32,034 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=123528.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 17:53:43,862 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.9155, 5.3008, 4.9909, 5.0643, 4.7796, 4.6876, 4.7099, 5.3665], device='cuda:1'), covar=tensor([0.1200, 0.0914, 0.1044, 0.0843, 0.0862, 0.0998, 0.1129, 0.0954], device='cuda:1'), in_proj_covar=tensor([0.0585, 0.0737, 0.0595, 0.0524, 0.0467, 0.0477, 0.0613, 0.0565], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-29 17:54:04,789 INFO [train.py:904] (1/8) Epoch 13, batch 1750, loss[loss=0.1955, simple_loss=0.2605, pruned_loss=0.06524, over 16910.00 frames. ], tot_loss[loss=0.1817, simple_loss=0.2643, pruned_loss=0.04962, over 3313784.34 frames. ], batch size: 109, lr: 5.32e-03, grad_scale: 8.0 2023-04-29 17:54:16,109 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.7572, 3.9107, 3.0504, 2.2713, 2.6203, 2.3857, 3.9550, 3.5110], device='cuda:1'), covar=tensor([0.2421, 0.0559, 0.1471, 0.2596, 0.2357, 0.1830, 0.0527, 0.1159], device='cuda:1'), in_proj_covar=tensor([0.0306, 0.0262, 0.0289, 0.0283, 0.0280, 0.0228, 0.0272, 0.0304], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-29 17:54:34,131 INFO [optim.py:368] (1/8) 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 17:54:37,988 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=123576.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 17:55:14,400 INFO [train.py:904] (1/8) Epoch 13, batch 1800, loss[loss=0.185, simple_loss=0.276, pruned_loss=0.04702, over 17037.00 frames. ], tot_loss[loss=0.1827, simple_loss=0.266, pruned_loss=0.04967, over 3300667.41 frames. ], batch size: 55, lr: 5.32e-03, grad_scale: 8.0 2023-04-29 17:55:16,009 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.3996, 4.3586, 4.3821, 3.8519, 4.3612, 1.7917, 4.0973, 4.0997], device='cuda:1'), covar=tensor([0.0111, 0.0099, 0.0139, 0.0291, 0.0108, 0.2406, 0.0147, 0.0206], device='cuda:1'), in_proj_covar=tensor([0.0142, 0.0131, 0.0176, 0.0165, 0.0149, 0.0191, 0.0165, 0.0164], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-29 17:56:23,372 INFO [train.py:904] (1/8) Epoch 13, batch 1850, loss[loss=0.2054, simple_loss=0.2918, pruned_loss=0.0595, over 16887.00 frames. ], tot_loss[loss=0.1827, simple_loss=0.2664, pruned_loss=0.04954, over 3302405.33 frames. ], batch size: 116, lr: 5.32e-03, grad_scale: 8.0 2023-04-29 17:56:47,436 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=123670.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 17:56:52,510 INFO [optim.py:368] (1/8) 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 17:57:33,303 INFO [train.py:904] (1/8) Epoch 13, batch 1900, loss[loss=0.163, simple_loss=0.2582, pruned_loss=0.03387, over 17035.00 frames. ], tot_loss[loss=0.1813, simple_loss=0.2655, pruned_loss=0.04858, over 3316527.15 frames. ], batch size: 50, lr: 5.32e-03, grad_scale: 8.0 2023-04-29 17:58:29,984 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.9155, 4.1249, 2.5481, 4.6320, 3.1335, 4.6029, 2.6117, 3.1865], device='cuda:1'), covar=tensor([0.0275, 0.0365, 0.1419, 0.0211, 0.0774, 0.0467, 0.1344, 0.0676], device='cuda:1'), in_proj_covar=tensor([0.0157, 0.0167, 0.0188, 0.0142, 0.0170, 0.0210, 0.0197, 0.0174], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-04-29 17:58:39,328 INFO [train.py:904] (1/8) Epoch 13, batch 1950, loss[loss=0.1732, simple_loss=0.2669, pruned_loss=0.03982, over 16716.00 frames. ], tot_loss[loss=0.1816, simple_loss=0.2659, pruned_loss=0.04865, over 3316187.02 frames. ], batch size: 57, lr: 5.32e-03, grad_scale: 8.0 2023-04-29 17:58:56,747 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.8586, 4.0421, 3.0315, 2.3740, 2.7621, 2.3931, 4.2454, 3.6916], device='cuda:1'), covar=tensor([0.2386, 0.0643, 0.1564, 0.2514, 0.2573, 0.1865, 0.0468, 0.1107], device='cuda:1'), in_proj_covar=tensor([0.0305, 0.0262, 0.0289, 0.0284, 0.0282, 0.0228, 0.0272, 0.0306], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-29 17:59:09,939 INFO [optim.py:368] (1/8) 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 17:59:22,474 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([1.8576, 2.9187, 2.5376, 4.3057, 3.6162, 4.1625, 1.4564, 2.9273], device='cuda:1'), covar=tensor([0.1317, 0.0551, 0.1110, 0.0153, 0.0188, 0.0345, 0.1520, 0.0747], device='cuda:1'), in_proj_covar=tensor([0.0156, 0.0161, 0.0184, 0.0155, 0.0196, 0.0211, 0.0183, 0.0182], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0003, 0.0004], device='cuda:1') 2023-04-29 17:59:23,815 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.21 vs. limit=2.0 2023-04-29 17:59:25,258 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=123784.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 17:59:48,915 INFO [train.py:904] (1/8) Epoch 13, batch 2000, loss[loss=0.1954, simple_loss=0.2564, pruned_loss=0.06724, over 16873.00 frames. ], tot_loss[loss=0.1811, simple_loss=0.2655, pruned_loss=0.04833, over 3319035.94 frames. ], batch size: 96, lr: 5.31e-03, grad_scale: 8.0 2023-04-29 18:00:31,320 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.4669, 4.4562, 4.6107, 4.4271, 4.4628, 5.0921, 4.6377, 4.3387], device='cuda:1'), covar=tensor([0.1576, 0.2089, 0.2381, 0.2250, 0.3100, 0.1160, 0.1608, 0.2666], device='cuda:1'), in_proj_covar=tensor([0.0376, 0.0537, 0.0583, 0.0467, 0.0624, 0.0610, 0.0463, 0.0616], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-04-29 18:00:59,629 INFO [train.py:904] (1/8) Epoch 13, batch 2050, loss[loss=0.1873, simple_loss=0.2799, pruned_loss=0.04731, over 16725.00 frames. ], tot_loss[loss=0.1822, simple_loss=0.2656, pruned_loss=0.04938, over 3313242.88 frames. ], batch size: 57, lr: 5.31e-03, grad_scale: 8.0 2023-04-29 18:01:28,828 INFO [zipformer.py:625] (1/8) 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] (1/8) Clipping_scale=2.0, grad-norm quartiles 1.552e+02 2.367e+02 2.767e+02 3.495e+02 6.657e+02, threshold=5.534e+02, percent-clipped=3.0 2023-04-29 18:02:09,973 INFO [train.py:904] (1/8) Epoch 13, batch 2100, loss[loss=0.2101, simple_loss=0.2985, pruned_loss=0.06086, over 17059.00 frames. ], tot_loss[loss=0.1819, simple_loss=0.2655, pruned_loss=0.04908, over 3324520.19 frames. ], batch size: 53, lr: 5.31e-03, grad_scale: 8.0 2023-04-29 18:02:41,421 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.2086, 4.1755, 4.0727, 3.8585, 3.8145, 4.1593, 3.8582, 3.9541], device='cuda:1'), covar=tensor([0.0659, 0.0647, 0.0276, 0.0259, 0.0779, 0.0465, 0.0866, 0.0568], device='cuda:1'), in_proj_covar=tensor([0.0271, 0.0364, 0.0322, 0.0301, 0.0344, 0.0346, 0.0218, 0.0378], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-04-29 18:02:54,464 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=123933.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 18:03:20,325 INFO [train.py:904] (1/8) Epoch 13, batch 2150, loss[loss=0.1972, simple_loss=0.274, pruned_loss=0.06023, over 16216.00 frames. ], tot_loss[loss=0.1833, simple_loss=0.2667, pruned_loss=0.04994, over 3319726.86 frames. ], batch size: 165, lr: 5.31e-03, grad_scale: 8.0 2023-04-29 18:03:44,958 INFO [zipformer.py:625] (1/8) 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] (1/8) 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 18:04:30,742 INFO [train.py:904] (1/8) Epoch 13, batch 2200, loss[loss=0.1818, simple_loss=0.276, pruned_loss=0.04382, over 17074.00 frames. ], tot_loss[loss=0.1826, simple_loss=0.2663, pruned_loss=0.04941, over 3324873.51 frames. ], batch size: 53, lr: 5.31e-03, grad_scale: 8.0 2023-04-29 18:04:53,398 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=124018.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 18:05:19,530 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=124036.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 18:05:36,226 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=4.26 vs. limit=5.0 2023-04-29 18:05:36,332 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.95 vs. limit=2.0 2023-04-29 18:05:40,213 INFO [train.py:904] (1/8) Epoch 13, batch 2250, loss[loss=0.1903, simple_loss=0.2784, pruned_loss=0.05113, over 16713.00 frames. ], tot_loss[loss=0.1824, simple_loss=0.2662, pruned_loss=0.04933, over 3324875.56 frames. ], batch size: 62, lr: 5.31e-03, grad_scale: 8.0 2023-04-29 18:05:53,696 INFO [zipformer.py:625] (1/8) 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] (1/8) 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 18:06:22,668 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=124084.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 18:06:42,876 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=124097.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 18:06:48,019 INFO [train.py:904] (1/8) Epoch 13, batch 2300, loss[loss=0.1701, simple_loss=0.256, pruned_loss=0.04211, over 17201.00 frames. ], tot_loss[loss=0.1831, simple_loss=0.2669, pruned_loss=0.0496, over 3319948.26 frames. ], batch size: 46, lr: 5.31e-03, grad_scale: 8.0 2023-04-29 18:07:16,670 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.9279, 4.7442, 4.9594, 5.1691, 5.3568, 4.6272, 5.3513, 5.3385], device='cuda:1'), covar=tensor([0.1591, 0.1217, 0.1470, 0.0640, 0.0466, 0.0854, 0.0518, 0.0501], device='cuda:1'), in_proj_covar=tensor([0.0573, 0.0715, 0.0862, 0.0723, 0.0541, 0.0570, 0.0576, 0.0661], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-29 18:07:17,812 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=124123.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 18:07:22,836 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.17 vs. limit=2.0 2023-04-29 18:07:29,439 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 13, batch 2350, loss[loss=0.2066, simple_loss=0.2829, pruned_loss=0.06516, over 16704.00 frames. ], tot_loss[loss=0.1838, simple_loss=0.2675, pruned_loss=0.05002, over 3322118.58 frames. ], batch size: 134, lr: 5.31e-03, grad_scale: 8.0 2023-04-29 18:08:26,417 INFO [optim.py:368] (1/8) 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 18:08:32,536 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.39 vs. limit=2.0 2023-04-29 18:09:06,167 INFO [train.py:904] (1/8) Epoch 13, batch 2400, loss[loss=0.1969, simple_loss=0.2755, pruned_loss=0.05917, over 16274.00 frames. ], tot_loss[loss=0.1847, simple_loss=0.2689, pruned_loss=0.05022, over 3324504.10 frames. ], batch size: 165, lr: 5.31e-03, grad_scale: 8.0 2023-04-29 18:09:07,810 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.0071, 4.3963, 4.4835, 3.3255, 3.7040, 4.2911, 3.9466, 2.5168], device='cuda:1'), covar=tensor([0.0378, 0.0066, 0.0030, 0.0266, 0.0102, 0.0081, 0.0071, 0.0387], device='cuda:1'), in_proj_covar=tensor([0.0129, 0.0072, 0.0071, 0.0127, 0.0083, 0.0092, 0.0082, 0.0121], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:1') 2023-04-29 18:09:42,796 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=124228.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 18:09:53,129 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.36 vs. limit=2.0 2023-04-29 18:10:03,740 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([1.8267, 3.1928, 2.5509, 4.5354, 3.5632, 4.2144, 1.6706, 3.0626], device='cuda:1'), covar=tensor([0.1483, 0.0632, 0.1126, 0.0191, 0.0284, 0.0371, 0.1680, 0.0804], device='cuda:1'), in_proj_covar=tensor([0.0156, 0.0161, 0.0184, 0.0157, 0.0197, 0.0210, 0.0183, 0.0182], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0003, 0.0004], device='cuda:1') 2023-04-29 18:10:15,680 INFO [train.py:904] (1/8) Epoch 13, batch 2450, loss[loss=0.19, simple_loss=0.2659, pruned_loss=0.05705, over 16875.00 frames. ], tot_loss[loss=0.1854, simple_loss=0.27, pruned_loss=0.05037, over 3322850.09 frames. ], batch size: 96, lr: 5.31e-03, grad_scale: 8.0 2023-04-29 18:10:43,317 INFO [zipformer.py:625] (1/8) 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] (1/8) Clipping_scale=2.0, grad-norm quartiles 1.567e+02 2.400e+02 2.819e+02 3.354e+02 7.582e+02, threshold=5.638e+02, percent-clipped=2.0 2023-04-29 18:11:11,575 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.94 vs. limit=2.0 2023-04-29 18:11:24,236 INFO [train.py:904] (1/8) Epoch 13, batch 2500, loss[loss=0.2003, simple_loss=0.2891, pruned_loss=0.05576, over 17009.00 frames. ], tot_loss[loss=0.1853, simple_loss=0.2696, pruned_loss=0.05049, over 3322879.54 frames. ], batch size: 50, lr: 5.30e-03, grad_scale: 8.0 2023-04-29 18:11:49,564 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.23 vs. limit=2.0 2023-04-29 18:12:09,430 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=124333.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 18:12:31,486 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=124348.0, num_to_drop=1, layers_to_drop={0} 2023-04-29 18:12:37,101 INFO [train.py:904] (1/8) Epoch 13, batch 2550, loss[loss=0.1748, simple_loss=0.2493, pruned_loss=0.05011, over 16714.00 frames. ], tot_loss[loss=0.1863, simple_loss=0.2707, pruned_loss=0.05098, over 3304857.66 frames. ], batch size: 89, lr: 5.30e-03, grad_scale: 8.0 2023-04-29 18:13:06,417 INFO [optim.py:368] (1/8) 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 18:13:26,941 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=3.21 vs. limit=5.0 2023-04-29 18:13:31,385 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=124392.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 18:13:44,302 INFO [train.py:904] (1/8) Epoch 13, batch 2600, loss[loss=0.1859, simple_loss=0.2792, pruned_loss=0.04626, over 16702.00 frames. ], tot_loss[loss=0.1853, simple_loss=0.2697, pruned_loss=0.05042, over 3307823.42 frames. ], batch size: 57, lr: 5.30e-03, grad_scale: 8.0 2023-04-29 18:13:51,350 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.6153, 3.7707, 2.8913, 2.2400, 2.4618, 2.2484, 3.7595, 3.3393], device='cuda:1'), covar=tensor([0.2437, 0.0612, 0.1419, 0.2603, 0.2504, 0.1870, 0.0500, 0.1296], device='cuda:1'), in_proj_covar=tensor([0.0307, 0.0262, 0.0288, 0.0285, 0.0284, 0.0229, 0.0272, 0.0308], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:1') 2023-04-29 18:13:54,824 INFO [zipformer.py:625] (1/8) 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] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=124418.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 18:14:26,633 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=124432.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 18:14:48,664 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.8629, 4.8776, 5.3149, 5.2992, 5.3617, 4.9512, 4.9395, 4.6942], device='cuda:1'), covar=tensor([0.0270, 0.0486, 0.0399, 0.0425, 0.0392, 0.0323, 0.0805, 0.0363], device='cuda:1'), in_proj_covar=tensor([0.0363, 0.0382, 0.0382, 0.0361, 0.0432, 0.0406, 0.0504, 0.0324], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:1') 2023-04-29 18:14:54,847 INFO [train.py:904] (1/8) Epoch 13, batch 2650, loss[loss=0.185, simple_loss=0.2788, pruned_loss=0.04564, over 17171.00 frames. ], tot_loss[loss=0.185, simple_loss=0.27, pruned_loss=0.04999, over 3318534.03 frames. ], batch size: 48, lr: 5.30e-03, grad_scale: 8.0 2023-04-29 18:15:26,000 INFO [optim.py:368] (1/8) 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 18:15:52,754 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 13, batch 2700, loss[loss=0.1898, simple_loss=0.2704, pruned_loss=0.05456, over 16759.00 frames. ], tot_loss[loss=0.1842, simple_loss=0.27, pruned_loss=0.04919, over 3325538.13 frames. ], batch size: 124, lr: 5.30e-03, grad_scale: 8.0 2023-04-29 18:16:40,349 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=124528.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 18:16:48,439 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.8398, 3.7481, 3.9377, 4.0360, 4.0688, 3.6384, 3.8795, 4.0950], device='cuda:1'), covar=tensor([0.1424, 0.0983, 0.1114, 0.0585, 0.0599, 0.1868, 0.1748, 0.0682], device='cuda:1'), in_proj_covar=tensor([0.0579, 0.0722, 0.0871, 0.0729, 0.0548, 0.0578, 0.0581, 0.0672], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-29 18:16:56,183 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.3684, 3.7014, 4.0477, 2.2352, 3.1836, 2.5672, 3.9102, 3.9323], device='cuda:1'), covar=tensor([0.0262, 0.0759, 0.0436, 0.1675, 0.0749, 0.0851, 0.0580, 0.0857], device='cuda:1'), in_proj_covar=tensor([0.0148, 0.0152, 0.0162, 0.0147, 0.0138, 0.0126, 0.0139, 0.0162], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:1') 2023-04-29 18:17:13,323 INFO [train.py:904] (1/8) Epoch 13, batch 2750, loss[loss=0.1716, simple_loss=0.2691, pruned_loss=0.03705, over 16998.00 frames. ], tot_loss[loss=0.1839, simple_loss=0.2698, pruned_loss=0.04895, over 3321375.82 frames. ], batch size: 50, lr: 5.30e-03, grad_scale: 8.0 2023-04-29 18:17:44,016 INFO [optim.py:368] (1/8) 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 18:17:46,595 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=124576.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 18:18:05,372 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.7415, 2.2459, 2.3267, 4.5513, 2.2275, 2.7484, 2.4370, 2.4732], device='cuda:1'), covar=tensor([0.0949, 0.3623, 0.2540, 0.0387, 0.3924, 0.2378, 0.3214, 0.3550], device='cuda:1'), in_proj_covar=tensor([0.0374, 0.0405, 0.0339, 0.0325, 0.0417, 0.0468, 0.0369, 0.0476], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-29 18:18:22,961 INFO [train.py:904] (1/8) Epoch 13, batch 2800, loss[loss=0.1779, simple_loss=0.2666, pruned_loss=0.04458, over 17111.00 frames. ], tot_loss[loss=0.1837, simple_loss=0.2694, pruned_loss=0.04898, over 3309580.62 frames. ], batch size: 47, lr: 5.30e-03, grad_scale: 8.0 2023-04-29 18:18:24,555 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.9905, 4.1335, 2.1506, 4.7428, 2.9158, 4.6143, 2.1620, 3.2809], device='cuda:1'), covar=tensor([0.0234, 0.0273, 0.1766, 0.0163, 0.0799, 0.0340, 0.1769, 0.0637], device='cuda:1'), in_proj_covar=tensor([0.0158, 0.0169, 0.0190, 0.0145, 0.0171, 0.0215, 0.0199, 0.0176], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-04-29 18:18:59,117 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=124628.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 18:19:12,806 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 13, batch 2850, loss[loss=0.1677, simple_loss=0.2652, pruned_loss=0.03506, over 17026.00 frames. ], tot_loss[loss=0.1824, simple_loss=0.2682, pruned_loss=0.04826, over 3310038.12 frames. ], batch size: 50, lr: 5.30e-03, grad_scale: 8.0 2023-04-29 18:19:46,030 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=124663.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 18:20:00,129 INFO [optim.py:368] (1/8) 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 18:20:04,888 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.2959, 4.1538, 4.3122, 4.4787, 4.5673, 4.1395, 4.3716, 4.5703], device='cuda:1'), covar=tensor([0.1234, 0.0910, 0.1177, 0.0601, 0.0548, 0.1132, 0.1636, 0.0579], device='cuda:1'), in_proj_covar=tensor([0.0578, 0.0722, 0.0873, 0.0733, 0.0548, 0.0581, 0.0583, 0.0672], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-29 18:20:24,517 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=124692.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 18:20:33,942 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=124699.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 18:20:37,525 INFO [train.py:904] (1/8) Epoch 13, batch 2900, loss[loss=0.2065, simple_loss=0.2837, pruned_loss=0.06458, over 16883.00 frames. ], tot_loss[loss=0.1828, simple_loss=0.2676, pruned_loss=0.04901, over 3313201.30 frames. ], batch size: 109, lr: 5.30e-03, grad_scale: 8.0 2023-04-29 18:20:40,110 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=124704.0, num_to_drop=1, layers_to_drop={3} 2023-04-29 18:20:58,976 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=124718.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 18:21:07,462 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=124724.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 18:21:08,697 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.2935, 3.5992, 3.6992, 2.2915, 3.1339, 2.5060, 3.8365, 3.7845], device='cuda:1'), covar=tensor([0.0241, 0.0809, 0.0475, 0.1657, 0.0715, 0.0919, 0.0484, 0.0859], device='cuda:1'), in_proj_covar=tensor([0.0148, 0.0151, 0.0161, 0.0146, 0.0138, 0.0126, 0.0138, 0.0162], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:1') 2023-04-29 18:21:29,782 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=124740.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 18:21:37,887 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.45 vs. limit=2.0 2023-04-29 18:21:46,237 INFO [train.py:904] (1/8) Epoch 13, batch 2950, loss[loss=0.1919, simple_loss=0.2654, pruned_loss=0.05926, over 16739.00 frames. ], tot_loss[loss=0.1818, simple_loss=0.2668, pruned_loss=0.04839, over 3323872.06 frames. ], batch size: 134, lr: 5.29e-03, grad_scale: 8.0 2023-04-29 18:22:05,796 INFO [zipformer.py:625] (1/8) 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] (1/8) 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 18:22:35,701 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=124788.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 18:22:39,656 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([5.4505, 5.3818, 5.2568, 4.8156, 4.8362, 5.2719, 5.3166, 4.9026], device='cuda:1'), covar=tensor([0.0529, 0.0420, 0.0244, 0.0280, 0.1132, 0.0392, 0.0235, 0.0643], device='cuda:1'), in_proj_covar=tensor([0.0271, 0.0366, 0.0324, 0.0305, 0.0348, 0.0350, 0.0221, 0.0380], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-04-29 18:22:48,051 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.7012, 3.6159, 4.0154, 2.1695, 4.0908, 4.0806, 3.2128, 3.1108], device='cuda:1'), covar=tensor([0.0662, 0.0197, 0.0148, 0.1003, 0.0063, 0.0157, 0.0318, 0.0401], device='cuda:1'), in_proj_covar=tensor([0.0144, 0.0101, 0.0090, 0.0138, 0.0070, 0.0112, 0.0122, 0.0127], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:1') 2023-04-29 18:22:55,199 INFO [train.py:904] (1/8) Epoch 13, batch 3000, loss[loss=0.2144, simple_loss=0.284, pruned_loss=0.07245, over 16861.00 frames. ], tot_loss[loss=0.183, simple_loss=0.267, pruned_loss=0.04944, over 3321574.55 frames. ], batch size: 109, lr: 5.29e-03, grad_scale: 8.0 2023-04-29 18:22:55,199 INFO [train.py:929] (1/8) Computing validation loss 2023-04-29 18:23:04,001 INFO [train.py:938] (1/8) Epoch 13, validation: loss=0.1391, simple_loss=0.2452, pruned_loss=0.01648, over 944034.00 frames. 2023-04-29 18:23:04,001 INFO [train.py:939] (1/8) Maximum memory allocated so far is 17845MB 2023-04-29 18:23:29,527 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=124820.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 18:24:14,167 INFO [train.py:904] (1/8) Epoch 13, batch 3050, loss[loss=0.1588, simple_loss=0.2525, pruned_loss=0.03262, over 17185.00 frames. ], tot_loss[loss=0.1818, simple_loss=0.2661, pruned_loss=0.04877, over 3326237.58 frames. ], batch size: 46, lr: 5.29e-03, grad_scale: 8.0 2023-04-29 18:24:44,775 INFO [optim.py:368] (1/8) Clipping_scale=2.0, grad-norm quartiles 1.534e+02 2.296e+02 3.000e+02 3.565e+02 8.414e+02, threshold=5.999e+02, percent-clipped=2.0 2023-04-29 18:24:55,371 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=124881.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 18:25:05,928 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([5.0514, 5.0691, 4.8509, 4.4783, 4.2070, 4.9654, 4.9454, 4.4642], device='cuda:1'), covar=tensor([0.0669, 0.0580, 0.0392, 0.0387, 0.1484, 0.0512, 0.0378, 0.0767], device='cuda:1'), in_proj_covar=tensor([0.0270, 0.0366, 0.0325, 0.0305, 0.0347, 0.0349, 0.0221, 0.0380], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-04-29 18:25:25,283 INFO [train.py:904] (1/8) Epoch 13, batch 3100, loss[loss=0.157, simple_loss=0.2456, pruned_loss=0.03417, over 17156.00 frames. ], tot_loss[loss=0.1808, simple_loss=0.2652, pruned_loss=0.04824, over 3323632.81 frames. ], batch size: 46, lr: 5.29e-03, grad_scale: 8.0 2023-04-29 18:26:01,301 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=124928.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 18:26:13,269 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.5979, 1.7470, 2.2185, 2.5226, 2.6088, 2.4692, 1.7638, 2.7150], device='cuda:1'), covar=tensor([0.0164, 0.0391, 0.0261, 0.0228, 0.0228, 0.0232, 0.0407, 0.0122], device='cuda:1'), in_proj_covar=tensor([0.0174, 0.0181, 0.0165, 0.0172, 0.0180, 0.0134, 0.0180, 0.0126], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-29 18:26:34,831 INFO [train.py:904] (1/8) Epoch 13, batch 3150, loss[loss=0.1584, simple_loss=0.2391, pruned_loss=0.03884, over 16484.00 frames. ], tot_loss[loss=0.1803, simple_loss=0.2646, pruned_loss=0.04796, over 3327083.59 frames. ], batch size: 75, lr: 5.29e-03, grad_scale: 8.0 2023-04-29 18:27:05,835 INFO [optim.py:368] (1/8) 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 18:27:09,070 INFO [zipformer.py:625] (1/8) 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] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=124994.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 18:27:45,737 INFO [train.py:904] (1/8) Epoch 13, batch 3200, loss[loss=0.1521, simple_loss=0.2451, pruned_loss=0.02954, over 17211.00 frames. ], tot_loss[loss=0.1791, simple_loss=0.2636, pruned_loss=0.04733, over 3326198.70 frames. ], batch size: 44, lr: 5.29e-03, grad_scale: 8.0 2023-04-29 18:27:48,497 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=125004.0, num_to_drop=1, layers_to_drop={0} 2023-04-29 18:27:52,268 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.5192, 3.3499, 3.7251, 1.9537, 3.7580, 3.8030, 3.0506, 2.8940], device='cuda:1'), covar=tensor([0.0698, 0.0211, 0.0130, 0.1085, 0.0073, 0.0143, 0.0358, 0.0419], device='cuda:1'), in_proj_covar=tensor([0.0144, 0.0102, 0.0090, 0.0139, 0.0071, 0.0112, 0.0122, 0.0128], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:1') 2023-04-29 18:28:10,523 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=125019.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 18:28:19,679 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([1.6921, 3.1094, 2.6953, 4.9824, 4.1200, 4.4754, 1.6187, 3.1680], device='cuda:1'), covar=tensor([0.1429, 0.0692, 0.1200, 0.0158, 0.0279, 0.0392, 0.1640, 0.0809], device='cuda:1'), in_proj_covar=tensor([0.0155, 0.0162, 0.0183, 0.0158, 0.0198, 0.0210, 0.0183, 0.0181], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0003, 0.0004], device='cuda:1') 2023-04-29 18:28:56,246 INFO [train.py:904] (1/8) Epoch 13, batch 3250, loss[loss=0.1641, simple_loss=0.2416, pruned_loss=0.04337, over 16797.00 frames. ], tot_loss[loss=0.1799, simple_loss=0.2639, pruned_loss=0.04797, over 3324863.52 frames. ], batch size: 39, lr: 5.29e-03, grad_scale: 8.0 2023-04-29 18:28:56,531 INFO [zipformer.py:625] (1/8) 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] (1/8) 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 18:29:46,160 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=125088.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 18:30:05,262 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=4.96 vs. limit=5.0 2023-04-29 18:30:05,563 INFO [train.py:904] (1/8) Epoch 13, batch 3300, loss[loss=0.1782, simple_loss=0.2722, pruned_loss=0.04215, over 17118.00 frames. ], tot_loss[loss=0.1812, simple_loss=0.2652, pruned_loss=0.04858, over 3316933.39 frames. ], batch size: 49, lr: 5.29e-03, grad_scale: 8.0 2023-04-29 18:30:11,924 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.86 vs. limit=2.0 2023-04-29 18:30:42,335 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=125128.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 18:30:53,997 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=125136.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 18:31:15,061 INFO [train.py:904] (1/8) Epoch 13, batch 3350, loss[loss=0.1875, simple_loss=0.2829, pruned_loss=0.04607, over 16658.00 frames. ], tot_loss[loss=0.1825, simple_loss=0.2668, pruned_loss=0.04915, over 3300933.69 frames. ], batch size: 57, lr: 5.29e-03, grad_scale: 8.0 2023-04-29 18:31:45,937 INFO [optim.py:368] (1/8) 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 18:31:48,365 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=125176.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 18:32:06,958 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=125189.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 18:32:09,086 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([5.1603, 5.5516, 5.3033, 5.3647, 4.9176, 4.9366, 4.9839, 5.6677], device='cuda:1'), covar=tensor([0.1194, 0.0908, 0.1070, 0.0757, 0.0884, 0.0808, 0.1182, 0.0883], device='cuda:1'), in_proj_covar=tensor([0.0593, 0.0749, 0.0606, 0.0530, 0.0474, 0.0478, 0.0622, 0.0572], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-29 18:32:24,429 INFO [train.py:904] (1/8) Epoch 13, batch 3400, loss[loss=0.1818, simple_loss=0.259, pruned_loss=0.05229, over 15548.00 frames. ], tot_loss[loss=0.182, simple_loss=0.2663, pruned_loss=0.04891, over 3307253.22 frames. ], batch size: 191, lr: 5.29e-03, grad_scale: 4.0 2023-04-29 18:33:35,396 INFO [train.py:904] (1/8) Epoch 13, batch 3450, loss[loss=0.179, simple_loss=0.2715, pruned_loss=0.04326, over 17049.00 frames. ], tot_loss[loss=0.1811, simple_loss=0.2652, pruned_loss=0.04851, over 3310016.01 frames. ], batch size: 53, lr: 5.28e-03, grad_scale: 4.0 2023-04-29 18:33:42,597 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.6730, 3.6273, 2.8283, 2.1884, 2.4527, 2.1387, 3.6907, 3.3500], device='cuda:1'), covar=tensor([0.2361, 0.0614, 0.1428, 0.2624, 0.2530, 0.2009, 0.0510, 0.1198], device='cuda:1'), in_proj_covar=tensor([0.0304, 0.0261, 0.0287, 0.0285, 0.0284, 0.0229, 0.0274, 0.0308], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:1') 2023-04-29 18:34:07,280 INFO [optim.py:368] (1/8) 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 18:34:34,781 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=125294.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 18:34:46,234 INFO [train.py:904] (1/8) Epoch 13, batch 3500, loss[loss=0.2024, simple_loss=0.2777, pruned_loss=0.06355, over 16824.00 frames. ], tot_loss[loss=0.1807, simple_loss=0.2646, pruned_loss=0.04839, over 3302906.53 frames. ], batch size: 116, lr: 5.28e-03, grad_scale: 4.0 2023-04-29 18:35:09,882 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=125319.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 18:35:42,973 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 13, batch 3550, loss[loss=0.164, simple_loss=0.2576, pruned_loss=0.0352, over 16684.00 frames. ], tot_loss[loss=0.18, simple_loss=0.2638, pruned_loss=0.04806, over 3298933.85 frames. ], batch size: 62, lr: 5.28e-03, grad_scale: 4.0 2023-04-29 18:36:03,230 INFO [zipformer.py:625] (1/8) 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] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=125367.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 18:36:20,550 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([5.4811, 5.4209, 5.2892, 4.7966, 5.3332, 2.5152, 5.0554, 5.2779], device='cuda:1'), covar=tensor([0.0062, 0.0055, 0.0134, 0.0313, 0.0064, 0.1928, 0.0102, 0.0135], device='cuda:1'), in_proj_covar=tensor([0.0144, 0.0134, 0.0181, 0.0170, 0.0152, 0.0191, 0.0170, 0.0168], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-29 18:36:30,595 INFO [optim.py:368] (1/8) 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] (1/8) 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] (1/8) Epoch 13, batch 3600, loss[loss=0.1877, simple_loss=0.2802, pruned_loss=0.04765, over 16757.00 frames. ], tot_loss[loss=0.1779, simple_loss=0.2613, pruned_loss=0.04719, over 3309358.75 frames. ], batch size: 57, lr: 5.28e-03, grad_scale: 8.0 2023-04-29 18:37:24,880 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=4.68 vs. limit=5.0 2023-04-29 18:37:31,283 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=125417.0, num_to_drop=1, layers_to_drop={2} 2023-04-29 18:37:35,962 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=125421.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 18:37:59,812 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 13, batch 3650, loss[loss=0.1906, simple_loss=0.2614, pruned_loss=0.05989, over 16761.00 frames. ], tot_loss[loss=0.1777, simple_loss=0.2603, pruned_loss=0.0475, over 3309762.92 frames. ], batch size: 124, lr: 5.28e-03, grad_scale: 8.0 2023-04-29 18:38:57,386 INFO [optim.py:368] (1/8) 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 18:38:59,582 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=125476.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 18:39:07,779 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=125482.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 18:39:10,723 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 13, batch 3700, loss[loss=0.1942, simple_loss=0.2659, pruned_loss=0.06124, over 15460.00 frames. ], tot_loss[loss=0.1795, simple_loss=0.2601, pruned_loss=0.04949, over 3279559.69 frames. ], batch size: 190, lr: 5.28e-03, grad_scale: 8.0 2023-04-29 18:40:10,903 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=125524.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 18:40:51,173 INFO [train.py:904] (1/8) Epoch 13, batch 3750, loss[loss=0.1949, simple_loss=0.2749, pruned_loss=0.0575, over 15463.00 frames. ], tot_loss[loss=0.1814, simple_loss=0.2606, pruned_loss=0.05109, over 3276349.96 frames. ], batch size: 190, lr: 5.28e-03, grad_scale: 8.0 2023-04-29 18:40:56,497 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.3702, 2.1203, 2.2269, 4.0507, 2.2507, 2.5811, 2.2452, 2.3607], device='cuda:1'), covar=tensor([0.1034, 0.3393, 0.2326, 0.0421, 0.3176, 0.2262, 0.3098, 0.2600], device='cuda:1'), in_proj_covar=tensor([0.0374, 0.0409, 0.0340, 0.0327, 0.0421, 0.0472, 0.0372, 0.0479], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-29 18:41:24,163 INFO [optim.py:368] (1/8) 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] (1/8) Epoch 13, batch 3800, loss[loss=0.1639, simple_loss=0.2352, pruned_loss=0.04625, over 16785.00 frames. ], tot_loss[loss=0.183, simple_loss=0.2614, pruned_loss=0.05228, over 3284309.05 frames. ], batch size: 90, lr: 5.28e-03, grad_scale: 8.0 2023-04-29 18:42:07,414 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=125603.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 18:43:18,389 INFO [train.py:904] (1/8) Epoch 13, batch 3850, loss[loss=0.1987, simple_loss=0.2749, pruned_loss=0.06123, over 16303.00 frames. ], tot_loss[loss=0.1837, simple_loss=0.2615, pruned_loss=0.05296, over 3277429.06 frames. ], batch size: 165, lr: 5.28e-03, grad_scale: 8.0 2023-04-29 18:43:35,073 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=125664.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 18:43:43,787 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.7773, 4.9310, 5.0913, 5.0279, 5.0012, 5.5405, 5.0926, 4.8055], device='cuda:1'), covar=tensor([0.1250, 0.1837, 0.1734, 0.1829, 0.2331, 0.0888, 0.1351, 0.2434], device='cuda:1'), in_proj_covar=tensor([0.0378, 0.0533, 0.0579, 0.0463, 0.0621, 0.0601, 0.0458, 0.0613], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-04-29 18:43:50,561 INFO [optim.py:368] (1/8) 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 18:44:25,294 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.4230, 4.3773, 4.3607, 3.8024, 4.4085, 1.6030, 4.1306, 3.9999], device='cuda:1'), covar=tensor([0.0111, 0.0084, 0.0162, 0.0302, 0.0087, 0.2678, 0.0133, 0.0221], device='cuda:1'), in_proj_covar=tensor([0.0141, 0.0131, 0.0178, 0.0166, 0.0150, 0.0187, 0.0167, 0.0165], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-29 18:44:29,762 INFO [train.py:904] (1/8) Epoch 13, batch 3900, loss[loss=0.1823, simple_loss=0.2637, pruned_loss=0.05044, over 16679.00 frames. ], tot_loss[loss=0.1838, simple_loss=0.2609, pruned_loss=0.05331, over 3279186.48 frames. ], batch size: 57, lr: 5.27e-03, grad_scale: 8.0 2023-04-29 18:44:45,125 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=125712.0, num_to_drop=1, layers_to_drop={0} 2023-04-29 18:44:50,603 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.9188, 5.2218, 5.0260, 4.9853, 4.6895, 4.6943, 4.6772, 5.2912], device='cuda:1'), covar=tensor([0.1018, 0.0787, 0.0814, 0.0651, 0.0732, 0.0855, 0.0956, 0.0793], device='cuda:1'), in_proj_covar=tensor([0.0583, 0.0737, 0.0593, 0.0523, 0.0465, 0.0470, 0.0614, 0.0565], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:1') 2023-04-29 18:44:53,067 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([1.7568, 2.6793, 2.7078, 1.8463, 2.6847, 2.8087, 2.6694, 1.8003], device='cuda:1'), covar=tensor([0.0446, 0.0077, 0.0055, 0.0358, 0.0087, 0.0094, 0.0085, 0.0370], device='cuda:1'), in_proj_covar=tensor([0.0131, 0.0072, 0.0072, 0.0127, 0.0083, 0.0092, 0.0083, 0.0122], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:1') 2023-04-29 18:45:03,284 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=125725.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 18:45:14,704 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 13, batch 3950, loss[loss=0.1871, simple_loss=0.2564, pruned_loss=0.05895, over 16759.00 frames. ], tot_loss[loss=0.1846, simple_loss=0.2611, pruned_loss=0.054, over 3273029.55 frames. ], batch size: 124, lr: 5.27e-03, grad_scale: 8.0 2023-04-29 18:46:16,760 INFO [optim.py:368] (1/8) 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 18:46:20,132 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=125777.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 18:46:28,648 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=5.09 vs. limit=5.0 2023-04-29 18:46:29,422 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=125784.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 18:46:32,665 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=125786.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 18:46:41,500 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.3946, 4.1258, 4.3290, 2.7200, 3.8073, 4.2464, 3.8463, 2.4661], device='cuda:1'), covar=tensor([0.0468, 0.0075, 0.0028, 0.0317, 0.0062, 0.0072, 0.0059, 0.0348], device='cuda:1'), in_proj_covar=tensor([0.0132, 0.0073, 0.0072, 0.0128, 0.0084, 0.0093, 0.0084, 0.0123], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:1') 2023-04-29 18:46:55,491 INFO [train.py:904] (1/8) Epoch 13, batch 4000, loss[loss=0.1681, simple_loss=0.2496, pruned_loss=0.0433, over 17000.00 frames. ], tot_loss[loss=0.1838, simple_loss=0.2603, pruned_loss=0.05361, over 3286558.07 frames. ], batch size: 50, lr: 5.27e-03, grad_scale: 8.0 2023-04-29 18:47:33,748 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.2072, 4.3168, 2.8066, 4.9883, 3.4672, 5.0271, 3.0056, 3.3902], device='cuda:1'), covar=tensor([0.0159, 0.0247, 0.1308, 0.0058, 0.0612, 0.0157, 0.1157, 0.0570], device='cuda:1'), in_proj_covar=tensor([0.0157, 0.0169, 0.0189, 0.0144, 0.0169, 0.0213, 0.0198, 0.0173], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-04-29 18:47:37,206 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 13, batch 4050, loss[loss=0.1625, simple_loss=0.2438, pruned_loss=0.04056, over 17253.00 frames. ], tot_loss[loss=0.1821, simple_loss=0.26, pruned_loss=0.05205, over 3283782.30 frames. ], batch size: 52, lr: 5.27e-03, grad_scale: 8.0 2023-04-29 18:48:36,494 INFO [optim.py:368] (1/8) 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 18:48:45,120 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.8699, 3.2440, 3.2331, 1.9125, 2.7297, 2.1830, 3.4196, 3.3664], device='cuda:1'), covar=tensor([0.0234, 0.0640, 0.0631, 0.1902, 0.0856, 0.0969, 0.0551, 0.0777], device='cuda:1'), in_proj_covar=tensor([0.0147, 0.0152, 0.0160, 0.0146, 0.0138, 0.0126, 0.0137, 0.0162], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:1') 2023-04-29 18:49:15,115 INFO [train.py:904] (1/8) Epoch 13, batch 4100, loss[loss=0.2068, simple_loss=0.285, pruned_loss=0.06432, over 12189.00 frames. ], tot_loss[loss=0.1825, simple_loss=0.2619, pruned_loss=0.05156, over 3273954.45 frames. ], batch size: 246, lr: 5.27e-03, grad_scale: 8.0 2023-04-29 18:49:16,151 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=4.82 vs. limit=5.0 2023-04-29 18:50:13,228 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.8499, 4.0633, 2.7952, 2.4856, 2.8450, 2.3560, 4.2139, 3.6775], device='cuda:1'), covar=tensor([0.2535, 0.0739, 0.1801, 0.2099, 0.2284, 0.1965, 0.0548, 0.0934], device='cuda:1'), in_proj_covar=tensor([0.0304, 0.0260, 0.0290, 0.0287, 0.0288, 0.0229, 0.0275, 0.0307], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-29 18:50:30,153 INFO [train.py:904] (1/8) Epoch 13, batch 4150, loss[loss=0.2223, simple_loss=0.306, pruned_loss=0.06926, over 16926.00 frames. ], tot_loss[loss=0.1899, simple_loss=0.2699, pruned_loss=0.05489, over 3226231.47 frames. ], batch size: 109, lr: 5.27e-03, grad_scale: 8.0 2023-04-29 18:50:36,428 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.0777, 1.4345, 1.8037, 2.0591, 2.1956, 2.2834, 1.6291, 2.2562], device='cuda:1'), covar=tensor([0.0186, 0.0411, 0.0220, 0.0259, 0.0217, 0.0148, 0.0368, 0.0080], device='cuda:1'), in_proj_covar=tensor([0.0169, 0.0178, 0.0163, 0.0168, 0.0178, 0.0132, 0.0178, 0.0123], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-29 18:50:42,648 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=125959.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 18:50:43,024 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.37 vs. limit=2.0 2023-04-29 18:50:53,863 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.8445, 2.2505, 1.7110, 2.0116, 2.5679, 2.2591, 2.6517, 2.8235], device='cuda:1'), covar=tensor([0.0121, 0.0335, 0.0469, 0.0417, 0.0199, 0.0331, 0.0178, 0.0213], device='cuda:1'), in_proj_covar=tensor([0.0171, 0.0214, 0.0207, 0.0208, 0.0214, 0.0212, 0.0223, 0.0205], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-29 18:51:05,850 INFO [optim.py:368] (1/8) 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 18:51:08,695 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([5.4859, 5.3641, 5.3854, 5.0592, 5.0391, 5.3048, 5.3600, 5.0602], device='cuda:1'), covar=tensor([0.0519, 0.0313, 0.0205, 0.0224, 0.0805, 0.0329, 0.0214, 0.0560], device='cuda:1'), in_proj_covar=tensor([0.0261, 0.0351, 0.0310, 0.0290, 0.0332, 0.0337, 0.0209, 0.0362], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002], device='cuda:1') 2023-04-29 18:51:30,167 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.2014, 1.4975, 1.8028, 2.1759, 2.2809, 2.4753, 1.6638, 2.3754], device='cuda:1'), covar=tensor([0.0191, 0.0399, 0.0246, 0.0239, 0.0228, 0.0141, 0.0374, 0.0083], device='cuda:1'), in_proj_covar=tensor([0.0169, 0.0177, 0.0162, 0.0167, 0.0177, 0.0132, 0.0177, 0.0123], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-29 18:51:34,127 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.3775, 2.1266, 1.7233, 1.8217, 2.3516, 2.0175, 2.3009, 2.5029], device='cuda:1'), covar=tensor([0.0138, 0.0268, 0.0387, 0.0383, 0.0181, 0.0306, 0.0161, 0.0197], device='cuda:1'), in_proj_covar=tensor([0.0172, 0.0215, 0.0208, 0.0209, 0.0215, 0.0214, 0.0224, 0.0206], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-29 18:51:40,604 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=125998.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 18:51:48,696 INFO [train.py:904] (1/8) Epoch 13, batch 4200, loss[loss=0.1961, simple_loss=0.297, pruned_loss=0.04758, over 16790.00 frames. ], tot_loss[loss=0.1957, simple_loss=0.2776, pruned_loss=0.05688, over 3204703.75 frames. ], batch size: 102, lr: 5.27e-03, grad_scale: 8.0 2023-04-29 18:52:04,559 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=126012.0, num_to_drop=1, layers_to_drop={1} 2023-04-29 18:52:34,265 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=126032.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 18:53:02,528 INFO [train.py:904] (1/8) Epoch 13, batch 4250, loss[loss=0.1861, simple_loss=0.2733, pruned_loss=0.04944, over 16239.00 frames. ], tot_loss[loss=0.1974, simple_loss=0.2808, pruned_loss=0.057, over 3177082.54 frames. ], batch size: 165, lr: 5.27e-03, grad_scale: 8.0 2023-04-29 18:53:13,100 INFO [zipformer.py:625] (1/8) 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] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=126060.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 18:53:26,389 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.4365, 3.3874, 2.4764, 2.0885, 2.3297, 2.0342, 3.4577, 3.1481], device='cuda:1'), covar=tensor([0.2825, 0.0931, 0.1926, 0.2470, 0.2526, 0.2208, 0.0653, 0.1176], device='cuda:1'), in_proj_covar=tensor([0.0299, 0.0256, 0.0285, 0.0282, 0.0282, 0.0225, 0.0270, 0.0300], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-29 18:53:36,045 INFO [optim.py:368] (1/8) 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 18:53:36,627 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([1.6020, 2.2954, 2.1772, 3.8047, 2.4895, 3.8277, 1.4000, 2.7331], device='cuda:1'), covar=tensor([0.1340, 0.0906, 0.1429, 0.0177, 0.0250, 0.0400, 0.1635, 0.0908], device='cuda:1'), in_proj_covar=tensor([0.0153, 0.0160, 0.0182, 0.0156, 0.0196, 0.0206, 0.0181, 0.0181], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0003, 0.0004], device='cuda:1') 2023-04-29 18:53:39,685 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=126077.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 18:53:43,862 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=126080.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 18:53:45,099 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=126081.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 18:54:06,591 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.42 vs. limit=2.0 2023-04-29 18:54:16,795 INFO [train.py:904] (1/8) Epoch 13, batch 4300, loss[loss=0.2034, simple_loss=0.2982, pruned_loss=0.0543, over 16509.00 frames. ], tot_loss[loss=0.1969, simple_loss=0.2822, pruned_loss=0.05581, over 3190865.42 frames. ], batch size: 75, lr: 5.27e-03, grad_scale: 8.0 2023-04-29 18:54:51,666 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=126125.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 18:55:07,299 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.62 vs. limit=2.0 2023-04-29 18:55:30,684 INFO [train.py:904] (1/8) Epoch 13, batch 4350, loss[loss=0.209, simple_loss=0.2835, pruned_loss=0.06723, over 11576.00 frames. ], tot_loss[loss=0.1994, simple_loss=0.2852, pruned_loss=0.05677, over 3188151.99 frames. ], batch size: 248, lr: 5.27e-03, grad_scale: 8.0 2023-04-29 18:56:06,024 INFO [optim.py:368] (1/8) 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 18:56:28,999 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.40 vs. limit=2.0 2023-04-29 18:56:46,664 INFO [train.py:904] (1/8) Epoch 13, batch 4400, loss[loss=0.1937, simple_loss=0.2803, pruned_loss=0.05355, over 16576.00 frames. ], tot_loss[loss=0.2016, simple_loss=0.2873, pruned_loss=0.05792, over 3183691.69 frames. ], batch size: 62, lr: 5.26e-03, grad_scale: 8.0 2023-04-29 18:57:59,272 INFO [train.py:904] (1/8) Epoch 13, batch 4450, loss[loss=0.2049, simple_loss=0.2921, pruned_loss=0.05884, over 16987.00 frames. ], tot_loss[loss=0.2039, simple_loss=0.2905, pruned_loss=0.05861, over 3211638.44 frames. ], batch size: 41, lr: 5.26e-03, grad_scale: 8.0 2023-04-29 18:58:10,094 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=126259.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 18:58:20,791 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([5.2843, 5.2659, 5.1315, 4.8220, 4.8056, 5.1901, 5.0423, 4.8119], device='cuda:1'), covar=tensor([0.0444, 0.0259, 0.0171, 0.0198, 0.0771, 0.0235, 0.0259, 0.0513], device='cuda:1'), in_proj_covar=tensor([0.0255, 0.0343, 0.0302, 0.0283, 0.0325, 0.0325, 0.0207, 0.0353], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-29 18:58:33,037 INFO [optim.py:368] (1/8) 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] (1/8) Epoch 13, batch 4500, loss[loss=0.1969, simple_loss=0.2895, pruned_loss=0.0522, over 16812.00 frames. ], tot_loss[loss=0.2043, simple_loss=0.2906, pruned_loss=0.05905, over 3203479.61 frames. ], batch size: 39, lr: 5.26e-03, grad_scale: 8.0 2023-04-29 18:59:21,401 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 13, batch 4550, loss[loss=0.2098, simple_loss=0.3003, pruned_loss=0.05965, over 16771.00 frames. ], tot_loss[loss=0.2058, simple_loss=0.2917, pruned_loss=0.05991, over 3228749.08 frames. ], batch size: 83, lr: 5.26e-03, grad_scale: 8.0 2023-04-29 19:00:28,790 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=126354.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 19:00:59,197 INFO [optim.py:368] (1/8) 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 19:01:06,170 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=126380.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 19:01:07,334 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=126381.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 19:01:13,727 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.7310, 3.6900, 2.0653, 4.4087, 2.7158, 4.2642, 2.4497, 2.8373], device='cuda:1'), covar=tensor([0.0196, 0.0304, 0.1645, 0.0100, 0.0812, 0.0399, 0.1288, 0.0722], device='cuda:1'), in_proj_covar=tensor([0.0153, 0.0165, 0.0188, 0.0138, 0.0166, 0.0209, 0.0195, 0.0171], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0003, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-04-29 19:01:35,453 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([5.3110, 5.3065, 5.1847, 4.7922, 4.7574, 5.2178, 5.0755, 4.8445], device='cuda:1'), covar=tensor([0.0434, 0.0232, 0.0171, 0.0207, 0.0879, 0.0225, 0.0231, 0.0503], device='cuda:1'), in_proj_covar=tensor([0.0253, 0.0340, 0.0299, 0.0281, 0.0324, 0.0324, 0.0206, 0.0351], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-29 19:01:36,158 INFO [train.py:904] (1/8) Epoch 13, batch 4600, loss[loss=0.211, simple_loss=0.2959, pruned_loss=0.06307, over 17036.00 frames. ], tot_loss[loss=0.2071, simple_loss=0.2926, pruned_loss=0.06073, over 3226161.86 frames. ], batch size: 50, lr: 5.26e-03, grad_scale: 8.0 2023-04-29 19:02:00,558 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([1.5110, 2.4158, 2.4058, 3.2569, 2.5016, 3.5297, 1.5295, 2.6161], device='cuda:1'), covar=tensor([0.1525, 0.0750, 0.1166, 0.0157, 0.0223, 0.0353, 0.1703, 0.0820], device='cuda:1'), in_proj_covar=tensor([0.0154, 0.0161, 0.0183, 0.0157, 0.0198, 0.0206, 0.0182, 0.0181], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0003, 0.0004], device='cuda:1') 2023-04-29 19:02:22,725 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=126429.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 19:02:23,064 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.5969, 2.6238, 2.4877, 4.7427, 2.4445, 3.0702, 2.6285, 2.6801], device='cuda:1'), covar=tensor([0.0925, 0.2581, 0.2061, 0.0284, 0.3270, 0.1767, 0.2497, 0.2701], device='cuda:1'), in_proj_covar=tensor([0.0372, 0.0405, 0.0336, 0.0319, 0.0419, 0.0470, 0.0369, 0.0473], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-29 19:02:26,954 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=4.80 vs. limit=5.0 2023-04-29 19:02:38,623 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=126441.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 19:02:53,208 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=126450.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 19:02:55,164 INFO [train.py:904] (1/8) Epoch 13, batch 4650, loss[loss=0.2324, simple_loss=0.3071, pruned_loss=0.07881, over 11937.00 frames. ], tot_loss[loss=0.207, simple_loss=0.2919, pruned_loss=0.06101, over 3203542.99 frames. ], batch size: 248, lr: 5.26e-03, grad_scale: 8.0 2023-04-29 19:03:27,524 INFO [optim.py:368] (1/8) 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 19:03:35,801 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.18 vs. limit=2.0 2023-04-29 19:04:07,652 INFO [train.py:904] (1/8) Epoch 13, batch 4700, loss[loss=0.1728, simple_loss=0.2634, pruned_loss=0.04108, over 16920.00 frames. ], tot_loss[loss=0.2042, simple_loss=0.2891, pruned_loss=0.05966, over 3209988.69 frames. ], batch size: 90, lr: 5.26e-03, grad_scale: 8.0 2023-04-29 19:04:21,546 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=126511.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 19:05:13,240 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=4.68 vs. limit=5.0 2023-04-29 19:05:21,228 INFO [train.py:904] (1/8) Epoch 13, batch 4750, loss[loss=0.2319, simple_loss=0.3032, pruned_loss=0.08026, over 11678.00 frames. ], tot_loss[loss=0.1998, simple_loss=0.2848, pruned_loss=0.05735, over 3220947.15 frames. ], batch size: 248, lr: 5.26e-03, grad_scale: 8.0 2023-04-29 19:05:53,853 INFO [optim.py:368] (1/8) 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] (1/8) Epoch 13, batch 4800, loss[loss=0.1758, simple_loss=0.2702, pruned_loss=0.04069, over 15384.00 frames. ], tot_loss[loss=0.1964, simple_loss=0.2814, pruned_loss=0.05572, over 3205324.20 frames. ], batch size: 190, lr: 5.26e-03, grad_scale: 8.0 2023-04-29 19:07:26,280 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.79 vs. limit=2.0 2023-04-29 19:07:43,510 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([1.8743, 3.0223, 2.6283, 5.0676, 3.9260, 4.3947, 1.7077, 3.1549], device='cuda:1'), covar=tensor([0.1230, 0.0696, 0.1190, 0.0104, 0.0316, 0.0333, 0.1446, 0.0764], device='cuda:1'), in_proj_covar=tensor([0.0156, 0.0163, 0.0185, 0.0159, 0.0201, 0.0208, 0.0184, 0.0184], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-04-29 19:07:46,139 INFO [train.py:904] (1/8) Epoch 13, batch 4850, loss[loss=0.1867, simple_loss=0.2814, pruned_loss=0.04603, over 15446.00 frames. ], tot_loss[loss=0.1965, simple_loss=0.2823, pruned_loss=0.05534, over 3179160.64 frames. ], batch size: 191, lr: 5.25e-03, grad_scale: 8.0 2023-04-29 19:07:50,129 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=126654.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 19:08:19,806 INFO [optim.py:368] (1/8) 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 19:08:59,235 INFO [train.py:904] (1/8) Epoch 13, batch 4900, loss[loss=0.1896, simple_loss=0.2903, pruned_loss=0.04451, over 15290.00 frames. ], tot_loss[loss=0.195, simple_loss=0.2817, pruned_loss=0.05415, over 3147246.91 frames. ], batch size: 190, lr: 5.25e-03, grad_scale: 8.0 2023-04-29 19:08:59,539 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=126702.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 19:09:49,771 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=126736.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 19:10:12,984 INFO [train.py:904] (1/8) Epoch 13, batch 4950, loss[loss=0.2015, simple_loss=0.2952, pruned_loss=0.05392, over 15316.00 frames. ], tot_loss[loss=0.1938, simple_loss=0.2811, pruned_loss=0.05319, over 3157805.26 frames. ], batch size: 190, lr: 5.25e-03, grad_scale: 8.0 2023-04-29 19:10:39,198 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.5343, 3.4865, 3.4577, 2.8438, 3.3704, 1.9487, 3.1228, 2.7353], device='cuda:1'), covar=tensor([0.0118, 0.0113, 0.0135, 0.0267, 0.0087, 0.2230, 0.0120, 0.0215], device='cuda:1'), in_proj_covar=tensor([0.0134, 0.0124, 0.0168, 0.0159, 0.0141, 0.0180, 0.0158, 0.0155], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-29 19:10:45,731 INFO [optim.py:368] (1/8) 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] (1/8) Epoch 13, batch 5000, loss[loss=0.2028, simple_loss=0.2975, pruned_loss=0.0541, over 16283.00 frames. ], tot_loss[loss=0.1945, simple_loss=0.2826, pruned_loss=0.05315, over 3173726.62 frames. ], batch size: 165, lr: 5.25e-03, grad_scale: 8.0 2023-04-29 19:11:32,726 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=126806.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 19:11:47,335 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 13, batch 5050, loss[loss=0.1839, simple_loss=0.2822, pruned_loss=0.04285, over 16879.00 frames. ], tot_loss[loss=0.1943, simple_loss=0.283, pruned_loss=0.05274, over 3195434.23 frames. ], batch size: 96, lr: 5.25e-03, grad_scale: 8.0 2023-04-29 19:13:11,123 INFO [optim.py:368] (1/8) 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 19:13:14,593 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 13, batch 5100, loss[loss=0.1896, simple_loss=0.2754, pruned_loss=0.05192, over 15410.00 frames. ], tot_loss[loss=0.1911, simple_loss=0.2798, pruned_loss=0.05123, over 3208739.79 frames. ], batch size: 190, lr: 5.25e-03, grad_scale: 8.0 2023-04-29 19:14:02,103 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=4.95 vs. limit=5.0 2023-04-29 19:15:04,900 INFO [train.py:904] (1/8) Epoch 13, batch 5150, loss[loss=0.2185, simple_loss=0.3, pruned_loss=0.06846, over 12267.00 frames. ], tot_loss[loss=0.1904, simple_loss=0.2798, pruned_loss=0.05054, over 3192534.34 frames. ], batch size: 247, lr: 5.25e-03, grad_scale: 8.0 2023-04-29 19:15:14,474 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([1.9989, 3.5265, 3.6145, 2.3124, 3.2360, 3.5592, 3.3250, 2.1030], device='cuda:1'), covar=tensor([0.0479, 0.0036, 0.0035, 0.0338, 0.0081, 0.0096, 0.0069, 0.0377], device='cuda:1'), in_proj_covar=tensor([0.0132, 0.0071, 0.0072, 0.0128, 0.0085, 0.0093, 0.0083, 0.0122], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:1') 2023-04-29 19:15:37,786 INFO [optim.py:368] (1/8) 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 19:16:05,385 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([1.7650, 1.3751, 1.5722, 1.7252, 1.8563, 1.8925, 1.5304, 1.8888], device='cuda:1'), covar=tensor([0.0183, 0.0333, 0.0172, 0.0218, 0.0216, 0.0150, 0.0360, 0.0088], device='cuda:1'), in_proj_covar=tensor([0.0166, 0.0177, 0.0162, 0.0165, 0.0176, 0.0131, 0.0177, 0.0123], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-29 19:16:17,737 INFO [train.py:904] (1/8) Epoch 13, batch 5200, loss[loss=0.1686, simple_loss=0.2584, pruned_loss=0.03937, over 16525.00 frames. ], tot_loss[loss=0.189, simple_loss=0.2783, pruned_loss=0.04984, over 3197657.95 frames. ], batch size: 75, lr: 5.25e-03, grad_scale: 8.0 2023-04-29 19:16:25,737 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.6033, 3.7796, 4.3765, 2.1202, 4.4319, 4.4567, 3.0360, 3.2586], device='cuda:1'), covar=tensor([0.0753, 0.0191, 0.0080, 0.1064, 0.0039, 0.0066, 0.0336, 0.0390], device='cuda:1'), in_proj_covar=tensor([0.0144, 0.0103, 0.0088, 0.0137, 0.0070, 0.0108, 0.0122, 0.0127], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:1') 2023-04-29 19:16:58,471 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([5.0458, 5.3408, 5.0484, 5.0973, 4.7501, 4.7287, 4.7316, 5.3900], device='cuda:1'), covar=tensor([0.1142, 0.0832, 0.1024, 0.0730, 0.0892, 0.0860, 0.1015, 0.0932], device='cuda:1'), in_proj_covar=tensor([0.0562, 0.0705, 0.0571, 0.0498, 0.0447, 0.0451, 0.0586, 0.0543], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:1') 2023-04-29 19:17:08,359 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 13, batch 5250, loss[loss=0.1868, simple_loss=0.2794, pruned_loss=0.04709, over 15453.00 frames. ], tot_loss[loss=0.1874, simple_loss=0.2759, pruned_loss=0.04946, over 3211794.27 frames. ], batch size: 190, lr: 5.25e-03, grad_scale: 8.0 2023-04-29 19:18:04,188 INFO [optim.py:368] (1/8) 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 19:18:18,850 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 13, batch 5300, loss[loss=0.2093, simple_loss=0.2838, pruned_loss=0.06738, over 12379.00 frames. ], tot_loss[loss=0.1854, simple_loss=0.2732, pruned_loss=0.04876, over 3199811.17 frames. ], batch size: 248, lr: 5.25e-03, grad_scale: 8.0 2023-04-29 19:18:50,618 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=127106.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 19:18:56,543 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.17 vs. limit=2.0 2023-04-29 19:19:50,715 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.1825, 4.1477, 4.0679, 3.3890, 4.0930, 1.6936, 3.8452, 3.6299], device='cuda:1'), covar=tensor([0.0090, 0.0081, 0.0124, 0.0344, 0.0084, 0.2455, 0.0123, 0.0218], device='cuda:1'), in_proj_covar=tensor([0.0136, 0.0125, 0.0170, 0.0162, 0.0143, 0.0183, 0.0160, 0.0158], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-29 19:19:58,315 INFO [train.py:904] (1/8) Epoch 13, batch 5350, loss[loss=0.1788, simple_loss=0.2738, pruned_loss=0.0419, over 16893.00 frames. ], tot_loss[loss=0.1839, simple_loss=0.2716, pruned_loss=0.04813, over 3199645.24 frames. ], batch size: 116, lr: 5.24e-03, grad_scale: 4.0 2023-04-29 19:20:01,842 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=127154.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 19:20:27,731 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=127172.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 19:20:32,660 INFO [optim.py:368] (1/8) 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] (1/8) Epoch 13, batch 5400, loss[loss=0.1854, simple_loss=0.2807, pruned_loss=0.04503, over 16209.00 frames. ], tot_loss[loss=0.1858, simple_loss=0.2743, pruned_loss=0.04866, over 3210507.64 frames. ], batch size: 165, lr: 5.24e-03, grad_scale: 4.0 2023-04-29 19:22:20,861 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=127249.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 19:22:26,303 INFO [train.py:904] (1/8) Epoch 13, batch 5450, loss[loss=0.2378, simple_loss=0.3209, pruned_loss=0.07738, over 16355.00 frames. ], tot_loss[loss=0.189, simple_loss=0.2773, pruned_loss=0.05036, over 3207761.25 frames. ], batch size: 146, lr: 5.24e-03, grad_scale: 4.0 2023-04-29 19:23:02,325 INFO [optim.py:368] (1/8) 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 19:23:28,915 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=4.82 vs. limit=5.0 2023-04-29 19:23:43,865 INFO [train.py:904] (1/8) Epoch 13, batch 5500, loss[loss=0.2662, simple_loss=0.3337, pruned_loss=0.09931, over 15194.00 frames. ], tot_loss[loss=0.1976, simple_loss=0.285, pruned_loss=0.05511, over 3191257.11 frames. ], batch size: 190, lr: 5.24e-03, grad_scale: 4.0 2023-04-29 19:23:57,064 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=127310.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 19:24:43,251 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.8361, 3.9923, 1.9882, 4.6079, 2.9079, 4.4390, 2.4296, 2.8318], device='cuda:1'), covar=tensor([0.0218, 0.0299, 0.1827, 0.0140, 0.0796, 0.0414, 0.1475, 0.0828], device='cuda:1'), in_proj_covar=tensor([0.0155, 0.0166, 0.0190, 0.0137, 0.0168, 0.0207, 0.0195, 0.0172], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0003, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-04-29 19:24:46,018 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.0062, 3.9691, 3.9230, 3.2596, 3.9646, 1.7342, 3.7610, 3.5458], device='cuda:1'), covar=tensor([0.0109, 0.0095, 0.0159, 0.0302, 0.0090, 0.2362, 0.0133, 0.0225], device='cuda:1'), in_proj_covar=tensor([0.0138, 0.0126, 0.0172, 0.0163, 0.0144, 0.0185, 0.0161, 0.0158], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-29 19:25:00,716 INFO [train.py:904] (1/8) Epoch 13, batch 5550, loss[loss=0.2056, simple_loss=0.2937, pruned_loss=0.05876, over 16504.00 frames. ], tot_loss[loss=0.2073, simple_loss=0.2926, pruned_loss=0.06106, over 3153500.75 frames. ], batch size: 68, lr: 5.24e-03, grad_scale: 4.0 2023-04-29 19:25:38,574 INFO [optim.py:368] (1/8) 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 19:26:09,115 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.1949, 3.3929, 3.5686, 3.5325, 3.5524, 3.3880, 3.4074, 3.4460], device='cuda:1'), covar=tensor([0.0434, 0.0722, 0.0457, 0.0501, 0.0499, 0.0547, 0.0824, 0.0522], device='cuda:1'), in_proj_covar=tensor([0.0345, 0.0360, 0.0359, 0.0349, 0.0410, 0.0384, 0.0480, 0.0308], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:1') 2023-04-29 19:26:20,876 INFO [train.py:904] (1/8) Epoch 13, batch 5600, loss[loss=0.2401, simple_loss=0.3179, pruned_loss=0.0811, over 16179.00 frames. ], tot_loss[loss=0.2152, simple_loss=0.2982, pruned_loss=0.06614, over 3111300.77 frames. ], batch size: 165, lr: 5.24e-03, grad_scale: 8.0 2023-04-29 19:27:02,215 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=127427.0, num_to_drop=1, layers_to_drop={1} 2023-04-29 19:27:41,567 INFO [train.py:904] (1/8) Epoch 13, batch 5650, loss[loss=0.2377, simple_loss=0.3163, pruned_loss=0.07958, over 17039.00 frames. ], tot_loss[loss=0.2227, simple_loss=0.3037, pruned_loss=0.0708, over 3076790.09 frames. ], batch size: 53, lr: 5.24e-03, grad_scale: 8.0 2023-04-29 19:28:12,104 INFO [zipformer.py:625] (1/8) 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] (1/8) 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 19:28:36,976 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=127488.0, num_to_drop=1, layers_to_drop={0} 2023-04-29 19:28:58,352 INFO [train.py:904] (1/8) Epoch 13, batch 5700, loss[loss=0.2074, simple_loss=0.311, pruned_loss=0.05196, over 16758.00 frames. ], tot_loss[loss=0.2236, simple_loss=0.3042, pruned_loss=0.07147, over 3069992.78 frames. ], batch size: 83, lr: 5.24e-03, grad_scale: 8.0 2023-04-29 19:29:14,302 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=2.76 vs. limit=5.0 2023-04-29 19:29:27,679 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=127520.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 19:29:42,024 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=127529.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 19:30:18,041 INFO [train.py:904] (1/8) Epoch 13, batch 5750, loss[loss=0.2105, simple_loss=0.3024, pruned_loss=0.05933, over 16482.00 frames. ], tot_loss[loss=0.2264, simple_loss=0.3071, pruned_loss=0.0728, over 3072079.18 frames. ], batch size: 75, lr: 5.24e-03, grad_scale: 8.0 2023-04-29 19:30:20,776 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.62 vs. limit=2.0 2023-04-29 19:30:21,794 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.48 vs. limit=2.0 2023-04-29 19:30:56,349 INFO [optim.py:368] (1/8) 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 19:31:21,790 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=127590.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 19:31:39,684 INFO [train.py:904] (1/8) Epoch 13, batch 5800, loss[loss=0.2041, simple_loss=0.2931, pruned_loss=0.05755, over 15492.00 frames. ], tot_loss[loss=0.2266, simple_loss=0.3078, pruned_loss=0.07273, over 3035687.99 frames. ], batch size: 192, lr: 5.24e-03, grad_scale: 4.0 2023-04-29 19:31:44,375 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=127605.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 19:32:21,167 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.4526, 1.6240, 2.0489, 2.3204, 2.4401, 2.7175, 1.7464, 2.6391], device='cuda:1'), covar=tensor([0.0157, 0.0423, 0.0257, 0.0252, 0.0237, 0.0146, 0.0399, 0.0100], device='cuda:1'), in_proj_covar=tensor([0.0164, 0.0176, 0.0160, 0.0163, 0.0174, 0.0130, 0.0175, 0.0121], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001], device='cuda:1') 2023-04-29 19:32:46,332 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.5473, 2.5757, 1.6739, 2.7200, 2.1129, 2.7218, 1.8486, 2.2314], device='cuda:1'), covar=tensor([0.0295, 0.0378, 0.1582, 0.0253, 0.0642, 0.0513, 0.1465, 0.0655], device='cuda:1'), in_proj_covar=tensor([0.0155, 0.0165, 0.0188, 0.0136, 0.0167, 0.0205, 0.0194, 0.0172], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0003, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-04-29 19:32:57,618 INFO [train.py:904] (1/8) Epoch 13, batch 5850, loss[loss=0.1867, simple_loss=0.2736, pruned_loss=0.04993, over 16616.00 frames. ], tot_loss[loss=0.223, simple_loss=0.3051, pruned_loss=0.07045, over 3052328.05 frames. ], batch size: 57, lr: 5.23e-03, grad_scale: 4.0 2023-04-29 19:33:15,127 INFO [zipformer.py:625] (1/8) 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] (1/8) 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] (1/8) Epoch 13, batch 5900, loss[loss=0.2113, simple_loss=0.3062, pruned_loss=0.05822, over 17000.00 frames. ], tot_loss[loss=0.2225, simple_loss=0.3044, pruned_loss=0.07027, over 3058707.17 frames. ], batch size: 41, lr: 5.23e-03, grad_scale: 4.0 2023-04-29 19:34:57,921 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=127723.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 19:35:30,418 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.9610, 2.8145, 2.7748, 2.0421, 2.6110, 2.0941, 2.7861, 2.9353], device='cuda:1'), covar=tensor([0.0253, 0.0634, 0.0501, 0.1716, 0.0776, 0.0904, 0.0572, 0.0682], device='cuda:1'), in_proj_covar=tensor([0.0146, 0.0148, 0.0159, 0.0145, 0.0139, 0.0126, 0.0139, 0.0158], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:1') 2023-04-29 19:35:42,724 INFO [train.py:904] (1/8) Epoch 13, batch 5950, loss[loss=0.225, simple_loss=0.3069, pruned_loss=0.07158, over 16912.00 frames. ], tot_loss[loss=0.2225, simple_loss=0.3058, pruned_loss=0.0696, over 3056029.74 frames. ], batch size: 109, lr: 5.23e-03, grad_scale: 4.0 2023-04-29 19:36:21,828 INFO [optim.py:368] (1/8) 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 19:36:26,841 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=127779.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 19:36:30,583 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.2924, 2.5680, 1.9343, 2.2218, 2.9219, 2.4843, 3.0318, 3.0726], device='cuda:1'), covar=tensor([0.0082, 0.0305, 0.0441, 0.0370, 0.0191, 0.0317, 0.0192, 0.0191], device='cuda:1'), in_proj_covar=tensor([0.0162, 0.0208, 0.0202, 0.0204, 0.0207, 0.0207, 0.0213, 0.0199], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-29 19:36:32,792 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=127783.0, num_to_drop=1, layers_to_drop={2} 2023-04-29 19:36:39,358 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.4865, 2.2603, 2.3131, 4.1867, 2.1472, 2.6287, 2.3972, 2.4404], device='cuda:1'), covar=tensor([0.0994, 0.3316, 0.2376, 0.0453, 0.3839, 0.2244, 0.3029, 0.3182], device='cuda:1'), in_proj_covar=tensor([0.0365, 0.0400, 0.0332, 0.0316, 0.0413, 0.0461, 0.0366, 0.0466], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-29 19:36:55,911 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.42 vs. limit=2.0 2023-04-29 19:37:04,075 INFO [train.py:904] (1/8) Epoch 13, batch 6000, loss[loss=0.2242, simple_loss=0.293, pruned_loss=0.07775, over 11269.00 frames. ], tot_loss[loss=0.2218, simple_loss=0.3049, pruned_loss=0.06933, over 3050542.72 frames. ], batch size: 246, lr: 5.23e-03, grad_scale: 8.0 2023-04-29 19:37:04,075 INFO [train.py:929] (1/8) Computing validation loss 2023-04-29 19:37:14,216 INFO [train.py:938] (1/8) Epoch 13, validation: loss=0.1599, simple_loss=0.2726, pruned_loss=0.02359, over 944034.00 frames. 2023-04-29 19:37:14,217 INFO [train.py:939] (1/8) Maximum memory allocated so far is 17845MB 2023-04-29 19:37:23,862 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.0679, 3.8999, 4.0853, 4.2696, 4.3885, 3.9647, 4.3254, 4.3849], device='cuda:1'), covar=tensor([0.1566, 0.1131, 0.1531, 0.0646, 0.0549, 0.1260, 0.0633, 0.0637], device='cuda:1'), in_proj_covar=tensor([0.0539, 0.0673, 0.0803, 0.0682, 0.0517, 0.0531, 0.0545, 0.0623], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-29 19:38:13,316 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=127840.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 19:38:32,207 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 13, batch 6050, loss[loss=0.2086, simple_loss=0.2962, pruned_loss=0.06048, over 16859.00 frames. ], tot_loss[loss=0.2203, simple_loss=0.3035, pruned_loss=0.06851, over 3058220.65 frames. ], batch size: 116, lr: 5.23e-03, grad_scale: 8.0 2023-04-29 19:38:44,379 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=4.81 vs. limit=5.0 2023-04-29 19:39:07,466 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=127872.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 19:39:14,939 INFO [optim.py:368] (1/8) 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] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=127885.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 19:39:55,361 INFO [train.py:904] (1/8) Epoch 13, batch 6100, loss[loss=0.1963, simple_loss=0.2825, pruned_loss=0.05501, over 16873.00 frames. ], tot_loss[loss=0.2188, simple_loss=0.3029, pruned_loss=0.06735, over 3072779.68 frames. ], batch size: 116, lr: 5.23e-03, grad_scale: 8.0 2023-04-29 19:40:01,518 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=127905.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 19:40:12,615 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=127911.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 19:40:23,012 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.84 vs. limit=2.0 2023-04-29 19:40:47,077 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 13, batch 6150, loss[loss=0.1958, simple_loss=0.2805, pruned_loss=0.0556, over 16787.00 frames. ], tot_loss[loss=0.2168, simple_loss=0.3007, pruned_loss=0.06641, over 3073555.76 frames. ], batch size: 124, lr: 5.23e-03, grad_scale: 8.0 2023-04-29 19:41:17,425 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=127953.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 19:41:56,817 INFO [optim.py:368] (1/8) 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 19:42:18,103 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([1.9784, 2.5381, 2.6458, 1.8068, 2.7552, 2.8348, 2.4131, 2.3444], device='cuda:1'), covar=tensor([0.0779, 0.0219, 0.0188, 0.1005, 0.0087, 0.0195, 0.0442, 0.0439], device='cuda:1'), in_proj_covar=tensor([0.0144, 0.0102, 0.0088, 0.0138, 0.0069, 0.0109, 0.0121, 0.0126], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:1') 2023-04-29 19:42:19,357 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.7569, 3.8143, 3.9319, 3.7305, 3.8504, 4.2445, 3.9245, 3.6478], device='cuda:1'), covar=tensor([0.2139, 0.2169, 0.2020, 0.2332, 0.2795, 0.1580, 0.1593, 0.2722], device='cuda:1'), in_proj_covar=tensor([0.0367, 0.0515, 0.0559, 0.0439, 0.0596, 0.0583, 0.0448, 0.0595], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-04-29 19:42:39,407 INFO [train.py:904] (1/8) Epoch 13, batch 6200, loss[loss=0.1877, simple_loss=0.2783, pruned_loss=0.04856, over 16940.00 frames. ], tot_loss[loss=0.2158, simple_loss=0.2989, pruned_loss=0.06629, over 3058365.17 frames. ], batch size: 96, lr: 5.23e-03, grad_scale: 4.0 2023-04-29 19:42:39,906 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.8217, 1.9701, 2.3364, 2.7562, 2.7253, 3.0939, 1.9206, 3.1439], device='cuda:1'), covar=tensor([0.0167, 0.0381, 0.0262, 0.0222, 0.0231, 0.0154, 0.0412, 0.0091], device='cuda:1'), in_proj_covar=tensor([0.0164, 0.0176, 0.0160, 0.0163, 0.0174, 0.0130, 0.0176, 0.0121], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001], device='cuda:1') 2023-04-29 19:43:05,227 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 13, batch 6250, loss[loss=0.227, simple_loss=0.2983, pruned_loss=0.07785, over 11368.00 frames. ], tot_loss[loss=0.2141, simple_loss=0.2979, pruned_loss=0.06516, over 3088201.99 frames. ], batch size: 247, lr: 5.23e-03, grad_scale: 4.0 2023-04-29 19:44:08,617 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.3300, 3.4163, 1.8273, 3.6949, 2.4544, 3.6550, 2.0124, 2.5895], device='cuda:1'), covar=tensor([0.0236, 0.0345, 0.1692, 0.0210, 0.0822, 0.0691, 0.1552, 0.0764], device='cuda:1'), in_proj_covar=tensor([0.0155, 0.0167, 0.0190, 0.0137, 0.0168, 0.0207, 0.0197, 0.0173], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0003, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-04-29 19:44:37,187 INFO [optim.py:368] (1/8) 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 19:44:44,548 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=128083.0, num_to_drop=1, layers_to_drop={0} 2023-04-29 19:45:13,033 INFO [train.py:904] (1/8) Epoch 13, batch 6300, loss[loss=0.2118, simple_loss=0.2946, pruned_loss=0.06449, over 16922.00 frames. ], tot_loss[loss=0.2136, simple_loss=0.2977, pruned_loss=0.06475, over 3112708.54 frames. ], batch size: 109, lr: 5.23e-03, grad_scale: 4.0 2023-04-29 19:45:51,376 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=2.36 vs. limit=5.0 2023-04-29 19:46:02,062 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=128131.0, num_to_drop=1, layers_to_drop={0} 2023-04-29 19:46:08,187 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=128135.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 19:46:14,647 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.5139, 3.6010, 2.0953, 3.9768, 2.5285, 3.9573, 2.2427, 2.7649], device='cuda:1'), covar=tensor([0.0246, 0.0349, 0.1541, 0.0169, 0.0840, 0.0487, 0.1508, 0.0781], device='cuda:1'), in_proj_covar=tensor([0.0157, 0.0169, 0.0191, 0.0138, 0.0169, 0.0208, 0.0199, 0.0175], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-04-29 19:46:33,871 INFO [train.py:904] (1/8) Epoch 13, batch 6350, loss[loss=0.2453, simple_loss=0.3233, pruned_loss=0.08366, over 15284.00 frames. ], tot_loss[loss=0.2151, simple_loss=0.2983, pruned_loss=0.06594, over 3102885.00 frames. ], batch size: 190, lr: 5.22e-03, grad_scale: 4.0 2023-04-29 19:47:13,674 INFO [optim.py:368] (1/8) 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 19:47:24,804 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 13, batch 6400, loss[loss=0.303, simple_loss=0.3595, pruned_loss=0.1233, over 11268.00 frames. ], tot_loss[loss=0.2165, simple_loss=0.299, pruned_loss=0.06699, over 3100629.84 frames. ], batch size: 246, lr: 5.22e-03, grad_scale: 8.0 2023-04-29 19:47:56,273 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=128206.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 19:48:29,397 INFO [zipformer.py:625] (1/8) 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] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=128233.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 19:49:04,023 INFO [train.py:904] (1/8) Epoch 13, batch 6450, loss[loss=0.2255, simple_loss=0.3124, pruned_loss=0.06931, over 16619.00 frames. ], tot_loss[loss=0.2154, simple_loss=0.2987, pruned_loss=0.06604, over 3108249.73 frames. ], batch size: 62, lr: 5.22e-03, grad_scale: 2.0 2023-04-29 19:49:48,881 INFO [optim.py:368] (1/8) Clipping_scale=2.0, grad-norm quartiles 1.913e+02 2.912e+02 3.353e+02 4.052e+02 7.402e+02, threshold=6.705e+02, percent-clipped=0.0 2023-04-29 19:50:21,885 INFO [train.py:904] (1/8) Epoch 13, batch 6500, loss[loss=0.2106, simple_loss=0.2959, pruned_loss=0.06269, over 16676.00 frames. ], tot_loss[loss=0.2135, simple_loss=0.2966, pruned_loss=0.0652, over 3105658.44 frames. ], batch size: 134, lr: 5.22e-03, grad_scale: 2.0 2023-04-29 19:50:37,314 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.2067, 1.5017, 1.8982, 2.1536, 2.2968, 2.4798, 1.5572, 2.3795], device='cuda:1'), covar=tensor([0.0170, 0.0389, 0.0226, 0.0236, 0.0223, 0.0162, 0.0421, 0.0110], device='cuda:1'), in_proj_covar=tensor([0.0164, 0.0176, 0.0160, 0.0162, 0.0174, 0.0131, 0.0176, 0.0121], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001], device='cuda:1') 2023-04-29 19:50:45,332 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=128318.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 19:50:47,717 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 13, batch 6550, loss[loss=0.1968, simple_loss=0.2977, pruned_loss=0.04797, over 16664.00 frames. ], tot_loss[loss=0.2153, simple_loss=0.2989, pruned_loss=0.0658, over 3114513.13 frames. ], batch size: 76, lr: 5.22e-03, grad_scale: 2.0 2023-04-29 19:52:01,583 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=128366.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 19:52:22,963 INFO [optim.py:368] (1/8) 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 19:52:24,979 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=128381.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 19:52:56,141 INFO [train.py:904] (1/8) Epoch 13, batch 6600, loss[loss=0.2314, simple_loss=0.3145, pruned_loss=0.0742, over 15351.00 frames. ], tot_loss[loss=0.2183, simple_loss=0.3022, pruned_loss=0.06721, over 3088646.79 frames. ], batch size: 190, lr: 5.22e-03, grad_scale: 2.0 2023-04-29 19:53:13,395 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=128413.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 19:53:15,384 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.2597, 4.3478, 4.1531, 3.9163, 3.8430, 4.2612, 3.9930, 4.0228], device='cuda:1'), covar=tensor([0.0651, 0.0622, 0.0271, 0.0265, 0.0843, 0.0422, 0.0630, 0.0620], device='cuda:1'), in_proj_covar=tensor([0.0248, 0.0341, 0.0297, 0.0278, 0.0317, 0.0322, 0.0202, 0.0347], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-29 19:53:44,984 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([1.6898, 2.4204, 2.1729, 3.2431, 2.3550, 3.5581, 1.4283, 2.7363], device='cuda:1'), covar=tensor([0.1299, 0.0724, 0.1247, 0.0182, 0.0235, 0.0410, 0.1580, 0.0782], device='cuda:1'), in_proj_covar=tensor([0.0154, 0.0162, 0.0182, 0.0155, 0.0201, 0.0207, 0.0183, 0.0181], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0003, 0.0004], device='cuda:1') 2023-04-29 19:53:48,805 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=128435.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 19:54:14,487 INFO [train.py:904] (1/8) Epoch 13, batch 6650, loss[loss=0.2047, simple_loss=0.2885, pruned_loss=0.06045, over 16703.00 frames. ], tot_loss[loss=0.2195, simple_loss=0.3028, pruned_loss=0.06811, over 3075921.54 frames. ], batch size: 76, lr: 5.22e-03, grad_scale: 2.0 2023-04-29 19:54:19,826 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.0805, 3.3761, 3.5159, 3.4721, 3.4950, 3.3418, 3.2066, 3.3984], device='cuda:1'), covar=tensor([0.0622, 0.0776, 0.0576, 0.0729, 0.0732, 0.0693, 0.1225, 0.0593], device='cuda:1'), in_proj_covar=tensor([0.0345, 0.0362, 0.0362, 0.0348, 0.0411, 0.0384, 0.0481, 0.0309], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:1') 2023-04-29 19:54:47,937 INFO [zipformer.py:625] (1/8) 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] (1/8) 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] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=128483.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 19:55:30,391 INFO [train.py:904] (1/8) Epoch 13, batch 6700, loss[loss=0.1843, simple_loss=0.2696, pruned_loss=0.04952, over 17027.00 frames. ], tot_loss[loss=0.2192, simple_loss=0.3017, pruned_loss=0.06834, over 3058372.43 frames. ], batch size: 55, lr: 5.22e-03, grad_scale: 2.0 2023-04-29 19:55:36,517 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=128506.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 19:56:09,907 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=128528.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 19:56:46,042 INFO [train.py:904] (1/8) Epoch 13, batch 6750, loss[loss=0.2147, simple_loss=0.2916, pruned_loss=0.06894, over 16615.00 frames. ], tot_loss[loss=0.2184, simple_loss=0.3001, pruned_loss=0.06836, over 3069677.29 frames. ], batch size: 62, lr: 5.22e-03, grad_scale: 2.0 2023-04-29 19:56:49,432 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=128554.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 19:57:22,974 INFO [zipformer.py:625] (1/8) 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] (1/8) 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 19:57:38,491 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.2305, 2.1548, 2.1825, 4.1809, 2.0346, 2.5437, 2.1836, 2.2802], device='cuda:1'), covar=tensor([0.1087, 0.3293, 0.2515, 0.0392, 0.3823, 0.2244, 0.3078, 0.3230], device='cuda:1'), in_proj_covar=tensor([0.0366, 0.0401, 0.0334, 0.0315, 0.0416, 0.0462, 0.0368, 0.0468], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-29 19:58:01,114 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.6916, 3.7288, 2.2572, 4.3790, 2.7710, 4.2788, 2.4495, 2.9359], device='cuda:1'), covar=tensor([0.0220, 0.0302, 0.1458, 0.0119, 0.0764, 0.0408, 0.1295, 0.0717], device='cuda:1'), in_proj_covar=tensor([0.0155, 0.0166, 0.0189, 0.0136, 0.0168, 0.0207, 0.0196, 0.0173], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0003, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-04-29 19:58:01,812 INFO [train.py:904] (1/8) Epoch 13, batch 6800, loss[loss=0.2049, simple_loss=0.2902, pruned_loss=0.05978, over 16475.00 frames. ], tot_loss[loss=0.2171, simple_loss=0.2992, pruned_loss=0.06748, over 3081636.84 frames. ], batch size: 68, lr: 5.21e-03, grad_scale: 4.0 2023-04-29 19:58:57,064 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.0256, 3.0178, 1.7946, 3.2424, 2.2826, 3.2775, 1.9827, 2.5186], device='cuda:1'), covar=tensor([0.0247, 0.0385, 0.1524, 0.0206, 0.0780, 0.0608, 0.1501, 0.0687], device='cuda:1'), in_proj_covar=tensor([0.0155, 0.0167, 0.0189, 0.0137, 0.0168, 0.0207, 0.0196, 0.0173], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0003, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-04-29 19:59:19,100 INFO [train.py:904] (1/8) Epoch 13, batch 6850, loss[loss=0.2424, simple_loss=0.3344, pruned_loss=0.07517, over 16942.00 frames. ], tot_loss[loss=0.2179, simple_loss=0.3004, pruned_loss=0.06769, over 3088282.52 frames. ], batch size: 109, lr: 5.21e-03, grad_scale: 4.0 2023-04-29 19:59:29,331 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.9079, 5.2328, 4.9717, 5.0374, 4.8142, 4.7551, 4.5682, 5.3117], device='cuda:1'), covar=tensor([0.1051, 0.0858, 0.1013, 0.0752, 0.0766, 0.0763, 0.1122, 0.0877], device='cuda:1'), in_proj_covar=tensor([0.0567, 0.0700, 0.0575, 0.0500, 0.0440, 0.0450, 0.0587, 0.0538], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:1') 2023-04-29 19:59:54,880 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=128676.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 20:00:00,830 INFO [optim.py:368] (1/8) 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 20:00:34,569 INFO [train.py:904] (1/8) Epoch 13, batch 6900, loss[loss=0.2159, simple_loss=0.3064, pruned_loss=0.06264, over 16764.00 frames. ], tot_loss[loss=0.2189, simple_loss=0.303, pruned_loss=0.06738, over 3099400.38 frames. ], batch size: 83, lr: 5.21e-03, grad_scale: 4.0 2023-04-29 20:01:13,617 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.7558, 4.7597, 4.6135, 4.2979, 4.2156, 4.6258, 4.5898, 4.3377], device='cuda:1'), covar=tensor([0.0613, 0.0573, 0.0303, 0.0303, 0.1060, 0.0513, 0.0348, 0.0677], device='cuda:1'), in_proj_covar=tensor([0.0246, 0.0338, 0.0296, 0.0275, 0.0315, 0.0321, 0.0201, 0.0346], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-29 20:01:52,386 INFO [train.py:904] (1/8) Epoch 13, batch 6950, loss[loss=0.2479, simple_loss=0.3214, pruned_loss=0.08722, over 15304.00 frames. ], tot_loss[loss=0.2211, simple_loss=0.3044, pruned_loss=0.06889, over 3084987.58 frames. ], batch size: 190, lr: 5.21e-03, grad_scale: 4.0 2023-04-29 20:02:18,898 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=128769.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 20:02:36,165 INFO [optim.py:368] (1/8) 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 20:02:38,376 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=128781.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 20:03:08,750 INFO [train.py:904] (1/8) Epoch 13, batch 7000, loss[loss=0.2274, simple_loss=0.3153, pruned_loss=0.06971, over 16438.00 frames. ], tot_loss[loss=0.2199, simple_loss=0.3039, pruned_loss=0.06797, over 3078327.03 frames. ], batch size: 68, lr: 5.21e-03, grad_scale: 4.0 2023-04-29 20:04:07,306 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=128842.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 20:04:20,571 INFO [train.py:904] (1/8) Epoch 13, batch 7050, loss[loss=0.2033, simple_loss=0.2881, pruned_loss=0.05921, over 16716.00 frames. ], tot_loss[loss=0.2203, simple_loss=0.3047, pruned_loss=0.0679, over 3082376.16 frames. ], batch size: 57, lr: 5.21e-03, grad_scale: 4.0 2023-04-29 20:05:02,586 INFO [optim.py:368] (1/8) Clipping_scale=2.0, grad-norm quartiles 1.996e+02 2.845e+02 3.584e+02 4.333e+02 9.135e+02, threshold=7.167e+02, percent-clipped=3.0 2023-04-29 20:05:19,772 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=128890.0, num_to_drop=1, layers_to_drop={1} 2023-04-29 20:05:38,298 INFO [train.py:904] (1/8) Epoch 13, batch 7100, loss[loss=0.1804, simple_loss=0.2689, pruned_loss=0.04598, over 16571.00 frames. ], tot_loss[loss=0.2187, simple_loss=0.3026, pruned_loss=0.06738, over 3071532.26 frames. ], batch size: 75, lr: 5.21e-03, grad_scale: 4.0 2023-04-29 20:06:56,765 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=128951.0, num_to_drop=1, layers_to_drop={1} 2023-04-29 20:06:57,448 INFO [train.py:904] (1/8) Epoch 13, batch 7150, loss[loss=0.2176, simple_loss=0.3042, pruned_loss=0.06551, over 16539.00 frames. ], tot_loss[loss=0.218, simple_loss=0.3012, pruned_loss=0.06743, over 3063036.82 frames. ], batch size: 75, lr: 5.21e-03, grad_scale: 4.0 2023-04-29 20:07:34,044 INFO [zipformer.py:625] (1/8) 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] (1/8) 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] (1/8) Epoch 13, batch 7200, loss[loss=0.1915, simple_loss=0.2768, pruned_loss=0.05314, over 16629.00 frames. ], tot_loss[loss=0.2157, simple_loss=0.2994, pruned_loss=0.06603, over 3076297.80 frames. ], batch size: 62, lr: 5.21e-03, grad_scale: 8.0 2023-04-29 20:08:44,526 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 13, batch 7250, loss[loss=0.2073, simple_loss=0.2799, pruned_loss=0.06735, over 16625.00 frames. ], tot_loss[loss=0.2139, simple_loss=0.2976, pruned_loss=0.06514, over 3076242.02 frames. ], batch size: 134, lr: 5.21e-03, grad_scale: 8.0 2023-04-29 20:09:56,884 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=129069.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 20:10:12,179 INFO [optim.py:368] (1/8) 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 20:10:45,713 INFO [train.py:904] (1/8) Epoch 13, batch 7300, loss[loss=0.2392, simple_loss=0.3067, pruned_loss=0.08583, over 11341.00 frames. ], tot_loss[loss=0.215, simple_loss=0.2982, pruned_loss=0.06585, over 3054870.56 frames. ], batch size: 249, lr: 5.20e-03, grad_scale: 8.0 2023-04-29 20:11:09,070 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=129117.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 20:11:40,245 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 13, batch 7350, loss[loss=0.2086, simple_loss=0.2962, pruned_loss=0.06051, over 16727.00 frames. ], tot_loss[loss=0.2162, simple_loss=0.299, pruned_loss=0.06665, over 3037172.89 frames. ], batch size: 83, lr: 5.20e-03, grad_scale: 8.0 2023-04-29 20:12:46,220 INFO [optim.py:368] (1/8) 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 20:13:01,470 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=2.91 vs. limit=5.0 2023-04-29 20:13:18,285 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([1.7669, 1.2918, 1.6544, 1.6255, 1.7467, 1.8492, 1.5537, 1.7145], device='cuda:1'), covar=tensor([0.0200, 0.0289, 0.0140, 0.0196, 0.0189, 0.0125, 0.0291, 0.0086], device='cuda:1'), in_proj_covar=tensor([0.0162, 0.0172, 0.0157, 0.0159, 0.0171, 0.0127, 0.0172, 0.0118], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001], device='cuda:1') 2023-04-29 20:13:21,036 INFO [train.py:904] (1/8) Epoch 13, batch 7400, loss[loss=0.2585, simple_loss=0.327, pruned_loss=0.09496, over 11555.00 frames. ], tot_loss[loss=0.2162, simple_loss=0.2993, pruned_loss=0.06653, over 3050441.29 frames. ], batch size: 248, lr: 5.20e-03, grad_scale: 8.0 2023-04-29 20:13:21,736 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=2.34 vs. limit=5.0 2023-04-29 20:13:26,126 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.14 vs. limit=2.0 2023-04-29 20:14:32,918 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=129246.0, num_to_drop=1, layers_to_drop={3} 2023-04-29 20:14:41,688 INFO [train.py:904] (1/8) Epoch 13, batch 7450, loss[loss=0.2584, simple_loss=0.319, pruned_loss=0.09887, over 11427.00 frames. ], tot_loss[loss=0.2175, simple_loss=0.3001, pruned_loss=0.06745, over 3052609.96 frames. ], batch size: 248, lr: 5.20e-03, grad_scale: 8.0 2023-04-29 20:15:10,989 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.28 vs. limit=2.0 2023-04-29 20:15:30,917 INFO [optim.py:368] (1/8) Clipping_scale=2.0, grad-norm quartiles 1.914e+02 3.121e+02 3.842e+02 4.432e+02 7.351e+02, threshold=7.685e+02, percent-clipped=1.0 2023-04-29 20:16:05,626 INFO [train.py:904] (1/8) Epoch 13, batch 7500, loss[loss=0.2511, simple_loss=0.3179, pruned_loss=0.09216, over 11222.00 frames. ], tot_loss[loss=0.2169, simple_loss=0.3003, pruned_loss=0.06675, over 3058299.66 frames. ], batch size: 248, lr: 5.20e-03, grad_scale: 8.0 2023-04-29 20:17:24,560 INFO [train.py:904] (1/8) Epoch 13, batch 7550, loss[loss=0.2405, simple_loss=0.3007, pruned_loss=0.09013, over 11305.00 frames. ], tot_loss[loss=0.2156, simple_loss=0.2987, pruned_loss=0.0662, over 3067490.56 frames. ], batch size: 247, lr: 5.20e-03, grad_scale: 8.0 2023-04-29 20:18:07,708 INFO [optim.py:368] (1/8) 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 20:18:41,433 INFO [train.py:904] (1/8) Epoch 13, batch 7600, loss[loss=0.1971, simple_loss=0.2837, pruned_loss=0.05525, over 16820.00 frames. ], tot_loss[loss=0.2147, simple_loss=0.2973, pruned_loss=0.06607, over 3074016.56 frames. ], batch size: 102, lr: 5.20e-03, grad_scale: 8.0 2023-04-29 20:18:59,698 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=129413.0, num_to_drop=1, layers_to_drop={0} 2023-04-29 20:19:29,437 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.3104, 1.6026, 1.9847, 2.2839, 2.3470, 2.5396, 1.6670, 2.4797], device='cuda:1'), covar=tensor([0.0192, 0.0439, 0.0245, 0.0259, 0.0254, 0.0176, 0.0434, 0.0138], device='cuda:1'), in_proj_covar=tensor([0.0163, 0.0174, 0.0158, 0.0162, 0.0173, 0.0128, 0.0174, 0.0119], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001], device='cuda:1') 2023-04-29 20:19:30,701 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([1.9378, 3.0936, 3.1700, 1.9611, 2.9427, 3.1687, 3.0227, 1.8949], device='cuda:1'), covar=tensor([0.0473, 0.0053, 0.0049, 0.0404, 0.0100, 0.0105, 0.0087, 0.0410], device='cuda:1'), in_proj_covar=tensor([0.0133, 0.0072, 0.0074, 0.0131, 0.0085, 0.0096, 0.0084, 0.0124], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-04-29 20:19:37,484 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=129437.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 20:20:00,007 INFO [train.py:904] (1/8) Epoch 13, batch 7650, loss[loss=0.2139, simple_loss=0.2909, pruned_loss=0.06848, over 17033.00 frames. ], tot_loss[loss=0.2166, simple_loss=0.2985, pruned_loss=0.06732, over 3068411.42 frames. ], batch size: 55, lr: 5.20e-03, grad_scale: 8.0 2023-04-29 20:20:35,776 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=129474.0, num_to_drop=1, layers_to_drop={0} 2023-04-29 20:20:44,164 INFO [optim.py:368] (1/8) 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 20:20:53,105 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=129485.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 20:21:18,584 INFO [train.py:904] (1/8) Epoch 13, batch 7700, loss[loss=0.2375, simple_loss=0.3011, pruned_loss=0.08694, over 11423.00 frames. ], tot_loss[loss=0.2169, simple_loss=0.2987, pruned_loss=0.0676, over 3071847.90 frames. ], batch size: 247, lr: 5.20e-03, grad_scale: 4.0 2023-04-29 20:21:34,665 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.2860, 2.1144, 1.6333, 1.8910, 2.3848, 2.0878, 2.2983, 2.5323], device='cuda:1'), covar=tensor([0.0145, 0.0306, 0.0414, 0.0374, 0.0186, 0.0306, 0.0145, 0.0209], device='cuda:1'), in_proj_covar=tensor([0.0157, 0.0208, 0.0200, 0.0200, 0.0205, 0.0205, 0.0209, 0.0195], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-29 20:21:35,951 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([1.7907, 1.2860, 1.6632, 1.6685, 1.7767, 1.9006, 1.5603, 1.7693], device='cuda:1'), covar=tensor([0.0178, 0.0303, 0.0147, 0.0208, 0.0195, 0.0114, 0.0303, 0.0101], device='cuda:1'), in_proj_covar=tensor([0.0163, 0.0174, 0.0158, 0.0162, 0.0172, 0.0128, 0.0174, 0.0119], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001], device='cuda:1') 2023-04-29 20:22:06,596 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.0878, 3.2390, 1.7496, 3.4281, 2.3695, 3.4805, 2.0077, 2.5306], device='cuda:1'), covar=tensor([0.0304, 0.0375, 0.1779, 0.0195, 0.0869, 0.0557, 0.1486, 0.0726], device='cuda:1'), in_proj_covar=tensor([0.0155, 0.0166, 0.0190, 0.0136, 0.0168, 0.0208, 0.0196, 0.0174], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0003, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-04-29 20:22:12,925 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.33 vs. limit=2.0 2023-04-29 20:22:17,739 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.4491, 2.2985, 2.3794, 4.2383, 2.2302, 2.7027, 2.3098, 2.4792], device='cuda:1'), covar=tensor([0.0986, 0.3242, 0.2260, 0.0407, 0.3546, 0.2158, 0.3024, 0.2872], device='cuda:1'), in_proj_covar=tensor([0.0365, 0.0402, 0.0335, 0.0315, 0.0417, 0.0461, 0.0367, 0.0468], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-29 20:22:27,009 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=129546.0, num_to_drop=1, layers_to_drop={1} 2023-04-29 20:22:35,876 INFO [train.py:904] (1/8) Epoch 13, batch 7750, loss[loss=0.2081, simple_loss=0.2904, pruned_loss=0.06292, over 17052.00 frames. ], tot_loss[loss=0.2165, simple_loss=0.2987, pruned_loss=0.06717, over 3072973.37 frames. ], batch size: 53, lr: 5.20e-03, grad_scale: 4.0 2023-04-29 20:22:39,466 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.4247, 3.5189, 3.8688, 1.7789, 3.9934, 4.0900, 2.8786, 3.0237], device='cuda:1'), covar=tensor([0.0863, 0.0215, 0.0175, 0.1279, 0.0065, 0.0128, 0.0425, 0.0436], device='cuda:1'), in_proj_covar=tensor([0.0146, 0.0103, 0.0090, 0.0140, 0.0070, 0.0111, 0.0123, 0.0128], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:1') 2023-04-29 20:23:19,688 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.9969, 4.0597, 4.4084, 4.3614, 4.3869, 4.0853, 4.0931, 3.9955], device='cuda:1'), covar=tensor([0.0359, 0.0597, 0.0354, 0.0443, 0.0478, 0.0417, 0.0982, 0.0565], device='cuda:1'), in_proj_covar=tensor([0.0346, 0.0363, 0.0365, 0.0348, 0.0411, 0.0386, 0.0482, 0.0310], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:1') 2023-04-29 20:23:20,358 INFO [optim.py:368] (1/8) Clipping_scale=2.0, grad-norm quartiles 1.834e+02 3.197e+02 3.787e+02 4.587e+02 8.661e+02, threshold=7.574e+02, percent-clipped=1.0 2023-04-29 20:23:40,233 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=129594.0, num_to_drop=1, layers_to_drop={1} 2023-04-29 20:23:52,176 INFO [train.py:904] (1/8) Epoch 13, batch 7800, loss[loss=0.2236, simple_loss=0.3027, pruned_loss=0.07224, over 16298.00 frames. ], tot_loss[loss=0.2176, simple_loss=0.2999, pruned_loss=0.06768, over 3089931.23 frames. ], batch size: 165, lr: 5.19e-03, grad_scale: 4.0 2023-04-29 20:24:09,273 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 13, batch 7850, loss[loss=0.1981, simple_loss=0.2853, pruned_loss=0.0555, over 16455.00 frames. ], tot_loss[loss=0.2178, simple_loss=0.3007, pruned_loss=0.06746, over 3083151.03 frames. ], batch size: 68, lr: 5.19e-03, grad_scale: 4.0 2023-04-29 20:25:34,780 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.22 vs. limit=2.0 2023-04-29 20:25:39,032 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.7140, 3.7015, 4.0874, 2.0113, 4.1904, 4.2227, 3.0247, 3.1978], device='cuda:1'), covar=tensor([0.0704, 0.0201, 0.0130, 0.1119, 0.0052, 0.0112, 0.0364, 0.0404], device='cuda:1'), in_proj_covar=tensor([0.0145, 0.0102, 0.0089, 0.0139, 0.0069, 0.0110, 0.0122, 0.0127], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:1') 2023-04-29 20:25:40,269 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=129673.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 20:25:52,267 INFO [optim.py:368] (1/8) 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 20:26:22,715 INFO [train.py:904] (1/8) Epoch 13, batch 7900, loss[loss=0.2107, simple_loss=0.2998, pruned_loss=0.06078, over 16218.00 frames. ], tot_loss[loss=0.2171, simple_loss=0.2999, pruned_loss=0.06719, over 3066594.82 frames. ], batch size: 165, lr: 5.19e-03, grad_scale: 4.0 2023-04-29 20:26:24,436 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.6789, 4.7235, 4.5500, 4.2526, 4.1698, 4.5952, 4.4444, 4.2879], device='cuda:1'), covar=tensor([0.0552, 0.0448, 0.0255, 0.0272, 0.0904, 0.0435, 0.0362, 0.0661], device='cuda:1'), in_proj_covar=tensor([0.0245, 0.0336, 0.0293, 0.0274, 0.0311, 0.0317, 0.0201, 0.0342], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-29 20:26:36,795 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.5007, 3.5532, 3.2924, 3.0522, 3.0954, 3.4261, 3.2645, 3.2627], device='cuda:1'), covar=tensor([0.0533, 0.0497, 0.0268, 0.0238, 0.0534, 0.0431, 0.1109, 0.0492], device='cuda:1'), in_proj_covar=tensor([0.0245, 0.0337, 0.0293, 0.0274, 0.0311, 0.0317, 0.0201, 0.0343], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-29 20:27:14,815 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=3.14 vs. limit=5.0 2023-04-29 20:27:36,769 INFO [train.py:904] (1/8) Epoch 13, batch 7950, loss[loss=0.2657, simple_loss=0.317, pruned_loss=0.1072, over 11765.00 frames. ], tot_loss[loss=0.2182, simple_loss=0.3004, pruned_loss=0.06801, over 3062304.58 frames. ], batch size: 246, lr: 5.19e-03, grad_scale: 4.0 2023-04-29 20:28:01,700 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=129769.0, num_to_drop=1, layers_to_drop={3} 2023-04-29 20:28:18,207 INFO [optim.py:368] (1/8) 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 20:28:33,237 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 13, batch 8000, loss[loss=0.1859, simple_loss=0.276, pruned_loss=0.04783, over 16872.00 frames. ], tot_loss[loss=0.2195, simple_loss=0.3013, pruned_loss=0.06886, over 3064877.20 frames. ], batch size: 96, lr: 5.19e-03, grad_scale: 8.0 2023-04-29 20:30:02,332 INFO [train.py:904] (1/8) Epoch 13, batch 8050, loss[loss=0.2155, simple_loss=0.3025, pruned_loss=0.0643, over 16457.00 frames. ], tot_loss[loss=0.2187, simple_loss=0.3008, pruned_loss=0.06835, over 3069963.98 frames. ], batch size: 68, lr: 5.19e-03, grad_scale: 4.0 2023-04-29 20:30:02,925 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=129852.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 20:30:45,831 INFO [optim.py:368] (1/8) 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 20:30:46,922 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.45 vs. limit=2.0 2023-04-29 20:31:02,181 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=3.06 vs. limit=5.0 2023-04-29 20:31:15,197 INFO [train.py:904] (1/8) Epoch 13, batch 8100, loss[loss=0.1994, simple_loss=0.2888, pruned_loss=0.05502, over 16507.00 frames. ], tot_loss[loss=0.2176, simple_loss=0.3005, pruned_loss=0.06742, over 3081567.64 frames. ], batch size: 75, lr: 5.19e-03, grad_scale: 4.0 2023-04-29 20:32:29,503 INFO [train.py:904] (1/8) Epoch 13, batch 8150, loss[loss=0.2173, simple_loss=0.2912, pruned_loss=0.07165, over 11421.00 frames. ], tot_loss[loss=0.2157, simple_loss=0.2981, pruned_loss=0.06668, over 3078566.45 frames. ], batch size: 246, lr: 5.19e-03, grad_scale: 4.0 2023-04-29 20:32:53,000 INFO [zipformer.py:625] (1/8) 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] (1/8) 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 20:33:28,578 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.7825, 4.5914, 4.7669, 4.9797, 5.1618, 4.6107, 5.1419, 5.0871], device='cuda:1'), covar=tensor([0.1691, 0.1116, 0.1617, 0.0657, 0.0516, 0.0819, 0.0520, 0.0676], device='cuda:1'), in_proj_covar=tensor([0.0544, 0.0672, 0.0802, 0.0684, 0.0524, 0.0530, 0.0551, 0.0638], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-29 20:33:46,569 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 13, batch 8200, loss[loss=0.2009, simple_loss=0.2871, pruned_loss=0.05733, over 16708.00 frames. ], tot_loss[loss=0.2135, simple_loss=0.2954, pruned_loss=0.0658, over 3088989.86 frames. ], batch size: 124, lr: 5.19e-03, grad_scale: 4.0 2023-04-29 20:33:49,226 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=130002.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 20:33:52,952 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.4339, 3.2083, 3.4323, 1.7095, 3.5922, 3.6682, 2.8767, 2.7736], device='cuda:1'), covar=tensor([0.0682, 0.0221, 0.0195, 0.1227, 0.0067, 0.0175, 0.0389, 0.0423], device='cuda:1'), in_proj_covar=tensor([0.0144, 0.0102, 0.0089, 0.0139, 0.0069, 0.0110, 0.0121, 0.0127], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:1') 2023-04-29 20:35:09,258 INFO [train.py:904] (1/8) Epoch 13, batch 8250, loss[loss=0.2032, simple_loss=0.2932, pruned_loss=0.05662, over 15260.00 frames. ], tot_loss[loss=0.2111, simple_loss=0.2949, pruned_loss=0.06368, over 3063484.48 frames. ], batch size: 190, lr: 5.19e-03, grad_scale: 4.0 2023-04-29 20:35:23,984 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=130061.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 20:35:27,840 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=130063.0, num_to_drop=1, layers_to_drop={3} 2023-04-29 20:35:37,102 INFO [zipformer.py:625] (1/8) 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] (1/8) Clipping_scale=2.0, grad-norm quartiles 1.490e+02 2.705e+02 3.349e+02 4.037e+02 8.257e+02, threshold=6.697e+02, percent-clipped=3.0 2023-04-29 20:36:29,992 INFO [train.py:904] (1/8) Epoch 13, batch 8300, loss[loss=0.1747, simple_loss=0.2574, pruned_loss=0.04601, over 12278.00 frames. ], tot_loss[loss=0.2061, simple_loss=0.2916, pruned_loss=0.0603, over 3060484.94 frames. ], batch size: 248, lr: 5.18e-03, grad_scale: 4.0 2023-04-29 20:36:55,410 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=130117.0, num_to_drop=1, layers_to_drop={1} 2023-04-29 20:37:43,871 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=130147.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 20:37:47,693 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.6975, 4.4919, 4.6866, 4.8540, 5.0440, 4.5075, 5.0245, 4.9920], device='cuda:1'), covar=tensor([0.1494, 0.1002, 0.1331, 0.0599, 0.0468, 0.0809, 0.0444, 0.0571], device='cuda:1'), in_proj_covar=tensor([0.0534, 0.0662, 0.0790, 0.0678, 0.0517, 0.0525, 0.0542, 0.0631], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-29 20:37:48,341 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.94 vs. limit=2.0 2023-04-29 20:37:51,986 INFO [train.py:904] (1/8) Epoch 13, batch 8350, loss[loss=0.1889, simple_loss=0.2766, pruned_loss=0.05061, over 17192.00 frames. ], tot_loss[loss=0.2036, simple_loss=0.2906, pruned_loss=0.05835, over 3063148.95 frames. ], batch size: 44, lr: 5.18e-03, grad_scale: 4.0 2023-04-29 20:38:39,964 INFO [optim.py:368] (1/8) 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 20:39:12,413 INFO [train.py:904] (1/8) Epoch 13, batch 8400, loss[loss=0.2066, simple_loss=0.274, pruned_loss=0.06962, over 12521.00 frames. ], tot_loss[loss=0.1999, simple_loss=0.2875, pruned_loss=0.0561, over 3055034.30 frames. ], batch size: 248, lr: 5.18e-03, grad_scale: 8.0 2023-04-29 20:40:03,693 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.44 vs. limit=2.0 2023-04-29 20:40:29,223 INFO [train.py:904] (1/8) Epoch 13, batch 8450, loss[loss=0.1736, simple_loss=0.2596, pruned_loss=0.04382, over 12663.00 frames. ], tot_loss[loss=0.1971, simple_loss=0.2856, pruned_loss=0.05437, over 3055971.23 frames. ], batch size: 248, lr: 5.18e-03, grad_scale: 8.0 2023-04-29 20:40:55,985 INFO [zipformer.py:625] (1/8) 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] (1/8) 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 20:41:49,391 INFO [train.py:904] (1/8) Epoch 13, batch 8500, loss[loss=0.1624, simple_loss=0.2493, pruned_loss=0.03777, over 16910.00 frames. ], tot_loss[loss=0.1925, simple_loss=0.2815, pruned_loss=0.05173, over 3057067.43 frames. ], batch size: 109, lr: 5.18e-03, grad_scale: 8.0 2023-04-29 20:42:12,542 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 13, batch 8550, loss[loss=0.1786, simple_loss=0.2572, pruned_loss=0.05001, over 12002.00 frames. ], tot_loss[loss=0.1905, simple_loss=0.2794, pruned_loss=0.05076, over 3026969.64 frames. ], batch size: 249, lr: 5.18e-03, grad_scale: 8.0 2023-04-29 20:43:19,677 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=130356.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 20:43:23,659 INFO [zipformer.py:625] (1/8) 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] (1/8) 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 20:44:12,684 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=130384.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 20:44:34,833 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.2325, 3.5495, 3.2460, 5.4441, 4.3554, 4.8670, 2.1247, 3.7341], device='cuda:1'), covar=tensor([0.1146, 0.0557, 0.0842, 0.0132, 0.0188, 0.0257, 0.1288, 0.0516], device='cuda:1'), in_proj_covar=tensor([0.0157, 0.0163, 0.0183, 0.0155, 0.0200, 0.0208, 0.0186, 0.0183], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-04-29 20:44:45,147 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([1.9777, 1.7709, 1.5694, 1.5099, 1.8968, 1.6143, 1.6621, 1.9368], device='cuda:1'), covar=tensor([0.0151, 0.0267, 0.0358, 0.0325, 0.0206, 0.0249, 0.0193, 0.0200], device='cuda:1'), in_proj_covar=tensor([0.0154, 0.0204, 0.0198, 0.0199, 0.0203, 0.0203, 0.0204, 0.0192], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-29 20:44:50,594 INFO [train.py:904] (1/8) Epoch 13, batch 8600, loss[loss=0.1756, simple_loss=0.27, pruned_loss=0.04061, over 16591.00 frames. ], tot_loss[loss=0.1901, simple_loss=0.2799, pruned_loss=0.05014, over 3011104.20 frames. ], batch size: 57, lr: 5.18e-03, grad_scale: 4.0 2023-04-29 20:45:03,327 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.5070, 4.2807, 4.5348, 4.6785, 4.8540, 4.4076, 4.8543, 4.8168], device='cuda:1'), covar=tensor([0.1601, 0.1214, 0.1498, 0.0713, 0.0551, 0.0882, 0.0440, 0.0572], device='cuda:1'), in_proj_covar=tensor([0.0528, 0.0653, 0.0780, 0.0671, 0.0511, 0.0518, 0.0534, 0.0622], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-29 20:45:22,263 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.6497, 2.1900, 2.4196, 4.4528, 2.1976, 2.6906, 2.2762, 2.4743], device='cuda:1'), covar=tensor([0.0890, 0.3740, 0.2427, 0.0304, 0.3952, 0.2293, 0.3482, 0.3200], device='cuda:1'), in_proj_covar=tensor([0.0360, 0.0396, 0.0329, 0.0309, 0.0409, 0.0451, 0.0360, 0.0459], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-29 20:45:51,016 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=130432.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 20:46:04,168 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=130438.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 20:46:16,907 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=130445.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 20:46:20,534 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 13, batch 8650, loss[loss=0.1811, simple_loss=0.275, pruned_loss=0.04358, over 16229.00 frames. ], tot_loss[loss=0.1869, simple_loss=0.2776, pruned_loss=0.04814, over 3031489.53 frames. ], batch size: 165, lr: 5.18e-03, grad_scale: 4.0 2023-04-29 20:47:20,236 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.4775, 4.2379, 4.1904, 4.6095, 4.8134, 4.3456, 4.7289, 4.7256], device='cuda:1'), covar=tensor([0.1576, 0.1254, 0.2712, 0.1025, 0.0753, 0.1184, 0.0907, 0.1061], device='cuda:1'), in_proj_covar=tensor([0.0524, 0.0648, 0.0773, 0.0664, 0.0505, 0.0513, 0.0529, 0.0617], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-29 20:47:40,993 INFO [optim.py:368] (1/8) 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 20:48:01,214 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=130493.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 20:48:04,695 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=130495.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 20:48:12,877 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 13, batch 8700, loss[loss=0.1868, simple_loss=0.2736, pruned_loss=0.05001, over 16577.00 frames. ], tot_loss[loss=0.184, simple_loss=0.2749, pruned_loss=0.04657, over 3064282.82 frames. ], batch size: 57, lr: 5.18e-03, grad_scale: 4.0 2023-04-29 20:49:46,612 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 13, batch 8750, loss[loss=0.1912, simple_loss=0.2868, pruned_loss=0.04781, over 15546.00 frames. ], tot_loss[loss=0.1832, simple_loss=0.2745, pruned_loss=0.04594, over 3070040.79 frames. ], batch size: 191, lr: 5.18e-03, grad_scale: 4.0 2023-04-29 20:50:30,138 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=130565.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 20:50:44,984 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.3824, 3.0378, 2.6868, 2.2038, 2.1975, 2.1952, 3.0571, 2.8771], device='cuda:1'), covar=tensor([0.2482, 0.0775, 0.1411, 0.2299, 0.2319, 0.1882, 0.0501, 0.1177], device='cuda:1'), in_proj_covar=tensor([0.0302, 0.0252, 0.0283, 0.0279, 0.0269, 0.0224, 0.0266, 0.0293], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-29 20:51:07,892 INFO [optim.py:368] (1/8) 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 20:51:48,218 INFO [train.py:904] (1/8) Epoch 13, batch 8800, loss[loss=0.1997, simple_loss=0.2882, pruned_loss=0.05559, over 16683.00 frames. ], tot_loss[loss=0.1811, simple_loss=0.2727, pruned_loss=0.04477, over 3071365.35 frames. ], batch size: 134, lr: 5.17e-03, grad_scale: 8.0 2023-04-29 20:52:02,433 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=130609.0, num_to_drop=1, layers_to_drop={1} 2023-04-29 20:52:39,093 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=130626.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 20:53:32,460 INFO [train.py:904] (1/8) Epoch 13, batch 8850, loss[loss=0.181, simple_loss=0.2825, pruned_loss=0.03976, over 15308.00 frames. ], tot_loss[loss=0.1812, simple_loss=0.2748, pruned_loss=0.04381, over 3066282.77 frames. ], batch size: 190, lr: 5.17e-03, grad_scale: 8.0 2023-04-29 20:53:41,312 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=130656.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 20:53:45,783 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=130658.0, num_to_drop=1, layers_to_drop={1} 2023-04-29 20:54:37,529 INFO [optim.py:368] (1/8) 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] (1/8) Epoch 13, batch 8900, loss[loss=0.184, simple_loss=0.2897, pruned_loss=0.03919, over 16861.00 frames. ], tot_loss[loss=0.1813, simple_loss=0.2757, pruned_loss=0.04343, over 3076516.42 frames. ], batch size: 96, lr: 5.17e-03, grad_scale: 8.0 2023-04-29 20:55:22,704 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=130704.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 20:55:23,028 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=2.50 vs. limit=5.0 2023-04-29 20:55:24,655 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([1.8608, 2.1405, 2.2799, 3.0337, 1.6718, 3.3536, 1.6068, 2.6519], device='cuda:1'), covar=tensor([0.1421, 0.0746, 0.1151, 0.0159, 0.0105, 0.0321, 0.1698, 0.0792], device='cuda:1'), in_proj_covar=tensor([0.0158, 0.0162, 0.0183, 0.0155, 0.0197, 0.0206, 0.0187, 0.0183], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-04-29 20:55:26,794 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=130706.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 20:55:48,929 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.3852, 3.2310, 3.4161, 2.0207, 3.6766, 3.7220, 2.9234, 2.8179], device='cuda:1'), covar=tensor([0.0685, 0.0209, 0.0196, 0.0960, 0.0047, 0.0102, 0.0318, 0.0395], device='cuda:1'), in_proj_covar=tensor([0.0138, 0.0096, 0.0084, 0.0132, 0.0065, 0.0103, 0.0115, 0.0121], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0003, 0.0002, 0.0004, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:1') 2023-04-29 20:56:54,227 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=130740.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 20:56:56,851 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.75 vs. limit=2.0 2023-04-29 20:57:21,721 INFO [train.py:904] (1/8) Epoch 13, batch 8950, loss[loss=0.1665, simple_loss=0.2621, pruned_loss=0.03542, over 16735.00 frames. ], tot_loss[loss=0.1818, simple_loss=0.2754, pruned_loss=0.04408, over 3074812.90 frames. ], batch size: 134, lr: 5.17e-03, grad_scale: 8.0 2023-04-29 20:58:29,463 INFO [optim.py:368] (1/8) Clipping_scale=2.0, grad-norm quartiles 1.466e+02 2.201e+02 2.684e+02 3.167e+02 7.959e+02, threshold=5.368e+02, percent-clipped=1.0 2023-04-29 20:58:41,704 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=130788.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 20:58:54,642 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=130794.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 20:59:11,359 INFO [train.py:904] (1/8) Epoch 13, batch 9000, loss[loss=0.1881, simple_loss=0.267, pruned_loss=0.0546, over 11746.00 frames. ], tot_loss[loss=0.1785, simple_loss=0.2719, pruned_loss=0.04255, over 3088217.45 frames. ], batch size: 248, lr: 5.17e-03, grad_scale: 8.0 2023-04-29 20:59:11,359 INFO [train.py:929] (1/8) Computing validation loss 2023-04-29 20:59:22,054 INFO [train.py:938] (1/8) Epoch 13, validation: loss=0.1517, simple_loss=0.2561, pruned_loss=0.02371, over 944034.00 frames. 2023-04-29 20:59:22,055 INFO [train.py:939] (1/8) Maximum memory allocated so far is 17845MB 2023-04-29 20:59:41,010 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.6719, 2.3277, 2.3981, 4.4448, 2.2399, 2.8051, 2.4203, 2.5111], device='cuda:1'), covar=tensor([0.0874, 0.3392, 0.2469, 0.0346, 0.3807, 0.2185, 0.3310, 0.3070], device='cuda:1'), in_proj_covar=tensor([0.0361, 0.0397, 0.0332, 0.0310, 0.0410, 0.0451, 0.0362, 0.0460], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-29 21:00:12,220 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.9904, 3.4971, 3.5202, 1.8734, 2.8848, 2.3383, 3.4633, 3.5471], device='cuda:1'), covar=tensor([0.0235, 0.0662, 0.0475, 0.1928, 0.0755, 0.0922, 0.0619, 0.0859], device='cuda:1'), in_proj_covar=tensor([0.0140, 0.0140, 0.0153, 0.0141, 0.0133, 0.0122, 0.0133, 0.0152], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:1') 2023-04-29 21:00:38,151 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.9786, 4.2693, 4.1065, 4.1126, 3.7302, 3.8789, 3.8857, 4.2502], device='cuda:1'), covar=tensor([0.0994, 0.0848, 0.0895, 0.0705, 0.0742, 0.1481, 0.0889, 0.0997], device='cuda:1'), in_proj_covar=tensor([0.0556, 0.0689, 0.0563, 0.0493, 0.0433, 0.0447, 0.0577, 0.0525], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:1') 2023-04-29 21:01:06,017 INFO [train.py:904] (1/8) Epoch 13, batch 9050, loss[loss=0.1662, simple_loss=0.2659, pruned_loss=0.0333, over 16696.00 frames. ], tot_loss[loss=0.1796, simple_loss=0.2727, pruned_loss=0.04331, over 3075360.94 frames. ], batch size: 76, lr: 5.17e-03, grad_scale: 8.0 2023-04-29 21:02:07,080 INFO [optim.py:368] (1/8) 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 21:02:41,258 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.9831, 3.8766, 4.0300, 4.1401, 4.2735, 3.8286, 4.2540, 4.2823], device='cuda:1'), covar=tensor([0.1402, 0.0913, 0.1197, 0.0616, 0.0524, 0.1373, 0.0520, 0.0606], device='cuda:1'), in_proj_covar=tensor([0.0526, 0.0648, 0.0773, 0.0666, 0.0505, 0.0509, 0.0528, 0.0613], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:1') 2023-04-29 21:02:52,476 INFO [train.py:904] (1/8) Epoch 13, batch 9100, loss[loss=0.1722, simple_loss=0.2763, pruned_loss=0.03408, over 16725.00 frames. ], tot_loss[loss=0.18, simple_loss=0.2723, pruned_loss=0.04384, over 3084256.29 frames. ], batch size: 89, lr: 5.17e-03, grad_scale: 8.0 2023-04-29 21:02:58,112 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=130904.0, num_to_drop=1, layers_to_drop={1} 2023-04-29 21:03:33,374 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 13, batch 9150, loss[loss=0.199, simple_loss=0.2778, pruned_loss=0.06008, over 11950.00 frames. ], tot_loss[loss=0.1798, simple_loss=0.2727, pruned_loss=0.04338, over 3079744.49 frames. ], batch size: 250, lr: 5.17e-03, grad_scale: 8.0 2023-04-29 21:05:52,920 INFO [optim.py:368] (1/8) 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] (1/8) Epoch 13, batch 9200, loss[loss=0.1494, simple_loss=0.2352, pruned_loss=0.03181, over 12499.00 frames. ], tot_loss[loss=0.1762, simple_loss=0.2683, pruned_loss=0.04211, over 3068979.98 frames. ], batch size: 248, lr: 5.17e-03, grad_scale: 8.0 2023-04-29 21:07:02,885 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.7950, 3.7071, 3.8710, 3.9756, 4.0705, 3.6497, 4.0496, 4.0954], device='cuda:1'), covar=tensor([0.1339, 0.1017, 0.1220, 0.0686, 0.0577, 0.1696, 0.0663, 0.0626], device='cuda:1'), in_proj_covar=tensor([0.0526, 0.0646, 0.0772, 0.0663, 0.0503, 0.0509, 0.0530, 0.0611], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:1') 2023-04-29 21:07:19,037 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.83 vs. limit=2.0 2023-04-29 21:07:43,463 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 13, batch 9250, loss[loss=0.1768, simple_loss=0.2684, pruned_loss=0.04258, over 16244.00 frames. ], tot_loss[loss=0.1763, simple_loss=0.2681, pruned_loss=0.04225, over 3064201.29 frames. ], batch size: 165, lr: 5.17e-03, grad_scale: 8.0 2023-04-29 21:09:12,849 INFO [optim.py:368] (1/8) Clipping_scale=2.0, grad-norm quartiles 1.311e+02 2.375e+02 2.887e+02 3.471e+02 9.563e+02, threshold=5.774e+02, percent-clipped=3.0 2023-04-29 21:09:24,456 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=131088.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 21:09:24,514 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=131088.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 21:09:39,397 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 13, batch 9300, loss[loss=0.1485, simple_loss=0.2472, pruned_loss=0.02483, over 16833.00 frames. ], tot_loss[loss=0.1751, simple_loss=0.2667, pruned_loss=0.04178, over 3060840.83 frames. ], batch size: 90, lr: 5.17e-03, grad_scale: 8.0 2023-04-29 21:10:00,628 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.2052, 4.0594, 4.3209, 4.4026, 4.5500, 4.0840, 4.5426, 4.5680], device='cuda:1'), covar=tensor([0.1694, 0.1143, 0.1196, 0.0642, 0.0516, 0.1118, 0.0469, 0.0543], device='cuda:1'), in_proj_covar=tensor([0.0525, 0.0643, 0.0769, 0.0662, 0.0501, 0.0507, 0.0528, 0.0609], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:1') 2023-04-29 21:11:11,075 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=131136.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 21:11:23,564 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=131142.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 21:11:28,590 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.73 vs. limit=2.0 2023-04-29 21:11:40,550 INFO [train.py:904] (1/8) Epoch 13, batch 9350, loss[loss=0.1912, simple_loss=0.2835, pruned_loss=0.04948, over 16206.00 frames. ], tot_loss[loss=0.1751, simple_loss=0.2667, pruned_loss=0.04171, over 3077499.64 frames. ], batch size: 165, lr: 5.16e-03, grad_scale: 8.0 2023-04-29 21:12:24,952 INFO [zipformer.py:625] (1/8) 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] (1/8) 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 21:12:42,633 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.6118, 4.6557, 4.4818, 4.1529, 4.1667, 4.5638, 4.3517, 4.2324], device='cuda:1'), covar=tensor([0.0527, 0.0441, 0.0267, 0.0269, 0.0799, 0.0506, 0.0420, 0.0642], device='cuda:1'), in_proj_covar=tensor([0.0237, 0.0324, 0.0287, 0.0265, 0.0296, 0.0308, 0.0195, 0.0331], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-29 21:13:20,691 INFO [train.py:904] (1/8) Epoch 13, batch 9400, loss[loss=0.1747, simple_loss=0.2745, pruned_loss=0.03744, over 15498.00 frames. ], tot_loss[loss=0.1747, simple_loss=0.2667, pruned_loss=0.04135, over 3087237.79 frames. ], batch size: 190, lr: 5.16e-03, grad_scale: 8.0 2023-04-29 21:13:25,620 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=131204.0, num_to_drop=1, layers_to_drop={0} 2023-04-29 21:13:59,482 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=131221.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 21:14:27,942 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=131235.0, num_to_drop=1, layers_to_drop={0} 2023-04-29 21:14:59,708 INFO [train.py:904] (1/8) Epoch 13, batch 9450, loss[loss=0.1843, simple_loss=0.2735, pruned_loss=0.04756, over 16817.00 frames. ], tot_loss[loss=0.1761, simple_loss=0.2688, pruned_loss=0.04173, over 3097307.87 frames. ], batch size: 124, lr: 5.16e-03, grad_scale: 8.0 2023-04-29 21:15:00,215 INFO [zipformer.py:625] (1/8) 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] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=131269.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 21:16:03,420 INFO [optim.py:368] (1/8) 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 21:16:40,123 INFO [train.py:904] (1/8) Epoch 13, batch 9500, loss[loss=0.1589, simple_loss=0.2563, pruned_loss=0.03072, over 16185.00 frames. ], tot_loss[loss=0.1753, simple_loss=0.2678, pruned_loss=0.04142, over 3071657.15 frames. ], batch size: 165, lr: 5.16e-03, grad_scale: 8.0 2023-04-29 21:17:44,527 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.47 vs. limit=2.0 2023-04-29 21:18:25,254 INFO [train.py:904] (1/8) Epoch 13, batch 9550, loss[loss=0.1772, simple_loss=0.2774, pruned_loss=0.03849, over 15463.00 frames. ], tot_loss[loss=0.175, simple_loss=0.2674, pruned_loss=0.04132, over 3060499.73 frames. ], batch size: 191, lr: 5.16e-03, grad_scale: 8.0 2023-04-29 21:19:29,575 INFO [optim.py:368] (1/8) 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 21:19:39,583 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.5339, 3.5228, 2.7458, 2.0773, 2.2486, 2.1475, 3.7204, 3.2415], device='cuda:1'), covar=tensor([0.2602, 0.0631, 0.1589, 0.2702, 0.2517, 0.2052, 0.0428, 0.1151], device='cuda:1'), in_proj_covar=tensor([0.0297, 0.0249, 0.0279, 0.0275, 0.0259, 0.0221, 0.0261, 0.0288], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-29 21:19:45,400 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.2473, 4.3087, 2.4768, 4.9260, 3.3608, 4.7666, 2.4132, 3.4550], device='cuda:1'), covar=tensor([0.0173, 0.0210, 0.1527, 0.0125, 0.0598, 0.0305, 0.1502, 0.0521], device='cuda:1'), in_proj_covar=tensor([0.0150, 0.0159, 0.0184, 0.0130, 0.0164, 0.0196, 0.0192, 0.0168], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0003, 0.0004, 0.0004, 0.0003], device='cuda:1') 2023-04-29 21:19:58,232 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.5841, 4.5315, 4.3333, 3.7750, 4.3692, 1.6618, 4.1357, 4.1345], device='cuda:1'), covar=tensor([0.0094, 0.0087, 0.0206, 0.0301, 0.0110, 0.2571, 0.0145, 0.0240], device='cuda:1'), in_proj_covar=tensor([0.0129, 0.0118, 0.0159, 0.0146, 0.0134, 0.0179, 0.0150, 0.0145], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-29 21:20:03,686 INFO [train.py:904] (1/8) Epoch 13, batch 9600, loss[loss=0.1947, simple_loss=0.2739, pruned_loss=0.05769, over 12646.00 frames. ], tot_loss[loss=0.1766, simple_loss=0.2689, pruned_loss=0.04218, over 3062515.87 frames. ], batch size: 247, lr: 5.16e-03, grad_scale: 8.0 2023-04-29 21:20:18,010 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=131409.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 21:20:29,748 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=131416.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 21:21:49,829 INFO [train.py:904] (1/8) Epoch 13, batch 9650, loss[loss=0.1789, simple_loss=0.2734, pruned_loss=0.04217, over 16783.00 frames. ], tot_loss[loss=0.1782, simple_loss=0.2711, pruned_loss=0.04264, over 3067108.91 frames. ], batch size: 76, lr: 5.16e-03, grad_scale: 8.0 2023-04-29 21:22:29,299 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.6424, 3.7704, 2.5727, 2.1862, 2.3434, 2.1471, 3.9460, 3.2721], device='cuda:1'), covar=tensor([0.2860, 0.0751, 0.1937, 0.2624, 0.2788, 0.2191, 0.0459, 0.1118], device='cuda:1'), in_proj_covar=tensor([0.0297, 0.0249, 0.0279, 0.0275, 0.0259, 0.0221, 0.0261, 0.0288], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-29 21:22:33,782 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=131470.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 21:22:47,959 INFO [zipformer.py:625] (1/8) 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] (1/8) 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 21:23:18,214 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 13, batch 9700, loss[loss=0.1666, simple_loss=0.2636, pruned_loss=0.03479, over 16813.00 frames. ], tot_loss[loss=0.1777, simple_loss=0.2704, pruned_loss=0.04253, over 3068649.08 frames. ], batch size: 83, lr: 5.16e-03, grad_scale: 8.0 2023-04-29 21:24:36,246 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 13, batch 9750, loss[loss=0.1764, simple_loss=0.2736, pruned_loss=0.0396, over 16761.00 frames. ], tot_loss[loss=0.1774, simple_loss=0.2692, pruned_loss=0.04276, over 3068512.03 frames. ], batch size: 124, lr: 5.16e-03, grad_scale: 8.0 2023-04-29 21:25:22,392 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=131553.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 21:25:53,996 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.6564, 4.9518, 4.7467, 4.7461, 4.4807, 4.4499, 4.3821, 5.0165], device='cuda:1'), covar=tensor([0.1003, 0.0892, 0.0950, 0.0788, 0.0780, 0.1021, 0.1137, 0.0863], device='cuda:1'), in_proj_covar=tensor([0.0550, 0.0691, 0.0554, 0.0488, 0.0431, 0.0442, 0.0571, 0.0525], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:1') 2023-04-29 21:26:22,842 INFO [optim.py:368] (1/8) 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] (1/8) Epoch 13, batch 9800, loss[loss=0.1911, simple_loss=0.2965, pruned_loss=0.0429, over 16214.00 frames. ], tot_loss[loss=0.1763, simple_loss=0.2692, pruned_loss=0.04167, over 3072760.67 frames. ], batch size: 165, lr: 5.16e-03, grad_scale: 8.0 2023-04-29 21:27:35,039 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=4.74 vs. limit=5.0 2023-04-29 21:28:39,800 INFO [train.py:904] (1/8) Epoch 13, batch 9850, loss[loss=0.1768, simple_loss=0.2691, pruned_loss=0.04219, over 16795.00 frames. ], tot_loss[loss=0.1766, simple_loss=0.2703, pruned_loss=0.04149, over 3066567.11 frames. ], batch size: 124, lr: 5.15e-03, grad_scale: 4.0 2023-04-29 21:29:46,601 INFO [optim.py:368] (1/8) Clipping_scale=2.0, grad-norm quartiles 1.743e+02 2.305e+02 2.925e+02 3.439e+02 6.228e+02, threshold=5.850e+02, percent-clipped=2.0 2023-04-29 21:30:32,565 INFO [train.py:904] (1/8) Epoch 13, batch 9900, loss[loss=0.1719, simple_loss=0.2738, pruned_loss=0.03504, over 15259.00 frames. ], tot_loss[loss=0.1762, simple_loss=0.2703, pruned_loss=0.04108, over 3077424.59 frames. ], batch size: 190, lr: 5.15e-03, grad_scale: 4.0 2023-04-29 21:31:22,544 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.45 vs. limit=2.0 2023-04-29 21:32:30,669 INFO [train.py:904] (1/8) Epoch 13, batch 9950, loss[loss=0.1896, simple_loss=0.2844, pruned_loss=0.04743, over 15340.00 frames. ], tot_loss[loss=0.1782, simple_loss=0.2734, pruned_loss=0.04157, over 3098463.51 frames. ], batch size: 191, lr: 5.15e-03, grad_scale: 4.0 2023-04-29 21:33:02,011 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=131765.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 21:33:20,075 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=131772.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 21:33:47,202 INFO [optim.py:368] (1/8) 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 21:34:31,163 INFO [train.py:904] (1/8) Epoch 13, batch 10000, loss[loss=0.1591, simple_loss=0.261, pruned_loss=0.02858, over 16681.00 frames. ], tot_loss[loss=0.1763, simple_loss=0.2709, pruned_loss=0.0408, over 3110399.30 frames. ], batch size: 89, lr: 5.15e-03, grad_scale: 8.0 2023-04-29 21:35:27,023 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=131830.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 21:36:06,298 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=131848.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 21:36:11,631 INFO [train.py:904] (1/8) Epoch 13, batch 10050, loss[loss=0.1874, simple_loss=0.2847, pruned_loss=0.04505, over 16234.00 frames. ], tot_loss[loss=0.176, simple_loss=0.2707, pruned_loss=0.0407, over 3093003.58 frames. ], batch size: 165, lr: 5.15e-03, grad_scale: 8.0 2023-04-29 21:36:23,363 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=131857.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 21:37:04,201 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=131878.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 21:37:14,089 INFO [optim.py:368] (1/8) 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 21:37:18,277 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.0177, 1.8288, 1.6057, 1.4836, 1.9495, 1.6846, 1.6620, 1.9417], device='cuda:1'), covar=tensor([0.0134, 0.0290, 0.0364, 0.0331, 0.0194, 0.0249, 0.0181, 0.0184], device='cuda:1'), in_proj_covar=tensor([0.0155, 0.0207, 0.0202, 0.0202, 0.0206, 0.0206, 0.0203, 0.0194], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-29 21:37:33,275 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.3005, 3.4450, 3.6771, 3.6519, 3.6485, 3.4738, 3.4753, 3.5169], device='cuda:1'), covar=tensor([0.0435, 0.0802, 0.0421, 0.0443, 0.0481, 0.0531, 0.0730, 0.0468], device='cuda:1'), in_proj_covar=tensor([0.0327, 0.0341, 0.0340, 0.0330, 0.0387, 0.0367, 0.0445, 0.0293], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0002], device='cuda:1') 2023-04-29 21:37:45,822 INFO [train.py:904] (1/8) Epoch 13, batch 10100, loss[loss=0.1712, simple_loss=0.2681, pruned_loss=0.03719, over 16595.00 frames. ], tot_loss[loss=0.1767, simple_loss=0.2711, pruned_loss=0.04112, over 3071518.05 frames. ], batch size: 134, lr: 5.15e-03, grad_scale: 8.0 2023-04-29 21:37:48,560 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.5628, 3.5792, 3.4218, 3.0641, 3.2088, 3.5084, 3.2934, 3.2446], device='cuda:1'), covar=tensor([0.0540, 0.0637, 0.0272, 0.0235, 0.0545, 0.0923, 0.1125, 0.0450], device='cuda:1'), in_proj_covar=tensor([0.0234, 0.0317, 0.0284, 0.0259, 0.0290, 0.0303, 0.0192, 0.0323], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002], device='cuda:1') 2023-04-29 21:37:51,895 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.7716, 3.2591, 3.1189, 1.8629, 2.6812, 1.9413, 3.2509, 3.4373], device='cuda:1'), covar=tensor([0.0291, 0.0614, 0.0631, 0.2125, 0.0942, 0.1184, 0.0733, 0.0760], device='cuda:1'), in_proj_covar=tensor([0.0141, 0.0140, 0.0156, 0.0143, 0.0135, 0.0124, 0.0134, 0.0152], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:1') 2023-04-29 21:38:03,332 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=3.07 vs. limit=5.0 2023-04-29 21:38:16,831 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=131918.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 21:39:29,413 INFO [train.py:904] (1/8) Epoch 14, batch 0, loss[loss=0.2576, simple_loss=0.3085, pruned_loss=0.1033, over 16899.00 frames. ], tot_loss[loss=0.2576, simple_loss=0.3085, pruned_loss=0.1033, over 16899.00 frames. ], batch size: 96, lr: 4.96e-03, grad_scale: 8.0 2023-04-29 21:39:29,413 INFO [train.py:929] (1/8) Computing validation loss 2023-04-29 21:39:36,899 INFO [train.py:938] (1/8) Epoch 14, validation: loss=0.1515, simple_loss=0.2551, pruned_loss=0.024, over 944034.00 frames. 2023-04-29 21:39:36,899 INFO [train.py:939] (1/8) Maximum memory allocated so far is 17845MB 2023-04-29 21:40:22,370 INFO [optim.py:368] (1/8) 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] (1/8) Epoch 14, batch 50, loss[loss=0.1848, simple_loss=0.2817, pruned_loss=0.04389, over 16610.00 frames. ], tot_loss[loss=0.2021, simple_loss=0.2846, pruned_loss=0.05979, over 748944.59 frames. ], batch size: 57, lr: 4.96e-03, grad_scale: 1.0 2023-04-29 21:41:27,140 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 14, batch 100, loss[loss=0.1611, simple_loss=0.2523, pruned_loss=0.03491, over 17206.00 frames. ], tot_loss[loss=0.1985, simple_loss=0.2808, pruned_loss=0.05808, over 1313988.52 frames. ], batch size: 44, lr: 4.96e-03, grad_scale: 1.0 2023-04-29 21:41:58,471 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=132054.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 21:42:04,235 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.25 vs. limit=2.0 2023-04-29 21:42:13,323 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=132065.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 21:42:23,832 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=132072.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 21:42:28,011 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.8510, 3.8726, 4.2705, 4.2463, 4.2768, 3.9413, 3.9951, 3.9272], device='cuda:1'), covar=tensor([0.0377, 0.0547, 0.0377, 0.0395, 0.0428, 0.0422, 0.0798, 0.0495], device='cuda:1'), in_proj_covar=tensor([0.0337, 0.0350, 0.0351, 0.0338, 0.0395, 0.0378, 0.0462, 0.0301], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003], device='cuda:1') 2023-04-29 21:42:44,827 INFO [optim.py:368] (1/8) 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] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=132092.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 21:43:04,249 INFO [train.py:904] (1/8) Epoch 14, batch 150, loss[loss=0.1918, simple_loss=0.2674, pruned_loss=0.05811, over 16859.00 frames. ], tot_loss[loss=0.1937, simple_loss=0.276, pruned_loss=0.05573, over 1763964.27 frames. ], batch size: 116, lr: 4.96e-03, grad_scale: 1.0 2023-04-29 21:43:19,328 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.8725, 2.3411, 2.3748, 4.9043, 2.4105, 2.7846, 2.4395, 2.5920], device='cuda:1'), covar=tensor([0.0905, 0.3576, 0.2630, 0.0295, 0.3753, 0.2385, 0.3126, 0.3396], device='cuda:1'), in_proj_covar=tensor([0.0368, 0.0402, 0.0338, 0.0316, 0.0415, 0.0456, 0.0366, 0.0466], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-29 21:43:20,791 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=132113.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 21:43:23,928 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=132115.0, num_to_drop=1, layers_to_drop={1} 2023-04-29 21:43:30,552 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=132120.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 21:43:33,535 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.56 vs. limit=2.0 2023-04-29 21:43:44,011 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.0718, 2.0857, 2.6459, 3.0628, 2.8831, 3.4139, 2.1727, 3.4938], device='cuda:1'), covar=tensor([0.0178, 0.0411, 0.0261, 0.0212, 0.0240, 0.0160, 0.0395, 0.0111], device='cuda:1'), in_proj_covar=tensor([0.0165, 0.0176, 0.0159, 0.0162, 0.0173, 0.0129, 0.0178, 0.0119], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001], device='cuda:1') 2023-04-29 21:44:07,521 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.9669, 3.7863, 4.0229, 4.1501, 4.2168, 3.8214, 4.0434, 4.2441], device='cuda:1'), covar=tensor([0.1401, 0.1014, 0.1129, 0.0636, 0.0544, 0.1434, 0.1705, 0.0575], device='cuda:1'), in_proj_covar=tensor([0.0543, 0.0666, 0.0799, 0.0682, 0.0515, 0.0523, 0.0545, 0.0628], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-29 21:44:08,568 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=132148.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 21:44:15,118 INFO [train.py:904] (1/8) Epoch 14, batch 200, loss[loss=0.1882, simple_loss=0.2656, pruned_loss=0.05547, over 16782.00 frames. ], tot_loss[loss=0.1924, simple_loss=0.2746, pruned_loss=0.05515, over 2108263.80 frames. ], batch size: 102, lr: 4.96e-03, grad_scale: 1.0 2023-04-29 21:44:15,680 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.0442, 1.9230, 2.5818, 3.0231, 2.7360, 3.3389, 2.1266, 3.3745], device='cuda:1'), covar=tensor([0.0197, 0.0438, 0.0278, 0.0225, 0.0258, 0.0175, 0.0417, 0.0115], device='cuda:1'), in_proj_covar=tensor([0.0165, 0.0177, 0.0159, 0.0162, 0.0173, 0.0129, 0.0178, 0.0119], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001], device='cuda:1') 2023-04-29 21:44:27,949 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.43 vs. limit=2.0 2023-04-29 21:45:02,143 INFO [optim.py:368] (1/8) 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] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=132196.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 21:45:21,428 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.28 vs. limit=2.0 2023-04-29 21:45:23,095 INFO [train.py:904] (1/8) Epoch 14, batch 250, loss[loss=0.1703, simple_loss=0.2635, pruned_loss=0.03848, over 17145.00 frames. ], tot_loss[loss=0.1913, simple_loss=0.2731, pruned_loss=0.05473, over 2369452.41 frames. ], batch size: 48, lr: 4.96e-03, grad_scale: 1.0 2023-04-29 21:45:38,784 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=132213.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 21:46:33,334 INFO [train.py:904] (1/8) Epoch 14, batch 300, loss[loss=0.1856, simple_loss=0.2526, pruned_loss=0.05929, over 16864.00 frames. ], tot_loss[loss=0.1882, simple_loss=0.2701, pruned_loss=0.05316, over 2571471.51 frames. ], batch size: 96, lr: 4.96e-03, grad_scale: 1.0 2023-04-29 21:46:53,648 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.9197, 4.8674, 5.3589, 5.3402, 5.3725, 5.0134, 5.0246, 4.7008], device='cuda:1'), covar=tensor([0.0308, 0.0475, 0.0477, 0.0460, 0.0479, 0.0389, 0.0871, 0.0408], device='cuda:1'), in_proj_covar=tensor([0.0348, 0.0363, 0.0364, 0.0350, 0.0406, 0.0390, 0.0478, 0.0311], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003], device='cuda:1') 2023-04-29 21:47:06,413 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.8285, 4.0468, 3.0450, 2.3757, 2.6833, 2.4428, 4.2600, 3.5975], device='cuda:1'), covar=tensor([0.2490, 0.0631, 0.1662, 0.2415, 0.2399, 0.1852, 0.0426, 0.1187], device='cuda:1'), in_proj_covar=tensor([0.0302, 0.0255, 0.0285, 0.0281, 0.0269, 0.0226, 0.0268, 0.0299], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-29 21:47:22,675 INFO [optim.py:368] (1/8) 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 21:47:38,656 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.1843, 1.9882, 2.2314, 3.7090, 2.0558, 2.3568, 2.1100, 2.1621], device='cuda:1'), covar=tensor([0.1122, 0.3467, 0.2450, 0.0532, 0.3665, 0.2309, 0.3333, 0.3192], device='cuda:1'), in_proj_covar=tensor([0.0369, 0.0403, 0.0339, 0.0318, 0.0417, 0.0460, 0.0368, 0.0470], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-29 21:47:43,455 INFO [train.py:904] (1/8) Epoch 14, batch 350, loss[loss=0.1572, simple_loss=0.244, pruned_loss=0.03521, over 16821.00 frames. ], tot_loss[loss=0.184, simple_loss=0.2665, pruned_loss=0.05072, over 2745350.51 frames. ], batch size: 42, lr: 4.95e-03, grad_scale: 1.0 2023-04-29 21:47:50,355 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.5417, 3.5398, 3.4967, 3.0173, 3.4556, 2.0107, 3.2296, 2.9151], device='cuda:1'), covar=tensor([0.0143, 0.0104, 0.0144, 0.0223, 0.0084, 0.2187, 0.0133, 0.0197], device='cuda:1'), in_proj_covar=tensor([0.0138, 0.0126, 0.0168, 0.0155, 0.0141, 0.0189, 0.0159, 0.0155], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-29 21:47:55,958 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.7599, 2.1884, 2.2660, 4.7153, 2.2461, 2.7061, 2.2805, 2.3995], device='cuda:1'), covar=tensor([0.0915, 0.3582, 0.2585, 0.0314, 0.3842, 0.2512, 0.3339, 0.3567], device='cuda:1'), in_proj_covar=tensor([0.0370, 0.0403, 0.0340, 0.0318, 0.0417, 0.0461, 0.0368, 0.0470], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-29 21:48:51,146 INFO [train.py:904] (1/8) Epoch 14, batch 400, loss[loss=0.1622, simple_loss=0.248, pruned_loss=0.03824, over 17164.00 frames. ], tot_loss[loss=0.1824, simple_loss=0.2648, pruned_loss=0.04997, over 2866077.35 frames. ], batch size: 46, lr: 4.95e-03, grad_scale: 2.0 2023-04-29 21:49:26,842 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.20 vs. limit=2.0 2023-04-29 21:49:38,425 INFO [optim.py:368] (1/8) 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 21:49:39,468 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=132387.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 21:50:00,122 INFO [train.py:904] (1/8) Epoch 14, batch 450, loss[loss=0.2, simple_loss=0.2705, pruned_loss=0.06473, over 16905.00 frames. ], tot_loss[loss=0.18, simple_loss=0.2629, pruned_loss=0.04858, over 2968992.45 frames. ], batch size: 109, lr: 4.95e-03, grad_scale: 2.0 2023-04-29 21:50:11,960 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=132410.0, num_to_drop=1, layers_to_drop={2} 2023-04-29 21:50:35,471 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.0069, 2.5548, 2.0269, 2.3938, 2.9527, 2.7659, 3.1169, 3.0530], device='cuda:1'), covar=tensor([0.0183, 0.0297, 0.0423, 0.0352, 0.0183, 0.0259, 0.0214, 0.0205], device='cuda:1'), in_proj_covar=tensor([0.0167, 0.0216, 0.0210, 0.0209, 0.0214, 0.0215, 0.0218, 0.0205], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-29 21:50:55,009 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=4.60 vs. limit=5.0 2023-04-29 21:51:11,603 INFO [train.py:904] (1/8) Epoch 14, batch 500, loss[loss=0.1885, simple_loss=0.284, pruned_loss=0.04654, over 17037.00 frames. ], tot_loss[loss=0.1785, simple_loss=0.2614, pruned_loss=0.04783, over 3049610.45 frames. ], batch size: 55, lr: 4.95e-03, grad_scale: 2.0 2023-04-29 21:51:58,582 INFO [optim.py:368] (1/8) 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] (1/8) Epoch 14, batch 550, loss[loss=0.1988, simple_loss=0.2721, pruned_loss=0.06282, over 16813.00 frames. ], tot_loss[loss=0.1781, simple_loss=0.2613, pruned_loss=0.04744, over 3121718.60 frames. ], batch size: 102, lr: 4.95e-03, grad_scale: 2.0 2023-04-29 21:52:34,268 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=132513.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 21:53:28,147 INFO [train.py:904] (1/8) Epoch 14, batch 600, loss[loss=0.1712, simple_loss=0.2408, pruned_loss=0.05082, over 16758.00 frames. ], tot_loss[loss=0.1776, simple_loss=0.2605, pruned_loss=0.04731, over 3169629.93 frames. ], batch size: 83, lr: 4.95e-03, grad_scale: 2.0 2023-04-29 21:53:41,854 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=132561.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 21:54:17,820 INFO [optim.py:368] (1/8) 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 21:54:39,140 INFO [train.py:904] (1/8) Epoch 14, batch 650, loss[loss=0.1553, simple_loss=0.233, pruned_loss=0.0388, over 16786.00 frames. ], tot_loss[loss=0.1758, simple_loss=0.2588, pruned_loss=0.04646, over 3214401.87 frames. ], batch size: 102, lr: 4.95e-03, grad_scale: 2.0 2023-04-29 21:55:49,510 INFO [train.py:904] (1/8) Epoch 14, batch 700, loss[loss=0.1747, simple_loss=0.252, pruned_loss=0.04871, over 16757.00 frames. ], tot_loss[loss=0.1753, simple_loss=0.2585, pruned_loss=0.0461, over 3238856.03 frames. ], batch size: 89, lr: 4.95e-03, grad_scale: 2.0 2023-04-29 21:56:01,350 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=132660.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 21:56:38,039 INFO [optim.py:368] (1/8) 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 21:56:38,482 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 14, batch 750, loss[loss=0.1487, simple_loss=0.2312, pruned_loss=0.03314, over 16779.00 frames. ], tot_loss[loss=0.1752, simple_loss=0.2585, pruned_loss=0.04597, over 3240529.57 frames. ], batch size: 39, lr: 4.95e-03, grad_scale: 2.0 2023-04-29 21:57:09,168 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=132710.0, num_to_drop=1, layers_to_drop={1} 2023-04-29 21:57:24,375 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=132721.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 21:57:43,137 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=132735.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 21:57:59,758 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 14, batch 800, loss[loss=0.1829, simple_loss=0.2714, pruned_loss=0.0472, over 16613.00 frames. ], tot_loss[loss=0.1753, simple_loss=0.2584, pruned_loss=0.04608, over 3253581.79 frames. ], batch size: 62, lr: 4.95e-03, grad_scale: 4.0 2023-04-29 21:58:16,058 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=132758.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 21:58:51,788 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.8883, 5.2015, 5.2880, 5.1715, 5.0896, 5.7575, 5.2576, 4.9695], device='cuda:1'), covar=tensor([0.1094, 0.1865, 0.2259, 0.1883, 0.3032, 0.1035, 0.1542, 0.2517], device='cuda:1'), in_proj_covar=tensor([0.0370, 0.0525, 0.0578, 0.0443, 0.0604, 0.0601, 0.0457, 0.0597], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-04-29 21:58:54,909 INFO [optim.py:368] (1/8) 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] (1/8) Epoch 14, batch 850, loss[loss=0.1761, simple_loss=0.2521, pruned_loss=0.05004, over 16925.00 frames. ], tot_loss[loss=0.1748, simple_loss=0.2576, pruned_loss=0.04605, over 3269649.77 frames. ], batch size: 109, lr: 4.94e-03, grad_scale: 4.0 2023-04-29 21:59:23,472 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=132807.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 22:00:07,619 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.38 vs. limit=2.0 2023-04-29 22:00:24,478 INFO [train.py:904] (1/8) Epoch 14, batch 900, loss[loss=0.1813, simple_loss=0.2554, pruned_loss=0.05362, over 16881.00 frames. ], tot_loss[loss=0.1747, simple_loss=0.2575, pruned_loss=0.046, over 3285470.11 frames. ], batch size: 96, lr: 4.94e-03, grad_scale: 4.0 2023-04-29 22:01:14,371 INFO [optim.py:368] (1/8) 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 22:01:34,127 INFO [train.py:904] (1/8) Epoch 14, batch 950, loss[loss=0.1833, simple_loss=0.2523, pruned_loss=0.05711, over 12053.00 frames. ], tot_loss[loss=0.175, simple_loss=0.2573, pruned_loss=0.04637, over 3287988.10 frames. ], batch size: 248, lr: 4.94e-03, grad_scale: 4.0 2023-04-29 22:02:33,231 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.5863, 4.5601, 4.5126, 4.0079, 4.5247, 1.7334, 4.2671, 4.1984], device='cuda:1'), covar=tensor([0.0112, 0.0095, 0.0148, 0.0315, 0.0088, 0.2437, 0.0137, 0.0176], device='cuda:1'), in_proj_covar=tensor([0.0143, 0.0131, 0.0176, 0.0162, 0.0147, 0.0192, 0.0166, 0.0161], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-29 22:02:41,755 INFO [train.py:904] (1/8) Epoch 14, batch 1000, loss[loss=0.1467, simple_loss=0.2316, pruned_loss=0.03088, over 16778.00 frames. ], tot_loss[loss=0.1747, simple_loss=0.2567, pruned_loss=0.04638, over 3296649.83 frames. ], batch size: 39, lr: 4.94e-03, grad_scale: 4.0 2023-04-29 22:03:29,371 INFO [optim.py:368] (1/8) 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 22:03:48,018 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([5.1816, 5.6131, 5.1494, 5.6310, 5.1615, 4.8288, 5.1644, 5.7309], device='cuda:1'), covar=tensor([0.2323, 0.1982, 0.2655, 0.1289, 0.1531, 0.1405, 0.2246, 0.1991], device='cuda:1'), in_proj_covar=tensor([0.0598, 0.0747, 0.0607, 0.0531, 0.0476, 0.0477, 0.0625, 0.0571], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:1') 2023-04-29 22:03:50,011 INFO [train.py:904] (1/8) Epoch 14, batch 1050, loss[loss=0.1952, simple_loss=0.263, pruned_loss=0.06368, over 16753.00 frames. ], tot_loss[loss=0.175, simple_loss=0.2572, pruned_loss=0.04642, over 3311350.49 frames. ], batch size: 124, lr: 4.94e-03, grad_scale: 4.0 2023-04-29 22:04:06,809 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([5.3672, 5.3193, 5.1240, 4.5657, 5.1552, 1.9588, 4.9041, 5.0778], device='cuda:1'), covar=tensor([0.0078, 0.0079, 0.0152, 0.0352, 0.0082, 0.2461, 0.0123, 0.0150], device='cuda:1'), in_proj_covar=tensor([0.0143, 0.0132, 0.0177, 0.0164, 0.0148, 0.0193, 0.0167, 0.0162], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-29 22:04:10,796 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=133016.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 22:04:59,791 INFO [train.py:904] (1/8) Epoch 14, batch 1100, loss[loss=0.1739, simple_loss=0.2669, pruned_loss=0.04045, over 17076.00 frames. ], tot_loss[loss=0.174, simple_loss=0.2566, pruned_loss=0.04567, over 3300548.11 frames. ], batch size: 55, lr: 4.94e-03, grad_scale: 4.0 2023-04-29 22:05:47,408 INFO [optim.py:368] (1/8) 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 22:06:08,352 INFO [train.py:904] (1/8) Epoch 14, batch 1150, loss[loss=0.1808, simple_loss=0.2633, pruned_loss=0.04914, over 16819.00 frames. ], tot_loss[loss=0.1734, simple_loss=0.2562, pruned_loss=0.0453, over 3312510.79 frames. ], batch size: 102, lr: 4.94e-03, grad_scale: 4.0 2023-04-29 22:06:08,687 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=133102.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 22:06:25,336 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.18 vs. limit=2.0 2023-04-29 22:07:03,965 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([1.6129, 1.7439, 1.5798, 1.5033, 1.8669, 1.6084, 1.6183, 1.8746], device='cuda:1'), covar=tensor([0.0184, 0.0246, 0.0353, 0.0325, 0.0177, 0.0250, 0.0175, 0.0179], device='cuda:1'), in_proj_covar=tensor([0.0173, 0.0218, 0.0212, 0.0211, 0.0218, 0.0220, 0.0224, 0.0211], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-29 22:07:16,496 INFO [train.py:904] (1/8) Epoch 14, batch 1200, loss[loss=0.2008, simple_loss=0.2914, pruned_loss=0.05514, over 17048.00 frames. ], tot_loss[loss=0.1731, simple_loss=0.2557, pruned_loss=0.04525, over 3308120.59 frames. ], batch size: 55, lr: 4.94e-03, grad_scale: 8.0 2023-04-29 22:08:05,977 INFO [optim.py:368] (1/8) 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 22:08:28,179 INFO [train.py:904] (1/8) Epoch 14, batch 1250, loss[loss=0.1541, simple_loss=0.2519, pruned_loss=0.02815, over 17128.00 frames. ], tot_loss[loss=0.1739, simple_loss=0.2562, pruned_loss=0.04578, over 3311359.43 frames. ], batch size: 48, lr: 4.94e-03, grad_scale: 8.0 2023-04-29 22:09:33,453 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.1106, 2.0873, 2.4585, 2.9910, 2.8063, 3.3860, 2.3469, 3.3065], device='cuda:1'), covar=tensor([0.0178, 0.0394, 0.0287, 0.0227, 0.0245, 0.0159, 0.0358, 0.0168], device='cuda:1'), in_proj_covar=tensor([0.0169, 0.0178, 0.0163, 0.0166, 0.0175, 0.0133, 0.0178, 0.0123], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-29 22:09:37,992 INFO [train.py:904] (1/8) Epoch 14, batch 1300, loss[loss=0.1901, simple_loss=0.2626, pruned_loss=0.05878, over 16884.00 frames. ], tot_loss[loss=0.1738, simple_loss=0.2558, pruned_loss=0.04587, over 3302695.72 frames. ], batch size: 96, lr: 4.94e-03, grad_scale: 8.0 2023-04-29 22:10:27,131 INFO [optim.py:368] (1/8) 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] (1/8) Epoch 14, batch 1350, loss[loss=0.1903, simple_loss=0.2608, pruned_loss=0.05987, over 16572.00 frames. ], tot_loss[loss=0.1736, simple_loss=0.2559, pruned_loss=0.04563, over 3304634.61 frames. ], batch size: 68, lr: 4.94e-03, grad_scale: 8.0 2023-04-29 22:11:06,797 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=133316.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 22:11:29,382 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.2199, 4.0287, 4.2283, 4.3943, 4.4764, 4.0147, 4.2233, 4.4370], device='cuda:1'), covar=tensor([0.1289, 0.0972, 0.1206, 0.0591, 0.0530, 0.1305, 0.2080, 0.0638], device='cuda:1'), in_proj_covar=tensor([0.0581, 0.0717, 0.0861, 0.0735, 0.0552, 0.0561, 0.0577, 0.0676], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-29 22:11:33,356 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([5.0745, 5.0678, 5.5873, 5.6082, 5.5873, 5.2190, 5.1859, 4.9092], device='cuda:1'), covar=tensor([0.0278, 0.0446, 0.0333, 0.0352, 0.0421, 0.0367, 0.0937, 0.0393], device='cuda:1'), in_proj_covar=tensor([0.0370, 0.0389, 0.0388, 0.0370, 0.0431, 0.0415, 0.0506, 0.0329], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:1') 2023-04-29 22:11:56,451 INFO [train.py:904] (1/8) Epoch 14, batch 1400, loss[loss=0.1677, simple_loss=0.2658, pruned_loss=0.03483, over 17047.00 frames. ], tot_loss[loss=0.1734, simple_loss=0.2559, pruned_loss=0.04548, over 3309403.44 frames. ], batch size: 55, lr: 4.93e-03, grad_scale: 8.0 2023-04-29 22:12:12,798 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=133364.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 22:12:26,467 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=133374.0, num_to_drop=1, layers_to_drop={0} 2023-04-29 22:12:29,373 INFO [zipformer.py:625] (1/8) 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] (1/8) 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] (1/8) Epoch 14, batch 1450, loss[loss=0.1701, simple_loss=0.2452, pruned_loss=0.04746, over 16375.00 frames. ], tot_loss[loss=0.1733, simple_loss=0.2559, pruned_loss=0.04538, over 3306044.87 frames. ], batch size: 146, lr: 4.93e-03, grad_scale: 8.0 2023-04-29 22:13:06,529 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=133402.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 22:13:38,376 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.62 vs. limit=2.0 2023-04-29 22:13:52,409 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=133435.0, num_to_drop=1, layers_to_drop={3} 2023-04-29 22:13:55,206 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=133437.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 22:14:12,862 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 14, batch 1500, loss[loss=0.2034, simple_loss=0.271, pruned_loss=0.06789, over 16637.00 frames. ], tot_loss[loss=0.1735, simple_loss=0.2563, pruned_loss=0.04538, over 3313874.68 frames. ], batch size: 68, lr: 4.93e-03, grad_scale: 8.0 2023-04-29 22:14:46,584 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=133474.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 22:15:03,756 INFO [optim.py:368] (1/8) 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 22:15:23,303 INFO [train.py:904] (1/8) Epoch 14, batch 1550, loss[loss=0.1746, simple_loss=0.2628, pruned_loss=0.04317, over 16630.00 frames. ], tot_loss[loss=0.1763, simple_loss=0.258, pruned_loss=0.0473, over 3315096.03 frames. ], batch size: 62, lr: 4.93e-03, grad_scale: 8.0 2023-04-29 22:16:10,857 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=133535.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 22:16:33,568 INFO [train.py:904] (1/8) Epoch 14, batch 1600, loss[loss=0.2001, simple_loss=0.2808, pruned_loss=0.05973, over 16806.00 frames. ], tot_loss[loss=0.1779, simple_loss=0.2598, pruned_loss=0.04798, over 3303003.11 frames. ], batch size: 83, lr: 4.93e-03, grad_scale: 8.0 2023-04-29 22:17:21,877 INFO [optim.py:368] (1/8) 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 22:17:43,026 INFO [train.py:904] (1/8) Epoch 14, batch 1650, loss[loss=0.174, simple_loss=0.2609, pruned_loss=0.0436, over 17205.00 frames. ], tot_loss[loss=0.1793, simple_loss=0.2615, pruned_loss=0.04857, over 3306196.77 frames. ], batch size: 46, lr: 4.93e-03, grad_scale: 8.0 2023-04-29 22:18:12,358 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.23 vs. limit=2.0 2023-04-29 22:18:17,562 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.8487, 5.1819, 4.9674, 4.9375, 4.7272, 4.5889, 4.6614, 5.2693], device='cuda:1'), covar=tensor([0.1208, 0.0949, 0.1030, 0.0784, 0.0772, 0.1054, 0.1075, 0.0955], device='cuda:1'), in_proj_covar=tensor([0.0599, 0.0752, 0.0606, 0.0535, 0.0477, 0.0479, 0.0628, 0.0574], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:1') 2023-04-29 22:18:52,510 INFO [train.py:904] (1/8) Epoch 14, batch 1700, loss[loss=0.2321, simple_loss=0.3096, pruned_loss=0.07733, over 12357.00 frames. ], tot_loss[loss=0.1813, simple_loss=0.2636, pruned_loss=0.04953, over 3300315.70 frames. ], batch size: 248, lr: 4.93e-03, grad_scale: 8.0 2023-04-29 22:18:53,060 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=133652.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 22:19:42,412 INFO [optim.py:368] (1/8) 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 22:19:48,099 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([1.7086, 2.8549, 2.5089, 4.7124, 3.8302, 4.1823, 1.5230, 3.0720], device='cuda:1'), covar=tensor([0.1379, 0.0723, 0.1195, 0.0226, 0.0256, 0.0399, 0.1552, 0.0791], device='cuda:1'), in_proj_covar=tensor([0.0157, 0.0163, 0.0181, 0.0160, 0.0197, 0.0209, 0.0186, 0.0184], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-04-29 22:20:03,031 INFO [train.py:904] (1/8) Epoch 14, batch 1750, loss[loss=0.1892, simple_loss=0.2783, pruned_loss=0.05, over 16600.00 frames. ], tot_loss[loss=0.1811, simple_loss=0.2641, pruned_loss=0.04903, over 3306631.84 frames. ], batch size: 68, lr: 4.93e-03, grad_scale: 8.0 2023-04-29 22:20:08,926 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.69 vs. limit=2.0 2023-04-29 22:20:18,788 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.8662, 4.5816, 4.9155, 5.1032, 5.2929, 4.6613, 5.2943, 5.2931], device='cuda:1'), covar=tensor([0.1870, 0.1394, 0.1779, 0.0798, 0.0581, 0.0822, 0.0542, 0.0604], device='cuda:1'), in_proj_covar=tensor([0.0591, 0.0731, 0.0877, 0.0747, 0.0562, 0.0577, 0.0587, 0.0690], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-29 22:20:18,853 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=133713.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 22:20:37,192 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.9973, 4.9764, 5.4871, 5.4554, 5.4815, 5.0693, 5.0709, 4.7725], device='cuda:1'), covar=tensor([0.0305, 0.0485, 0.0366, 0.0400, 0.0398, 0.0377, 0.0927, 0.0444], device='cuda:1'), in_proj_covar=tensor([0.0374, 0.0391, 0.0391, 0.0373, 0.0433, 0.0418, 0.0511, 0.0332], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:1') 2023-04-29 22:20:42,468 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=133730.0, num_to_drop=1, layers_to_drop={2} 2023-04-29 22:20:44,901 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 14, batch 1800, loss[loss=0.1444, simple_loss=0.2237, pruned_loss=0.03256, over 16743.00 frames. ], tot_loss[loss=0.1799, simple_loss=0.2637, pruned_loss=0.04801, over 3317513.12 frames. ], batch size: 39, lr: 4.93e-03, grad_scale: 8.0 2023-04-29 22:21:37,189 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=4.75 vs. limit=5.0 2023-04-29 22:22:00,849 INFO [optim.py:368] (1/8) 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 22:22:22,390 INFO [train.py:904] (1/8) Epoch 14, batch 1850, loss[loss=0.1738, simple_loss=0.2706, pruned_loss=0.03848, over 17028.00 frames. ], tot_loss[loss=0.1802, simple_loss=0.2645, pruned_loss=0.04793, over 3323343.26 frames. ], batch size: 50, lr: 4.93e-03, grad_scale: 4.0 2023-04-29 22:23:02,958 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 14, batch 1900, loss[loss=0.1585, simple_loss=0.2456, pruned_loss=0.03569, over 17230.00 frames. ], tot_loss[loss=0.1791, simple_loss=0.2636, pruned_loss=0.04734, over 3326378.46 frames. ], batch size: 45, lr: 4.93e-03, grad_scale: 4.0 2023-04-29 22:23:44,604 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.7264, 4.6490, 4.6132, 4.3296, 4.3363, 4.6857, 4.4491, 4.3871], device='cuda:1'), covar=tensor([0.0578, 0.0565, 0.0240, 0.0250, 0.0749, 0.0432, 0.0469, 0.0599], device='cuda:1'), in_proj_covar=tensor([0.0271, 0.0372, 0.0328, 0.0308, 0.0343, 0.0357, 0.0222, 0.0382], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-04-29 22:24:15,922 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([5.0408, 4.9135, 4.9160, 4.5039, 4.5373, 4.9311, 4.8744, 4.5684], device='cuda:1'), covar=tensor([0.0575, 0.0529, 0.0270, 0.0297, 0.1038, 0.0393, 0.0377, 0.0754], device='cuda:1'), in_proj_covar=tensor([0.0272, 0.0373, 0.0328, 0.0308, 0.0344, 0.0358, 0.0222, 0.0383], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-04-29 22:24:23,480 INFO [optim.py:368] (1/8) 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 22:24:42,152 INFO [train.py:904] (1/8) Epoch 14, batch 1950, loss[loss=0.2265, simple_loss=0.3027, pruned_loss=0.07517, over 16279.00 frames. ], tot_loss[loss=0.1783, simple_loss=0.2635, pruned_loss=0.04662, over 3325194.89 frames. ], batch size: 165, lr: 4.92e-03, grad_scale: 4.0 2023-04-29 22:25:27,232 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.51 vs. limit=2.0 2023-04-29 22:25:52,905 INFO [train.py:904] (1/8) Epoch 14, batch 2000, loss[loss=0.2289, simple_loss=0.2961, pruned_loss=0.08088, over 16796.00 frames. ], tot_loss[loss=0.1773, simple_loss=0.2623, pruned_loss=0.04608, over 3328901.77 frames. ], batch size: 83, lr: 4.92e-03, grad_scale: 8.0 2023-04-29 22:25:53,851 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.61 vs. limit=2.0 2023-04-29 22:26:18,089 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.40 vs. limit=2.0 2023-04-29 22:26:29,257 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=3.64 vs. limit=5.0 2023-04-29 22:26:43,618 INFO [optim.py:368] (1/8) 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 22:26:45,876 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.9871, 3.8073, 4.0730, 4.1998, 4.2432, 3.8474, 4.0043, 4.2553], device='cuda:1'), covar=tensor([0.1341, 0.1012, 0.1102, 0.0605, 0.0533, 0.1573, 0.1804, 0.0619], device='cuda:1'), in_proj_covar=tensor([0.0597, 0.0739, 0.0887, 0.0757, 0.0568, 0.0586, 0.0597, 0.0700], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-29 22:26:57,046 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 14, batch 2050, loss[loss=0.1909, simple_loss=0.2811, pruned_loss=0.05037, over 17072.00 frames. ], tot_loss[loss=0.1774, simple_loss=0.2622, pruned_loss=0.04626, over 3326120.61 frames. ], batch size: 53, lr: 4.92e-03, grad_scale: 8.0 2023-04-29 22:27:15,545 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=134008.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 22:27:21,571 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.3234, 4.2723, 4.2256, 3.6937, 4.2533, 1.7407, 4.0192, 3.8209], device='cuda:1'), covar=tensor([0.0124, 0.0109, 0.0166, 0.0305, 0.0096, 0.2605, 0.0139, 0.0195], device='cuda:1'), in_proj_covar=tensor([0.0148, 0.0136, 0.0184, 0.0170, 0.0154, 0.0196, 0.0173, 0.0167], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-29 22:27:46,424 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=134030.0, num_to_drop=1, layers_to_drop={0} 2023-04-29 22:27:49,330 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 14, batch 2100, loss[loss=0.2133, simple_loss=0.2883, pruned_loss=0.06915, over 16502.00 frames. ], tot_loss[loss=0.1786, simple_loss=0.2632, pruned_loss=0.047, over 3320475.22 frames. ], batch size: 75, lr: 4.92e-03, grad_scale: 8.0 2023-04-29 22:28:25,605 INFO [zipformer.py:625] (1/8) 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] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=134078.0, num_to_drop=1, layers_to_drop={0} 2023-04-29 22:28:56,309 INFO [zipformer.py:625] (1/8) 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] (1/8) 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 22:29:26,681 INFO [train.py:904] (1/8) Epoch 14, batch 2150, loss[loss=0.1732, simple_loss=0.2676, pruned_loss=0.03936, over 17091.00 frames. ], tot_loss[loss=0.1791, simple_loss=0.2639, pruned_loss=0.04715, over 3323820.28 frames. ], batch size: 47, lr: 4.92e-03, grad_scale: 4.0 2023-04-29 22:29:59,855 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.19 vs. limit=2.0 2023-04-29 22:30:05,071 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([1.9526, 3.4971, 3.0489, 5.2268, 4.5145, 4.6192, 1.7841, 3.3537], device='cuda:1'), covar=tensor([0.1260, 0.0544, 0.0953, 0.0155, 0.0181, 0.0362, 0.1401, 0.0661], device='cuda:1'), in_proj_covar=tensor([0.0157, 0.0163, 0.0181, 0.0161, 0.0198, 0.0210, 0.0186, 0.0183], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-04-29 22:30:06,145 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=134130.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 22:30:30,223 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.8590, 2.3276, 2.4383, 4.6547, 2.2888, 2.7360, 2.3829, 2.5559], device='cuda:1'), covar=tensor([0.1035, 0.3564, 0.2639, 0.0395, 0.4031, 0.2427, 0.3213, 0.3571], device='cuda:1'), in_proj_covar=tensor([0.0380, 0.0412, 0.0345, 0.0328, 0.0421, 0.0476, 0.0378, 0.0484], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-29 22:30:35,960 INFO [train.py:904] (1/8) Epoch 14, batch 2200, loss[loss=0.1653, simple_loss=0.2571, pruned_loss=0.03672, over 17199.00 frames. ], tot_loss[loss=0.1795, simple_loss=0.2644, pruned_loss=0.0473, over 3329838.45 frames. ], batch size: 45, lr: 4.92e-03, grad_scale: 4.0 2023-04-29 22:31:12,755 INFO [zipformer.py:625] (1/8) 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] (1/8) 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 22:31:41,790 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.62 vs. limit=2.0 2023-04-29 22:31:45,970 INFO [train.py:904] (1/8) Epoch 14, batch 2250, loss[loss=0.1577, simple_loss=0.2489, pruned_loss=0.03325, over 17261.00 frames. ], tot_loss[loss=0.1808, simple_loss=0.2658, pruned_loss=0.04793, over 3328601.48 frames. ], batch size: 45, lr: 4.92e-03, grad_scale: 4.0 2023-04-29 22:32:53,953 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.5499, 4.3571, 4.5922, 4.7565, 4.9088, 4.4393, 4.7503, 4.8756], device='cuda:1'), covar=tensor([0.1566, 0.1177, 0.1411, 0.0680, 0.0551, 0.0976, 0.1597, 0.0624], device='cuda:1'), in_proj_covar=tensor([0.0601, 0.0743, 0.0890, 0.0764, 0.0572, 0.0588, 0.0602, 0.0702], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-29 22:32:56,451 INFO [train.py:904] (1/8) Epoch 14, batch 2300, loss[loss=0.2347, simple_loss=0.3027, pruned_loss=0.08331, over 11989.00 frames. ], tot_loss[loss=0.1815, simple_loss=0.266, pruned_loss=0.0485, over 3315098.80 frames. ], batch size: 247, lr: 4.92e-03, grad_scale: 4.0 2023-04-29 22:33:46,608 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.5474, 3.2618, 3.6000, 1.8131, 3.6982, 3.6707, 3.0091, 2.7681], device='cuda:1'), covar=tensor([0.0669, 0.0201, 0.0146, 0.1144, 0.0085, 0.0177, 0.0378, 0.0410], device='cuda:1'), in_proj_covar=tensor([0.0143, 0.0102, 0.0089, 0.0136, 0.0070, 0.0112, 0.0121, 0.0126], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:1') 2023-04-29 22:33:48,101 INFO [optim.py:368] (1/8) Clipping_scale=2.0, grad-norm quartiles 1.417e+02 2.438e+02 2.937e+02 3.292e+02 7.337e+02, threshold=5.875e+02, percent-clipped=1.0 2023-04-29 22:33:57,597 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=4.94 vs. limit=5.0 2023-04-29 22:33:58,803 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.32 vs. limit=2.0 2023-04-29 22:34:06,356 INFO [train.py:904] (1/8) Epoch 14, batch 2350, loss[loss=0.1532, simple_loss=0.2418, pruned_loss=0.03233, over 16939.00 frames. ], tot_loss[loss=0.1815, simple_loss=0.2659, pruned_loss=0.04855, over 3309285.23 frames. ], batch size: 41, lr: 4.92e-03, grad_scale: 4.0 2023-04-29 22:34:09,739 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=134304.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 22:34:15,101 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=134308.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 22:34:16,979 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=134309.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 22:34:38,262 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.3393, 4.6314, 4.4279, 4.4668, 4.1642, 4.0682, 4.1711, 4.6816], device='cuda:1'), covar=tensor([0.1120, 0.0934, 0.1022, 0.0736, 0.0805, 0.1448, 0.0996, 0.0915], device='cuda:1'), in_proj_covar=tensor([0.0600, 0.0752, 0.0611, 0.0538, 0.0477, 0.0477, 0.0625, 0.0577], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:1') 2023-04-29 22:35:17,014 INFO [train.py:904] (1/8) Epoch 14, batch 2400, loss[loss=0.1854, simple_loss=0.2756, pruned_loss=0.04757, over 16127.00 frames. ], tot_loss[loss=0.1822, simple_loss=0.2665, pruned_loss=0.04889, over 3312159.76 frames. ], batch size: 35, lr: 4.92e-03, grad_scale: 8.0 2023-04-29 22:35:19,256 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=134353.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 22:35:23,353 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=134356.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 22:35:35,560 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=134365.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 22:35:42,568 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=134370.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 22:35:46,753 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.7737, 2.7685, 2.2505, 2.5926, 3.0653, 2.9330, 3.5117, 3.3187], device='cuda:1'), covar=tensor([0.0100, 0.0309, 0.0393, 0.0343, 0.0223, 0.0287, 0.0197, 0.0205], device='cuda:1'), in_proj_covar=tensor([0.0179, 0.0222, 0.0215, 0.0214, 0.0222, 0.0223, 0.0230, 0.0216], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-29 22:36:08,842 INFO [optim.py:368] (1/8) 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] (1/8) Epoch 14, batch 2450, loss[loss=0.1769, simple_loss=0.2842, pruned_loss=0.03485, over 17313.00 frames. ], tot_loss[loss=0.1814, simple_loss=0.267, pruned_loss=0.04792, over 3325982.34 frames. ], batch size: 52, lr: 4.92e-03, grad_scale: 4.0 2023-04-29 22:37:28,772 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([5.1010, 4.8298, 5.1058, 5.3035, 5.5420, 4.9161, 5.4871, 5.4984], device='cuda:1'), covar=tensor([0.1611, 0.1283, 0.1699, 0.0726, 0.0487, 0.0733, 0.0465, 0.0525], device='cuda:1'), in_proj_covar=tensor([0.0605, 0.0747, 0.0899, 0.0767, 0.0574, 0.0592, 0.0605, 0.0705], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-29 22:37:35,056 INFO [train.py:904] (1/8) Epoch 14, batch 2500, loss[loss=0.164, simple_loss=0.2604, pruned_loss=0.03379, over 17142.00 frames. ], tot_loss[loss=0.1815, simple_loss=0.2672, pruned_loss=0.04794, over 3326286.91 frames. ], batch size: 48, lr: 4.91e-03, grad_scale: 4.0 2023-04-29 22:37:46,217 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([5.0106, 4.7678, 5.0230, 5.2288, 5.4513, 4.7541, 5.4036, 5.4034], device='cuda:1'), covar=tensor([0.1633, 0.1225, 0.1630, 0.0727, 0.0501, 0.0825, 0.0463, 0.0565], device='cuda:1'), in_proj_covar=tensor([0.0606, 0.0748, 0.0900, 0.0768, 0.0575, 0.0593, 0.0606, 0.0707], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-29 22:38:16,658 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.99 vs. limit=2.0 2023-04-29 22:38:28,207 INFO [optim.py:368] (1/8) 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 22:38:40,137 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.5486, 2.1967, 2.1598, 4.2999, 2.1774, 2.6345, 2.3069, 2.3184], device='cuda:1'), covar=tensor([0.1103, 0.3715, 0.2790, 0.0476, 0.4144, 0.2463, 0.3454, 0.3605], device='cuda:1'), in_proj_covar=tensor([0.0378, 0.0412, 0.0345, 0.0328, 0.0420, 0.0476, 0.0377, 0.0483], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-29 22:38:45,442 INFO [train.py:904] (1/8) Epoch 14, batch 2550, loss[loss=0.1682, simple_loss=0.2565, pruned_loss=0.03998, over 17135.00 frames. ], tot_loss[loss=0.1817, simple_loss=0.2673, pruned_loss=0.048, over 3325707.90 frames. ], batch size: 49, lr: 4.91e-03, grad_scale: 4.0 2023-04-29 22:39:55,150 INFO [train.py:904] (1/8) Epoch 14, batch 2600, loss[loss=0.1745, simple_loss=0.2584, pruned_loss=0.04524, over 15914.00 frames. ], tot_loss[loss=0.1812, simple_loss=0.2668, pruned_loss=0.04784, over 3323277.21 frames. ], batch size: 35, lr: 4.91e-03, grad_scale: 4.0 2023-04-29 22:40:27,110 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.8911, 4.6143, 4.8653, 5.0534, 5.2661, 4.5518, 5.2789, 5.2572], device='cuda:1'), covar=tensor([0.1599, 0.1222, 0.1638, 0.0752, 0.0509, 0.0943, 0.0519, 0.0504], device='cuda:1'), in_proj_covar=tensor([0.0603, 0.0744, 0.0896, 0.0765, 0.0574, 0.0592, 0.0603, 0.0703], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-29 22:40:46,906 INFO [optim.py:368] (1/8) 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] (1/8) Epoch 14, batch 2650, loss[loss=0.2, simple_loss=0.2934, pruned_loss=0.05325, over 16615.00 frames. ], tot_loss[loss=0.1806, simple_loss=0.2669, pruned_loss=0.04713, over 3331594.98 frames. ], batch size: 62, lr: 4.91e-03, grad_scale: 4.0 2023-04-29 22:41:35,445 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([1.8531, 3.2050, 2.9321, 5.0824, 4.2942, 4.6223, 1.5583, 3.4183], device='cuda:1'), covar=tensor([0.1300, 0.0677, 0.1008, 0.0147, 0.0224, 0.0346, 0.1539, 0.0639], device='cuda:1'), in_proj_covar=tensor([0.0156, 0.0162, 0.0181, 0.0162, 0.0197, 0.0210, 0.0185, 0.0182], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0003, 0.0004], device='cuda:1') 2023-04-29 22:42:12,178 INFO [train.py:904] (1/8) Epoch 14, batch 2700, loss[loss=0.1713, simple_loss=0.2788, pruned_loss=0.03183, over 16682.00 frames. ], tot_loss[loss=0.1801, simple_loss=0.267, pruned_loss=0.04662, over 3329716.10 frames. ], batch size: 62, lr: 4.91e-03, grad_scale: 4.0 2023-04-29 22:42:13,677 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=134653.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 22:42:23,680 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=134660.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 22:42:31,300 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=134665.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 22:42:35,110 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.3488, 3.7849, 3.8143, 2.0293, 3.0424, 2.4588, 3.7459, 3.7883], device='cuda:1'), covar=tensor([0.0312, 0.0744, 0.0536, 0.1899, 0.0777, 0.0930, 0.0692, 0.0980], device='cuda:1'), in_proj_covar=tensor([0.0150, 0.0153, 0.0161, 0.0147, 0.0140, 0.0126, 0.0139, 0.0165], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:1') 2023-04-29 22:42:58,735 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.23 vs. limit=2.0 2023-04-29 22:43:04,243 INFO [optim.py:368] (1/8) 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] (1/8) 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] (1/8) Epoch 14, batch 2750, loss[loss=0.1698, simple_loss=0.2663, pruned_loss=0.03659, over 17289.00 frames. ], tot_loss[loss=0.1799, simple_loss=0.2673, pruned_loss=0.04624, over 3331123.67 frames. ], batch size: 52, lr: 4.91e-03, grad_scale: 4.0 2023-04-29 22:43:47,821 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.9518, 2.0106, 2.4605, 2.9237, 2.7887, 3.3292, 2.1313, 3.3098], device='cuda:1'), covar=tensor([0.0214, 0.0406, 0.0287, 0.0277, 0.0261, 0.0169, 0.0420, 0.0139], device='cuda:1'), in_proj_covar=tensor([0.0175, 0.0181, 0.0166, 0.0172, 0.0180, 0.0137, 0.0182, 0.0128], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-29 22:43:53,621 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.2938, 3.0597, 3.2587, 1.6416, 3.3900, 3.4372, 2.8570, 2.6234], device='cuda:1'), covar=tensor([0.0750, 0.0230, 0.0192, 0.1292, 0.0100, 0.0180, 0.0405, 0.0456], device='cuda:1'), in_proj_covar=tensor([0.0146, 0.0104, 0.0089, 0.0139, 0.0072, 0.0115, 0.0122, 0.0128], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:1') 2023-04-29 22:44:29,010 INFO [train.py:904] (1/8) Epoch 14, batch 2800, loss[loss=0.1786, simple_loss=0.2601, pruned_loss=0.04852, over 16503.00 frames. ], tot_loss[loss=0.1802, simple_loss=0.2675, pruned_loss=0.04649, over 3334547.75 frames. ], batch size: 68, lr: 4.91e-03, grad_scale: 8.0 2023-04-29 22:45:20,154 INFO [optim.py:368] (1/8) Clipping_scale=2.0, grad-norm quartiles 1.581e+02 2.461e+02 2.848e+02 3.331e+02 6.185e+02, threshold=5.697e+02, percent-clipped=2.0 2023-04-29 22:45:37,175 INFO [train.py:904] (1/8) Epoch 14, batch 2850, loss[loss=0.1623, simple_loss=0.2541, pruned_loss=0.03523, over 17123.00 frames. ], tot_loss[loss=0.1798, simple_loss=0.2667, pruned_loss=0.04646, over 3321934.41 frames. ], batch size: 49, lr: 4.91e-03, grad_scale: 8.0 2023-04-29 22:46:44,996 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=4.83 vs. limit=5.0 2023-04-29 22:46:45,299 INFO [train.py:904] (1/8) Epoch 14, batch 2900, loss[loss=0.1619, simple_loss=0.2323, pruned_loss=0.04576, over 16745.00 frames. ], tot_loss[loss=0.1785, simple_loss=0.2647, pruned_loss=0.04617, over 3319383.31 frames. ], batch size: 83, lr: 4.91e-03, grad_scale: 8.0 2023-04-29 22:46:56,435 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.1782, 4.0041, 4.2396, 4.3778, 4.4435, 4.0197, 4.2304, 4.4546], device='cuda:1'), covar=tensor([0.1338, 0.1026, 0.1126, 0.0606, 0.0560, 0.1267, 0.1748, 0.0630], device='cuda:1'), in_proj_covar=tensor([0.0600, 0.0742, 0.0892, 0.0764, 0.0572, 0.0588, 0.0600, 0.0698], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-29 22:46:58,852 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=134862.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 22:47:36,338 INFO [optim.py:368] (1/8) 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] (1/8) Epoch 14, batch 2950, loss[loss=0.1682, simple_loss=0.2596, pruned_loss=0.03834, over 17132.00 frames. ], tot_loss[loss=0.1789, simple_loss=0.2638, pruned_loss=0.04696, over 3318825.83 frames. ], batch size: 48, lr: 4.91e-03, grad_scale: 8.0 2023-04-29 22:48:08,986 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.9208, 4.2694, 3.1720, 2.3195, 2.8779, 2.4918, 4.5812, 3.7733], device='cuda:1'), covar=tensor([0.2435, 0.0521, 0.1472, 0.2399, 0.2396, 0.1783, 0.0360, 0.1031], device='cuda:1'), in_proj_covar=tensor([0.0308, 0.0262, 0.0290, 0.0289, 0.0283, 0.0232, 0.0275, 0.0314], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:1') 2023-04-29 22:48:24,032 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=134923.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 22:49:02,686 INFO [train.py:904] (1/8) Epoch 14, batch 3000, loss[loss=0.1703, simple_loss=0.2725, pruned_loss=0.03408, over 17088.00 frames. ], tot_loss[loss=0.1788, simple_loss=0.2637, pruned_loss=0.04691, over 3328775.65 frames. ], batch size: 49, lr: 4.91e-03, grad_scale: 8.0 2023-04-29 22:49:02,686 INFO [train.py:929] (1/8) Computing validation loss 2023-04-29 22:49:12,420 INFO [train.py:938] (1/8) Epoch 14, validation: loss=0.1382, simple_loss=0.244, pruned_loss=0.01621, over 944034.00 frames. 2023-04-29 22:49:12,421 INFO [train.py:939] (1/8) Maximum memory allocated so far is 17845MB 2023-04-29 22:49:24,935 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=134960.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 22:49:31,407 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=134965.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 22:50:06,981 INFO [optim.py:368] (1/8) 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 22:50:24,228 INFO [train.py:904] (1/8) Epoch 14, batch 3050, loss[loss=0.1759, simple_loss=0.2686, pruned_loss=0.04159, over 17114.00 frames. ], tot_loss[loss=0.1784, simple_loss=0.2636, pruned_loss=0.04664, over 3334255.54 frames. ], batch size: 47, lr: 4.90e-03, grad_scale: 8.0 2023-04-29 22:50:33,223 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=135008.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 22:50:40,117 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=135013.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 22:50:59,621 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=4.75 vs. limit=5.0 2023-04-29 22:51:32,217 INFO [train.py:904] (1/8) Epoch 14, batch 3100, loss[loss=0.2017, simple_loss=0.2746, pruned_loss=0.06443, over 16751.00 frames. ], tot_loss[loss=0.1795, simple_loss=0.2641, pruned_loss=0.04746, over 3327956.28 frames. ], batch size: 124, lr: 4.90e-03, grad_scale: 8.0 2023-04-29 22:52:24,972 INFO [optim.py:368] (1/8) 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 22:52:40,973 INFO [train.py:904] (1/8) Epoch 14, batch 3150, loss[loss=0.1831, simple_loss=0.2818, pruned_loss=0.04219, over 17049.00 frames. ], tot_loss[loss=0.1781, simple_loss=0.2629, pruned_loss=0.04671, over 3334936.18 frames. ], batch size: 50, lr: 4.90e-03, grad_scale: 8.0 2023-04-29 22:53:50,739 INFO [train.py:904] (1/8) Epoch 14, batch 3200, loss[loss=0.182, simple_loss=0.2706, pruned_loss=0.04673, over 17104.00 frames. ], tot_loss[loss=0.1774, simple_loss=0.262, pruned_loss=0.04637, over 3336664.93 frames. ], batch size: 53, lr: 4.90e-03, grad_scale: 8.0 2023-04-29 22:54:17,417 INFO [zipformer.py:625] (1/8) 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] (1/8) 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 22:54:59,015 INFO [train.py:904] (1/8) Epoch 14, batch 3250, loss[loss=0.1843, simple_loss=0.2656, pruned_loss=0.05151, over 16748.00 frames. ], tot_loss[loss=0.1776, simple_loss=0.2621, pruned_loss=0.0466, over 3339397.93 frames. ], batch size: 83, lr: 4.90e-03, grad_scale: 8.0 2023-04-29 22:55:22,124 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=135218.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 22:55:27,979 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.9924, 2.3700, 2.3858, 4.7493, 2.3508, 2.8929, 2.4771, 2.5969], device='cuda:1'), covar=tensor([0.0923, 0.3353, 0.2654, 0.0321, 0.3864, 0.2313, 0.3216, 0.3441], device='cuda:1'), in_proj_covar=tensor([0.0380, 0.0414, 0.0346, 0.0329, 0.0421, 0.0478, 0.0378, 0.0485], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-29 22:55:41,129 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=135232.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 22:56:07,847 INFO [train.py:904] (1/8) Epoch 14, batch 3300, loss[loss=0.2006, simple_loss=0.287, pruned_loss=0.05715, over 16669.00 frames. ], tot_loss[loss=0.1782, simple_loss=0.2629, pruned_loss=0.04678, over 3326532.13 frames. ], batch size: 57, lr: 4.90e-03, grad_scale: 4.0 2023-04-29 22:56:56,878 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([5.0802, 5.0599, 4.8264, 4.2315, 4.9406, 1.7608, 4.6777, 4.7457], device='cuda:1'), covar=tensor([0.0091, 0.0085, 0.0191, 0.0421, 0.0098, 0.2642, 0.0138, 0.0200], device='cuda:1'), in_proj_covar=tensor([0.0148, 0.0137, 0.0184, 0.0172, 0.0155, 0.0196, 0.0174, 0.0168], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-29 22:57:01,614 INFO [optim.py:368] (1/8) 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 22:57:02,029 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.7619, 4.2287, 4.3407, 3.1512, 3.5887, 4.2599, 3.8825, 2.4787], device='cuda:1'), covar=tensor([0.0352, 0.0058, 0.0031, 0.0261, 0.0105, 0.0078, 0.0068, 0.0366], device='cuda:1'), in_proj_covar=tensor([0.0129, 0.0073, 0.0073, 0.0128, 0.0086, 0.0094, 0.0083, 0.0123], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:1') 2023-04-29 22:57:16,993 INFO [train.py:904] (1/8) Epoch 14, batch 3350, loss[loss=0.1744, simple_loss=0.2617, pruned_loss=0.04358, over 16519.00 frames. ], tot_loss[loss=0.1796, simple_loss=0.2646, pruned_loss=0.0473, over 3309783.67 frames. ], batch size: 75, lr: 4.90e-03, grad_scale: 4.0 2023-04-29 22:58:24,246 INFO [train.py:904] (1/8) Epoch 14, batch 3400, loss[loss=0.166, simple_loss=0.2588, pruned_loss=0.03664, over 17257.00 frames. ], tot_loss[loss=0.179, simple_loss=0.2641, pruned_loss=0.04694, over 3310604.85 frames. ], batch size: 52, lr: 4.90e-03, grad_scale: 4.0 2023-04-29 22:58:40,985 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.20 vs. limit=2.0 2023-04-29 22:59:16,983 INFO [optim.py:368] (1/8) 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] (1/8) Epoch 14, batch 3450, loss[loss=0.181, simple_loss=0.2622, pruned_loss=0.0499, over 16697.00 frames. ], tot_loss[loss=0.1783, simple_loss=0.2634, pruned_loss=0.04658, over 3318569.86 frames. ], batch size: 89, lr: 4.90e-03, grad_scale: 4.0 2023-04-29 22:59:51,164 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([1.7936, 2.9800, 2.5792, 4.2785, 3.5682, 4.1948, 1.5419, 3.0047], device='cuda:1'), covar=tensor([0.1354, 0.0587, 0.1027, 0.0166, 0.0155, 0.0350, 0.1488, 0.0733], device='cuda:1'), in_proj_covar=tensor([0.0158, 0.0165, 0.0184, 0.0165, 0.0203, 0.0213, 0.0188, 0.0185], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-04-29 23:00:17,159 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.18 vs. limit=2.0 2023-04-29 23:00:37,048 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.8903, 4.2241, 3.0442, 2.1925, 2.8057, 2.4752, 4.5205, 3.7011], device='cuda:1'), covar=tensor([0.2511, 0.0581, 0.1546, 0.2508, 0.2483, 0.1798, 0.0393, 0.1099], device='cuda:1'), in_proj_covar=tensor([0.0309, 0.0262, 0.0290, 0.0289, 0.0285, 0.0233, 0.0275, 0.0314], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:1') 2023-04-29 23:00:41,081 INFO [train.py:904] (1/8) Epoch 14, batch 3500, loss[loss=0.1747, simple_loss=0.2567, pruned_loss=0.04631, over 15424.00 frames. ], tot_loss[loss=0.1771, simple_loss=0.2621, pruned_loss=0.04603, over 3318276.12 frames. ], batch size: 190, lr: 4.90e-03, grad_scale: 4.0 2023-04-29 23:01:37,086 INFO [optim.py:368] (1/8) 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 23:01:51,742 INFO [train.py:904] (1/8) Epoch 14, batch 3550, loss[loss=0.1689, simple_loss=0.2621, pruned_loss=0.03789, over 17258.00 frames. ], tot_loss[loss=0.1767, simple_loss=0.2612, pruned_loss=0.04613, over 3314339.64 frames. ], batch size: 52, lr: 4.90e-03, grad_scale: 4.0 2023-04-29 23:02:15,151 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=135518.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 23:02:25,348 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.2859, 4.0991, 4.5024, 2.0769, 4.6960, 4.7630, 3.5488, 3.6278], device='cuda:1'), covar=tensor([0.0554, 0.0186, 0.0167, 0.1027, 0.0047, 0.0130, 0.0285, 0.0336], device='cuda:1'), in_proj_covar=tensor([0.0144, 0.0103, 0.0089, 0.0136, 0.0071, 0.0114, 0.0121, 0.0127], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:1') 2023-04-29 23:02:27,269 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=135527.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 23:02:31,607 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.8692, 4.0285, 3.0748, 2.3297, 2.6935, 2.5239, 4.1646, 3.5997], device='cuda:1'), covar=tensor([0.2504, 0.0596, 0.1562, 0.2563, 0.2535, 0.1767, 0.0510, 0.1219], device='cuda:1'), in_proj_covar=tensor([0.0308, 0.0262, 0.0289, 0.0288, 0.0285, 0.0232, 0.0274, 0.0313], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:1') 2023-04-29 23:03:02,250 INFO [train.py:904] (1/8) Epoch 14, batch 3600, loss[loss=0.1669, simple_loss=0.2581, pruned_loss=0.03787, over 17112.00 frames. ], tot_loss[loss=0.176, simple_loss=0.2601, pruned_loss=0.04599, over 3314363.72 frames. ], batch size: 47, lr: 4.89e-03, grad_scale: 8.0 2023-04-29 23:03:04,086 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.7769, 2.9988, 3.1860, 2.1113, 2.7128, 2.2935, 3.3566, 3.2531], device='cuda:1'), covar=tensor([0.0283, 0.0966, 0.0528, 0.1815, 0.0828, 0.0935, 0.0558, 0.0862], device='cuda:1'), in_proj_covar=tensor([0.0151, 0.0156, 0.0162, 0.0148, 0.0140, 0.0127, 0.0141, 0.0166], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-04-29 23:03:22,566 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=135566.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 23:03:25,636 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=4.57 vs. limit=5.0 2023-04-29 23:03:58,634 INFO [optim.py:368] (1/8) 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 23:04:14,252 INFO [train.py:904] (1/8) Epoch 14, batch 3650, loss[loss=0.1714, simple_loss=0.2373, pruned_loss=0.05274, over 16682.00 frames. ], tot_loss[loss=0.1761, simple_loss=0.2588, pruned_loss=0.04664, over 3316189.30 frames. ], batch size: 76, lr: 4.89e-03, grad_scale: 8.0 2023-04-29 23:04:57,258 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=135630.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 23:05:03,307 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.8769, 5.2522, 4.9136, 4.9768, 4.7317, 4.7029, 4.6249, 5.2684], device='cuda:1'), covar=tensor([0.1231, 0.0782, 0.1083, 0.0813, 0.0757, 0.0974, 0.1155, 0.0890], device='cuda:1'), in_proj_covar=tensor([0.0609, 0.0756, 0.0619, 0.0545, 0.0484, 0.0481, 0.0631, 0.0583], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:1') 2023-04-29 23:05:31,030 INFO [train.py:904] (1/8) Epoch 14, batch 3700, loss[loss=0.208, simple_loss=0.2791, pruned_loss=0.0685, over 11849.00 frames. ], tot_loss[loss=0.1769, simple_loss=0.2578, pruned_loss=0.04802, over 3285855.31 frames. ], batch size: 246, lr: 4.89e-03, grad_scale: 4.0 2023-04-29 23:06:00,373 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([1.8772, 2.6255, 2.5723, 1.9184, 2.5418, 2.6791, 2.5366, 1.9367], device='cuda:1'), covar=tensor([0.0346, 0.0073, 0.0054, 0.0306, 0.0093, 0.0104, 0.0093, 0.0310], device='cuda:1'), in_proj_covar=tensor([0.0129, 0.0073, 0.0073, 0.0128, 0.0086, 0.0094, 0.0083, 0.0123], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:1') 2023-04-29 23:06:30,675 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=135691.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 23:06:32,148 INFO [optim.py:368] (1/8) 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] (1/8) Epoch 14, batch 3750, loss[loss=0.186, simple_loss=0.2678, pruned_loss=0.05213, over 16347.00 frames. ], tot_loss[loss=0.1792, simple_loss=0.259, pruned_loss=0.04971, over 3271306.99 frames. ], batch size: 35, lr: 4.89e-03, grad_scale: 4.0 2023-04-29 23:08:01,578 INFO [train.py:904] (1/8) Epoch 14, batch 3800, loss[loss=0.2089, simple_loss=0.2775, pruned_loss=0.07015, over 16714.00 frames. ], tot_loss[loss=0.1811, simple_loss=0.2601, pruned_loss=0.05104, over 3279369.40 frames. ], batch size: 124, lr: 4.89e-03, grad_scale: 4.0 2023-04-29 23:08:28,218 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.8053, 5.2027, 4.9596, 4.9365, 4.6842, 4.6503, 4.6072, 5.3054], device='cuda:1'), covar=tensor([0.1314, 0.0900, 0.1106, 0.0831, 0.0854, 0.1047, 0.1114, 0.0900], device='cuda:1'), in_proj_covar=tensor([0.0605, 0.0748, 0.0611, 0.0542, 0.0480, 0.0479, 0.0626, 0.0577], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:1') 2023-04-29 23:09:00,855 INFO [optim.py:368] (1/8) Clipping_scale=2.0, grad-norm quartiles 1.772e+02 2.313e+02 2.548e+02 3.011e+02 5.523e+02, threshold=5.096e+02, percent-clipped=2.0 2023-04-29 23:09:16,026 INFO [train.py:904] (1/8) Epoch 14, batch 3850, loss[loss=0.1745, simple_loss=0.2437, pruned_loss=0.05262, over 16855.00 frames. ], tot_loss[loss=0.1818, simple_loss=0.2602, pruned_loss=0.05166, over 3284973.67 frames. ], batch size: 109, lr: 4.89e-03, grad_scale: 4.0 2023-04-29 23:09:53,648 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.4660, 4.5140, 4.8252, 4.8212, 4.8429, 4.5092, 4.5598, 4.4009], device='cuda:1'), covar=tensor([0.0348, 0.0574, 0.0411, 0.0440, 0.0523, 0.0429, 0.0884, 0.0500], device='cuda:1'), in_proj_covar=tensor([0.0374, 0.0395, 0.0391, 0.0371, 0.0437, 0.0416, 0.0511, 0.0329], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:1') 2023-04-29 23:09:53,676 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=135827.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 23:10:04,380 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=5.02 vs. limit=5.0 2023-04-29 23:10:30,031 INFO [train.py:904] (1/8) Epoch 14, batch 3900, loss[loss=0.1767, simple_loss=0.2592, pruned_loss=0.04709, over 16472.00 frames. ], tot_loss[loss=0.1826, simple_loss=0.2604, pruned_loss=0.05243, over 3276697.49 frames. ], batch size: 35, lr: 4.89e-03, grad_scale: 4.0 2023-04-29 23:11:02,869 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.46 vs. limit=2.0 2023-04-29 23:11:05,160 INFO [zipformer.py:625] (1/8) 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] (1/8) 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] (1/8) Epoch 14, batch 3950, loss[loss=0.2041, simple_loss=0.2827, pruned_loss=0.06279, over 12553.00 frames. ], tot_loss[loss=0.1822, simple_loss=0.2593, pruned_loss=0.05252, over 3281108.61 frames. ], batch size: 247, lr: 4.89e-03, grad_scale: 4.0 2023-04-29 23:11:51,942 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.18 vs. limit=2.0 2023-04-29 23:12:49,657 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.7150, 3.9470, 2.2873, 4.5219, 2.9740, 4.5334, 2.5299, 2.9457], device='cuda:1'), covar=tensor([0.0244, 0.0308, 0.1650, 0.0089, 0.0791, 0.0294, 0.1457, 0.0727], device='cuda:1'), in_proj_covar=tensor([0.0162, 0.0173, 0.0192, 0.0149, 0.0172, 0.0217, 0.0201, 0.0175], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-04-29 23:12:58,425 INFO [train.py:904] (1/8) Epoch 14, batch 4000, loss[loss=0.2007, simple_loss=0.2948, pruned_loss=0.05331, over 15513.00 frames. ], tot_loss[loss=0.1824, simple_loss=0.2595, pruned_loss=0.0527, over 3279790.22 frames. ], batch size: 190, lr: 4.89e-03, grad_scale: 8.0 2023-04-29 23:13:26,521 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([1.7710, 2.9228, 3.0257, 5.0783, 4.2332, 4.4517, 2.1289, 3.2469], device='cuda:1'), covar=tensor([0.1249, 0.0763, 0.1010, 0.0110, 0.0346, 0.0335, 0.1266, 0.0816], device='cuda:1'), in_proj_covar=tensor([0.0155, 0.0162, 0.0182, 0.0163, 0.0200, 0.0210, 0.0185, 0.0183], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0003, 0.0004], device='cuda:1') 2023-04-29 23:13:41,777 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=135982.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 23:13:47,606 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=135986.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 23:13:55,965 INFO [optim.py:368] (1/8) 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 23:13:58,982 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=135994.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 23:14:13,532 INFO [train.py:904] (1/8) Epoch 14, batch 4050, loss[loss=0.1682, simple_loss=0.2497, pruned_loss=0.04336, over 16178.00 frames. ], tot_loss[loss=0.182, simple_loss=0.2601, pruned_loss=0.05194, over 3268359.99 frames. ], batch size: 35, lr: 4.89e-03, grad_scale: 8.0 2023-04-29 23:15:12,398 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=136043.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 23:15:24,936 INFO [train.py:904] (1/8) Epoch 14, batch 4100, loss[loss=0.1961, simple_loss=0.2734, pruned_loss=0.05938, over 16559.00 frames. ], tot_loss[loss=0.1816, simple_loss=0.2611, pruned_loss=0.05103, over 3269728.32 frames. ], batch size: 62, lr: 4.89e-03, grad_scale: 8.0 2023-04-29 23:15:29,832 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=136055.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 23:16:24,202 INFO [optim.py:368] (1/8) 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] (1/8) Epoch 14, batch 4150, loss[loss=0.2128, simple_loss=0.305, pruned_loss=0.06025, over 16994.00 frames. ], tot_loss[loss=0.188, simple_loss=0.2684, pruned_loss=0.05379, over 3242161.31 frames. ], batch size: 41, lr: 4.88e-03, grad_scale: 8.0 2023-04-29 23:17:41,266 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=136142.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 23:17:55,763 INFO [train.py:904] (1/8) Epoch 14, batch 4200, loss[loss=0.2067, simple_loss=0.2898, pruned_loss=0.06181, over 16876.00 frames. ], tot_loss[loss=0.1944, simple_loss=0.2763, pruned_loss=0.05629, over 3206639.09 frames. ], batch size: 109, lr: 4.88e-03, grad_scale: 8.0 2023-04-29 23:18:26,748 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.9810, 4.9127, 4.6869, 4.0971, 4.7932, 1.9471, 4.5777, 4.6042], device='cuda:1'), covar=tensor([0.0141, 0.0134, 0.0190, 0.0390, 0.0155, 0.2456, 0.0164, 0.0189], device='cuda:1'), in_proj_covar=tensor([0.0146, 0.0134, 0.0180, 0.0169, 0.0153, 0.0193, 0.0171, 0.0165], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-29 23:18:55,501 INFO [optim.py:368] (1/8) 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 23:19:05,734 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([5.4087, 5.6863, 5.4009, 5.4810, 5.1517, 4.9169, 5.0560, 5.8100], device='cuda:1'), covar=tensor([0.1004, 0.0728, 0.0966, 0.0703, 0.0814, 0.0699, 0.1024, 0.0783], device='cuda:1'), in_proj_covar=tensor([0.0589, 0.0730, 0.0595, 0.0529, 0.0468, 0.0469, 0.0612, 0.0566], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:1') 2023-04-29 23:19:10,103 INFO [train.py:904] (1/8) Epoch 14, batch 4250, loss[loss=0.1953, simple_loss=0.2866, pruned_loss=0.05199, over 16590.00 frames. ], tot_loss[loss=0.1955, simple_loss=0.2795, pruned_loss=0.05577, over 3200866.90 frames. ], batch size: 62, lr: 4.88e-03, grad_scale: 8.0 2023-04-29 23:19:12,611 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=136203.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 23:20:23,645 INFO [train.py:904] (1/8) Epoch 14, batch 4300, loss[loss=0.1971, simple_loss=0.282, pruned_loss=0.05616, over 11611.00 frames. ], tot_loss[loss=0.1949, simple_loss=0.2804, pruned_loss=0.05473, over 3183882.69 frames. ], batch size: 246, lr: 4.88e-03, grad_scale: 8.0 2023-04-29 23:20:52,429 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.6145, 1.6818, 2.2048, 2.5998, 2.5550, 2.9743, 1.6268, 2.8037], device='cuda:1'), covar=tensor([0.0215, 0.0421, 0.0274, 0.0240, 0.0237, 0.0142, 0.0482, 0.0125], device='cuda:1'), in_proj_covar=tensor([0.0172, 0.0178, 0.0163, 0.0171, 0.0179, 0.0134, 0.0179, 0.0128], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-29 23:21:13,970 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=136286.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 23:21:16,414 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.6418, 3.6516, 4.0544, 2.0246, 4.2840, 4.3610, 2.9417, 3.1736], device='cuda:1'), covar=tensor([0.0793, 0.0224, 0.0195, 0.1105, 0.0043, 0.0078, 0.0444, 0.0419], device='cuda:1'), in_proj_covar=tensor([0.0146, 0.0104, 0.0089, 0.0137, 0.0071, 0.0113, 0.0122, 0.0127], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:1') 2023-04-29 23:21:23,663 INFO [optim.py:368] (1/8) 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] (1/8) Epoch 14, batch 4350, loss[loss=0.2109, simple_loss=0.3019, pruned_loss=0.05996, over 16479.00 frames. ], tot_loss[loss=0.1979, simple_loss=0.2838, pruned_loss=0.05601, over 3176832.79 frames. ], batch size: 75, lr: 4.88e-03, grad_scale: 8.0 2023-04-29 23:22:26,310 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=136334.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 23:22:28,596 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=136335.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 23:22:33,159 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=136338.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 23:22:51,136 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 14, batch 4400, loss[loss=0.2405, simple_loss=0.3093, pruned_loss=0.08583, over 11878.00 frames. ], tot_loss[loss=0.1999, simple_loss=0.2856, pruned_loss=0.05708, over 3169726.71 frames. ], batch size: 246, lr: 4.88e-03, grad_scale: 8.0 2023-04-29 23:23:46,005 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.4800, 1.6875, 2.1276, 2.4664, 2.4086, 2.8299, 1.6764, 2.5701], device='cuda:1'), covar=tensor([0.0182, 0.0425, 0.0246, 0.0239, 0.0241, 0.0141, 0.0452, 0.0137], device='cuda:1'), in_proj_covar=tensor([0.0173, 0.0179, 0.0164, 0.0172, 0.0179, 0.0135, 0.0181, 0.0129], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-29 23:23:51,148 INFO [optim.py:368] (1/8) 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 23:23:57,388 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 14, batch 4450, loss[loss=0.1962, simple_loss=0.2918, pruned_loss=0.05027, over 16669.00 frames. ], tot_loss[loss=0.2029, simple_loss=0.2891, pruned_loss=0.05837, over 3175892.21 frames. ], batch size: 89, lr: 4.88e-03, grad_scale: 8.0 2023-04-29 23:24:24,770 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.64 vs. limit=2.0 2023-04-29 23:25:18,608 INFO [train.py:904] (1/8) Epoch 14, batch 4500, loss[loss=0.197, simple_loss=0.2782, pruned_loss=0.05796, over 17073.00 frames. ], tot_loss[loss=0.2036, simple_loss=0.2896, pruned_loss=0.05876, over 3192646.51 frames. ], batch size: 53, lr: 4.88e-03, grad_scale: 8.0 2023-04-29 23:25:26,711 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.8376, 3.1521, 3.2838, 1.8557, 2.6888, 2.0431, 3.2992, 3.2695], device='cuda:1'), covar=tensor([0.0227, 0.0638, 0.0514, 0.1935, 0.0828, 0.0996, 0.0601, 0.0851], device='cuda:1'), in_proj_covar=tensor([0.0148, 0.0152, 0.0161, 0.0146, 0.0139, 0.0126, 0.0139, 0.0163], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:1') 2023-04-29 23:25:55,243 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=4.82 vs. limit=5.0 2023-04-29 23:26:18,140 INFO [optim.py:368] (1/8) 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 23:26:27,077 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 14, batch 4550, loss[loss=0.2026, simple_loss=0.2955, pruned_loss=0.05489, over 16700.00 frames. ], tot_loss[loss=0.2051, simple_loss=0.2908, pruned_loss=0.05967, over 3202090.87 frames. ], batch size: 83, lr: 4.88e-03, grad_scale: 8.0 2023-04-29 23:26:54,804 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.40 vs. limit=2.0 2023-04-29 23:27:12,680 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.1840, 4.2235, 1.8747, 4.9981, 3.1469, 4.6875, 2.1242, 3.1994], device='cuda:1'), covar=tensor([0.0159, 0.0215, 0.1832, 0.0091, 0.0641, 0.0267, 0.1665, 0.0593], device='cuda:1'), in_proj_covar=tensor([0.0158, 0.0169, 0.0188, 0.0141, 0.0168, 0.0210, 0.0196, 0.0171], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0003, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-04-29 23:27:45,693 INFO [train.py:904] (1/8) Epoch 14, batch 4600, loss[loss=0.2015, simple_loss=0.2868, pruned_loss=0.05816, over 17138.00 frames. ], tot_loss[loss=0.2056, simple_loss=0.2914, pruned_loss=0.05985, over 3206383.72 frames. ], batch size: 47, lr: 4.88e-03, grad_scale: 8.0 2023-04-29 23:28:42,904 INFO [optim.py:368] (1/8) 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 23:28:57,104 INFO [train.py:904] (1/8) Epoch 14, batch 4650, loss[loss=0.2005, simple_loss=0.2858, pruned_loss=0.05759, over 16655.00 frames. ], tot_loss[loss=0.205, simple_loss=0.2907, pruned_loss=0.05968, over 3211826.14 frames. ], batch size: 76, lr: 4.88e-03, grad_scale: 8.0 2023-04-29 23:29:51,580 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=136638.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 23:29:58,837 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=136643.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 23:30:07,988 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=136650.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 23:30:10,585 INFO [train.py:904] (1/8) Epoch 14, batch 4700, loss[loss=0.1978, simple_loss=0.2809, pruned_loss=0.0574, over 16857.00 frames. ], tot_loss[loss=0.2021, simple_loss=0.2876, pruned_loss=0.05825, over 3200806.16 frames. ], batch size: 42, lr: 4.87e-03, grad_scale: 8.0 2023-04-29 23:31:01,303 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=136686.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 23:31:08,480 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=136691.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 23:31:09,908 INFO [optim.py:368] (1/8) 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 23:31:14,761 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.4294, 3.2295, 2.6972, 2.1108, 2.1859, 2.1911, 3.3059, 3.0189], device='cuda:1'), covar=tensor([0.2749, 0.0829, 0.1593, 0.2406, 0.2331, 0.1914, 0.0548, 0.1147], device='cuda:1'), in_proj_covar=tensor([0.0309, 0.0260, 0.0290, 0.0289, 0.0285, 0.0231, 0.0276, 0.0311], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:1') 2023-04-29 23:31:18,047 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 14, batch 4750, loss[loss=0.2016, simple_loss=0.2758, pruned_loss=0.06374, over 16633.00 frames. ], tot_loss[loss=0.1983, simple_loss=0.284, pruned_loss=0.05631, over 3194739.16 frames. ], batch size: 62, lr: 4.87e-03, grad_scale: 8.0 2023-04-29 23:31:27,821 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=136704.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 23:32:36,917 INFO [train.py:904] (1/8) Epoch 14, batch 4800, loss[loss=0.1669, simple_loss=0.253, pruned_loss=0.04042, over 16591.00 frames. ], tot_loss[loss=0.1941, simple_loss=0.2798, pruned_loss=0.05427, over 3188676.56 frames. ], batch size: 57, lr: 4.87e-03, grad_scale: 8.0 2023-04-29 23:33:03,269 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([1.7451, 1.4161, 1.6692, 1.7346, 1.8247, 1.9389, 1.5645, 1.8288], device='cuda:1'), covar=tensor([0.0191, 0.0286, 0.0174, 0.0215, 0.0200, 0.0145, 0.0318, 0.0108], device='cuda:1'), in_proj_covar=tensor([0.0170, 0.0177, 0.0162, 0.0170, 0.0178, 0.0134, 0.0179, 0.0127], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-29 23:33:24,865 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.94 vs. limit=2.0 2023-04-29 23:33:36,432 INFO [optim.py:368] (1/8) 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 23:33:46,383 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 14, batch 4850, loss[loss=0.189, simple_loss=0.2827, pruned_loss=0.04766, over 16696.00 frames. ], tot_loss[loss=0.1926, simple_loss=0.2794, pruned_loss=0.05289, over 3185938.23 frames. ], batch size: 134, lr: 4.87e-03, grad_scale: 8.0 2023-04-29 23:33:52,950 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.79 vs. limit=2.0 2023-04-29 23:33:58,240 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=136806.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 23:34:56,905 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=136846.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 23:35:05,048 INFO [train.py:904] (1/8) Epoch 14, batch 4900, loss[loss=0.1855, simple_loss=0.2805, pruned_loss=0.04523, over 15385.00 frames. ], tot_loss[loss=0.1912, simple_loss=0.2789, pruned_loss=0.05176, over 3169840.18 frames. ], batch size: 190, lr: 4.87e-03, grad_scale: 8.0 2023-04-29 23:35:28,812 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=136867.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 23:35:48,976 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.5356, 3.6965, 1.9891, 4.2312, 2.7293, 4.1239, 2.1825, 2.7797], device='cuda:1'), covar=tensor([0.0281, 0.0309, 0.1780, 0.0107, 0.0859, 0.0402, 0.1572, 0.0796], device='cuda:1'), in_proj_covar=tensor([0.0159, 0.0170, 0.0191, 0.0141, 0.0171, 0.0211, 0.0198, 0.0173], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-04-29 23:36:02,817 INFO [optim.py:368] (1/8) 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 23:36:18,495 INFO [train.py:904] (1/8) Epoch 14, batch 4950, loss[loss=0.2183, simple_loss=0.303, pruned_loss=0.0668, over 12152.00 frames. ], tot_loss[loss=0.1903, simple_loss=0.2785, pruned_loss=0.05111, over 3181521.50 frames. ], batch size: 246, lr: 4.87e-03, grad_scale: 8.0 2023-04-29 23:36:48,232 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.5078, 3.5936, 3.3447, 3.0341, 3.1352, 3.4089, 3.3046, 3.2191], device='cuda:1'), covar=tensor([0.0496, 0.0441, 0.0264, 0.0258, 0.0525, 0.0350, 0.1168, 0.0435], device='cuda:1'), in_proj_covar=tensor([0.0249, 0.0346, 0.0304, 0.0284, 0.0320, 0.0330, 0.0206, 0.0356], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-29 23:37:04,039 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([5.6115, 5.9505, 5.6404, 5.8136, 5.3981, 5.1335, 5.3428, 6.0884], device='cuda:1'), covar=tensor([0.1195, 0.0770, 0.1033, 0.0694, 0.0827, 0.0617, 0.0979, 0.0817], device='cuda:1'), in_proj_covar=tensor([0.0574, 0.0708, 0.0582, 0.0512, 0.0455, 0.0458, 0.0596, 0.0553], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:1') 2023-04-29 23:37:24,312 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=2.92 vs. limit=5.0 2023-04-29 23:37:30,153 INFO [train.py:904] (1/8) Epoch 14, batch 5000, loss[loss=0.1774, simple_loss=0.2737, pruned_loss=0.04057, over 16752.00 frames. ], tot_loss[loss=0.1917, simple_loss=0.2806, pruned_loss=0.05145, over 3182331.19 frames. ], batch size: 89, lr: 4.87e-03, grad_scale: 8.0 2023-04-29 23:37:44,080 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=4.85 vs. limit=5.0 2023-04-29 23:37:48,273 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.2428, 1.5266, 1.9346, 2.2374, 2.3006, 2.5357, 1.6019, 2.3765], device='cuda:1'), covar=tensor([0.0200, 0.0434, 0.0291, 0.0274, 0.0248, 0.0148, 0.0446, 0.0106], device='cuda:1'), in_proj_covar=tensor([0.0170, 0.0177, 0.0162, 0.0170, 0.0177, 0.0134, 0.0179, 0.0127], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-29 23:38:24,317 INFO [zipformer.py:625] (1/8) 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] (1/8) 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 23:38:29,207 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.3369, 3.3577, 2.6401, 2.0256, 2.2571, 2.1472, 3.4596, 3.0405], device='cuda:1'), covar=tensor([0.2799, 0.0765, 0.1690, 0.2490, 0.2282, 0.1844, 0.0556, 0.1128], device='cuda:1'), in_proj_covar=tensor([0.0310, 0.0261, 0.0291, 0.0290, 0.0286, 0.0231, 0.0277, 0.0311], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:1') 2023-04-29 23:38:36,204 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 14, batch 5050, loss[loss=0.1901, simple_loss=0.2801, pruned_loss=0.05004, over 16268.00 frames. ], tot_loss[loss=0.1916, simple_loss=0.2809, pruned_loss=0.05118, over 3206113.82 frames. ], batch size: 165, lr: 4.87e-03, grad_scale: 8.0 2023-04-29 23:38:41,397 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([5.1661, 5.1263, 4.9982, 4.6015, 4.5288, 5.0435, 5.0575, 4.7835], device='cuda:1'), covar=tensor([0.0561, 0.0490, 0.0276, 0.0289, 0.1181, 0.0447, 0.0251, 0.0600], device='cuda:1'), in_proj_covar=tensor([0.0251, 0.0350, 0.0306, 0.0287, 0.0323, 0.0334, 0.0207, 0.0359], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-29 23:38:49,772 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.5907, 2.3843, 2.3743, 4.4589, 2.2514, 2.7410, 2.4659, 2.5721], device='cuda:1'), covar=tensor([0.0986, 0.3059, 0.2396, 0.0366, 0.3561, 0.2218, 0.2956, 0.2815], device='cuda:1'), in_proj_covar=tensor([0.0374, 0.0414, 0.0343, 0.0322, 0.0418, 0.0476, 0.0375, 0.0482], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-29 23:39:02,039 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.15 vs. limit=2.0 2023-04-29 23:39:04,492 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.36 vs. limit=2.0 2023-04-29 23:39:31,297 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=137039.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 23:39:32,846 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.1467, 1.9681, 1.7032, 1.8389, 2.2069, 1.9970, 2.0475, 2.3750], device='cuda:1'), covar=tensor([0.0125, 0.0336, 0.0411, 0.0348, 0.0185, 0.0298, 0.0135, 0.0208], device='cuda:1'), in_proj_covar=tensor([0.0170, 0.0212, 0.0207, 0.0206, 0.0212, 0.0213, 0.0218, 0.0207], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-29 23:39:49,978 INFO [train.py:904] (1/8) Epoch 14, batch 5100, loss[loss=0.1595, simple_loss=0.257, pruned_loss=0.03104, over 16816.00 frames. ], tot_loss[loss=0.1898, simple_loss=0.279, pruned_loss=0.05034, over 3216994.89 frames. ], batch size: 102, lr: 4.87e-03, grad_scale: 8.0 2023-04-29 23:40:13,669 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([1.9695, 3.1844, 3.2294, 2.0772, 2.8997, 3.2037, 3.0125, 1.8001], device='cuda:1'), covar=tensor([0.0462, 0.0043, 0.0044, 0.0380, 0.0088, 0.0085, 0.0081, 0.0451], device='cuda:1'), in_proj_covar=tensor([0.0130, 0.0073, 0.0073, 0.0129, 0.0086, 0.0095, 0.0084, 0.0123], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-04-29 23:40:45,951 INFO [optim.py:368] (1/8) 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] (1/8) Epoch 14, batch 5150, loss[loss=0.2023, simple_loss=0.3036, pruned_loss=0.05046, over 15423.00 frames. ], tot_loss[loss=0.1894, simple_loss=0.2794, pruned_loss=0.04966, over 3205611.00 frames. ], batch size: 190, lr: 4.87e-03, grad_scale: 8.0 2023-04-29 23:41:30,871 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.7134, 4.9212, 5.1486, 4.9000, 4.9374, 5.5168, 4.9636, 4.6478], device='cuda:1'), covar=tensor([0.0992, 0.1770, 0.1757, 0.1738, 0.2428, 0.0945, 0.1257, 0.2453], device='cuda:1'), in_proj_covar=tensor([0.0370, 0.0518, 0.0558, 0.0439, 0.0591, 0.0587, 0.0447, 0.0589], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-04-29 23:41:41,086 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.97 vs. limit=2.0 2023-04-29 23:42:08,569 INFO [train.py:904] (1/8) Epoch 14, batch 5200, loss[loss=0.2009, simple_loss=0.2824, pruned_loss=0.05966, over 16900.00 frames. ], tot_loss[loss=0.1884, simple_loss=0.2781, pruned_loss=0.04935, over 3204544.25 frames. ], batch size: 109, lr: 4.87e-03, grad_scale: 8.0 2023-04-29 23:42:22,360 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=137162.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 23:42:27,931 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.88 vs. limit=2.0 2023-04-29 23:42:37,919 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.44 vs. limit=2.0 2023-04-29 23:43:03,327 INFO [optim.py:368] (1/8) 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] (1/8) Epoch 14, batch 5250, loss[loss=0.1935, simple_loss=0.2716, pruned_loss=0.05768, over 12189.00 frames. ], tot_loss[loss=0.187, simple_loss=0.2757, pruned_loss=0.04919, over 3210907.64 frames. ], batch size: 247, lr: 4.87e-03, grad_scale: 8.0 2023-04-29 23:44:00,721 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([1.9353, 2.3958, 2.2552, 2.7787, 2.0310, 3.2658, 1.6406, 2.6917], device='cuda:1'), covar=tensor([0.1181, 0.0621, 0.1105, 0.0148, 0.0137, 0.0395, 0.1388, 0.0681], device='cuda:1'), in_proj_covar=tensor([0.0156, 0.0162, 0.0184, 0.0160, 0.0199, 0.0207, 0.0186, 0.0184], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-04-29 23:44:26,543 INFO [train.py:904] (1/8) Epoch 14, batch 5300, loss[loss=0.1777, simple_loss=0.2698, pruned_loss=0.04277, over 16665.00 frames. ], tot_loss[loss=0.1848, simple_loss=0.273, pruned_loss=0.04836, over 3189543.79 frames. ], batch size: 134, lr: 4.86e-03, grad_scale: 8.0 2023-04-29 23:44:32,751 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.5235, 4.4881, 4.9172, 4.8470, 4.8262, 4.5243, 4.5035, 4.3556], device='cuda:1'), covar=tensor([0.0231, 0.0654, 0.0257, 0.0316, 0.0462, 0.0321, 0.0844, 0.0471], device='cuda:1'), in_proj_covar=tensor([0.0353, 0.0373, 0.0370, 0.0353, 0.0418, 0.0394, 0.0488, 0.0315], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:1') 2023-04-29 23:44:49,604 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.4820, 4.3045, 4.5395, 4.7297, 4.8882, 4.3961, 4.8596, 4.8925], device='cuda:1'), covar=tensor([0.1657, 0.1241, 0.1585, 0.0674, 0.0461, 0.0920, 0.0464, 0.0544], device='cuda:1'), in_proj_covar=tensor([0.0563, 0.0693, 0.0831, 0.0708, 0.0534, 0.0549, 0.0551, 0.0654], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-29 23:45:23,604 INFO [optim.py:368] (1/8) 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 23:45:33,732 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=137299.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 23:45:35,158 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.23 vs. limit=2.0 2023-04-29 23:45:36,785 INFO [train.py:904] (1/8) Epoch 14, batch 5350, loss[loss=0.1858, simple_loss=0.2776, pruned_loss=0.04703, over 16681.00 frames. ], tot_loss[loss=0.1833, simple_loss=0.2712, pruned_loss=0.04771, over 3206172.67 frames. ], batch size: 134, lr: 4.86e-03, grad_scale: 8.0 2023-04-29 23:46:19,240 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.2766, 4.5053, 4.7576, 4.7126, 4.7242, 4.3889, 4.0710, 4.2340], device='cuda:1'), covar=tensor([0.0493, 0.0633, 0.0546, 0.0655, 0.0761, 0.0548, 0.1591, 0.0598], device='cuda:1'), in_proj_covar=tensor([0.0354, 0.0375, 0.0372, 0.0357, 0.0420, 0.0397, 0.0492, 0.0318], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:1') 2023-04-29 23:46:40,459 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 14, batch 5400, loss[loss=0.2369, simple_loss=0.3155, pruned_loss=0.07921, over 12398.00 frames. ], tot_loss[loss=0.1849, simple_loss=0.2737, pruned_loss=0.04806, over 3212508.82 frames. ], batch size: 247, lr: 4.86e-03, grad_scale: 8.0 2023-04-29 23:47:17,229 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.85 vs. limit=2.0 2023-04-29 23:47:37,424 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.0094, 2.5912, 2.6924, 1.8843, 2.8107, 2.8835, 2.4752, 2.3674], device='cuda:1'), covar=tensor([0.0675, 0.0196, 0.0166, 0.0900, 0.0080, 0.0145, 0.0383, 0.0399], device='cuda:1'), in_proj_covar=tensor([0.0148, 0.0105, 0.0089, 0.0139, 0.0072, 0.0113, 0.0124, 0.0129], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:1') 2023-04-29 23:47:46,868 INFO [optim.py:368] (1/8) 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 23:47:55,320 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([1.9966, 2.3452, 2.3261, 2.6475, 2.0774, 3.2586, 1.7476, 2.6978], device='cuda:1'), covar=tensor([0.1088, 0.0558, 0.0939, 0.0128, 0.0113, 0.0309, 0.1295, 0.0607], device='cuda:1'), in_proj_covar=tensor([0.0157, 0.0163, 0.0185, 0.0160, 0.0200, 0.0208, 0.0187, 0.0184], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-04-29 23:48:02,464 INFO [train.py:904] (1/8) Epoch 14, batch 5450, loss[loss=0.2373, simple_loss=0.3094, pruned_loss=0.08255, over 12187.00 frames. ], tot_loss[loss=0.1881, simple_loss=0.2766, pruned_loss=0.0498, over 3203932.65 frames. ], batch size: 246, lr: 4.86e-03, grad_scale: 4.0 2023-04-29 23:49:17,303 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.75 vs. limit=2.0 2023-04-29 23:49:19,961 INFO [train.py:904] (1/8) Epoch 14, batch 5500, loss[loss=0.2711, simple_loss=0.3333, pruned_loss=0.1044, over 11567.00 frames. ], tot_loss[loss=0.1976, simple_loss=0.2848, pruned_loss=0.05524, over 3155004.86 frames. ], batch size: 248, lr: 4.86e-03, grad_scale: 4.0 2023-04-29 23:49:35,231 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=137462.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 23:49:49,631 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.2794, 1.6394, 2.0570, 2.2797, 2.3291, 2.6026, 1.7124, 2.5296], device='cuda:1'), covar=tensor([0.0188, 0.0369, 0.0240, 0.0256, 0.0232, 0.0148, 0.0382, 0.0091], device='cuda:1'), in_proj_covar=tensor([0.0168, 0.0176, 0.0160, 0.0168, 0.0176, 0.0132, 0.0178, 0.0124], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-29 23:50:23,834 INFO [optim.py:368] (1/8) 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 23:50:36,993 INFO [train.py:904] (1/8) Epoch 14, batch 5550, loss[loss=0.3331, simple_loss=0.3774, pruned_loss=0.1444, over 11026.00 frames. ], tot_loss[loss=0.2062, simple_loss=0.2918, pruned_loss=0.06029, over 3146153.34 frames. ], batch size: 248, lr: 4.86e-03, grad_scale: 4.0 2023-04-29 23:50:51,278 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=137510.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 23:51:39,986 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.0528, 2.5107, 2.6291, 1.9276, 2.6894, 2.7874, 2.4353, 2.3801], device='cuda:1'), covar=tensor([0.0599, 0.0191, 0.0209, 0.0787, 0.0091, 0.0211, 0.0424, 0.0367], device='cuda:1'), in_proj_covar=tensor([0.0146, 0.0104, 0.0089, 0.0138, 0.0071, 0.0113, 0.0123, 0.0128], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:1') 2023-04-29 23:51:55,242 INFO [train.py:904] (1/8) Epoch 14, batch 5600, loss[loss=0.2932, simple_loss=0.3487, pruned_loss=0.1188, over 11194.00 frames. ], tot_loss[loss=0.2146, simple_loss=0.2975, pruned_loss=0.06588, over 3071888.09 frames. ], batch size: 248, lr: 4.86e-03, grad_scale: 8.0 2023-04-29 23:53:03,057 INFO [optim.py:368] (1/8) 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 23:53:17,473 INFO [train.py:904] (1/8) Epoch 14, batch 5650, loss[loss=0.2487, simple_loss=0.3391, pruned_loss=0.07914, over 16332.00 frames. ], tot_loss[loss=0.2221, simple_loss=0.3032, pruned_loss=0.0705, over 3050611.80 frames. ], batch size: 165, lr: 4.86e-03, grad_scale: 8.0 2023-04-29 23:53:43,322 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([1.5741, 2.6102, 2.4707, 3.9668, 2.5951, 3.8910, 1.3568, 2.7544], device='cuda:1'), covar=tensor([0.1475, 0.0792, 0.1225, 0.0185, 0.0288, 0.0422, 0.1773, 0.0890], device='cuda:1'), in_proj_covar=tensor([0.0156, 0.0162, 0.0184, 0.0160, 0.0200, 0.0208, 0.0187, 0.0184], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-04-29 23:54:38,720 INFO [train.py:904] (1/8) Epoch 14, batch 5700, loss[loss=0.2198, simple_loss=0.3164, pruned_loss=0.06165, over 16238.00 frames. ], tot_loss[loss=0.2243, simple_loss=0.3045, pruned_loss=0.07198, over 3037795.04 frames. ], batch size: 165, lr: 4.86e-03, grad_scale: 8.0 2023-04-29 23:55:45,155 INFO [optim.py:368] (1/8) 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 23:55:59,931 INFO [train.py:904] (1/8) Epoch 14, batch 5750, loss[loss=0.2716, simple_loss=0.3268, pruned_loss=0.1082, over 10982.00 frames. ], tot_loss[loss=0.2275, simple_loss=0.3074, pruned_loss=0.07379, over 3018425.22 frames. ], batch size: 247, lr: 4.86e-03, grad_scale: 8.0 2023-04-29 23:57:21,150 INFO [train.py:904] (1/8) Epoch 14, batch 5800, loss[loss=0.2302, simple_loss=0.315, pruned_loss=0.07272, over 16545.00 frames. ], tot_loss[loss=0.2253, simple_loss=0.3068, pruned_loss=0.07196, over 3026621.15 frames. ], batch size: 57, lr: 4.86e-03, grad_scale: 8.0 2023-04-29 23:57:37,717 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.8080, 4.6419, 4.8624, 5.0321, 5.1813, 4.6228, 5.1665, 5.1818], device='cuda:1'), covar=tensor([0.1701, 0.1182, 0.1390, 0.0600, 0.0490, 0.0773, 0.0490, 0.0526], device='cuda:1'), in_proj_covar=tensor([0.0556, 0.0687, 0.0819, 0.0700, 0.0527, 0.0547, 0.0546, 0.0647], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-29 23:58:10,945 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([1.9884, 2.4692, 2.6294, 1.9210, 2.6751, 2.7907, 2.4568, 2.3851], device='cuda:1'), covar=tensor([0.0708, 0.0220, 0.0216, 0.0881, 0.0106, 0.0240, 0.0424, 0.0417], device='cuda:1'), in_proj_covar=tensor([0.0148, 0.0105, 0.0089, 0.0139, 0.0072, 0.0115, 0.0124, 0.0130], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:1') 2023-04-29 23:58:26,263 INFO [optim.py:368] (1/8) 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] (1/8) Epoch 14, batch 5850, loss[loss=0.2237, simple_loss=0.3062, pruned_loss=0.07064, over 16371.00 frames. ], tot_loss[loss=0.2224, simple_loss=0.3045, pruned_loss=0.07017, over 3024198.95 frames. ], batch size: 146, lr: 4.85e-03, grad_scale: 8.0 2023-04-30 00:00:03,509 INFO [train.py:904] (1/8) Epoch 14, batch 5900, loss[loss=0.2136, simple_loss=0.2983, pruned_loss=0.06442, over 16642.00 frames. ], tot_loss[loss=0.2211, simple_loss=0.3037, pruned_loss=0.06926, over 3049818.37 frames. ], batch size: 62, lr: 4.85e-03, grad_scale: 8.0 2023-04-30 00:00:47,282 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.3743, 4.4169, 4.2083, 3.9459, 3.9123, 4.3429, 4.1153, 4.0157], device='cuda:1'), covar=tensor([0.0589, 0.0567, 0.0291, 0.0304, 0.0894, 0.0411, 0.0565, 0.0712], device='cuda:1'), in_proj_covar=tensor([0.0255, 0.0350, 0.0304, 0.0285, 0.0323, 0.0333, 0.0207, 0.0359], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-30 00:01:05,253 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.21 vs. limit=2.0 2023-04-30 00:01:10,332 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=137892.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 00:01:10,915 INFO [optim.py:368] (1/8) 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 00:01:15,078 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 14, batch 5950, loss[loss=0.2049, simple_loss=0.2944, pruned_loss=0.05768, over 17225.00 frames. ], tot_loss[loss=0.2206, simple_loss=0.3043, pruned_loss=0.06849, over 3038924.16 frames. ], batch size: 52, lr: 4.85e-03, grad_scale: 8.0 2023-04-30 00:02:44,552 INFO [train.py:904] (1/8) Epoch 14, batch 6000, loss[loss=0.2188, simple_loss=0.3024, pruned_loss=0.06766, over 16407.00 frames. ], tot_loss[loss=0.2188, simple_loss=0.3026, pruned_loss=0.06753, over 3049781.65 frames. ], batch size: 146, lr: 4.85e-03, grad_scale: 8.0 2023-04-30 00:02:44,552 INFO [train.py:929] (1/8) Computing validation loss 2023-04-30 00:02:55,350 INFO [train.py:938] (1/8) Epoch 14, validation: loss=0.1574, simple_loss=0.2706, pruned_loss=0.0221, over 944034.00 frames. 2023-04-30 00:02:55,351 INFO [train.py:939] (1/8) Maximum memory allocated so far is 17845MB 2023-04-30 00:02:56,889 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=137953.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 00:03:01,562 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=137956.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 00:03:59,081 INFO [optim.py:368] (1/8) 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] (1/8) Epoch 14, batch 6050, loss[loss=0.2135, simple_loss=0.3066, pruned_loss=0.06019, over 16838.00 frames. ], tot_loss[loss=0.2177, simple_loss=0.3012, pruned_loss=0.06705, over 3062666.88 frames. ], batch size: 116, lr: 4.85e-03, grad_scale: 8.0 2023-04-30 00:05:40,155 INFO [train.py:904] (1/8) Epoch 14, batch 6100, loss[loss=0.1931, simple_loss=0.2761, pruned_loss=0.05507, over 16541.00 frames. ], tot_loss[loss=0.2157, simple_loss=0.3003, pruned_loss=0.06556, over 3079082.79 frames. ], batch size: 68, lr: 4.85e-03, grad_scale: 8.0 2023-04-30 00:06:45,097 INFO [optim.py:368] (1/8) 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 00:06:59,048 INFO [train.py:904] (1/8) Epoch 14, batch 6150, loss[loss=0.1793, simple_loss=0.2703, pruned_loss=0.04413, over 16774.00 frames. ], tot_loss[loss=0.2147, simple_loss=0.2984, pruned_loss=0.06545, over 3069042.59 frames. ], batch size: 83, lr: 4.85e-03, grad_scale: 8.0 2023-04-30 00:07:15,817 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.0480, 3.5756, 3.5868, 2.3517, 3.2940, 3.6022, 3.3529, 2.0006], device='cuda:1'), covar=tensor([0.0497, 0.0037, 0.0041, 0.0341, 0.0081, 0.0083, 0.0067, 0.0403], device='cuda:1'), in_proj_covar=tensor([0.0130, 0.0072, 0.0073, 0.0127, 0.0085, 0.0095, 0.0082, 0.0123], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-04-30 00:07:50,346 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.3721, 3.7340, 3.2740, 5.4258, 4.3620, 4.5845, 2.2959, 3.3631], device='cuda:1'), covar=tensor([0.1008, 0.0494, 0.0864, 0.0104, 0.0284, 0.0355, 0.1131, 0.0726], device='cuda:1'), in_proj_covar=tensor([0.0158, 0.0164, 0.0187, 0.0163, 0.0204, 0.0211, 0.0189, 0.0187], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-04-30 00:08:16,203 INFO [train.py:904] (1/8) Epoch 14, batch 6200, loss[loss=0.189, simple_loss=0.2818, pruned_loss=0.04804, over 16622.00 frames. ], tot_loss[loss=0.2138, simple_loss=0.297, pruned_loss=0.06533, over 3060142.45 frames. ], batch size: 62, lr: 4.85e-03, grad_scale: 8.0 2023-04-30 00:09:10,054 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.6358, 2.1646, 1.7796, 1.9566, 2.5060, 2.2390, 2.4889, 2.7131], device='cuda:1'), covar=tensor([0.0135, 0.0339, 0.0450, 0.0407, 0.0196, 0.0315, 0.0175, 0.0223], device='cuda:1'), in_proj_covar=tensor([0.0168, 0.0211, 0.0207, 0.0207, 0.0212, 0.0213, 0.0218, 0.0206], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-30 00:09:12,880 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.41 vs. limit=2.0 2023-04-30 00:09:17,889 INFO [optim.py:368] (1/8) 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] (1/8) Epoch 14, batch 6250, loss[loss=0.2537, simple_loss=0.3213, pruned_loss=0.09299, over 11640.00 frames. ], tot_loss[loss=0.2127, simple_loss=0.2961, pruned_loss=0.06464, over 3073558.83 frames. ], batch size: 246, lr: 4.85e-03, grad_scale: 8.0 2023-04-30 00:10:07,636 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.8948, 2.0566, 2.3743, 3.1244, 2.1584, 2.2382, 2.2564, 2.1381], device='cuda:1'), covar=tensor([0.1138, 0.3133, 0.2040, 0.0614, 0.3664, 0.2315, 0.2937, 0.3169], device='cuda:1'), in_proj_covar=tensor([0.0372, 0.0409, 0.0340, 0.0318, 0.0417, 0.0471, 0.0374, 0.0477], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-30 00:10:42,264 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=138248.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 00:10:42,424 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.2431, 3.2034, 3.4807, 1.6393, 3.5858, 3.6688, 2.8604, 2.6907], device='cuda:1'), covar=tensor([0.0835, 0.0219, 0.0161, 0.1268, 0.0064, 0.0151, 0.0412, 0.0474], device='cuda:1'), in_proj_covar=tensor([0.0147, 0.0104, 0.0089, 0.0138, 0.0072, 0.0113, 0.0123, 0.0128], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:1') 2023-04-30 00:10:46,686 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 14, batch 6300, loss[loss=0.2082, simple_loss=0.2938, pruned_loss=0.06129, over 15173.00 frames. ], tot_loss[loss=0.2112, simple_loss=0.2955, pruned_loss=0.06349, over 3089558.88 frames. ], batch size: 190, lr: 4.85e-03, grad_scale: 8.0 2023-04-30 00:11:17,718 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=138270.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 00:11:23,561 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.3771, 4.4753, 4.6209, 4.4650, 4.4779, 5.0080, 4.5656, 4.2894], device='cuda:1'), covar=tensor([0.1515, 0.1783, 0.2118, 0.1816, 0.2610, 0.1059, 0.1569, 0.2617], device='cuda:1'), in_proj_covar=tensor([0.0374, 0.0524, 0.0575, 0.0444, 0.0599, 0.0592, 0.0453, 0.0602], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-04-30 00:11:52,435 INFO [optim.py:368] (1/8) 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 00:12:05,887 INFO [train.py:904] (1/8) Epoch 14, batch 6350, loss[loss=0.2035, simple_loss=0.2812, pruned_loss=0.06287, over 16990.00 frames. ], tot_loss[loss=0.2139, simple_loss=0.2974, pruned_loss=0.0652, over 3065572.21 frames. ], batch size: 41, lr: 4.85e-03, grad_scale: 8.0 2023-04-30 00:12:36,944 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=138322.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 00:12:50,066 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=138331.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 00:13:04,938 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.5189, 4.4811, 4.3859, 3.7094, 4.4449, 1.6542, 4.2059, 4.0737], device='cuda:1'), covar=tensor([0.0091, 0.0082, 0.0155, 0.0324, 0.0081, 0.2522, 0.0113, 0.0209], device='cuda:1'), in_proj_covar=tensor([0.0141, 0.0130, 0.0174, 0.0163, 0.0148, 0.0187, 0.0163, 0.0156], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-30 00:13:17,938 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.5358, 3.9082, 2.9371, 2.1544, 2.6102, 2.4107, 4.0697, 3.4879], device='cuda:1'), covar=tensor([0.2925, 0.0603, 0.1651, 0.2583, 0.2624, 0.1881, 0.0421, 0.1074], device='cuda:1'), in_proj_covar=tensor([0.0311, 0.0262, 0.0290, 0.0291, 0.0285, 0.0232, 0.0278, 0.0309], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-30 00:13:22,144 INFO [train.py:904] (1/8) Epoch 14, batch 6400, loss[loss=0.1786, simple_loss=0.2651, pruned_loss=0.04607, over 16943.00 frames. ], tot_loss[loss=0.2152, simple_loss=0.2977, pruned_loss=0.06632, over 3079856.22 frames. ], batch size: 58, lr: 4.84e-03, grad_scale: 8.0 2023-04-30 00:13:24,525 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([1.7143, 2.7716, 2.4051, 3.8180, 2.7769, 3.8663, 1.5388, 2.9060], device='cuda:1'), covar=tensor([0.1292, 0.0649, 0.1159, 0.0141, 0.0206, 0.0384, 0.1523, 0.0709], device='cuda:1'), in_proj_covar=tensor([0.0158, 0.0164, 0.0186, 0.0163, 0.0203, 0.0211, 0.0189, 0.0186], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-04-30 00:14:09,346 INFO [zipformer.py:625] (1/8) 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] (1/8) 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 00:14:37,174 INFO [train.py:904] (1/8) Epoch 14, batch 6450, loss[loss=0.2203, simple_loss=0.2876, pruned_loss=0.07651, over 11388.00 frames. ], tot_loss[loss=0.214, simple_loss=0.2971, pruned_loss=0.06539, over 3072598.99 frames. ], batch size: 247, lr: 4.84e-03, grad_scale: 8.0 2023-04-30 00:15:54,743 INFO [train.py:904] (1/8) Epoch 14, batch 6500, loss[loss=0.2224, simple_loss=0.3025, pruned_loss=0.07108, over 16539.00 frames. ], tot_loss[loss=0.2124, simple_loss=0.2953, pruned_loss=0.06474, over 3072597.96 frames. ], batch size: 75, lr: 4.84e-03, grad_scale: 4.0 2023-04-30 00:16:59,451 INFO [optim.py:368] (1/8) 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] (1/8) Epoch 14, batch 6550, loss[loss=0.1951, simple_loss=0.2968, pruned_loss=0.0467, over 17055.00 frames. ], tot_loss[loss=0.2134, simple_loss=0.2971, pruned_loss=0.06479, over 3086013.91 frames. ], batch size: 50, lr: 4.84e-03, grad_scale: 4.0 2023-04-30 00:18:22,333 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=138548.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 00:18:25,722 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=138551.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 00:18:26,468 INFO [train.py:904] (1/8) Epoch 14, batch 6600, loss[loss=0.2549, simple_loss=0.3234, pruned_loss=0.09322, over 11909.00 frames. ], tot_loss[loss=0.2155, simple_loss=0.2996, pruned_loss=0.06571, over 3081406.51 frames. ], batch size: 247, lr: 4.84e-03, grad_scale: 4.0 2023-04-30 00:19:29,572 INFO [optim.py:368] (1/8) Clipping_scale=2.0, grad-norm quartiles 2.028e+02 2.801e+02 3.547e+02 4.436e+02 1.538e+03, threshold=7.095e+02, percent-clipped=5.0 2023-04-30 00:19:33,985 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=138596.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 00:19:38,207 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 14, batch 6650, loss[loss=0.246, simple_loss=0.3144, pruned_loss=0.08878, over 11227.00 frames. ], tot_loss[loss=0.2169, simple_loss=0.3006, pruned_loss=0.06658, over 3077355.03 frames. ], batch size: 247, lr: 4.84e-03, grad_scale: 4.0 2023-04-30 00:19:50,684 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.4728, 4.2716, 4.2006, 2.7598, 3.7620, 4.2258, 3.7417, 2.1513], device='cuda:1'), covar=tensor([0.0497, 0.0023, 0.0031, 0.0352, 0.0077, 0.0067, 0.0065, 0.0442], device='cuda:1'), in_proj_covar=tensor([0.0133, 0.0074, 0.0074, 0.0132, 0.0087, 0.0097, 0.0085, 0.0126], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-04-30 00:20:19,709 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=138626.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 00:20:58,166 INFO [train.py:904] (1/8) Epoch 14, batch 6700, loss[loss=0.2376, simple_loss=0.3249, pruned_loss=0.07516, over 16356.00 frames. ], tot_loss[loss=0.2161, simple_loss=0.2991, pruned_loss=0.06659, over 3068719.13 frames. ], batch size: 146, lr: 4.84e-03, grad_scale: 4.0 2023-04-30 00:21:39,963 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=138678.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 00:21:44,875 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=138681.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 00:21:53,014 INFO [zipformer.py:625] (1/8) 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] (1/8) 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 00:22:06,949 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.3305, 2.1955, 2.2914, 4.1777, 2.1111, 2.6048, 2.2684, 2.3574], device='cuda:1'), covar=tensor([0.1023, 0.3230, 0.2542, 0.0404, 0.3878, 0.2174, 0.3156, 0.3179], device='cuda:1'), in_proj_covar=tensor([0.0370, 0.0408, 0.0339, 0.0317, 0.0417, 0.0470, 0.0373, 0.0476], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-30 00:22:08,604 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=138697.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 00:22:15,572 INFO [train.py:904] (1/8) Epoch 14, batch 6750, loss[loss=0.1894, simple_loss=0.2678, pruned_loss=0.0555, over 16697.00 frames. ], tot_loss[loss=0.2149, simple_loss=0.2975, pruned_loss=0.06613, over 3089133.74 frames. ], batch size: 57, lr: 4.84e-03, grad_scale: 4.0 2023-04-30 00:22:58,575 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.20 vs. limit=2.0 2023-04-30 00:23:14,916 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=138742.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 00:23:22,361 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=138747.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 00:23:29,044 INFO [train.py:904] (1/8) Epoch 14, batch 6800, loss[loss=0.2014, simple_loss=0.2924, pruned_loss=0.05519, over 16754.00 frames. ], tot_loss[loss=0.2162, simple_loss=0.2985, pruned_loss=0.06701, over 3070230.07 frames. ], batch size: 134, lr: 4.84e-03, grad_scale: 8.0 2023-04-30 00:23:38,444 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=138758.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 00:24:18,313 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([5.4619, 5.4008, 5.2984, 4.9651, 4.9215, 5.3754, 5.3442, 5.0155], device='cuda:1'), covar=tensor([0.0578, 0.0370, 0.0256, 0.0285, 0.0991, 0.0354, 0.0205, 0.0609], device='cuda:1'), in_proj_covar=tensor([0.0251, 0.0347, 0.0303, 0.0281, 0.0319, 0.0328, 0.0207, 0.0358], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-30 00:24:34,694 INFO [optim.py:368] (1/8) Clipping_scale=2.0, grad-norm quartiles 1.741e+02 2.812e+02 3.433e+02 4.059e+02 1.030e+03, threshold=6.866e+02, percent-clipped=2.0 2023-04-30 00:24:46,910 INFO [train.py:904] (1/8) Epoch 14, batch 6850, loss[loss=0.1955, simple_loss=0.3043, pruned_loss=0.04334, over 16459.00 frames. ], tot_loss[loss=0.218, simple_loss=0.3002, pruned_loss=0.06791, over 3059722.69 frames. ], batch size: 68, lr: 4.84e-03, grad_scale: 8.0 2023-04-30 00:26:02,477 INFO [train.py:904] (1/8) Epoch 14, batch 6900, loss[loss=0.242, simple_loss=0.3229, pruned_loss=0.08058, over 16636.00 frames. ], tot_loss[loss=0.2186, simple_loss=0.3027, pruned_loss=0.06726, over 3074800.27 frames. ], batch size: 134, lr: 4.84e-03, grad_scale: 8.0 2023-04-30 00:27:10,348 INFO [optim.py:368] (1/8) 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] (1/8) Epoch 14, batch 6950, loss[loss=0.2563, simple_loss=0.3207, pruned_loss=0.09599, over 11437.00 frames. ], tot_loss[loss=0.2214, simple_loss=0.3044, pruned_loss=0.06917, over 3053892.34 frames. ], batch size: 248, lr: 4.84e-03, grad_scale: 8.0 2023-04-30 00:27:33,762 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=138909.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 00:27:43,696 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=4.55 vs. limit=5.0 2023-04-30 00:27:59,934 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=138926.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 00:28:38,224 INFO [train.py:904] (1/8) Epoch 14, batch 7000, loss[loss=0.1779, simple_loss=0.2803, pruned_loss=0.03777, over 17025.00 frames. ], tot_loss[loss=0.2203, simple_loss=0.3038, pruned_loss=0.06841, over 3048350.41 frames. ], batch size: 50, lr: 4.83e-03, grad_scale: 8.0 2023-04-30 00:29:06,870 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=138970.0, num_to_drop=1, layers_to_drop={0} 2023-04-30 00:29:12,263 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=138974.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 00:29:17,817 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.44 vs. limit=2.0 2023-04-30 00:29:18,504 INFO [zipformer.py:625] (1/8) 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] (1/8) 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 00:29:55,148 INFO [train.py:904] (1/8) Epoch 14, batch 7050, loss[loss=0.1974, simple_loss=0.2875, pruned_loss=0.05368, over 16340.00 frames. ], tot_loss[loss=0.2196, simple_loss=0.3043, pruned_loss=0.06746, over 3070735.27 frames. ], batch size: 146, lr: 4.83e-03, grad_scale: 8.0 2023-04-30 00:30:33,407 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=139026.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 00:30:34,828 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.0820, 1.9283, 2.3547, 2.9206, 2.8683, 3.3388, 1.9552, 3.2245], device='cuda:1'), covar=tensor([0.0147, 0.0391, 0.0281, 0.0228, 0.0220, 0.0126, 0.0437, 0.0106], device='cuda:1'), in_proj_covar=tensor([0.0162, 0.0174, 0.0159, 0.0164, 0.0172, 0.0131, 0.0176, 0.0122], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001], device='cuda:1') 2023-04-30 00:30:45,031 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.76 vs. limit=2.0 2023-04-30 00:30:49,921 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=139037.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 00:30:58,294 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 14, batch 7100, loss[loss=0.2168, simple_loss=0.3036, pruned_loss=0.06504, over 15483.00 frames. ], tot_loss[loss=0.2187, simple_loss=0.303, pruned_loss=0.06724, over 3066146.87 frames. ], batch size: 191, lr: 4.83e-03, grad_scale: 8.0 2023-04-30 00:31:15,008 INFO [zipformer.py:625] (1/8) 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] (1/8) 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] (1/8) Epoch 14, batch 7150, loss[loss=0.2082, simple_loss=0.2946, pruned_loss=0.06088, over 16429.00 frames. ], tot_loss[loss=0.217, simple_loss=0.3007, pruned_loss=0.06659, over 3077404.25 frames. ], batch size: 146, lr: 4.83e-03, grad_scale: 8.0 2023-04-30 00:33:22,282 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.7769, 3.6875, 3.9341, 3.6929, 3.8892, 4.2376, 3.9262, 3.6785], device='cuda:1'), covar=tensor([0.2220, 0.2187, 0.2063, 0.2459, 0.2848, 0.1978, 0.1488, 0.2619], device='cuda:1'), in_proj_covar=tensor([0.0378, 0.0531, 0.0581, 0.0448, 0.0603, 0.0603, 0.0458, 0.0609], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-04-30 00:33:47,876 INFO [train.py:904] (1/8) Epoch 14, batch 7200, loss[loss=0.1651, simple_loss=0.2628, pruned_loss=0.03365, over 16782.00 frames. ], tot_loss[loss=0.2134, simple_loss=0.2981, pruned_loss=0.06436, over 3092471.49 frames. ], batch size: 89, lr: 4.83e-03, grad_scale: 8.0 2023-04-30 00:34:15,489 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.6135, 3.6858, 2.9181, 2.2330, 2.5379, 2.2323, 3.9581, 3.3910], device='cuda:1'), covar=tensor([0.2730, 0.0721, 0.1595, 0.2342, 0.2306, 0.1932, 0.0403, 0.1096], device='cuda:1'), in_proj_covar=tensor([0.0313, 0.0263, 0.0291, 0.0293, 0.0287, 0.0233, 0.0277, 0.0311], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:1') 2023-04-30 00:34:17,917 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.9100, 1.9882, 2.3634, 3.1534, 2.1359, 2.2320, 2.3056, 2.1149], device='cuda:1'), covar=tensor([0.1157, 0.3347, 0.2286, 0.0703, 0.4046, 0.2352, 0.2919, 0.3339], device='cuda:1'), in_proj_covar=tensor([0.0371, 0.0409, 0.0340, 0.0319, 0.0419, 0.0470, 0.0375, 0.0477], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-30 00:34:55,354 INFO [optim.py:368] (1/8) 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] (1/8) Epoch 14, batch 7250, loss[loss=0.2318, simple_loss=0.3014, pruned_loss=0.08111, over 11790.00 frames. ], tot_loss[loss=0.2121, simple_loss=0.2965, pruned_loss=0.06383, over 3081424.21 frames. ], batch size: 248, lr: 4.83e-03, grad_scale: 4.0 2023-04-30 00:35:56,943 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.43 vs. limit=2.0 2023-04-30 00:36:01,860 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=139238.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 00:36:03,286 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([1.7204, 2.7421, 2.3434, 4.3926, 3.2136, 3.9993, 1.6386, 2.6934], device='cuda:1'), covar=tensor([0.1257, 0.0697, 0.1257, 0.0149, 0.0322, 0.0472, 0.1401, 0.0959], device='cuda:1'), in_proj_covar=tensor([0.0159, 0.0164, 0.0186, 0.0162, 0.0203, 0.0210, 0.0188, 0.0187], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-04-30 00:36:22,183 INFO [train.py:904] (1/8) Epoch 14, batch 7300, loss[loss=0.2085, simple_loss=0.2983, pruned_loss=0.05933, over 16936.00 frames. ], tot_loss[loss=0.2119, simple_loss=0.2964, pruned_loss=0.06367, over 3091311.59 frames. ], batch size: 109, lr: 4.83e-03, grad_scale: 4.0 2023-04-30 00:36:42,737 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=139265.0, num_to_drop=1, layers_to_drop={0} 2023-04-30 00:37:00,576 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=139276.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 00:37:20,434 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=139289.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 00:37:29,626 INFO [optim.py:368] (1/8) 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 00:37:37,020 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=139299.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 00:37:38,923 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.3294, 3.2352, 2.6487, 2.0383, 2.2250, 2.2051, 3.2827, 3.0029], device='cuda:1'), covar=tensor([0.2967, 0.0752, 0.1709, 0.2354, 0.2372, 0.1952, 0.0524, 0.1140], device='cuda:1'), in_proj_covar=tensor([0.0312, 0.0262, 0.0291, 0.0292, 0.0287, 0.0233, 0.0277, 0.0310], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-30 00:37:40,533 INFO [train.py:904] (1/8) Epoch 14, batch 7350, loss[loss=0.2343, simple_loss=0.3113, pruned_loss=0.07865, over 16670.00 frames. ], tot_loss[loss=0.2126, simple_loss=0.2967, pruned_loss=0.06429, over 3081366.39 frames. ], batch size: 76, lr: 4.83e-03, grad_scale: 4.0 2023-04-30 00:38:36,865 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=139337.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 00:38:36,897 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=139337.0, num_to_drop=1, layers_to_drop={2} 2023-04-30 00:38:45,074 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=139342.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 00:38:51,901 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.7147, 4.9128, 5.0806, 4.9123, 4.9150, 5.4643, 4.9269, 4.6940], device='cuda:1'), covar=tensor([0.1018, 0.1704, 0.1895, 0.1757, 0.2355, 0.0883, 0.1469, 0.2329], device='cuda:1'), in_proj_covar=tensor([0.0375, 0.0523, 0.0577, 0.0444, 0.0595, 0.0599, 0.0453, 0.0604], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-04-30 00:38:56,603 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=139350.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 00:38:59,356 INFO [train.py:904] (1/8) Epoch 14, batch 7400, loss[loss=0.2163, simple_loss=0.3044, pruned_loss=0.06406, over 16505.00 frames. ], tot_loss[loss=0.2128, simple_loss=0.2969, pruned_loss=0.06432, over 3095660.33 frames. ], batch size: 68, lr: 4.83e-03, grad_scale: 4.0 2023-04-30 00:39:01,636 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=139353.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 00:39:49,158 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.31 vs. limit=2.0 2023-04-30 00:39:52,861 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=139385.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 00:40:00,700 INFO [zipformer.py:625] (1/8) 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] (1/8) 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] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=139401.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 00:40:19,477 INFO [train.py:904] (1/8) Epoch 14, batch 7450, loss[loss=0.2102, simple_loss=0.3031, pruned_loss=0.05871, over 16832.00 frames. ], tot_loss[loss=0.2144, simple_loss=0.2981, pruned_loss=0.0653, over 3097346.40 frames. ], batch size: 116, lr: 4.83e-03, grad_scale: 4.0 2023-04-30 00:41:42,946 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.86 vs. limit=2.0 2023-04-30 00:41:43,385 INFO [train.py:904] (1/8) Epoch 14, batch 7500, loss[loss=0.1814, simple_loss=0.2753, pruned_loss=0.04373, over 16685.00 frames. ], tot_loss[loss=0.2145, simple_loss=0.2987, pruned_loss=0.06514, over 3108295.22 frames. ], batch size: 89, lr: 4.83e-03, grad_scale: 4.0 2023-04-30 00:42:47,025 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([1.8489, 2.9682, 2.8218, 5.0117, 3.8944, 4.3346, 1.7355, 3.1856], device='cuda:1'), covar=tensor([0.1192, 0.0670, 0.1042, 0.0129, 0.0320, 0.0352, 0.1394, 0.0712], device='cuda:1'), in_proj_covar=tensor([0.0159, 0.0164, 0.0186, 0.0162, 0.0203, 0.0210, 0.0188, 0.0186], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-04-30 00:42:53,287 INFO [optim.py:368] (1/8) Clipping_scale=2.0, grad-norm quartiles 2.056e+02 3.028e+02 3.537e+02 4.351e+02 9.459e+02, threshold=7.073e+02, percent-clipped=3.0 2023-04-30 00:43:02,955 INFO [train.py:904] (1/8) Epoch 14, batch 7550, loss[loss=0.204, simple_loss=0.2903, pruned_loss=0.05886, over 16784.00 frames. ], tot_loss[loss=0.2146, simple_loss=0.2979, pruned_loss=0.06567, over 3100596.13 frames. ], batch size: 124, lr: 4.82e-03, grad_scale: 4.0 2023-04-30 00:43:16,540 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=4.48 vs. limit=5.0 2023-04-30 00:44:14,254 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.4438, 2.9838, 3.0011, 1.9686, 2.6894, 2.0910, 3.0742, 3.1937], device='cuda:1'), covar=tensor([0.0282, 0.0722, 0.0588, 0.1898, 0.0828, 0.0991, 0.0654, 0.0778], device='cuda:1'), in_proj_covar=tensor([0.0148, 0.0151, 0.0161, 0.0147, 0.0139, 0.0126, 0.0139, 0.0162], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:1') 2023-04-30 00:44:19,040 INFO [train.py:904] (1/8) Epoch 14, batch 7600, loss[loss=0.1994, simple_loss=0.2889, pruned_loss=0.05492, over 16529.00 frames. ], tot_loss[loss=0.2146, simple_loss=0.2973, pruned_loss=0.06589, over 3092325.93 frames. ], batch size: 75, lr: 4.82e-03, grad_scale: 8.0 2023-04-30 00:44:39,705 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=139565.0, num_to_drop=1, layers_to_drop={0} 2023-04-30 00:44:48,983 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=139571.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 00:44:55,039 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.6046, 3.7721, 2.8574, 2.1861, 2.5000, 2.3787, 3.8962, 3.3625], device='cuda:1'), covar=tensor([0.2775, 0.0695, 0.1644, 0.2473, 0.2422, 0.1933, 0.0462, 0.1123], device='cuda:1'), in_proj_covar=tensor([0.0313, 0.0263, 0.0292, 0.0293, 0.0287, 0.0234, 0.0276, 0.0311], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:1') 2023-04-30 00:45:05,773 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=139582.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 00:45:23,447 INFO [zipformer.py:625] (1/8) 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] (1/8) Clipping_scale=2.0, grad-norm quartiles 1.790e+02 2.745e+02 3.435e+02 4.599e+02 7.932e+02, threshold=6.869e+02, percent-clipped=2.0 2023-04-30 00:45:34,160 INFO [train.py:904] (1/8) Epoch 14, batch 7650, loss[loss=0.2082, simple_loss=0.3005, pruned_loss=0.05796, over 16741.00 frames. ], tot_loss[loss=0.2142, simple_loss=0.2973, pruned_loss=0.06549, over 3113308.26 frames. ], batch size: 83, lr: 4.82e-03, grad_scale: 8.0 2023-04-30 00:45:50,717 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=139613.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 00:46:17,153 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.46 vs. limit=2.0 2023-04-30 00:46:19,134 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=139632.0, num_to_drop=1, layers_to_drop={2} 2023-04-30 00:46:19,286 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=139632.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 00:46:30,108 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.0172, 2.3399, 2.2852, 2.8237, 1.9477, 3.2068, 1.8307, 2.7083], device='cuda:1'), covar=tensor([0.1155, 0.0627, 0.1076, 0.0181, 0.0153, 0.0381, 0.1359, 0.0671], device='cuda:1'), in_proj_covar=tensor([0.0160, 0.0165, 0.0187, 0.0163, 0.0205, 0.0211, 0.0189, 0.0187], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-04-30 00:46:33,427 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.4865, 4.5244, 4.8954, 4.8685, 4.8629, 4.5558, 4.5293, 4.3172], device='cuda:1'), covar=tensor([0.0284, 0.0442, 0.0328, 0.0355, 0.0418, 0.0326, 0.0899, 0.0534], device='cuda:1'), in_proj_covar=tensor([0.0352, 0.0369, 0.0370, 0.0354, 0.0415, 0.0395, 0.0482, 0.0314], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:1') 2023-04-30 00:46:36,597 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=139643.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 00:46:39,467 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=139645.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 00:46:43,116 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.17 vs. limit=2.0 2023-04-30 00:46:49,447 INFO [train.py:904] (1/8) Epoch 14, batch 7700, loss[loss=0.1919, simple_loss=0.2881, pruned_loss=0.04781, over 16906.00 frames. ], tot_loss[loss=0.2153, simple_loss=0.2979, pruned_loss=0.06631, over 3112510.14 frames. ], batch size: 102, lr: 4.82e-03, grad_scale: 8.0 2023-04-30 00:47:57,375 INFO [optim.py:368] (1/8) 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 00:48:01,254 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([1.9823, 2.3009, 2.3397, 2.7705, 2.0043, 3.2224, 1.7950, 2.6945], device='cuda:1'), covar=tensor([0.1102, 0.0661, 0.0994, 0.0154, 0.0127, 0.0350, 0.1319, 0.0672], device='cuda:1'), in_proj_covar=tensor([0.0159, 0.0164, 0.0186, 0.0162, 0.0204, 0.0210, 0.0188, 0.0186], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-04-30 00:48:06,799 INFO [train.py:904] (1/8) Epoch 14, batch 7750, loss[loss=0.2047, simple_loss=0.2936, pruned_loss=0.05791, over 16984.00 frames. ], tot_loss[loss=0.2146, simple_loss=0.2979, pruned_loss=0.06566, over 3117484.22 frames. ], batch size: 41, lr: 4.82e-03, grad_scale: 4.0 2023-04-30 00:49:10,246 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.1727, 4.2390, 2.8227, 4.9297, 3.2269, 4.8478, 2.7677, 3.4164], device='cuda:1'), covar=tensor([0.0207, 0.0287, 0.1390, 0.0174, 0.0683, 0.0402, 0.1362, 0.0648], device='cuda:1'), in_proj_covar=tensor([0.0159, 0.0169, 0.0190, 0.0139, 0.0166, 0.0209, 0.0196, 0.0173], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0003, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-04-30 00:49:24,608 INFO [train.py:904] (1/8) Epoch 14, batch 7800, loss[loss=0.1964, simple_loss=0.2847, pruned_loss=0.05405, over 16787.00 frames. ], tot_loss[loss=0.2163, simple_loss=0.2992, pruned_loss=0.06674, over 3109331.59 frames. ], batch size: 83, lr: 4.82e-03, grad_scale: 4.0 2023-04-30 00:49:51,693 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=139769.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 00:50:22,483 INFO [zipformer.py:625] (1/8) 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] (1/8) 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 00:50:41,155 INFO [train.py:904] (1/8) Epoch 14, batch 7850, loss[loss=0.2064, simple_loss=0.3006, pruned_loss=0.05613, over 16487.00 frames. ], tot_loss[loss=0.2168, simple_loss=0.3001, pruned_loss=0.06672, over 3105841.77 frames. ], batch size: 68, lr: 4.82e-03, grad_scale: 4.0 2023-04-30 00:50:45,419 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=139804.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 00:50:48,469 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=139806.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 00:51:24,536 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=139830.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 00:51:53,587 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=139850.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 00:51:53,685 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 14, batch 7900, loss[loss=0.238, simple_loss=0.3091, pruned_loss=0.08346, over 11411.00 frames. ], tot_loss[loss=0.2154, simple_loss=0.2987, pruned_loss=0.06603, over 3093436.02 frames. ], batch size: 246, lr: 4.82e-03, grad_scale: 4.0 2023-04-30 00:52:15,825 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=139865.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 00:52:18,320 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=139867.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 00:53:01,516 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=139894.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 00:53:03,656 INFO [optim.py:368] (1/8) 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] (1/8) Epoch 14, batch 7950, loss[loss=0.2225, simple_loss=0.304, pruned_loss=0.07048, over 15310.00 frames. ], tot_loss[loss=0.2156, simple_loss=0.2989, pruned_loss=0.06615, over 3096287.68 frames. ], batch size: 190, lr: 4.82e-03, grad_scale: 4.0 2023-04-30 00:53:27,380 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=139911.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 00:53:32,084 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([5.0288, 4.9792, 4.8278, 4.1361, 4.8556, 1.8312, 4.6197, 4.6968], device='cuda:1'), covar=tensor([0.0085, 0.0084, 0.0163, 0.0448, 0.0100, 0.2514, 0.0129, 0.0195], device='cuda:1'), in_proj_covar=tensor([0.0140, 0.0127, 0.0173, 0.0162, 0.0146, 0.0188, 0.0161, 0.0156], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-30 00:53:36,023 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.4476, 3.4021, 3.7994, 1.7317, 3.9108, 3.9824, 2.9224, 2.9190], device='cuda:1'), covar=tensor([0.0763, 0.0229, 0.0163, 0.1280, 0.0059, 0.0118, 0.0421, 0.0401], device='cuda:1'), in_proj_covar=tensor([0.0147, 0.0104, 0.0090, 0.0139, 0.0072, 0.0112, 0.0124, 0.0128], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:1') 2023-04-30 00:53:51,382 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=139927.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 00:53:59,019 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=139932.0, num_to_drop=1, layers_to_drop={2} 2023-04-30 00:54:08,379 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=139938.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 00:54:13,649 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=139942.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 00:54:17,744 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 14, batch 8000, loss[loss=0.228, simple_loss=0.3123, pruned_loss=0.0719, over 15273.00 frames. ], tot_loss[loss=0.2157, simple_loss=0.299, pruned_loss=0.06615, over 3110829.76 frames. ], batch size: 190, lr: 4.82e-03, grad_scale: 8.0 2023-04-30 00:55:10,467 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=139980.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 00:55:30,334 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=139993.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 00:55:34,105 INFO [optim.py:368] (1/8) 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] (1/8) Epoch 14, batch 8050, loss[loss=0.2077, simple_loss=0.2985, pruned_loss=0.05846, over 16316.00 frames. ], tot_loss[loss=0.2158, simple_loss=0.2992, pruned_loss=0.06618, over 3110809.43 frames. ], batch size: 146, lr: 4.82e-03, grad_scale: 8.0 2023-04-30 00:56:59,040 INFO [train.py:904] (1/8) Epoch 14, batch 8100, loss[loss=0.219, simple_loss=0.3064, pruned_loss=0.06579, over 16272.00 frames. ], tot_loss[loss=0.2156, simple_loss=0.2992, pruned_loss=0.06603, over 3108490.27 frames. ], batch size: 35, lr: 4.82e-03, grad_scale: 8.0 2023-04-30 00:58:06,571 INFO [optim.py:368] (1/8) 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 00:58:17,060 INFO [train.py:904] (1/8) Epoch 14, batch 8150, loss[loss=0.1803, simple_loss=0.267, pruned_loss=0.0468, over 16588.00 frames. ], tot_loss[loss=0.2127, simple_loss=0.2962, pruned_loss=0.06456, over 3129022.00 frames. ], batch size: 76, lr: 4.81e-03, grad_scale: 8.0 2023-04-30 00:58:18,980 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=140103.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 00:58:28,539 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.19 vs. limit=2.0 2023-04-30 00:58:29,384 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=140110.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 00:58:49,736 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.5496, 2.3946, 2.4471, 4.2146, 2.2717, 2.7461, 2.4570, 2.6028], device='cuda:1'), covar=tensor([0.1026, 0.3081, 0.2347, 0.0437, 0.3591, 0.2138, 0.2849, 0.2985], device='cuda:1'), in_proj_covar=tensor([0.0370, 0.0409, 0.0338, 0.0318, 0.0418, 0.0468, 0.0372, 0.0476], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-30 00:58:52,074 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=140125.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 00:59:23,723 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=140145.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 00:59:34,235 INFO [train.py:904] (1/8) Epoch 14, batch 8200, loss[loss=0.2193, simple_loss=0.3004, pruned_loss=0.06912, over 11258.00 frames. ], tot_loss[loss=0.2102, simple_loss=0.2935, pruned_loss=0.06345, over 3130441.86 frames. ], batch size: 246, lr: 4.81e-03, grad_scale: 8.0 2023-04-30 00:59:47,415 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=140160.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 00:59:50,114 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=140162.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 00:59:52,820 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=3.18 vs. limit=5.0 2023-04-30 00:59:53,860 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=140164.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 01:00:05,651 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=140171.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 01:00:44,843 INFO [optim.py:368] (1/8) 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 01:00:55,629 INFO [train.py:904] (1/8) Epoch 14, batch 8250, loss[loss=0.178, simple_loss=0.271, pruned_loss=0.04248, over 16893.00 frames. ], tot_loss[loss=0.2074, simple_loss=0.2925, pruned_loss=0.06114, over 3104770.40 frames. ], batch size: 116, lr: 4.81e-03, grad_scale: 8.0 2023-04-30 01:01:02,325 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=140206.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 01:01:37,259 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=140227.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 01:01:48,030 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.4159, 3.0458, 2.6903, 2.1618, 2.2030, 2.2057, 2.8101, 2.8560], device='cuda:1'), covar=tensor([0.2323, 0.0697, 0.1394, 0.2614, 0.2554, 0.2051, 0.0452, 0.1183], device='cuda:1'), in_proj_covar=tensor([0.0307, 0.0257, 0.0286, 0.0287, 0.0281, 0.0230, 0.0271, 0.0305], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-30 01:01:55,101 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 14, batch 8300, loss[loss=0.1751, simple_loss=0.2614, pruned_loss=0.04434, over 12236.00 frames. ], tot_loss[loss=0.2024, simple_loss=0.2889, pruned_loss=0.05794, over 3077407.89 frames. ], batch size: 248, lr: 4.81e-03, grad_scale: 8.0 2023-04-30 01:02:17,071 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.1131, 1.5499, 1.9222, 2.1148, 2.3263, 2.3994, 1.7045, 2.3241], device='cuda:1'), covar=tensor([0.0185, 0.0406, 0.0247, 0.0246, 0.0246, 0.0176, 0.0378, 0.0115], device='cuda:1'), in_proj_covar=tensor([0.0163, 0.0174, 0.0158, 0.0164, 0.0173, 0.0131, 0.0177, 0.0120], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001], device='cuda:1') 2023-04-30 01:02:17,388 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.27 vs. limit=2.0 2023-04-30 01:02:28,945 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=140259.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 01:02:54,964 INFO [zipformer.py:625] (1/8) 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] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=140286.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 01:03:12,990 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.4551, 4.3231, 4.8256, 2.4133, 5.0458, 5.0622, 3.7373, 3.9538], device='cuda:1'), covar=tensor([0.0541, 0.0170, 0.0114, 0.1032, 0.0027, 0.0067, 0.0254, 0.0290], device='cuda:1'), in_proj_covar=tensor([0.0146, 0.0104, 0.0090, 0.0138, 0.0072, 0.0112, 0.0123, 0.0127], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:1') 2023-04-30 01:03:27,882 INFO [optim.py:368] (1/8) 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 01:03:36,676 INFO [train.py:904] (1/8) Epoch 14, batch 8350, loss[loss=0.1718, simple_loss=0.2741, pruned_loss=0.03471, over 16856.00 frames. ], tot_loss[loss=0.1997, simple_loss=0.288, pruned_loss=0.05564, over 3091949.82 frames. ], batch size: 102, lr: 4.81e-03, grad_scale: 8.0 2023-04-30 01:03:37,639 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=4.83 vs. limit=5.0 2023-04-30 01:04:07,898 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=140320.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 01:04:57,945 INFO [train.py:904] (1/8) Epoch 14, batch 8400, loss[loss=0.1791, simple_loss=0.2613, pruned_loss=0.04845, over 12012.00 frames. ], tot_loss[loss=0.1968, simple_loss=0.2856, pruned_loss=0.05402, over 3072481.47 frames. ], batch size: 246, lr: 4.81e-03, grad_scale: 8.0 2023-04-30 01:05:40,920 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.9390, 4.2042, 4.0577, 4.0972, 3.7583, 3.8631, 3.8294, 4.1941], device='cuda:1'), covar=tensor([0.1102, 0.0925, 0.1037, 0.0744, 0.0818, 0.1490, 0.0979, 0.1097], device='cuda:1'), in_proj_covar=tensor([0.0574, 0.0702, 0.0579, 0.0506, 0.0444, 0.0458, 0.0589, 0.0536], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:1') 2023-04-30 01:05:54,553 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=140387.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 01:06:08,145 INFO [optim.py:368] (1/8) 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 01:06:18,702 INFO [train.py:904] (1/8) Epoch 14, batch 8450, loss[loss=0.1668, simple_loss=0.2637, pruned_loss=0.03497, over 16800.00 frames. ], tot_loss[loss=0.1935, simple_loss=0.2833, pruned_loss=0.05185, over 3076097.38 frames. ], batch size: 83, lr: 4.81e-03, grad_scale: 8.0 2023-04-30 01:06:55,883 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=140425.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 01:07:27,339 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=140445.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 01:07:32,300 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 14, batch 8500, loss[loss=0.1836, simple_loss=0.2776, pruned_loss=0.04479, over 16696.00 frames. ], tot_loss[loss=0.1899, simple_loss=0.2801, pruned_loss=0.0499, over 3078927.40 frames. ], batch size: 89, lr: 4.81e-03, grad_scale: 8.0 2023-04-30 01:07:50,011 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=140459.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 01:07:51,357 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=140460.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 01:07:54,707 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=140462.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 01:08:01,415 INFO [zipformer.py:625] (1/8) 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] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=140473.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 01:08:38,534 INFO [zipformer.py:625] (1/8) 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] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=140493.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 01:08:46,364 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.4196, 2.1246, 2.1194, 4.1085, 2.1066, 2.6023, 2.2529, 2.3487], device='cuda:1'), covar=tensor([0.0936, 0.3667, 0.2704, 0.0382, 0.4128, 0.2302, 0.3308, 0.3234], device='cuda:1'), in_proj_covar=tensor([0.0363, 0.0403, 0.0335, 0.0311, 0.0411, 0.0460, 0.0366, 0.0468], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-30 01:08:51,328 INFO [optim.py:368] (1/8) 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 01:09:01,486 INFO [train.py:904] (1/8) Epoch 14, batch 8550, loss[loss=0.1845, simple_loss=0.2748, pruned_loss=0.04712, over 16496.00 frames. ], tot_loss[loss=0.1875, simple_loss=0.2775, pruned_loss=0.04878, over 3066220.22 frames. ], batch size: 68, lr: 4.81e-03, grad_scale: 8.0 2023-04-30 01:09:11,939 INFO [zipformer.py:625] (1/8) 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] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=140508.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 01:09:17,236 INFO [zipformer.py:625] (1/8) 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] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=140510.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 01:10:39,010 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=140550.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 01:10:41,517 INFO [train.py:904] (1/8) Epoch 14, batch 8600, loss[loss=0.2074, simple_loss=0.2925, pruned_loss=0.06111, over 16966.00 frames. ], tot_loss[loss=0.1871, simple_loss=0.2783, pruned_loss=0.04792, over 3074003.47 frames. ], batch size: 109, lr: 4.81e-03, grad_scale: 8.0 2023-04-30 01:10:46,488 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=140554.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 01:11:19,566 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=140570.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 01:11:29,297 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.5910, 3.5865, 3.4993, 2.8288, 3.4543, 1.9793, 3.2490, 3.0377], device='cuda:1'), covar=tensor([0.0114, 0.0093, 0.0134, 0.0162, 0.0081, 0.2176, 0.0110, 0.0176], device='cuda:1'), in_proj_covar=tensor([0.0138, 0.0126, 0.0172, 0.0157, 0.0144, 0.0187, 0.0158, 0.0152], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-30 01:12:02,777 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.7890, 3.9268, 4.2257, 4.2025, 4.2091, 4.0048, 3.8259, 3.9646], device='cuda:1'), covar=tensor([0.0514, 0.0644, 0.0545, 0.0604, 0.0711, 0.0539, 0.1337, 0.0476], device='cuda:1'), in_proj_covar=tensor([0.0344, 0.0362, 0.0361, 0.0344, 0.0408, 0.0386, 0.0470, 0.0305], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003], device='cuda:1') 2023-04-30 01:12:06,767 INFO [optim.py:368] (1/8) 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] (1/8) Epoch 14, batch 8650, loss[loss=0.1831, simple_loss=0.2805, pruned_loss=0.04287, over 16825.00 frames. ], tot_loss[loss=0.1847, simple_loss=0.2763, pruned_loss=0.04651, over 3065055.87 frames. ], batch size: 116, lr: 4.81e-03, grad_scale: 8.0 2023-04-30 01:12:52,266 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 14, batch 8700, loss[loss=0.1883, simple_loss=0.2797, pruned_loss=0.04841, over 16313.00 frames. ], tot_loss[loss=0.1823, simple_loss=0.2737, pruned_loss=0.04541, over 3057776.21 frames. ], batch size: 146, lr: 4.81e-03, grad_scale: 8.0 2023-04-30 01:14:11,542 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.49 vs. limit=2.0 2023-04-30 01:15:05,210 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.33 vs. limit=2.0 2023-04-30 01:15:24,088 INFO [optim.py:368] (1/8) 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 01:15:25,635 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.0201, 1.8048, 1.6364, 1.5224, 1.9319, 1.5985, 1.6406, 1.9656], device='cuda:1'), covar=tensor([0.0138, 0.0247, 0.0329, 0.0313, 0.0193, 0.0257, 0.0140, 0.0178], device='cuda:1'), in_proj_covar=tensor([0.0160, 0.0207, 0.0201, 0.0203, 0.0206, 0.0207, 0.0207, 0.0197], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-30 01:15:38,741 INFO [train.py:904] (1/8) Epoch 14, batch 8750, loss[loss=0.193, simple_loss=0.2899, pruned_loss=0.04802, over 15400.00 frames. ], tot_loss[loss=0.1818, simple_loss=0.2738, pruned_loss=0.04495, over 3075866.89 frames. ], batch size: 191, lr: 4.80e-03, grad_scale: 8.0 2023-04-30 01:16:30,739 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=140723.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 01:17:11,079 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=140743.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 01:17:28,573 INFO [train.py:904] (1/8) Epoch 14, batch 8800, loss[loss=0.2079, simple_loss=0.2971, pruned_loss=0.05932, over 15368.00 frames. ], tot_loss[loss=0.1801, simple_loss=0.2722, pruned_loss=0.04401, over 3069088.04 frames. ], batch size: 190, lr: 4.80e-03, grad_scale: 8.0 2023-04-30 01:17:35,569 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.37 vs. limit=2.0 2023-04-30 01:17:43,083 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=140759.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 01:17:56,964 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=140766.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 01:18:35,747 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=140784.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 01:18:45,398 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.88 vs. limit=2.0 2023-04-30 01:19:00,060 INFO [optim.py:368] (1/8) Clipping_scale=2.0, grad-norm quartiles 1.423e+02 2.211e+02 2.894e+02 3.689e+02 7.406e+02, threshold=5.789e+02, percent-clipped=2.0 2023-04-30 01:19:12,350 INFO [train.py:904] (1/8) Epoch 14, batch 8850, loss[loss=0.1669, simple_loss=0.2586, pruned_loss=0.03757, over 12452.00 frames. ], tot_loss[loss=0.1809, simple_loss=0.2745, pruned_loss=0.04368, over 3052789.28 frames. ], batch size: 247, lr: 4.80e-03, grad_scale: 8.0 2023-04-30 01:19:23,456 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=140807.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 01:19:27,174 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=140809.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 01:19:37,088 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=140814.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 01:19:57,749 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.0221, 2.2392, 1.8949, 2.0083, 2.6155, 2.2753, 2.6380, 2.8290], device='cuda:1'), covar=tensor([0.0111, 0.0361, 0.0467, 0.0401, 0.0229, 0.0328, 0.0158, 0.0224], device='cuda:1'), in_proj_covar=tensor([0.0159, 0.0207, 0.0202, 0.0203, 0.0206, 0.0207, 0.0207, 0.0197], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-30 01:20:09,821 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.7139, 4.1738, 4.1133, 3.1948, 3.6391, 4.1803, 3.8474, 2.4357], device='cuda:1'), covar=tensor([0.0366, 0.0022, 0.0025, 0.0225, 0.0065, 0.0038, 0.0042, 0.0351], device='cuda:1'), in_proj_covar=tensor([0.0127, 0.0069, 0.0070, 0.0126, 0.0083, 0.0092, 0.0081, 0.0120], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:1') 2023-04-30 01:20:26,110 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=4.52 vs. limit=5.0 2023-04-30 01:20:43,599 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 14, batch 8900, loss[loss=0.1668, simple_loss=0.2548, pruned_loss=0.03933, over 12308.00 frames. ], tot_loss[loss=0.1802, simple_loss=0.2746, pruned_loss=0.04285, over 3061451.36 frames. ], batch size: 246, lr: 4.80e-03, grad_scale: 8.0 2023-04-30 01:21:25,589 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=140865.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 01:21:34,671 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=140870.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 01:22:31,873 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=2.43 vs. limit=5.0 2023-04-30 01:22:46,891 INFO [optim.py:368] (1/8) 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] (1/8) Epoch 14, batch 8950, loss[loss=0.1832, simple_loss=0.2706, pruned_loss=0.04789, over 12568.00 frames. ], tot_loss[loss=0.1807, simple_loss=0.2746, pruned_loss=0.04343, over 3066110.30 frames. ], batch size: 250, lr: 4.80e-03, grad_scale: 8.0 2023-04-30 01:23:12,514 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.80 vs. limit=2.0 2023-04-30 01:23:29,063 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=140915.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 01:23:32,851 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=140917.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 01:24:00,006 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.2204, 4.3169, 4.4375, 4.2695, 4.3538, 4.8346, 4.4996, 4.2026], device='cuda:1'), covar=tensor([0.1409, 0.1856, 0.1847, 0.1933, 0.2330, 0.0970, 0.1278, 0.2217], device='cuda:1'), in_proj_covar=tensor([0.0354, 0.0496, 0.0545, 0.0422, 0.0563, 0.0576, 0.0435, 0.0570], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-04-30 01:24:38,436 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 14, batch 9000, loss[loss=0.165, simple_loss=0.2527, pruned_loss=0.03862, over 16268.00 frames. ], tot_loss[loss=0.1774, simple_loss=0.2713, pruned_loss=0.04172, over 3084434.54 frames. ], batch size: 146, lr: 4.80e-03, grad_scale: 8.0 2023-04-30 01:24:48,129 INFO [train.py:929] (1/8) Computing validation loss 2023-04-30 01:24:58,094 INFO [train.py:938] (1/8) Epoch 14, validation: loss=0.1514, simple_loss=0.2553, pruned_loss=0.0237, over 944034.00 frames. 2023-04-30 01:24:58,095 INFO [train.py:939] (1/8) Maximum memory allocated so far is 17857MB 2023-04-30 01:25:21,238 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=140963.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 01:25:52,431 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=140978.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 01:26:29,789 INFO [optim.py:368] (1/8) 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 01:26:39,922 INFO [train.py:904] (1/8) Epoch 14, batch 9050, loss[loss=0.1701, simple_loss=0.2611, pruned_loss=0.03954, over 16589.00 frames. ], tot_loss[loss=0.1784, simple_loss=0.2724, pruned_loss=0.04218, over 3093567.07 frames. ], batch size: 62, lr: 4.80e-03, grad_scale: 8.0 2023-04-30 01:26:54,552 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=141008.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 01:28:07,473 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=141043.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 01:28:20,876 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([1.7332, 2.4809, 2.2705, 3.5034, 2.0197, 3.7170, 1.4058, 2.8959], device='cuda:1'), covar=tensor([0.1359, 0.0722, 0.1189, 0.0159, 0.0111, 0.0359, 0.1624, 0.0700], device='cuda:1'), in_proj_covar=tensor([0.0158, 0.0160, 0.0182, 0.0157, 0.0192, 0.0206, 0.0185, 0.0182], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0003, 0.0004], device='cuda:1') 2023-04-30 01:28:25,839 INFO [train.py:904] (1/8) Epoch 14, batch 9100, loss[loss=0.1741, simple_loss=0.2667, pruned_loss=0.04076, over 16600.00 frames. ], tot_loss[loss=0.1789, simple_loss=0.2721, pruned_loss=0.04285, over 3078897.74 frames. ], batch size: 62, lr: 4.80e-03, grad_scale: 8.0 2023-04-30 01:29:30,797 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=141079.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 01:29:58,702 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=141091.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 01:30:13,183 INFO [optim.py:368] (1/8) 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 01:30:22,850 INFO [train.py:904] (1/8) Epoch 14, batch 9150, loss[loss=0.1779, simple_loss=0.2634, pruned_loss=0.04623, over 12054.00 frames. ], tot_loss[loss=0.1788, simple_loss=0.2723, pruned_loss=0.04265, over 3070535.87 frames. ], batch size: 250, lr: 4.80e-03, grad_scale: 4.0 2023-04-30 01:31:51,276 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=141143.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 01:31:54,443 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 14, batch 9200, loss[loss=0.1918, simple_loss=0.2843, pruned_loss=0.04968, over 16680.00 frames. ], tot_loss[loss=0.1758, simple_loss=0.2681, pruned_loss=0.04177, over 3074891.35 frames. ], batch size: 57, lr: 4.80e-03, grad_scale: 8.0 2023-04-30 01:32:28,724 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=141165.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 01:32:28,776 INFO [zipformer.py:625] (1/8) 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] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=141193.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 01:33:31,528 INFO [optim.py:368] (1/8) 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 01:33:40,991 INFO [train.py:904] (1/8) Epoch 14, batch 9250, loss[loss=0.1766, simple_loss=0.266, pruned_loss=0.04363, over 15324.00 frames. ], tot_loss[loss=0.1761, simple_loss=0.2679, pruned_loss=0.04218, over 3047447.16 frames. ], batch size: 192, lr: 4.80e-03, grad_scale: 8.0 2023-04-30 01:33:46,060 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=141204.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 01:33:55,566 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.1937, 3.2893, 1.8938, 3.5546, 2.3379, 3.4533, 2.0854, 2.6136], device='cuda:1'), covar=tensor([0.0292, 0.0319, 0.1579, 0.0192, 0.0867, 0.0564, 0.1538, 0.0759], device='cuda:1'), in_proj_covar=tensor([0.0154, 0.0162, 0.0185, 0.0133, 0.0164, 0.0199, 0.0193, 0.0169], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0003, 0.0004, 0.0004, 0.0003], device='cuda:1') 2023-04-30 01:34:03,407 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=141213.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 01:35:30,115 INFO [train.py:904] (1/8) Epoch 14, batch 9300, loss[loss=0.1727, simple_loss=0.2591, pruned_loss=0.04312, over 12194.00 frames. ], tot_loss[loss=0.1751, simple_loss=0.2666, pruned_loss=0.04182, over 3037635.15 frames. ], batch size: 250, lr: 4.79e-03, grad_scale: 8.0 2023-04-30 01:35:30,908 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.9713, 5.2921, 5.0490, 5.0832, 4.7582, 4.7426, 4.6246, 5.3547], device='cuda:1'), covar=tensor([0.1006, 0.0704, 0.0834, 0.0657, 0.0694, 0.0724, 0.1016, 0.0673], device='cuda:1'), in_proj_covar=tensor([0.0558, 0.0687, 0.0563, 0.0495, 0.0437, 0.0449, 0.0578, 0.0523], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:1') 2023-04-30 01:36:20,192 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=141273.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 01:36:45,260 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.4934, 4.2804, 4.5553, 4.6922, 4.8382, 4.3597, 4.8041, 4.8480], device='cuda:1'), covar=tensor([0.1655, 0.1212, 0.1274, 0.0626, 0.0545, 0.0972, 0.0527, 0.0592], device='cuda:1'), in_proj_covar=tensor([0.0536, 0.0659, 0.0775, 0.0677, 0.0512, 0.0526, 0.0529, 0.0630], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-30 01:36:45,650 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.57 vs. limit=2.0 2023-04-30 01:37:05,690 INFO [optim.py:368] (1/8) 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 01:37:08,123 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.4961, 3.5788, 2.0801, 3.9358, 2.5739, 3.8290, 2.1659, 2.8660], device='cuda:1'), covar=tensor([0.0225, 0.0325, 0.1452, 0.0197, 0.0814, 0.0522, 0.1495, 0.0665], device='cuda:1'), in_proj_covar=tensor([0.0153, 0.0160, 0.0184, 0.0132, 0.0163, 0.0198, 0.0191, 0.0167], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0003, 0.0004, 0.0004, 0.0003], device='cuda:1') 2023-04-30 01:37:14,522 INFO [train.py:904] (1/8) Epoch 14, batch 9350, loss[loss=0.1744, simple_loss=0.2582, pruned_loss=0.04528, over 12377.00 frames. ], tot_loss[loss=0.1744, simple_loss=0.2659, pruned_loss=0.04142, over 3052852.23 frames. ], batch size: 247, lr: 4.79e-03, grad_scale: 8.0 2023-04-30 01:37:16,939 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=141303.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 01:38:08,751 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=141329.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 01:38:39,505 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.2204, 1.4994, 1.9146, 2.1794, 2.2402, 2.3791, 1.5466, 2.3493], device='cuda:1'), covar=tensor([0.0215, 0.0458, 0.0286, 0.0280, 0.0269, 0.0160, 0.0470, 0.0124], device='cuda:1'), in_proj_covar=tensor([0.0163, 0.0174, 0.0156, 0.0160, 0.0172, 0.0128, 0.0175, 0.0119], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001], device='cuda:1') 2023-04-30 01:38:54,830 INFO [train.py:904] (1/8) Epoch 14, batch 9400, loss[loss=0.1711, simple_loss=0.2525, pruned_loss=0.04483, over 12466.00 frames. ], tot_loss[loss=0.1737, simple_loss=0.2658, pruned_loss=0.04078, over 3061222.21 frames. ], batch size: 247, lr: 4.79e-03, grad_scale: 8.0 2023-04-30 01:39:49,364 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=141379.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 01:40:12,094 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=141390.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 01:40:25,323 INFO [optim.py:368] (1/8) Clipping_scale=2.0, grad-norm quartiles 1.384e+02 2.208e+02 2.590e+02 3.171e+02 6.862e+02, threshold=5.180e+02, percent-clipped=4.0 2023-04-30 01:40:33,410 INFO [train.py:904] (1/8) Epoch 14, batch 9450, loss[loss=0.17, simple_loss=0.2665, pruned_loss=0.03673, over 15466.00 frames. ], tot_loss[loss=0.1742, simple_loss=0.2671, pruned_loss=0.04066, over 3046979.45 frames. ], batch size: 191, lr: 4.79e-03, grad_scale: 8.0 2023-04-30 01:41:03,135 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=141417.0, num_to_drop=1, layers_to_drop={1} 2023-04-30 01:41:24,367 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=141427.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 01:41:54,107 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=141441.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 01:42:14,328 INFO [train.py:904] (1/8) Epoch 14, batch 9500, loss[loss=0.1676, simple_loss=0.2623, pruned_loss=0.03648, over 16669.00 frames. ], tot_loss[loss=0.173, simple_loss=0.2658, pruned_loss=0.04004, over 3052735.86 frames. ], batch size: 57, lr: 4.79e-03, grad_scale: 8.0 2023-04-30 01:42:43,598 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=141465.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 01:43:08,412 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=141478.0, num_to_drop=1, layers_to_drop={3} 2023-04-30 01:43:47,717 INFO [optim.py:368] (1/8) 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 01:43:55,177 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 14, batch 9550, loss[loss=0.1933, simple_loss=0.2894, pruned_loss=0.04859, over 16329.00 frames. ], tot_loss[loss=0.173, simple_loss=0.2658, pruned_loss=0.04009, over 3063547.72 frames. ], batch size: 146, lr: 4.79e-03, grad_scale: 8.0 2023-04-30 01:44:00,571 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=141502.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 01:44:23,850 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=141513.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 01:45:41,083 INFO [train.py:904] (1/8) Epoch 14, batch 9600, loss[loss=0.195, simple_loss=0.2915, pruned_loss=0.04929, over 15347.00 frames. ], tot_loss[loss=0.1749, simple_loss=0.2671, pruned_loss=0.04129, over 3053413.75 frames. ], batch size: 191, lr: 4.79e-03, grad_scale: 8.0 2023-04-30 01:45:44,606 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.3960, 2.1533, 2.0594, 4.0633, 1.9883, 2.5685, 2.2441, 2.2961], device='cuda:1'), covar=tensor([0.0947, 0.3501, 0.2800, 0.0379, 0.4112, 0.2181, 0.3111, 0.3148], device='cuda:1'), in_proj_covar=tensor([0.0358, 0.0395, 0.0332, 0.0308, 0.0407, 0.0449, 0.0361, 0.0459], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-30 01:46:22,397 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=141573.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 01:47:19,263 INFO [optim.py:368] (1/8) 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] (1/8) Epoch 14, batch 9650, loss[loss=0.174, simple_loss=0.2604, pruned_loss=0.04377, over 16575.00 frames. ], tot_loss[loss=0.1766, simple_loss=0.2695, pruned_loss=0.04184, over 3067162.90 frames. ], batch size: 57, lr: 4.79e-03, grad_scale: 4.0 2023-04-30 01:47:34,634 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=141603.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 01:48:16,938 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=141621.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 01:49:16,990 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 14, batch 9700, loss[loss=0.1731, simple_loss=0.2643, pruned_loss=0.04095, over 15229.00 frames. ], tot_loss[loss=0.1762, simple_loss=0.2688, pruned_loss=0.04186, over 3066094.31 frames. ], batch size: 190, lr: 4.79e-03, grad_scale: 4.0 2023-04-30 01:50:27,991 INFO [zipformer.py:625] (1/8) 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] (1/8) 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] (1/8) Epoch 14, batch 9750, loss[loss=0.1757, simple_loss=0.2654, pruned_loss=0.04299, over 16827.00 frames. ], tot_loss[loss=0.1761, simple_loss=0.2684, pruned_loss=0.04196, over 3077307.54 frames. ], batch size: 124, lr: 4.79e-03, grad_scale: 4.0 2023-04-30 01:51:43,826 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.0646, 2.5113, 2.6427, 1.8044, 2.8187, 2.8891, 2.5022, 2.4316], device='cuda:1'), covar=tensor([0.0687, 0.0199, 0.0175, 0.1021, 0.0088, 0.0193, 0.0447, 0.0427], device='cuda:1'), in_proj_covar=tensor([0.0140, 0.0099, 0.0084, 0.0134, 0.0068, 0.0106, 0.0119, 0.0123], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0003, 0.0002, 0.0004, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:1') 2023-04-30 01:52:38,197 INFO [train.py:904] (1/8) Epoch 14, batch 9800, loss[loss=0.1949, simple_loss=0.2958, pruned_loss=0.04701, over 16690.00 frames. ], tot_loss[loss=0.1754, simple_loss=0.2688, pruned_loss=0.04102, over 3091286.87 frames. ], batch size: 134, lr: 4.79e-03, grad_scale: 4.0 2023-04-30 01:53:18,069 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=141773.0, num_to_drop=1, layers_to_drop={0} 2023-04-30 01:53:28,076 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([1.8190, 2.5883, 2.3281, 3.6395, 2.3555, 3.7352, 1.5348, 2.7648], device='cuda:1'), covar=tensor([0.1248, 0.0666, 0.1117, 0.0155, 0.0100, 0.0468, 0.1491, 0.0777], device='cuda:1'), in_proj_covar=tensor([0.0159, 0.0162, 0.0184, 0.0157, 0.0191, 0.0206, 0.0188, 0.0184], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-04-30 01:53:53,738 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=2.98 vs. limit=5.0 2023-04-30 01:54:13,229 INFO [zipformer.py:625] (1/8) 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] (1/8) 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 01:54:17,231 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 14, batch 9850, loss[loss=0.1798, simple_loss=0.2799, pruned_loss=0.03987, over 16799.00 frames. ], tot_loss[loss=0.1753, simple_loss=0.2692, pruned_loss=0.04069, over 3070752.39 frames. ], batch size: 83, lr: 4.79e-03, grad_scale: 4.0 2023-04-30 01:55:56,295 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.4300, 3.0135, 2.6826, 2.1992, 2.1575, 2.1857, 2.8732, 2.8171], device='cuda:1'), covar=tensor([0.2487, 0.0750, 0.1497, 0.2680, 0.2757, 0.2058, 0.0465, 0.1236], device='cuda:1'), in_proj_covar=tensor([0.0303, 0.0253, 0.0284, 0.0283, 0.0266, 0.0228, 0.0264, 0.0295], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-30 01:56:04,787 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 14, batch 9900, loss[loss=0.1834, simple_loss=0.2851, pruned_loss=0.0409, over 16833.00 frames. ], tot_loss[loss=0.1753, simple_loss=0.2692, pruned_loss=0.04064, over 3048339.66 frames. ], batch size: 124, lr: 4.78e-03, grad_scale: 4.0 2023-04-30 01:58:03,761 INFO [optim.py:368] (1/8) Clipping_scale=2.0, grad-norm quartiles 1.629e+02 2.142e+02 2.825e+02 3.311e+02 8.255e+02, threshold=5.650e+02, percent-clipped=2.0 2023-04-30 01:58:05,830 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.9789, 3.8245, 4.0602, 4.1649, 4.2769, 3.8447, 4.2342, 4.2897], device='cuda:1'), covar=tensor([0.1506, 0.1005, 0.1125, 0.0605, 0.0532, 0.1383, 0.0600, 0.0564], device='cuda:1'), in_proj_covar=tensor([0.0534, 0.0658, 0.0776, 0.0677, 0.0508, 0.0526, 0.0533, 0.0630], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-30 01:58:13,367 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.0434, 1.9486, 2.0903, 3.4990, 1.9083, 2.2138, 2.0935, 2.0403], device='cuda:1'), covar=tensor([0.1131, 0.3827, 0.2677, 0.0550, 0.4460, 0.2609, 0.3818, 0.3466], device='cuda:1'), in_proj_covar=tensor([0.0359, 0.0396, 0.0333, 0.0309, 0.0408, 0.0451, 0.0362, 0.0460], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-30 01:58:13,944 INFO [train.py:904] (1/8) Epoch 14, batch 9950, loss[loss=0.2041, simple_loss=0.2976, pruned_loss=0.0553, over 15406.00 frames. ], tot_loss[loss=0.1773, simple_loss=0.272, pruned_loss=0.04128, over 3062244.77 frames. ], batch size: 191, lr: 4.78e-03, grad_scale: 4.0 2023-04-30 01:58:48,304 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=141916.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 01:59:21,077 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=141928.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 02:00:16,391 INFO [train.py:904] (1/8) Epoch 14, batch 10000, loss[loss=0.1433, simple_loss=0.2348, pruned_loss=0.02593, over 17199.00 frames. ], tot_loss[loss=0.1759, simple_loss=0.2702, pruned_loss=0.04081, over 3073290.86 frames. ], batch size: 44, lr: 4.78e-03, grad_scale: 8.0 2023-04-30 02:00:21,776 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.9232, 2.0675, 2.2904, 3.2346, 2.0663, 2.2310, 2.2221, 2.1778], device='cuda:1'), covar=tensor([0.1092, 0.3280, 0.2419, 0.0612, 0.3989, 0.2333, 0.3289, 0.3329], device='cuda:1'), in_proj_covar=tensor([0.0358, 0.0396, 0.0333, 0.0308, 0.0407, 0.0449, 0.0361, 0.0458], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-30 02:00:29,288 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=3.37 vs. limit=5.0 2023-04-30 02:01:07,136 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=141977.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 02:01:26,024 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=141985.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 02:01:34,661 INFO [zipformer.py:625] (1/8) 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] (1/8) Clipping_scale=2.0, grad-norm quartiles 1.584e+02 2.061e+02 2.551e+02 3.100e+02 8.241e+02, threshold=5.102e+02, percent-clipped=1.0 2023-04-30 02:02:01,159 INFO [train.py:904] (1/8) Epoch 14, batch 10050, loss[loss=0.1755, simple_loss=0.2697, pruned_loss=0.04067, over 16500.00 frames. ], tot_loss[loss=0.1752, simple_loss=0.2695, pruned_loss=0.04047, over 3071024.79 frames. ], batch size: 62, lr: 4.78e-03, grad_scale: 8.0 2023-04-30 02:03:00,956 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=142033.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 02:03:15,493 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.3234, 2.1242, 2.1977, 3.8940, 2.1320, 2.5004, 2.2485, 2.3041], device='cuda:1'), covar=tensor([0.0990, 0.3518, 0.2675, 0.0411, 0.3877, 0.2360, 0.3223, 0.3240], device='cuda:1'), in_proj_covar=tensor([0.0359, 0.0395, 0.0333, 0.0306, 0.0406, 0.0448, 0.0360, 0.0457], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-30 02:03:25,946 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.6457, 3.6641, 2.8690, 2.1953, 2.3124, 2.2980, 3.9383, 3.3307], device='cuda:1'), covar=tensor([0.2696, 0.0687, 0.1604, 0.2561, 0.2580, 0.1895, 0.0384, 0.1164], device='cuda:1'), in_proj_covar=tensor([0.0299, 0.0251, 0.0282, 0.0279, 0.0262, 0.0225, 0.0262, 0.0292], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-30 02:03:27,363 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.9765, 2.0916, 2.3407, 3.2584, 2.1880, 2.3038, 2.2565, 2.1786], device='cuda:1'), covar=tensor([0.1036, 0.3475, 0.2218, 0.0528, 0.3707, 0.2423, 0.2997, 0.3332], device='cuda:1'), in_proj_covar=tensor([0.0358, 0.0395, 0.0333, 0.0306, 0.0406, 0.0448, 0.0360, 0.0456], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-30 02:03:34,640 INFO [train.py:904] (1/8) Epoch 14, batch 10100, loss[loss=0.1768, simple_loss=0.2662, pruned_loss=0.04369, over 12711.00 frames. ], tot_loss[loss=0.1756, simple_loss=0.2697, pruned_loss=0.04076, over 3067703.58 frames. ], batch size: 248, lr: 4.78e-03, grad_scale: 8.0 2023-04-30 02:04:17,838 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.2994, 3.4095, 3.6258, 3.6192, 3.6198, 3.4753, 3.4627, 3.4791], device='cuda:1'), covar=tensor([0.0477, 0.1054, 0.0573, 0.0472, 0.0550, 0.0656, 0.0865, 0.0653], device='cuda:1'), in_proj_covar=tensor([0.0327, 0.0345, 0.0348, 0.0329, 0.0389, 0.0369, 0.0444, 0.0295], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0002], device='cuda:1') 2023-04-30 02:04:19,922 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=142073.0, num_to_drop=1, layers_to_drop={1} 2023-04-30 02:04:34,878 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.6670, 4.6581, 4.5036, 4.1385, 4.2029, 4.5745, 4.3867, 4.2491], device='cuda:1'), covar=tensor([0.0468, 0.0480, 0.0267, 0.0284, 0.0775, 0.0479, 0.0401, 0.0601], device='cuda:1'), in_proj_covar=tensor([0.0237, 0.0322, 0.0283, 0.0264, 0.0292, 0.0307, 0.0194, 0.0330], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002], device='cuda:1') 2023-04-30 02:04:49,789 INFO [zipformer.py:625] (1/8) 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] (1/8) 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] (1/8) Epoch 15, batch 0, loss[loss=0.1857, simple_loss=0.2742, pruned_loss=0.04861, over 17034.00 frames. ], tot_loss[loss=0.1857, simple_loss=0.2742, pruned_loss=0.04861, over 17034.00 frames. ], batch size: 50, lr: 4.62e-03, grad_scale: 8.0 2023-04-30 02:05:19,913 INFO [train.py:929] (1/8) Computing validation loss 2023-04-30 02:05:27,351 INFO [train.py:938] (1/8) Epoch 15, validation: loss=0.1501, simple_loss=0.2536, pruned_loss=0.02333, over 944034.00 frames. 2023-04-30 02:05:27,351 INFO [train.py:939] (1/8) Maximum memory allocated so far is 17857MB 2023-04-30 02:05:53,971 INFO [zipformer.py:625] (1/8) 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] (1/8) 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] (1/8) Epoch 15, batch 50, loss[loss=0.1556, simple_loss=0.2494, pruned_loss=0.03083, over 17067.00 frames. ], tot_loss[loss=0.1934, simple_loss=0.2755, pruned_loss=0.05568, over 746757.45 frames. ], batch size: 47, lr: 4.62e-03, grad_scale: 2.0 2023-04-30 02:07:44,550 INFO [optim.py:368] (1/8) 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 02:07:47,915 INFO [train.py:904] (1/8) Epoch 15, batch 100, loss[loss=0.1494, simple_loss=0.2302, pruned_loss=0.03432, over 16715.00 frames. ], tot_loss[loss=0.193, simple_loss=0.2752, pruned_loss=0.05537, over 1319531.69 frames. ], batch size: 83, lr: 4.62e-03, grad_scale: 2.0 2023-04-30 02:08:56,686 INFO [train.py:904] (1/8) Epoch 15, batch 150, loss[loss=0.2323, simple_loss=0.3149, pruned_loss=0.07484, over 12131.00 frames. ], tot_loss[loss=0.1904, simple_loss=0.2734, pruned_loss=0.05371, over 1756296.90 frames. ], batch size: 246, lr: 4.62e-03, grad_scale: 2.0 2023-04-30 02:09:11,298 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.9365, 1.9096, 2.3820, 2.8356, 2.6486, 3.2986, 2.1490, 3.2731], device='cuda:1'), covar=tensor([0.0217, 0.0423, 0.0303, 0.0258, 0.0272, 0.0156, 0.0401, 0.0141], device='cuda:1'), in_proj_covar=tensor([0.0164, 0.0176, 0.0160, 0.0164, 0.0173, 0.0130, 0.0177, 0.0120], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001], device='cuda:1') 2023-04-30 02:09:25,348 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=142272.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 02:09:42,879 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=142284.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 02:09:49,407 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([5.1242, 5.2397, 4.9798, 4.6528, 4.1831, 5.1665, 5.1978, 4.7037], device='cuda:1'), covar=tensor([0.0866, 0.0568, 0.0550, 0.0417, 0.2050, 0.0524, 0.0324, 0.0743], device='cuda:1'), in_proj_covar=tensor([0.0250, 0.0342, 0.0298, 0.0278, 0.0310, 0.0324, 0.0206, 0.0350], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-30 02:10:04,596 INFO [optim.py:368] (1/8) 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] (1/8) Epoch 15, batch 200, loss[loss=0.2056, simple_loss=0.2717, pruned_loss=0.0698, over 16834.00 frames. ], tot_loss[loss=0.1878, simple_loss=0.2712, pruned_loss=0.05219, over 2107628.98 frames. ], batch size: 90, lr: 4.61e-03, grad_scale: 2.0 2023-04-30 02:11:17,109 INFO [train.py:904] (1/8) Epoch 15, batch 250, loss[loss=0.1722, simple_loss=0.2678, pruned_loss=0.03828, over 17118.00 frames. ], tot_loss[loss=0.1868, simple_loss=0.2694, pruned_loss=0.05204, over 2376378.34 frames. ], batch size: 53, lr: 4.61e-03, grad_scale: 2.0 2023-04-30 02:12:22,521 INFO [optim.py:368] (1/8) 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 02:12:25,436 INFO [train.py:904] (1/8) Epoch 15, batch 300, loss[loss=0.2004, simple_loss=0.2747, pruned_loss=0.06311, over 16454.00 frames. ], tot_loss[loss=0.1849, simple_loss=0.2672, pruned_loss=0.05133, over 2576148.00 frames. ], batch size: 146, lr: 4.61e-03, grad_scale: 2.0 2023-04-30 02:13:18,281 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.6789, 4.5641, 4.5245, 4.2208, 4.2002, 4.5550, 4.4498, 4.3301], device='cuda:1'), covar=tensor([0.0640, 0.0707, 0.0341, 0.0315, 0.0961, 0.0488, 0.0476, 0.0627], device='cuda:1'), in_proj_covar=tensor([0.0255, 0.0348, 0.0304, 0.0283, 0.0317, 0.0331, 0.0209, 0.0358], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-30 02:13:29,739 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.7707, 3.5796, 3.8200, 2.9340, 3.4689, 3.9827, 3.7278, 2.2529], device='cuda:1'), covar=tensor([0.0396, 0.0215, 0.0047, 0.0281, 0.0103, 0.0085, 0.0072, 0.0419], device='cuda:1'), in_proj_covar=tensor([0.0131, 0.0074, 0.0074, 0.0130, 0.0086, 0.0096, 0.0084, 0.0125], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-04-30 02:13:36,228 INFO [train.py:904] (1/8) Epoch 15, batch 350, loss[loss=0.2175, simple_loss=0.2868, pruned_loss=0.07411, over 16880.00 frames. ], tot_loss[loss=0.1818, simple_loss=0.2642, pruned_loss=0.04973, over 2740757.52 frames. ], batch size: 109, lr: 4.61e-03, grad_scale: 2.0 2023-04-30 02:14:42,900 INFO [optim.py:368] (1/8) 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] (1/8) Epoch 15, batch 400, loss[loss=0.2003, simple_loss=0.2689, pruned_loss=0.06583, over 16923.00 frames. ], tot_loss[loss=0.1806, simple_loss=0.2627, pruned_loss=0.04928, over 2868836.27 frames. ], batch size: 109, lr: 4.61e-03, grad_scale: 4.0 2023-04-30 02:15:44,633 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.90 vs. limit=2.0 2023-04-30 02:15:54,199 INFO [train.py:904] (1/8) Epoch 15, batch 450, loss[loss=0.1827, simple_loss=0.2609, pruned_loss=0.05223, over 12534.00 frames. ], tot_loss[loss=0.1791, simple_loss=0.2615, pruned_loss=0.0483, over 2964044.47 frames. ], batch size: 246, lr: 4.61e-03, grad_scale: 4.0 2023-04-30 02:16:23,386 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=142572.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 02:16:24,606 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=142573.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 02:16:40,905 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=142584.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 02:17:01,111 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.9167, 1.8775, 2.3807, 2.7980, 2.7346, 2.8755, 1.9586, 3.0290], device='cuda:1'), covar=tensor([0.0167, 0.0392, 0.0281, 0.0229, 0.0220, 0.0198, 0.0411, 0.0136], device='cuda:1'), in_proj_covar=tensor([0.0167, 0.0179, 0.0163, 0.0168, 0.0176, 0.0133, 0.0180, 0.0123], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001], device='cuda:1') 2023-04-30 02:17:03,317 INFO [optim.py:368] (1/8) 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 02:17:05,222 INFO [train.py:904] (1/8) Epoch 15, batch 500, loss[loss=0.1685, simple_loss=0.2689, pruned_loss=0.03408, over 17258.00 frames. ], tot_loss[loss=0.1769, simple_loss=0.2597, pruned_loss=0.04706, over 3053854.76 frames. ], batch size: 52, lr: 4.61e-03, grad_scale: 2.0 2023-04-30 02:17:28,778 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=142620.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 02:17:45,634 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=142632.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 02:17:49,297 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=142634.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 02:17:52,994 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.7007, 2.4010, 2.4375, 4.7020, 2.3700, 2.8431, 2.5013, 2.6128], device='cuda:1'), covar=tensor([0.1017, 0.3481, 0.2593, 0.0361, 0.3796, 0.2354, 0.3161, 0.3337], device='cuda:1'), in_proj_covar=tensor([0.0373, 0.0409, 0.0343, 0.0320, 0.0419, 0.0469, 0.0375, 0.0478], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-30 02:18:05,902 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.22 vs. limit=2.0 2023-04-30 02:18:13,783 INFO [train.py:904] (1/8) Epoch 15, batch 550, loss[loss=0.2174, simple_loss=0.292, pruned_loss=0.07142, over 16271.00 frames. ], tot_loss[loss=0.1765, simple_loss=0.2594, pruned_loss=0.04686, over 3113320.96 frames. ], batch size: 165, lr: 4.61e-03, grad_scale: 2.0 2023-04-30 02:19:02,231 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.6927, 4.5217, 4.7427, 4.9351, 5.0629, 4.4995, 4.9635, 5.0308], device='cuda:1'), covar=tensor([0.1891, 0.1449, 0.1523, 0.0846, 0.0698, 0.1034, 0.1068, 0.1076], device='cuda:1'), in_proj_covar=tensor([0.0585, 0.0723, 0.0857, 0.0739, 0.0554, 0.0574, 0.0586, 0.0685], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-30 02:19:22,093 INFO [optim.py:368] (1/8) 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 02:19:23,200 INFO [train.py:904] (1/8) Epoch 15, batch 600, loss[loss=0.1536, simple_loss=0.2399, pruned_loss=0.0337, over 17231.00 frames. ], tot_loss[loss=0.1764, simple_loss=0.2589, pruned_loss=0.04694, over 3168724.03 frames. ], batch size: 44, lr: 4.61e-03, grad_scale: 2.0 2023-04-30 02:19:53,142 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=142724.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 02:20:32,860 INFO [train.py:904] (1/8) Epoch 15, batch 650, loss[loss=0.1854, simple_loss=0.2588, pruned_loss=0.05596, over 16701.00 frames. ], tot_loss[loss=0.1745, simple_loss=0.257, pruned_loss=0.046, over 3205014.03 frames. ], batch size: 134, lr: 4.61e-03, grad_scale: 2.0 2023-04-30 02:21:17,825 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=142785.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 02:21:40,880 INFO [optim.py:368] (1/8) 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] (1/8) Epoch 15, batch 700, loss[loss=0.1615, simple_loss=0.2514, pruned_loss=0.03585, over 17226.00 frames. ], tot_loss[loss=0.1741, simple_loss=0.2568, pruned_loss=0.04575, over 3229730.48 frames. ], batch size: 45, lr: 4.61e-03, grad_scale: 2.0 2023-04-30 02:22:50,283 INFO [train.py:904] (1/8) Epoch 15, batch 750, loss[loss=0.1685, simple_loss=0.2629, pruned_loss=0.03706, over 16755.00 frames. ], tot_loss[loss=0.1739, simple_loss=0.2568, pruned_loss=0.04551, over 3257956.74 frames. ], batch size: 62, lr: 4.61e-03, grad_scale: 2.0 2023-04-30 02:23:09,841 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([5.3764, 5.3516, 5.1466, 4.5899, 5.1334, 1.9320, 4.9230, 5.0603], device='cuda:1'), covar=tensor([0.0067, 0.0059, 0.0170, 0.0338, 0.0078, 0.2437, 0.0117, 0.0176], device='cuda:1'), in_proj_covar=tensor([0.0145, 0.0134, 0.0180, 0.0164, 0.0152, 0.0196, 0.0168, 0.0163], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-30 02:23:57,713 INFO [optim.py:368] (1/8) 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] (1/8) Epoch 15, batch 800, loss[loss=0.1729, simple_loss=0.2541, pruned_loss=0.04587, over 16869.00 frames. ], tot_loss[loss=0.1742, simple_loss=0.257, pruned_loss=0.04567, over 3276144.44 frames. ], batch size: 96, lr: 4.60e-03, grad_scale: 4.0 2023-04-30 02:24:36,439 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=142929.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 02:25:08,647 INFO [train.py:904] (1/8) Epoch 15, batch 850, loss[loss=0.1525, simple_loss=0.2353, pruned_loss=0.03486, over 16986.00 frames. ], tot_loss[loss=0.1734, simple_loss=0.2567, pruned_loss=0.04505, over 3286994.39 frames. ], batch size: 41, lr: 4.60e-03, grad_scale: 4.0 2023-04-30 02:26:15,121 INFO [optim.py:368] (1/8) 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] (1/8) Epoch 15, batch 900, loss[loss=0.1675, simple_loss=0.252, pruned_loss=0.0415, over 16466.00 frames. ], tot_loss[loss=0.1727, simple_loss=0.2557, pruned_loss=0.04481, over 3295726.67 frames. ], batch size: 75, lr: 4.60e-03, grad_scale: 4.0 2023-04-30 02:26:32,134 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.9493, 2.1025, 2.4088, 3.2022, 2.1927, 2.2987, 2.2890, 2.1978], device='cuda:1'), covar=tensor([0.1113, 0.3021, 0.2269, 0.0724, 0.3732, 0.2310, 0.2793, 0.3226], device='cuda:1'), in_proj_covar=tensor([0.0375, 0.0413, 0.0345, 0.0323, 0.0420, 0.0473, 0.0377, 0.0482], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-30 02:27:24,767 INFO [train.py:904] (1/8) Epoch 15, batch 950, loss[loss=0.1406, simple_loss=0.225, pruned_loss=0.02807, over 17216.00 frames. ], tot_loss[loss=0.1729, simple_loss=0.2559, pruned_loss=0.04497, over 3298481.14 frames. ], batch size: 43, lr: 4.60e-03, grad_scale: 4.0 2023-04-30 02:27:44,686 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.52 vs. limit=2.0 2023-04-30 02:28:02,445 INFO [zipformer.py:625] (1/8) 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] (1/8) 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] (1/8) Epoch 15, batch 1000, loss[loss=0.1798, simple_loss=0.2707, pruned_loss=0.04444, over 17025.00 frames. ], tot_loss[loss=0.1728, simple_loss=0.2551, pruned_loss=0.04529, over 3305415.35 frames. ], batch size: 50, lr: 4.60e-03, grad_scale: 4.0 2023-04-30 02:29:03,804 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([5.0376, 5.4380, 5.6034, 5.2678, 5.3907, 5.9890, 5.4727, 5.1871], device='cuda:1'), covar=tensor([0.0843, 0.1877, 0.2408, 0.2024, 0.2632, 0.1006, 0.1539, 0.2277], device='cuda:1'), in_proj_covar=tensor([0.0385, 0.0545, 0.0603, 0.0465, 0.0624, 0.0627, 0.0477, 0.0618], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-04-30 02:29:41,403 INFO [train.py:904] (1/8) Epoch 15, batch 1050, loss[loss=0.1699, simple_loss=0.2658, pruned_loss=0.03698, over 17046.00 frames. ], tot_loss[loss=0.1726, simple_loss=0.2553, pruned_loss=0.04497, over 3316180.14 frames. ], batch size: 53, lr: 4.60e-03, grad_scale: 4.0 2023-04-30 02:30:28,872 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.9155, 4.1579, 3.2124, 2.3118, 2.8296, 2.5494, 4.4335, 3.6745], device='cuda:1'), covar=tensor([0.2510, 0.0620, 0.1503, 0.2530, 0.2606, 0.1927, 0.0391, 0.1151], device='cuda:1'), in_proj_covar=tensor([0.0312, 0.0263, 0.0294, 0.0291, 0.0281, 0.0238, 0.0277, 0.0314], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:1') 2023-04-30 02:30:46,963 INFO [optim.py:368] (1/8) 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] (1/8) Epoch 15, batch 1100, loss[loss=0.1698, simple_loss=0.2476, pruned_loss=0.04596, over 16565.00 frames. ], tot_loss[loss=0.1722, simple_loss=0.255, pruned_loss=0.04475, over 3310735.28 frames. ], batch size: 68, lr: 4.60e-03, grad_scale: 4.0 2023-04-30 02:30:49,788 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.91 vs. limit=2.0 2023-04-30 02:31:25,789 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=143229.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 02:31:58,385 INFO [train.py:904] (1/8) Epoch 15, batch 1150, loss[loss=0.1794, simple_loss=0.2747, pruned_loss=0.04209, over 17072.00 frames. ], tot_loss[loss=0.1711, simple_loss=0.2543, pruned_loss=0.04393, over 3310634.26 frames. ], batch size: 55, lr: 4.60e-03, grad_scale: 4.0 2023-04-30 02:32:34,543 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=143277.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 02:32:49,016 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.6001, 2.5083, 2.1222, 2.3587, 2.8831, 2.6281, 3.3015, 3.1951], device='cuda:1'), covar=tensor([0.0130, 0.0423, 0.0502, 0.0439, 0.0283, 0.0373, 0.0299, 0.0260], device='cuda:1'), in_proj_covar=tensor([0.0179, 0.0224, 0.0215, 0.0216, 0.0224, 0.0224, 0.0229, 0.0216], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-30 02:33:07,201 INFO [optim.py:368] (1/8) 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] (1/8) Epoch 15, batch 1200, loss[loss=0.1847, simple_loss=0.2528, pruned_loss=0.05832, over 16745.00 frames. ], tot_loss[loss=0.1701, simple_loss=0.2532, pruned_loss=0.04348, over 3318813.71 frames. ], batch size: 83, lr: 4.60e-03, grad_scale: 8.0 2023-04-30 02:33:41,316 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.6859, 2.4631, 2.0351, 2.3635, 2.9211, 2.6465, 3.4177, 3.1934], device='cuda:1'), covar=tensor([0.0104, 0.0421, 0.0490, 0.0431, 0.0260, 0.0373, 0.0205, 0.0247], device='cuda:1'), in_proj_covar=tensor([0.0179, 0.0224, 0.0215, 0.0216, 0.0223, 0.0222, 0.0228, 0.0216], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-30 02:34:16,291 INFO [train.py:904] (1/8) Epoch 15, batch 1250, loss[loss=0.1621, simple_loss=0.2583, pruned_loss=0.0329, over 17138.00 frames. ], tot_loss[loss=0.1704, simple_loss=0.2536, pruned_loss=0.04359, over 3322093.00 frames. ], batch size: 48, lr: 4.60e-03, grad_scale: 8.0 2023-04-30 02:34:30,366 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.2651, 4.0762, 4.2841, 4.4564, 4.5525, 4.1311, 4.3306, 4.5521], device='cuda:1'), covar=tensor([0.1389, 0.1154, 0.1302, 0.0644, 0.0575, 0.1198, 0.1881, 0.0689], device='cuda:1'), in_proj_covar=tensor([0.0594, 0.0730, 0.0874, 0.0746, 0.0559, 0.0580, 0.0591, 0.0696], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-30 02:34:56,686 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=143380.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 02:35:25,441 INFO [zipformer.py:625] (1/8) 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] (1/8) 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] (1/8) Epoch 15, batch 1300, loss[loss=0.1865, simple_loss=0.2588, pruned_loss=0.0571, over 16824.00 frames. ], tot_loss[loss=0.1707, simple_loss=0.2536, pruned_loss=0.04386, over 3319245.60 frames. ], batch size: 116, lr: 4.60e-03, grad_scale: 8.0 2023-04-30 02:36:03,858 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=143428.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 02:36:13,561 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.92 vs. limit=2.0 2023-04-30 02:36:37,209 INFO [train.py:904] (1/8) Epoch 15, batch 1350, loss[loss=0.1454, simple_loss=0.2314, pruned_loss=0.02975, over 17210.00 frames. ], tot_loss[loss=0.17, simple_loss=0.2534, pruned_loss=0.04328, over 3321842.68 frames. ], batch size: 43, lr: 4.60e-03, grad_scale: 8.0 2023-04-30 02:36:49,864 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=143461.0, num_to_drop=1, layers_to_drop={2} 2023-04-30 02:36:59,334 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([1.7704, 2.9304, 2.7368, 5.0485, 4.1481, 4.5514, 1.6058, 3.2335], device='cuda:1'), covar=tensor([0.1372, 0.0732, 0.1170, 0.0151, 0.0242, 0.0370, 0.1585, 0.0721], device='cuda:1'), in_proj_covar=tensor([0.0160, 0.0165, 0.0186, 0.0167, 0.0197, 0.0213, 0.0190, 0.0185], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-04-30 02:37:06,523 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=143473.0, num_to_drop=1, layers_to_drop={0} 2023-04-30 02:37:45,743 INFO [optim.py:368] (1/8) 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] (1/8) Epoch 15, batch 1400, loss[loss=0.1603, simple_loss=0.2554, pruned_loss=0.03255, over 17127.00 frames. ], tot_loss[loss=0.1696, simple_loss=0.2536, pruned_loss=0.04279, over 3324922.27 frames. ], batch size: 48, lr: 4.59e-03, grad_scale: 8.0 2023-04-30 02:38:14,249 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.8684, 4.4148, 4.4066, 3.3194, 3.6690, 4.4616, 3.9156, 2.5620], device='cuda:1'), covar=tensor([0.0438, 0.0049, 0.0036, 0.0279, 0.0096, 0.0054, 0.0068, 0.0417], device='cuda:1'), in_proj_covar=tensor([0.0132, 0.0076, 0.0075, 0.0131, 0.0088, 0.0098, 0.0086, 0.0126], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-04-30 02:38:31,362 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=143534.0, num_to_drop=1, layers_to_drop={1} 2023-04-30 02:38:37,235 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.88 vs. limit=2.0 2023-04-30 02:38:39,732 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.8569, 3.9732, 3.0703, 2.3169, 2.6720, 2.4499, 3.9886, 3.5803], device='cuda:1'), covar=tensor([0.2372, 0.0547, 0.1417, 0.2582, 0.2332, 0.1862, 0.0514, 0.1228], device='cuda:1'), in_proj_covar=tensor([0.0311, 0.0262, 0.0293, 0.0290, 0.0283, 0.0238, 0.0276, 0.0313], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:1') 2023-04-30 02:38:55,992 INFO [train.py:904] (1/8) Epoch 15, batch 1450, loss[loss=0.1499, simple_loss=0.224, pruned_loss=0.03786, over 12624.00 frames. ], tot_loss[loss=0.1692, simple_loss=0.2527, pruned_loss=0.04285, over 3318817.62 frames. ], batch size: 246, lr: 4.59e-03, grad_scale: 8.0 2023-04-30 02:40:05,552 INFO [optim.py:368] (1/8) 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] (1/8) Epoch 15, batch 1500, loss[loss=0.1817, simple_loss=0.2507, pruned_loss=0.05632, over 16516.00 frames. ], tot_loss[loss=0.1687, simple_loss=0.2523, pruned_loss=0.04252, over 3319387.96 frames. ], batch size: 146, lr: 4.59e-03, grad_scale: 8.0 2023-04-30 02:41:01,667 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=4.69 vs. limit=5.0 2023-04-30 02:41:14,536 INFO [train.py:904] (1/8) Epoch 15, batch 1550, loss[loss=0.2056, simple_loss=0.27, pruned_loss=0.07062, over 16901.00 frames. ], tot_loss[loss=0.1713, simple_loss=0.2536, pruned_loss=0.04451, over 3313242.89 frames. ], batch size: 109, lr: 4.59e-03, grad_scale: 8.0 2023-04-30 02:41:46,727 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([5.2183, 5.0272, 4.9244, 4.4482, 4.5588, 4.9990, 4.9918, 4.6426], device='cuda:1'), covar=tensor([0.0522, 0.0502, 0.0346, 0.0363, 0.1196, 0.0461, 0.0352, 0.0749], device='cuda:1'), in_proj_covar=tensor([0.0272, 0.0375, 0.0326, 0.0306, 0.0341, 0.0356, 0.0222, 0.0386], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002], device='cuda:1') 2023-04-30 02:42:22,873 INFO [optim.py:368] (1/8) 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 02:42:24,078 INFO [train.py:904] (1/8) Epoch 15, batch 1600, loss[loss=0.1728, simple_loss=0.2701, pruned_loss=0.03772, over 17143.00 frames. ], tot_loss[loss=0.1726, simple_loss=0.2554, pruned_loss=0.04496, over 3323050.44 frames. ], batch size: 49, lr: 4.59e-03, grad_scale: 8.0 2023-04-30 02:43:02,275 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.89 vs. limit=2.0 2023-04-30 02:43:35,418 INFO [train.py:904] (1/8) Epoch 15, batch 1650, loss[loss=0.185, simple_loss=0.2561, pruned_loss=0.05693, over 16914.00 frames. ], tot_loss[loss=0.1751, simple_loss=0.2576, pruned_loss=0.04627, over 3305153.33 frames. ], batch size: 116, lr: 4.59e-03, grad_scale: 8.0 2023-04-30 02:43:40,913 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=143756.0, num_to_drop=1, layers_to_drop={0} 2023-04-30 02:44:22,504 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.2769, 3.1701, 3.3153, 2.3362, 3.0581, 3.4297, 3.2028, 1.6192], device='cuda:1'), covar=tensor([0.0434, 0.0112, 0.0070, 0.0355, 0.0125, 0.0109, 0.0110, 0.0547], device='cuda:1'), in_proj_covar=tensor([0.0131, 0.0075, 0.0074, 0.0130, 0.0087, 0.0098, 0.0085, 0.0124], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-04-30 02:44:46,124 INFO [optim.py:368] (1/8) 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] (1/8) Epoch 15, batch 1700, loss[loss=0.1868, simple_loss=0.2693, pruned_loss=0.05216, over 16847.00 frames. ], tot_loss[loss=0.1763, simple_loss=0.2594, pruned_loss=0.04661, over 3314440.02 frames. ], batch size: 96, lr: 4.59e-03, grad_scale: 4.0 2023-04-30 02:44:53,825 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.89 vs. limit=2.0 2023-04-30 02:45:22,464 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=143829.0, num_to_drop=1, layers_to_drop={1} 2023-04-30 02:45:23,045 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.90 vs. limit=2.0 2023-04-30 02:45:31,692 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([1.8234, 3.0327, 2.5734, 4.5014, 3.6224, 4.2751, 1.6776, 2.9542], device='cuda:1'), covar=tensor([0.1394, 0.0665, 0.1117, 0.0158, 0.0214, 0.0398, 0.1542, 0.0803], device='cuda:1'), in_proj_covar=tensor([0.0159, 0.0164, 0.0184, 0.0167, 0.0197, 0.0213, 0.0188, 0.0186], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-04-30 02:45:40,031 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.0802, 2.0679, 2.2456, 3.6646, 2.1225, 2.3838, 2.1808, 2.2563], device='cuda:1'), covar=tensor([0.1265, 0.3549, 0.2615, 0.0596, 0.3833, 0.2503, 0.3685, 0.3044], device='cuda:1'), in_proj_covar=tensor([0.0379, 0.0417, 0.0349, 0.0328, 0.0425, 0.0482, 0.0383, 0.0489], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-30 02:45:51,847 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.4211, 4.2369, 4.6614, 2.5302, 4.8655, 4.9000, 3.4473, 3.9736], device='cuda:1'), covar=tensor([0.0554, 0.0180, 0.0171, 0.1016, 0.0071, 0.0161, 0.0367, 0.0302], device='cuda:1'), in_proj_covar=tensor([0.0149, 0.0106, 0.0093, 0.0143, 0.0075, 0.0119, 0.0128, 0.0132], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0004, 0.0004], device='cuda:1') 2023-04-30 02:45:54,336 INFO [train.py:904] (1/8) Epoch 15, batch 1750, loss[loss=0.1513, simple_loss=0.2373, pruned_loss=0.03268, over 16860.00 frames. ], tot_loss[loss=0.1765, simple_loss=0.2601, pruned_loss=0.04639, over 3315309.54 frames. ], batch size: 42, lr: 4.59e-03, grad_scale: 4.0 2023-04-30 02:46:39,390 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.8088, 3.9271, 2.3283, 4.5126, 2.7474, 4.4940, 2.6053, 3.0076], device='cuda:1'), covar=tensor([0.0234, 0.0336, 0.1494, 0.0216, 0.0829, 0.0380, 0.1258, 0.0719], device='cuda:1'), in_proj_covar=tensor([0.0164, 0.0173, 0.0194, 0.0150, 0.0172, 0.0216, 0.0201, 0.0177], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-04-30 02:46:42,745 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.89 vs. limit=2.0 2023-04-30 02:46:45,424 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([1.9565, 3.3488, 2.6101, 5.1245, 4.3013, 4.5149, 1.6776, 3.0300], device='cuda:1'), covar=tensor([0.1166, 0.0546, 0.1104, 0.0128, 0.0212, 0.0400, 0.1370, 0.0793], device='cuda:1'), in_proj_covar=tensor([0.0157, 0.0163, 0.0183, 0.0166, 0.0196, 0.0211, 0.0187, 0.0184], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-04-30 02:46:52,088 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.6644, 2.9943, 3.0505, 2.0308, 2.6553, 2.1911, 3.1587, 3.3246], device='cuda:1'), covar=tensor([0.0297, 0.0813, 0.0595, 0.1782, 0.0887, 0.0967, 0.0634, 0.0774], device='cuda:1'), in_proj_covar=tensor([0.0149, 0.0151, 0.0161, 0.0146, 0.0138, 0.0125, 0.0138, 0.0163], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:1') 2023-04-30 02:47:05,601 INFO [optim.py:368] (1/8) 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 02:47:05,616 INFO [train.py:904] (1/8) Epoch 15, batch 1800, loss[loss=0.1651, simple_loss=0.2439, pruned_loss=0.0431, over 16210.00 frames. ], tot_loss[loss=0.1768, simple_loss=0.2613, pruned_loss=0.04611, over 3322432.77 frames. ], batch size: 36, lr: 4.59e-03, grad_scale: 4.0 2023-04-30 02:47:54,500 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.1259, 4.5349, 3.4038, 2.3885, 2.9246, 2.6895, 4.8404, 3.7967], device='cuda:1'), covar=tensor([0.2363, 0.0531, 0.1511, 0.2397, 0.2555, 0.1809, 0.0306, 0.1176], device='cuda:1'), in_proj_covar=tensor([0.0311, 0.0262, 0.0292, 0.0290, 0.0282, 0.0237, 0.0276, 0.0313], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:1') 2023-04-30 02:48:15,535 INFO [train.py:904] (1/8) Epoch 15, batch 1850, loss[loss=0.158, simple_loss=0.254, pruned_loss=0.03097, over 17126.00 frames. ], tot_loss[loss=0.1769, simple_loss=0.2621, pruned_loss=0.04591, over 3308413.98 frames. ], batch size: 48, lr: 4.59e-03, grad_scale: 2.0 2023-04-30 02:48:57,985 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.9298, 2.0112, 2.4254, 2.9296, 2.8810, 3.4424, 2.0928, 3.3076], device='cuda:1'), covar=tensor([0.0218, 0.0431, 0.0302, 0.0251, 0.0251, 0.0147, 0.0423, 0.0126], device='cuda:1'), in_proj_covar=tensor([0.0171, 0.0182, 0.0166, 0.0170, 0.0179, 0.0135, 0.0182, 0.0128], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-30 02:49:30,765 INFO [train.py:904] (1/8) Epoch 15, batch 1900, loss[loss=0.1897, simple_loss=0.2721, pruned_loss=0.05368, over 16879.00 frames. ], tot_loss[loss=0.1762, simple_loss=0.2611, pruned_loss=0.04567, over 3317124.06 frames. ], batch size: 116, lr: 4.59e-03, grad_scale: 2.0 2023-04-30 02:49:31,845 INFO [optim.py:368] (1/8) 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 02:49:39,716 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.6399, 3.9125, 4.1651, 2.2074, 3.4147, 2.7691, 4.0601, 4.1060], device='cuda:1'), covar=tensor([0.0252, 0.0808, 0.0440, 0.1809, 0.0707, 0.0879, 0.0554, 0.1031], device='cuda:1'), in_proj_covar=tensor([0.0150, 0.0153, 0.0162, 0.0146, 0.0139, 0.0126, 0.0139, 0.0164], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:1') 2023-04-30 02:50:03,273 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.6793, 1.7168, 2.1469, 2.5348, 2.6156, 2.6010, 1.6913, 2.7785], device='cuda:1'), covar=tensor([0.0143, 0.0389, 0.0265, 0.0207, 0.0199, 0.0216, 0.0419, 0.0119], device='cuda:1'), in_proj_covar=tensor([0.0171, 0.0182, 0.0165, 0.0170, 0.0179, 0.0136, 0.0182, 0.0128], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-30 02:50:39,887 INFO [train.py:904] (1/8) Epoch 15, batch 1950, loss[loss=0.1617, simple_loss=0.2508, pruned_loss=0.03635, over 17166.00 frames. ], tot_loss[loss=0.1757, simple_loss=0.2611, pruned_loss=0.04512, over 3315226.09 frames. ], batch size: 46, lr: 4.59e-03, grad_scale: 2.0 2023-04-30 02:50:46,105 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=144056.0, num_to_drop=1, layers_to_drop={2} 2023-04-30 02:51:49,541 INFO [train.py:904] (1/8) Epoch 15, batch 2000, loss[loss=0.1769, simple_loss=0.2515, pruned_loss=0.05115, over 16744.00 frames. ], tot_loss[loss=0.1751, simple_loss=0.2603, pruned_loss=0.0449, over 3318887.18 frames. ], batch size: 83, lr: 4.59e-03, grad_scale: 4.0 2023-04-30 02:51:51,363 INFO [optim.py:368] (1/8) 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] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=144104.0, num_to_drop=1, layers_to_drop={1} 2023-04-30 02:52:27,338 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=144129.0, num_to_drop=1, layers_to_drop={2} 2023-04-30 02:52:29,049 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.32 vs. limit=2.0 2023-04-30 02:52:50,812 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.60 vs. limit=2.0 2023-04-30 02:52:56,128 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=5.03 vs. limit=5.0 2023-04-30 02:52:58,038 INFO [train.py:904] (1/8) Epoch 15, batch 2050, loss[loss=0.1635, simple_loss=0.2543, pruned_loss=0.03634, over 17015.00 frames. ], tot_loss[loss=0.1755, simple_loss=0.2605, pruned_loss=0.04521, over 3323875.19 frames. ], batch size: 50, lr: 4.58e-03, grad_scale: 4.0 2023-04-30 02:53:06,669 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.9079, 4.1397, 4.0289, 3.2514, 3.5806, 4.0506, 3.7987, 2.0899], device='cuda:1'), covar=tensor([0.0426, 0.0073, 0.0064, 0.0300, 0.0124, 0.0166, 0.0162, 0.0560], device='cuda:1'), in_proj_covar=tensor([0.0133, 0.0076, 0.0075, 0.0131, 0.0088, 0.0100, 0.0086, 0.0126], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-04-30 02:53:32,905 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=144177.0, num_to_drop=1, layers_to_drop={1} 2023-04-30 02:53:35,611 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.37 vs. limit=2.0 2023-04-30 02:54:07,933 INFO [train.py:904] (1/8) Epoch 15, batch 2100, loss[loss=0.219, simple_loss=0.2878, pruned_loss=0.07506, over 16685.00 frames. ], tot_loss[loss=0.1774, simple_loss=0.2622, pruned_loss=0.04634, over 3326723.27 frames. ], batch size: 134, lr: 4.58e-03, grad_scale: 4.0 2023-04-30 02:54:08,980 INFO [optim.py:368] (1/8) 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 02:54:09,476 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([5.1933, 5.2181, 4.9812, 4.4922, 5.0307, 1.9923, 4.8297, 4.9690], device='cuda:1'), covar=tensor([0.0080, 0.0069, 0.0165, 0.0325, 0.0078, 0.2531, 0.0110, 0.0157], device='cuda:1'), in_proj_covar=tensor([0.0149, 0.0138, 0.0187, 0.0170, 0.0158, 0.0198, 0.0173, 0.0168], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-30 02:54:24,735 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.7807, 3.8602, 2.9556, 2.2551, 2.5639, 2.3636, 3.9770, 3.4598], device='cuda:1'), covar=tensor([0.2443, 0.0579, 0.1538, 0.2625, 0.2443, 0.1943, 0.0466, 0.1227], device='cuda:1'), in_proj_covar=tensor([0.0313, 0.0264, 0.0293, 0.0292, 0.0285, 0.0238, 0.0277, 0.0317], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:1') 2023-04-30 02:55:17,937 INFO [train.py:904] (1/8) Epoch 15, batch 2150, loss[loss=0.1835, simple_loss=0.2773, pruned_loss=0.04485, over 17083.00 frames. ], tot_loss[loss=0.1781, simple_loss=0.2626, pruned_loss=0.04676, over 3327462.56 frames. ], batch size: 53, lr: 4.58e-03, grad_scale: 4.0 2023-04-30 02:55:21,286 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.0095, 1.9752, 2.5237, 2.9649, 2.7398, 3.4804, 2.1694, 3.3601], device='cuda:1'), covar=tensor([0.0199, 0.0413, 0.0281, 0.0252, 0.0261, 0.0155, 0.0386, 0.0137], device='cuda:1'), in_proj_covar=tensor([0.0172, 0.0183, 0.0167, 0.0171, 0.0181, 0.0137, 0.0183, 0.0129], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-30 02:56:06,470 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=3.30 vs. limit=5.0 2023-04-30 02:56:25,351 INFO [train.py:904] (1/8) Epoch 15, batch 2200, loss[loss=0.2014, simple_loss=0.2803, pruned_loss=0.06131, over 15859.00 frames. ], tot_loss[loss=0.1787, simple_loss=0.2634, pruned_loss=0.04703, over 3315436.56 frames. ], batch size: 35, lr: 4.58e-03, grad_scale: 4.0 2023-04-30 02:56:27,073 INFO [optim.py:368] (1/8) 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 02:56:43,634 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([5.2916, 5.3404, 5.0695, 4.6001, 5.0897, 2.2712, 4.8635, 5.0593], device='cuda:1'), covar=tensor([0.0086, 0.0060, 0.0168, 0.0357, 0.0099, 0.2223, 0.0123, 0.0162], device='cuda:1'), in_proj_covar=tensor([0.0148, 0.0137, 0.0186, 0.0170, 0.0157, 0.0197, 0.0173, 0.0167], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-30 02:56:56,469 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.1555, 3.3687, 3.4607, 2.2518, 2.9443, 2.4825, 3.6128, 3.6745], device='cuda:1'), covar=tensor([0.0230, 0.0783, 0.0500, 0.1578, 0.0719, 0.0848, 0.0461, 0.0715], device='cuda:1'), in_proj_covar=tensor([0.0151, 0.0154, 0.0162, 0.0147, 0.0140, 0.0126, 0.0139, 0.0166], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:1') 2023-04-30 02:57:36,224 INFO [train.py:904] (1/8) Epoch 15, batch 2250, loss[loss=0.1727, simple_loss=0.2702, pruned_loss=0.03761, over 17109.00 frames. ], tot_loss[loss=0.1803, simple_loss=0.2646, pruned_loss=0.048, over 3304133.39 frames. ], batch size: 47, lr: 4.58e-03, grad_scale: 4.0 2023-04-30 02:58:46,690 INFO [train.py:904] (1/8) Epoch 15, batch 2300, loss[loss=0.2606, simple_loss=0.3255, pruned_loss=0.09784, over 11948.00 frames. ], tot_loss[loss=0.1807, simple_loss=0.2648, pruned_loss=0.04835, over 3306282.72 frames. ], batch size: 246, lr: 4.58e-03, grad_scale: 4.0 2023-04-30 02:58:47,892 INFO [optim.py:368] (1/8) 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 02:58:58,059 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=144411.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 02:59:10,022 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=4.96 vs. limit=5.0 2023-04-30 02:59:47,976 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.9267, 1.9231, 2.4526, 2.8429, 2.6231, 3.3991, 2.1696, 3.2330], device='cuda:1'), covar=tensor([0.0211, 0.0443, 0.0286, 0.0284, 0.0291, 0.0157, 0.0411, 0.0153], device='cuda:1'), in_proj_covar=tensor([0.0174, 0.0184, 0.0170, 0.0173, 0.0183, 0.0138, 0.0185, 0.0131], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-30 02:59:53,215 INFO [train.py:904] (1/8) Epoch 15, batch 2350, loss[loss=0.1704, simple_loss=0.2627, pruned_loss=0.039, over 17048.00 frames. ], tot_loss[loss=0.1811, simple_loss=0.2654, pruned_loss=0.04837, over 3301761.05 frames. ], batch size: 50, lr: 4.58e-03, grad_scale: 4.0 2023-04-30 03:00:20,708 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=144472.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 03:00:39,253 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.7419, 2.3126, 2.2911, 4.6188, 2.2858, 2.7820, 2.4141, 2.5087], device='cuda:1'), covar=tensor([0.0995, 0.3461, 0.2740, 0.0376, 0.3817, 0.2475, 0.3269, 0.3321], device='cuda:1'), in_proj_covar=tensor([0.0380, 0.0418, 0.0350, 0.0328, 0.0424, 0.0481, 0.0383, 0.0489], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-30 03:00:47,832 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.43 vs. limit=2.0 2023-04-30 03:01:02,749 INFO [train.py:904] (1/8) Epoch 15, batch 2400, loss[loss=0.1407, simple_loss=0.2283, pruned_loss=0.02657, over 16780.00 frames. ], tot_loss[loss=0.1825, simple_loss=0.267, pruned_loss=0.04896, over 3292362.44 frames. ], batch size: 39, lr: 4.58e-03, grad_scale: 8.0 2023-04-30 03:01:04,724 INFO [optim.py:368] (1/8) Clipping_scale=2.0, grad-norm quartiles 1.656e+02 2.391e+02 2.805e+02 3.317e+02 7.772e+02, threshold=5.609e+02, percent-clipped=1.0 2023-04-30 03:01:24,740 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.3531, 4.2753, 4.5960, 2.4392, 4.8321, 4.8463, 3.4187, 3.9009], device='cuda:1'), covar=tensor([0.0607, 0.0173, 0.0179, 0.1015, 0.0057, 0.0153, 0.0370, 0.0307], device='cuda:1'), in_proj_covar=tensor([0.0146, 0.0105, 0.0091, 0.0139, 0.0073, 0.0117, 0.0125, 0.0129], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:1') 2023-04-30 03:02:10,317 INFO [train.py:904] (1/8) Epoch 15, batch 2450, loss[loss=0.1948, simple_loss=0.2747, pruned_loss=0.05745, over 16787.00 frames. ], tot_loss[loss=0.182, simple_loss=0.2669, pruned_loss=0.04853, over 3305134.01 frames. ], batch size: 102, lr: 4.58e-03, grad_scale: 8.0 2023-04-30 03:02:14,199 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.7332, 1.8363, 2.2493, 2.5590, 2.6669, 2.6305, 1.8725, 2.8764], device='cuda:1'), covar=tensor([0.0151, 0.0388, 0.0291, 0.0218, 0.0211, 0.0254, 0.0415, 0.0116], device='cuda:1'), in_proj_covar=tensor([0.0173, 0.0183, 0.0168, 0.0172, 0.0181, 0.0138, 0.0183, 0.0130], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-30 03:03:17,679 INFO [train.py:904] (1/8) Epoch 15, batch 2500, loss[loss=0.173, simple_loss=0.2614, pruned_loss=0.04231, over 17214.00 frames. ], tot_loss[loss=0.182, simple_loss=0.267, pruned_loss=0.04845, over 3289619.29 frames. ], batch size: 44, lr: 4.58e-03, grad_scale: 8.0 2023-04-30 03:03:18,675 INFO [optim.py:368] (1/8) Clipping_scale=2.0, grad-norm quartiles 1.596e+02 2.258e+02 2.678e+02 3.424e+02 5.626e+02, threshold=5.355e+02, percent-clipped=1.0 2023-04-30 03:04:26,864 INFO [train.py:904] (1/8) Epoch 15, batch 2550, loss[loss=0.1834, simple_loss=0.2756, pruned_loss=0.04564, over 16705.00 frames. ], tot_loss[loss=0.1818, simple_loss=0.2668, pruned_loss=0.04844, over 3296811.92 frames. ], batch size: 57, lr: 4.58e-03, grad_scale: 8.0 2023-04-30 03:05:18,914 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([5.1541, 5.5656, 5.7819, 5.4932, 5.5717, 6.2072, 5.7306, 5.3705], device='cuda:1'), covar=tensor([0.0814, 0.1830, 0.2079, 0.2095, 0.2631, 0.0925, 0.1290, 0.2299], device='cuda:1'), in_proj_covar=tensor([0.0390, 0.0558, 0.0612, 0.0472, 0.0634, 0.0638, 0.0482, 0.0624], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-04-30 03:05:34,894 INFO [train.py:904] (1/8) Epoch 15, batch 2600, loss[loss=0.1863, simple_loss=0.2692, pruned_loss=0.05171, over 16535.00 frames. ], tot_loss[loss=0.181, simple_loss=0.2662, pruned_loss=0.04786, over 3310705.54 frames. ], batch size: 75, lr: 4.58e-03, grad_scale: 8.0 2023-04-30 03:05:36,052 INFO [optim.py:368] (1/8) 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 03:05:39,955 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.5148, 2.9302, 2.9219, 1.9937, 2.5511, 2.1574, 3.0441, 3.2273], device='cuda:1'), covar=tensor([0.0305, 0.0797, 0.0682, 0.1806, 0.0903, 0.0975, 0.0664, 0.0751], device='cuda:1'), in_proj_covar=tensor([0.0151, 0.0155, 0.0163, 0.0147, 0.0140, 0.0126, 0.0140, 0.0166], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:1') 2023-04-30 03:05:59,926 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([5.4013, 5.7654, 5.5118, 5.5977, 5.3065, 5.0730, 5.1474, 5.9057], device='cuda:1'), covar=tensor([0.1292, 0.0943, 0.1130, 0.0775, 0.0811, 0.0761, 0.1205, 0.0912], device='cuda:1'), in_proj_covar=tensor([0.0623, 0.0775, 0.0632, 0.0554, 0.0488, 0.0493, 0.0643, 0.0586], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:1') 2023-04-30 03:06:01,276 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.8016, 3.7554, 2.9164, 2.2418, 2.4813, 2.3328, 3.8389, 3.4125], device='cuda:1'), covar=tensor([0.2255, 0.0571, 0.1468, 0.2453, 0.2401, 0.1784, 0.0472, 0.1210], device='cuda:1'), in_proj_covar=tensor([0.0312, 0.0265, 0.0295, 0.0293, 0.0287, 0.0238, 0.0279, 0.0318], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:1') 2023-04-30 03:06:43,554 INFO [train.py:904] (1/8) Epoch 15, batch 2650, loss[loss=0.1698, simple_loss=0.2553, pruned_loss=0.04214, over 16752.00 frames. ], tot_loss[loss=0.1801, simple_loss=0.2658, pruned_loss=0.04715, over 3320145.48 frames. ], batch size: 83, lr: 4.58e-03, grad_scale: 8.0 2023-04-30 03:06:55,338 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.7477, 4.9279, 5.1954, 5.0149, 4.9909, 5.6702, 5.1647, 4.8297], device='cuda:1'), covar=tensor([0.1357, 0.2049, 0.2215, 0.2271, 0.2939, 0.1122, 0.1587, 0.2540], device='cuda:1'), in_proj_covar=tensor([0.0394, 0.0560, 0.0614, 0.0474, 0.0639, 0.0643, 0.0487, 0.0629], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-04-30 03:07:05,862 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=144767.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 03:07:53,565 INFO [train.py:904] (1/8) Epoch 15, batch 2700, loss[loss=0.1789, simple_loss=0.2732, pruned_loss=0.04225, over 16768.00 frames. ], tot_loss[loss=0.1805, simple_loss=0.2669, pruned_loss=0.04705, over 3322588.04 frames. ], batch size: 57, lr: 4.57e-03, grad_scale: 8.0 2023-04-30 03:07:54,730 INFO [optim.py:368] (1/8) Clipping_scale=2.0, grad-norm quartiles 1.560e+02 2.168e+02 2.529e+02 3.023e+02 4.642e+02, threshold=5.059e+02, percent-clipped=0.0 2023-04-30 03:08:22,792 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.58 vs. limit=2.0 2023-04-30 03:08:45,705 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.3691, 3.3911, 3.6039, 2.4584, 3.2397, 3.7191, 3.4385, 2.1886], device='cuda:1'), covar=tensor([0.0446, 0.0114, 0.0043, 0.0355, 0.0093, 0.0068, 0.0081, 0.0369], device='cuda:1'), in_proj_covar=tensor([0.0132, 0.0076, 0.0075, 0.0130, 0.0088, 0.0099, 0.0086, 0.0125], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-04-30 03:09:02,445 INFO [train.py:904] (1/8) Epoch 15, batch 2750, loss[loss=0.1875, simple_loss=0.2614, pruned_loss=0.05675, over 16534.00 frames. ], tot_loss[loss=0.1804, simple_loss=0.2667, pruned_loss=0.04704, over 3318551.02 frames. ], batch size: 146, lr: 4.57e-03, grad_scale: 8.0 2023-04-30 03:09:26,368 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([5.0647, 5.0767, 5.5986, 5.5875, 5.6145, 5.1853, 5.1420, 4.8610], device='cuda:1'), covar=tensor([0.0393, 0.0551, 0.0352, 0.0383, 0.0460, 0.0366, 0.1025, 0.0431], device='cuda:1'), in_proj_covar=tensor([0.0379, 0.0403, 0.0400, 0.0382, 0.0445, 0.0424, 0.0517, 0.0336], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:1') 2023-04-30 03:10:11,028 INFO [train.py:904] (1/8) Epoch 15, batch 2800, loss[loss=0.172, simple_loss=0.2579, pruned_loss=0.04308, over 17165.00 frames. ], tot_loss[loss=0.1785, simple_loss=0.2654, pruned_loss=0.0458, over 3317482.85 frames. ], batch size: 46, lr: 4.57e-03, grad_scale: 8.0 2023-04-30 03:10:12,144 INFO [optim.py:368] (1/8) 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 03:10:52,925 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([1.9266, 2.8225, 2.5019, 4.3029, 3.6163, 4.2308, 1.5462, 2.8166], device='cuda:1'), covar=tensor([0.1232, 0.0605, 0.1045, 0.0149, 0.0161, 0.0330, 0.1388, 0.0774], device='cuda:1'), in_proj_covar=tensor([0.0158, 0.0165, 0.0185, 0.0169, 0.0199, 0.0212, 0.0188, 0.0184], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-04-30 03:10:54,953 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.0624, 3.1104, 3.2944, 2.1413, 2.8021, 2.4147, 3.5286, 3.4414], device='cuda:1'), covar=tensor([0.0200, 0.0840, 0.0605, 0.1667, 0.0819, 0.0893, 0.0490, 0.0785], device='cuda:1'), in_proj_covar=tensor([0.0152, 0.0156, 0.0164, 0.0148, 0.0141, 0.0127, 0.0141, 0.0167], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:1') 2023-04-30 03:11:21,054 INFO [train.py:904] (1/8) Epoch 15, batch 2850, loss[loss=0.1972, simple_loss=0.2773, pruned_loss=0.05858, over 16305.00 frames. ], tot_loss[loss=0.1779, simple_loss=0.2647, pruned_loss=0.04556, over 3327944.29 frames. ], batch size: 165, lr: 4.57e-03, grad_scale: 8.0 2023-04-30 03:12:22,134 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=144996.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 03:12:31,819 INFO [train.py:904] (1/8) Epoch 15, batch 2900, loss[loss=0.2251, simple_loss=0.2946, pruned_loss=0.07779, over 15438.00 frames. ], tot_loss[loss=0.1782, simple_loss=0.2641, pruned_loss=0.04618, over 3332364.80 frames. ], batch size: 190, lr: 4.57e-03, grad_scale: 8.0 2023-04-30 03:12:33,008 INFO [optim.py:368] (1/8) Clipping_scale=2.0, grad-norm quartiles 1.595e+02 2.445e+02 2.844e+02 3.300e+02 6.709e+02, threshold=5.687e+02, percent-clipped=6.0 2023-04-30 03:12:45,145 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.34 vs. limit=2.0 2023-04-30 03:13:02,682 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.3696, 3.1570, 3.3854, 1.9029, 3.4790, 3.4367, 2.7497, 2.6712], device='cuda:1'), covar=tensor([0.0674, 0.0197, 0.0172, 0.0981, 0.0086, 0.0210, 0.0439, 0.0417], device='cuda:1'), in_proj_covar=tensor([0.0145, 0.0105, 0.0092, 0.0139, 0.0074, 0.0118, 0.0125, 0.0129], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:1') 2023-04-30 03:13:06,355 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=145027.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 03:13:38,256 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.87 vs. limit=2.0 2023-04-30 03:13:40,919 INFO [train.py:904] (1/8) Epoch 15, batch 2950, loss[loss=0.2343, simple_loss=0.301, pruned_loss=0.08384, over 15361.00 frames. ], tot_loss[loss=0.1779, simple_loss=0.2632, pruned_loss=0.04636, over 3333349.20 frames. ], batch size: 190, lr: 4.57e-03, grad_scale: 8.0 2023-04-30 03:13:47,661 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=145057.0, num_to_drop=1, layers_to_drop={1} 2023-04-30 03:14:01,011 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=145066.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 03:14:02,157 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=145067.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 03:14:31,822 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=145088.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 03:14:49,683 INFO [train.py:904] (1/8) Epoch 15, batch 3000, loss[loss=0.2031, simple_loss=0.2807, pruned_loss=0.06279, over 15454.00 frames. ], tot_loss[loss=0.1792, simple_loss=0.2642, pruned_loss=0.0471, over 3332114.99 frames. ], batch size: 190, lr: 4.57e-03, grad_scale: 8.0 2023-04-30 03:14:49,683 INFO [train.py:929] (1/8) Computing validation loss 2023-04-30 03:14:56,096 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.3687, 3.0569, 3.3424, 5.3196, 3.1471, 3.4379, 3.2054, 3.3321], device='cuda:1'), covar=tensor([0.0801, 0.2988, 0.2111, 0.0201, 0.3003, 0.1860, 0.2835, 0.2422], device='cuda:1'), in_proj_covar=tensor([0.0381, 0.0418, 0.0348, 0.0326, 0.0423, 0.0480, 0.0381, 0.0489], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-30 03:14:56,335 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([5.0920, 5.0371, 4.9277, 4.4350, 4.5460, 4.9496, 5.2100, 4.6546], device='cuda:1'), covar=tensor([0.0502, 0.0415, 0.0292, 0.0361, 0.1034, 0.0424, 0.0156, 0.0817], device='cuda:1'), in_proj_covar=tensor([0.0280, 0.0385, 0.0332, 0.0317, 0.0347, 0.0363, 0.0225, 0.0393], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-04-30 03:14:58,803 INFO [train.py:938] (1/8) Epoch 15, validation: loss=0.138, simple_loss=0.2438, pruned_loss=0.01616, over 944034.00 frames. 2023-04-30 03:14:58,804 INFO [train.py:939] (1/8) Maximum memory allocated so far is 17857MB 2023-04-30 03:15:00,800 INFO [optim.py:368] (1/8) 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 03:15:16,809 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=145115.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 03:15:34,538 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=145127.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 03:15:55,221 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.70 vs. limit=2.0 2023-04-30 03:16:07,556 INFO [train.py:904] (1/8) Epoch 15, batch 3050, loss[loss=0.174, simple_loss=0.2574, pruned_loss=0.04529, over 16500.00 frames. ], tot_loss[loss=0.1801, simple_loss=0.2649, pruned_loss=0.04767, over 3321738.54 frames. ], batch size: 68, lr: 4.57e-03, grad_scale: 8.0 2023-04-30 03:16:07,920 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.8741, 5.2287, 4.9882, 4.9939, 4.7554, 4.6372, 4.6373, 5.2841], device='cuda:1'), covar=tensor([0.1121, 0.0811, 0.0970, 0.0700, 0.0783, 0.0984, 0.1233, 0.0768], device='cuda:1'), in_proj_covar=tensor([0.0630, 0.0781, 0.0642, 0.0560, 0.0497, 0.0498, 0.0651, 0.0595], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-30 03:17:18,167 INFO [train.py:904] (1/8) Epoch 15, batch 3100, loss[loss=0.1761, simple_loss=0.2665, pruned_loss=0.04281, over 17107.00 frames. ], tot_loss[loss=0.1801, simple_loss=0.2645, pruned_loss=0.04785, over 3309774.73 frames. ], batch size: 47, lr: 4.57e-03, grad_scale: 8.0 2023-04-30 03:17:19,336 INFO [optim.py:368] (1/8) Clipping_scale=2.0, grad-norm quartiles 1.571e+02 2.445e+02 2.806e+02 3.389e+02 5.168e+02, threshold=5.611e+02, percent-clipped=0.0 2023-04-30 03:17:58,988 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.0420, 4.7772, 4.5944, 3.5405, 4.0822, 4.7326, 4.1804, 3.0757], device='cuda:1'), covar=tensor([0.0388, 0.0043, 0.0037, 0.0265, 0.0075, 0.0062, 0.0061, 0.0327], device='cuda:1'), in_proj_covar=tensor([0.0133, 0.0077, 0.0076, 0.0132, 0.0089, 0.0101, 0.0087, 0.0127], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-04-30 03:18:19,350 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=3.55 vs. limit=5.0 2023-04-30 03:18:28,439 INFO [train.py:904] (1/8) Epoch 15, batch 3150, loss[loss=0.1649, simple_loss=0.2652, pruned_loss=0.03232, over 17108.00 frames. ], tot_loss[loss=0.18, simple_loss=0.2639, pruned_loss=0.04806, over 3311894.95 frames. ], batch size: 49, lr: 4.57e-03, grad_scale: 8.0 2023-04-30 03:18:40,544 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.76 vs. limit=2.0 2023-04-30 03:19:37,246 INFO [train.py:904] (1/8) Epoch 15, batch 3200, loss[loss=0.1455, simple_loss=0.228, pruned_loss=0.03151, over 16801.00 frames. ], tot_loss[loss=0.1778, simple_loss=0.262, pruned_loss=0.04685, over 3309784.24 frames. ], batch size: 39, lr: 4.57e-03, grad_scale: 8.0 2023-04-30 03:19:38,468 INFO [optim.py:368] (1/8) Clipping_scale=2.0, grad-norm quartiles 1.650e+02 2.241e+02 2.735e+02 3.234e+02 5.514e+02, threshold=5.469e+02, percent-clipped=0.0 2023-04-30 03:19:49,681 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=4.67 vs. limit=5.0 2023-04-30 03:20:46,509 INFO [train.py:904] (1/8) Epoch 15, batch 3250, loss[loss=0.1625, simple_loss=0.2482, pruned_loss=0.03842, over 17020.00 frames. ], tot_loss[loss=0.1775, simple_loss=0.2615, pruned_loss=0.0468, over 3314431.18 frames. ], batch size: 41, lr: 4.57e-03, grad_scale: 8.0 2023-04-30 03:20:46,744 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=145352.0, num_to_drop=1, layers_to_drop={2} 2023-04-30 03:21:13,339 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=145371.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 03:21:25,709 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.9730, 4.5436, 3.3437, 2.4080, 2.9127, 2.6631, 4.8757, 3.8821], device='cuda:1'), covar=tensor([0.2571, 0.0582, 0.1530, 0.2669, 0.2893, 0.1895, 0.0283, 0.1292], device='cuda:1'), in_proj_covar=tensor([0.0314, 0.0267, 0.0297, 0.0296, 0.0291, 0.0240, 0.0281, 0.0321], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:1') 2023-04-30 03:21:30,554 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=145383.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 03:21:57,380 INFO [train.py:904] (1/8) Epoch 15, batch 3300, loss[loss=0.1673, simple_loss=0.2605, pruned_loss=0.03705, over 17143.00 frames. ], tot_loss[loss=0.1779, simple_loss=0.2623, pruned_loss=0.04679, over 3319173.04 frames. ], batch size: 48, lr: 4.56e-03, grad_scale: 8.0 2023-04-30 03:21:58,627 INFO [optim.py:368] (1/8) 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] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=145422.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 03:22:38,834 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=145432.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 03:23:06,152 INFO [train.py:904] (1/8) Epoch 15, batch 3350, loss[loss=0.1507, simple_loss=0.2297, pruned_loss=0.03588, over 16989.00 frames. ], tot_loss[loss=0.1776, simple_loss=0.2624, pruned_loss=0.04642, over 3324152.09 frames. ], batch size: 41, lr: 4.56e-03, grad_scale: 8.0 2023-04-30 03:24:17,515 INFO [train.py:904] (1/8) Epoch 15, batch 3400, loss[loss=0.178, simple_loss=0.2785, pruned_loss=0.03872, over 16774.00 frames. ], tot_loss[loss=0.178, simple_loss=0.2627, pruned_loss=0.04661, over 3323471.19 frames. ], batch size: 57, lr: 4.56e-03, grad_scale: 8.0 2023-04-30 03:24:18,607 INFO [optim.py:368] (1/8) 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 03:25:28,529 INFO [train.py:904] (1/8) Epoch 15, batch 3450, loss[loss=0.1835, simple_loss=0.2831, pruned_loss=0.04201, over 17064.00 frames. ], tot_loss[loss=0.1773, simple_loss=0.2617, pruned_loss=0.04646, over 3318656.41 frames. ], batch size: 50, lr: 4.56e-03, grad_scale: 8.0 2023-04-30 03:25:58,310 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.43 vs. limit=2.0 2023-04-30 03:26:25,325 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([5.2627, 5.3535, 5.1242, 4.6792, 4.4724, 5.2503, 5.2061, 4.7775], device='cuda:1'), covar=tensor([0.0720, 0.0499, 0.0358, 0.0380, 0.1452, 0.0480, 0.0257, 0.0811], device='cuda:1'), in_proj_covar=tensor([0.0283, 0.0388, 0.0334, 0.0319, 0.0352, 0.0363, 0.0226, 0.0397], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-04-30 03:26:27,199 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=145594.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 03:26:38,101 INFO [train.py:904] (1/8) Epoch 15, batch 3500, loss[loss=0.1503, simple_loss=0.2413, pruned_loss=0.02962, over 17244.00 frames. ], tot_loss[loss=0.1772, simple_loss=0.2615, pruned_loss=0.04647, over 3318411.50 frames. ], batch size: 45, lr: 4.56e-03, grad_scale: 8.0 2023-04-30 03:26:39,230 INFO [optim.py:368] (1/8) 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 03:26:40,201 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.26 vs. limit=2.0 2023-04-30 03:27:36,043 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 15, batch 3550, loss[loss=0.1591, simple_loss=0.2463, pruned_loss=0.03589, over 16860.00 frames. ], tot_loss[loss=0.1758, simple_loss=0.2602, pruned_loss=0.04573, over 3313923.74 frames. ], batch size: 42, lr: 4.56e-03, grad_scale: 8.0 2023-04-30 03:27:48,260 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=145652.0, num_to_drop=1, layers_to_drop={2} 2023-04-30 03:27:52,703 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=145655.0, num_to_drop=1, layers_to_drop={1} 2023-04-30 03:27:55,957 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.8832, 2.0606, 2.3550, 3.1638, 2.1712, 2.2349, 2.3406, 2.1927], device='cuda:1'), covar=tensor([0.1173, 0.3126, 0.2204, 0.0617, 0.3654, 0.2294, 0.2599, 0.3344], device='cuda:1'), in_proj_covar=tensor([0.0383, 0.0420, 0.0350, 0.0328, 0.0425, 0.0484, 0.0384, 0.0492], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-30 03:28:08,404 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.1117, 4.6692, 4.5007, 3.5890, 4.0056, 4.5921, 4.2162, 2.8025], device='cuda:1'), covar=tensor([0.0357, 0.0036, 0.0037, 0.0233, 0.0076, 0.0083, 0.0056, 0.0361], device='cuda:1'), in_proj_covar=tensor([0.0130, 0.0075, 0.0074, 0.0129, 0.0087, 0.0099, 0.0085, 0.0124], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-04-30 03:28:31,750 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=145683.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 03:28:54,823 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=145700.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 03:28:57,586 INFO [train.py:904] (1/8) Epoch 15, batch 3600, loss[loss=0.2052, simple_loss=0.2771, pruned_loss=0.06667, over 16432.00 frames. ], tot_loss[loss=0.1754, simple_loss=0.2591, pruned_loss=0.04588, over 3303304.04 frames. ], batch size: 165, lr: 4.56e-03, grad_scale: 8.0 2023-04-30 03:28:58,725 INFO [optim.py:368] (1/8) 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 03:29:00,396 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=145704.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 03:29:26,943 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=145722.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 03:29:34,303 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=145727.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 03:29:39,795 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=145731.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 03:30:10,698 INFO [train.py:904] (1/8) Epoch 15, batch 3650, loss[loss=0.1655, simple_loss=0.2377, pruned_loss=0.04665, over 16798.00 frames. ], tot_loss[loss=0.175, simple_loss=0.2578, pruned_loss=0.04608, over 3302270.39 frames. ], batch size: 124, lr: 4.56e-03, grad_scale: 8.0 2023-04-30 03:30:38,217 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=145770.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 03:31:24,496 INFO [train.py:904] (1/8) Epoch 15, batch 3700, loss[loss=0.1796, simple_loss=0.2532, pruned_loss=0.05303, over 15433.00 frames. ], tot_loss[loss=0.1753, simple_loss=0.2564, pruned_loss=0.04715, over 3291940.45 frames. ], batch size: 191, lr: 4.56e-03, grad_scale: 8.0 2023-04-30 03:31:26,279 INFO [optim.py:368] (1/8) Clipping_scale=2.0, grad-norm quartiles 1.472e+02 2.174e+02 2.701e+02 3.170e+02 5.249e+02, threshold=5.403e+02, percent-clipped=2.0 2023-04-30 03:31:53,379 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=145821.0, num_to_drop=1, layers_to_drop={0} 2023-04-30 03:31:55,268 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([1.6702, 1.7863, 1.6003, 1.5682, 1.9322, 1.5994, 1.6719, 1.9610], device='cuda:1'), covar=tensor([0.0162, 0.0237, 0.0327, 0.0309, 0.0178, 0.0216, 0.0148, 0.0164], device='cuda:1'), in_proj_covar=tensor([0.0183, 0.0222, 0.0213, 0.0216, 0.0224, 0.0223, 0.0232, 0.0218], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-30 03:32:38,908 INFO [train.py:904] (1/8) Epoch 15, batch 3750, loss[loss=0.1994, simple_loss=0.2603, pruned_loss=0.06927, over 16839.00 frames. ], tot_loss[loss=0.1769, simple_loss=0.2569, pruned_loss=0.04844, over 3279123.13 frames. ], batch size: 116, lr: 4.56e-03, grad_scale: 8.0 2023-04-30 03:33:23,431 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=145882.0, num_to_drop=1, layers_to_drop={2} 2023-04-30 03:33:51,795 INFO [train.py:904] (1/8) Epoch 15, batch 3800, loss[loss=0.1757, simple_loss=0.257, pruned_loss=0.04718, over 16663.00 frames. ], tot_loss[loss=0.1791, simple_loss=0.2582, pruned_loss=0.05, over 3280243.35 frames. ], batch size: 57, lr: 4.56e-03, grad_scale: 8.0 2023-04-30 03:33:53,678 INFO [optim.py:368] (1/8) 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 03:34:22,137 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.4158, 3.7388, 3.8813, 2.8079, 3.5522, 3.9890, 3.6652, 2.2035], device='cuda:1'), covar=tensor([0.0434, 0.0107, 0.0044, 0.0304, 0.0082, 0.0088, 0.0084, 0.0405], device='cuda:1'), in_proj_covar=tensor([0.0130, 0.0075, 0.0074, 0.0129, 0.0088, 0.0099, 0.0086, 0.0125], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-04-30 03:35:00,906 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=145950.0, num_to_drop=1, layers_to_drop={2} 2023-04-30 03:35:04,375 INFO [train.py:904] (1/8) Epoch 15, batch 3850, loss[loss=0.1913, simple_loss=0.2701, pruned_loss=0.05623, over 15542.00 frames. ], tot_loss[loss=0.1795, simple_loss=0.2583, pruned_loss=0.05037, over 3273540.61 frames. ], batch size: 191, lr: 4.56e-03, grad_scale: 16.0 2023-04-30 03:36:13,424 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=145999.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 03:36:21,018 INFO [train.py:904] (1/8) Epoch 15, batch 3900, loss[loss=0.1766, simple_loss=0.2527, pruned_loss=0.0502, over 16676.00 frames. ], tot_loss[loss=0.1803, simple_loss=0.2587, pruned_loss=0.05097, over 3273530.02 frames. ], batch size: 134, lr: 4.56e-03, grad_scale: 16.0 2023-04-30 03:36:22,204 INFO [optim.py:368] (1/8) 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 03:36:57,961 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=146027.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 03:37:32,700 INFO [train.py:904] (1/8) Epoch 15, batch 3950, loss[loss=0.1764, simple_loss=0.2579, pruned_loss=0.04748, over 16726.00 frames. ], tot_loss[loss=0.1808, simple_loss=0.2585, pruned_loss=0.05158, over 3272865.75 frames. ], batch size: 39, lr: 4.55e-03, grad_scale: 16.0 2023-04-30 03:38:06,890 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=146075.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 03:38:32,470 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.0786, 3.2610, 3.2834, 2.0603, 2.8189, 2.3444, 3.5947, 3.5779], device='cuda:1'), covar=tensor([0.0229, 0.0793, 0.0604, 0.1738, 0.0806, 0.0912, 0.0461, 0.0713], device='cuda:1'), in_proj_covar=tensor([0.0150, 0.0156, 0.0161, 0.0147, 0.0139, 0.0126, 0.0139, 0.0167], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:1') 2023-04-30 03:38:46,133 INFO [train.py:904] (1/8) Epoch 15, batch 4000, loss[loss=0.182, simple_loss=0.2584, pruned_loss=0.05278, over 16928.00 frames. ], tot_loss[loss=0.1814, simple_loss=0.2585, pruned_loss=0.05218, over 3268902.86 frames. ], batch size: 109, lr: 4.55e-03, grad_scale: 16.0 2023-04-30 03:38:47,409 INFO [optim.py:368] (1/8) 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 03:39:22,092 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=4.92 vs. limit=5.0 2023-04-30 03:39:56,184 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=146149.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 03:39:57,463 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.2811, 2.3163, 2.7088, 3.2394, 3.1215, 3.7945, 2.4217, 3.6106], device='cuda:1'), covar=tensor([0.0154, 0.0394, 0.0269, 0.0231, 0.0228, 0.0091, 0.0398, 0.0079], device='cuda:1'), in_proj_covar=tensor([0.0175, 0.0183, 0.0169, 0.0173, 0.0182, 0.0139, 0.0184, 0.0131], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-30 03:39:59,983 INFO [train.py:904] (1/8) Epoch 15, batch 4050, loss[loss=0.1798, simple_loss=0.2621, pruned_loss=0.04878, over 16552.00 frames. ], tot_loss[loss=0.1806, simple_loss=0.2591, pruned_loss=0.05104, over 3263595.36 frames. ], batch size: 62, lr: 4.55e-03, grad_scale: 16.0 2023-04-30 03:40:26,543 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.8980, 4.0440, 4.2885, 4.2608, 4.2688, 4.0054, 4.0244, 3.9486], device='cuda:1'), covar=tensor([0.0335, 0.0523, 0.0396, 0.0417, 0.0463, 0.0411, 0.0856, 0.0534], device='cuda:1'), in_proj_covar=tensor([0.0376, 0.0406, 0.0397, 0.0379, 0.0447, 0.0420, 0.0515, 0.0336], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:1') 2023-04-30 03:40:37,002 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=146177.0, num_to_drop=1, layers_to_drop={2} 2023-04-30 03:41:13,945 INFO [train.py:904] (1/8) Epoch 15, batch 4100, loss[loss=0.1922, simple_loss=0.2752, pruned_loss=0.05464, over 16428.00 frames. ], tot_loss[loss=0.1794, simple_loss=0.2594, pruned_loss=0.04969, over 3271923.16 frames. ], batch size: 146, lr: 4.55e-03, grad_scale: 16.0 2023-04-30 03:41:15,745 INFO [optim.py:368] (1/8) 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 03:41:24,233 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.9290, 3.3525, 3.3921, 1.9087, 2.8751, 2.2716, 3.4072, 3.5456], device='cuda:1'), covar=tensor([0.0255, 0.0727, 0.0545, 0.1966, 0.0871, 0.0933, 0.0617, 0.0888], device='cuda:1'), in_proj_covar=tensor([0.0150, 0.0156, 0.0161, 0.0147, 0.0139, 0.0127, 0.0139, 0.0167], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:1') 2023-04-30 03:41:26,660 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=146210.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 03:41:37,545 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.0282, 2.9194, 2.2656, 2.7605, 3.3191, 2.8592, 3.6137, 3.6125], device='cuda:1'), covar=tensor([0.0050, 0.0336, 0.0513, 0.0380, 0.0216, 0.0325, 0.0174, 0.0172], device='cuda:1'), in_proj_covar=tensor([0.0181, 0.0221, 0.0213, 0.0215, 0.0222, 0.0220, 0.0230, 0.0216], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-30 03:42:30,105 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=146250.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 03:42:32,939 INFO [train.py:904] (1/8) Epoch 15, batch 4150, loss[loss=0.2082, simple_loss=0.2948, pruned_loss=0.06081, over 16868.00 frames. ], tot_loss[loss=0.1855, simple_loss=0.2667, pruned_loss=0.05208, over 3253289.86 frames. ], batch size: 96, lr: 4.55e-03, grad_scale: 8.0 2023-04-30 03:43:45,065 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=146298.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 03:43:46,710 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=146299.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 03:43:50,411 INFO [train.py:904] (1/8) Epoch 15, batch 4200, loss[loss=0.2043, simple_loss=0.2949, pruned_loss=0.05679, over 16677.00 frames. ], tot_loss[loss=0.1904, simple_loss=0.2735, pruned_loss=0.05359, over 3242074.31 frames. ], batch size: 134, lr: 4.55e-03, grad_scale: 8.0 2023-04-30 03:43:53,464 INFO [optim.py:368] (1/8) 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 03:44:21,164 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.1506, 3.4113, 3.5104, 3.5094, 3.5101, 3.3680, 3.3797, 3.3885], device='cuda:1'), covar=tensor([0.0392, 0.0612, 0.0490, 0.0441, 0.0561, 0.0525, 0.0814, 0.0565], device='cuda:1'), in_proj_covar=tensor([0.0367, 0.0396, 0.0390, 0.0369, 0.0437, 0.0409, 0.0504, 0.0328], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:1') 2023-04-30 03:44:58,658 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 15, batch 4250, loss[loss=0.1942, simple_loss=0.2813, pruned_loss=0.05359, over 16655.00 frames. ], tot_loss[loss=0.192, simple_loss=0.277, pruned_loss=0.05344, over 3234864.70 frames. ], batch size: 62, lr: 4.55e-03, grad_scale: 4.0 2023-04-30 03:45:38,835 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.5618, 4.0043, 3.5376, 3.9346, 3.5430, 3.6060, 3.5332, 3.9516], device='cuda:1'), covar=tensor([0.3125, 0.1890, 0.3255, 0.1491, 0.2123, 0.3305, 0.2921, 0.2180], device='cuda:1'), in_proj_covar=tensor([0.0610, 0.0761, 0.0627, 0.0548, 0.0479, 0.0488, 0.0635, 0.0578], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:1') 2023-04-30 03:46:19,358 INFO [train.py:904] (1/8) Epoch 15, batch 4300, loss[loss=0.1936, simple_loss=0.2868, pruned_loss=0.05015, over 16683.00 frames. ], tot_loss[loss=0.1911, simple_loss=0.2779, pruned_loss=0.05217, over 3234083.48 frames. ], batch size: 134, lr: 4.55e-03, grad_scale: 4.0 2023-04-30 03:46:23,348 INFO [optim.py:368] (1/8) 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 03:46:40,136 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=4.50 vs. limit=5.0 2023-04-30 03:46:45,295 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.2317, 3.9973, 3.8791, 2.5225, 3.5656, 3.9889, 3.6323, 2.0788], device='cuda:1'), covar=tensor([0.0475, 0.0027, 0.0038, 0.0347, 0.0075, 0.0066, 0.0072, 0.0392], device='cuda:1'), in_proj_covar=tensor([0.0130, 0.0074, 0.0074, 0.0130, 0.0088, 0.0098, 0.0086, 0.0123], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-04-30 03:46:47,698 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.7464, 4.7533, 4.5592, 3.9376, 4.7033, 1.6642, 4.4518, 4.3098], device='cuda:1'), covar=tensor([0.0069, 0.0055, 0.0152, 0.0298, 0.0068, 0.2792, 0.0096, 0.0204], device='cuda:1'), in_proj_covar=tensor([0.0149, 0.0137, 0.0187, 0.0173, 0.0158, 0.0197, 0.0175, 0.0168], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-30 03:47:08,217 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.8620, 4.8691, 4.7052, 3.7751, 4.8123, 1.6336, 4.5092, 4.3814], device='cuda:1'), covar=tensor([0.0088, 0.0062, 0.0150, 0.0428, 0.0081, 0.3082, 0.0109, 0.0258], device='cuda:1'), in_proj_covar=tensor([0.0149, 0.0137, 0.0187, 0.0173, 0.0158, 0.0197, 0.0175, 0.0168], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-30 03:47:31,145 INFO [train.py:904] (1/8) Epoch 15, batch 4350, loss[loss=0.197, simple_loss=0.2824, pruned_loss=0.05583, over 17125.00 frames. ], tot_loss[loss=0.1947, simple_loss=0.2817, pruned_loss=0.05384, over 3210596.42 frames. ], batch size: 47, lr: 4.55e-03, grad_scale: 4.0 2023-04-30 03:48:08,892 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=146477.0, num_to_drop=1, layers_to_drop={2} 2023-04-30 03:48:45,757 INFO [train.py:904] (1/8) Epoch 15, batch 4400, loss[loss=0.1964, simple_loss=0.2811, pruned_loss=0.05581, over 16725.00 frames. ], tot_loss[loss=0.197, simple_loss=0.2836, pruned_loss=0.05518, over 3177107.29 frames. ], batch size: 134, lr: 4.55e-03, grad_scale: 8.0 2023-04-30 03:48:50,397 INFO [optim.py:368] (1/8) 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 03:48:50,743 INFO [zipformer.py:625] (1/8) 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] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=146525.0, num_to_drop=1, layers_to_drop={0} 2023-04-30 03:49:50,042 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.6794, 2.6615, 1.8078, 2.7453, 2.1506, 2.7952, 2.0951, 2.3472], device='cuda:1'), covar=tensor([0.0261, 0.0326, 0.1312, 0.0165, 0.0688, 0.0417, 0.1149, 0.0619], device='cuda:1'), in_proj_covar=tensor([0.0162, 0.0172, 0.0191, 0.0146, 0.0168, 0.0212, 0.0198, 0.0173], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0003, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-04-30 03:49:58,633 INFO [train.py:904] (1/8) Epoch 15, batch 4450, loss[loss=0.2127, simple_loss=0.3029, pruned_loss=0.06125, over 17002.00 frames. ], tot_loss[loss=0.2001, simple_loss=0.2871, pruned_loss=0.05653, over 3185154.17 frames. ], batch size: 41, lr: 4.55e-03, grad_scale: 8.0 2023-04-30 03:50:57,074 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 15, batch 4500, loss[loss=0.2081, simple_loss=0.2913, pruned_loss=0.06243, over 15471.00 frames. ], tot_loss[loss=0.2011, simple_loss=0.2877, pruned_loss=0.05724, over 3195477.40 frames. ], batch size: 191, lr: 4.55e-03, grad_scale: 8.0 2023-04-30 03:51:16,066 INFO [optim.py:368] (1/8) 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 03:51:48,040 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 15, batch 4550, loss[loss=0.2125, simple_loss=0.2978, pruned_loss=0.06365, over 16752.00 frames. ], tot_loss[loss=0.203, simple_loss=0.2891, pruned_loss=0.05841, over 3211149.95 frames. ], batch size: 83, lr: 4.55e-03, grad_scale: 8.0 2023-04-30 03:52:25,892 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=146652.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 03:52:45,636 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.7008, 4.5412, 4.7697, 4.9283, 5.0640, 4.5264, 5.0230, 5.0889], device='cuda:1'), covar=tensor([0.1435, 0.1014, 0.1236, 0.0507, 0.0388, 0.0836, 0.0458, 0.0440], device='cuda:1'), in_proj_covar=tensor([0.0565, 0.0702, 0.0843, 0.0718, 0.0536, 0.0564, 0.0569, 0.0662], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-30 03:53:15,982 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=146687.0, num_to_drop=1, layers_to_drop={2} 2023-04-30 03:53:37,390 INFO [train.py:904] (1/8) Epoch 15, batch 4600, loss[loss=0.1887, simple_loss=0.2787, pruned_loss=0.0493, over 16187.00 frames. ], tot_loss[loss=0.2034, simple_loss=0.2898, pruned_loss=0.05853, over 3225289.80 frames. ], batch size: 165, lr: 4.54e-03, grad_scale: 8.0 2023-04-30 03:53:41,728 INFO [optim.py:368] (1/8) 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 03:54:20,859 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.6953, 3.7450, 4.2189, 1.9816, 4.4333, 4.4643, 3.1566, 3.2344], device='cuda:1'), covar=tensor([0.0740, 0.0247, 0.0140, 0.1140, 0.0055, 0.0083, 0.0413, 0.0445], device='cuda:1'), in_proj_covar=tensor([0.0145, 0.0105, 0.0091, 0.0137, 0.0073, 0.0115, 0.0122, 0.0129], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:1') 2023-04-30 03:54:49,172 INFO [train.py:904] (1/8) Epoch 15, batch 4650, loss[loss=0.2203, simple_loss=0.3151, pruned_loss=0.06277, over 15278.00 frames. ], tot_loss[loss=0.2025, simple_loss=0.2885, pruned_loss=0.0582, over 3228383.24 frames. ], batch size: 190, lr: 4.54e-03, grad_scale: 8.0 2023-04-30 03:55:38,088 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=4.83 vs. limit=5.0 2023-04-30 03:56:03,159 INFO [train.py:904] (1/8) Epoch 15, batch 4700, loss[loss=0.1839, simple_loss=0.2758, pruned_loss=0.04604, over 16686.00 frames. ], tot_loss[loss=0.2001, simple_loss=0.2859, pruned_loss=0.05711, over 3217512.82 frames. ], batch size: 134, lr: 4.54e-03, grad_scale: 8.0 2023-04-30 03:56:07,886 INFO [optim.py:368] (1/8) 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 03:56:08,256 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=146805.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 03:57:07,591 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=146844.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 03:57:11,625 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.8121, 5.0740, 5.2353, 4.9880, 5.0297, 5.6380, 5.0881, 4.8016], device='cuda:1'), covar=tensor([0.0880, 0.1674, 0.1845, 0.1711, 0.2358, 0.0908, 0.1319, 0.2231], device='cuda:1'), in_proj_covar=tensor([0.0378, 0.0531, 0.0579, 0.0453, 0.0602, 0.0609, 0.0459, 0.0607], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-04-30 03:57:18,584 INFO [train.py:904] (1/8) Epoch 15, batch 4750, loss[loss=0.1717, simple_loss=0.2596, pruned_loss=0.0419, over 16738.00 frames. ], tot_loss[loss=0.1956, simple_loss=0.2815, pruned_loss=0.05484, over 3224341.29 frames. ], batch size: 76, lr: 4.54e-03, grad_scale: 8.0 2023-04-30 03:57:20,527 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=146853.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 03:57:27,367 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.5576, 4.5956, 4.4095, 4.1030, 4.0340, 4.5004, 4.2688, 4.1895], device='cuda:1'), covar=tensor([0.0544, 0.0526, 0.0278, 0.0279, 0.0975, 0.0519, 0.0530, 0.0567], device='cuda:1'), in_proj_covar=tensor([0.0262, 0.0360, 0.0313, 0.0294, 0.0326, 0.0339, 0.0210, 0.0368], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-30 03:57:37,448 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.5925, 3.8349, 4.2684, 1.8470, 4.4275, 4.4675, 3.0969, 3.2254], device='cuda:1'), covar=tensor([0.0830, 0.0220, 0.0173, 0.1290, 0.0048, 0.0078, 0.0379, 0.0452], device='cuda:1'), in_proj_covar=tensor([0.0144, 0.0104, 0.0091, 0.0137, 0.0072, 0.0115, 0.0122, 0.0128], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:1') 2023-04-30 03:57:47,158 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.23 vs. limit=2.0 2023-04-30 03:58:31,078 INFO [train.py:904] (1/8) Epoch 15, batch 4800, loss[loss=0.1824, simple_loss=0.2686, pruned_loss=0.04813, over 16480.00 frames. ], tot_loss[loss=0.1918, simple_loss=0.2777, pruned_loss=0.05295, over 3210213.55 frames. ], batch size: 68, lr: 4.54e-03, grad_scale: 8.0 2023-04-30 03:58:36,177 INFO [optim.py:368] (1/8) 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 03:58:36,846 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=146905.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 03:59:40,649 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 15, batch 4850, loss[loss=0.2356, simple_loss=0.3193, pruned_loss=0.07591, over 12052.00 frames. ], tot_loss[loss=0.1924, simple_loss=0.2792, pruned_loss=0.05277, over 3184847.36 frames. ], batch size: 246, lr: 4.54e-03, grad_scale: 8.0 2023-04-30 04:00:34,143 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=146982.0, num_to_drop=1, layers_to_drop={1} 2023-04-30 04:01:03,665 INFO [train.py:904] (1/8) Epoch 15, batch 4900, loss[loss=0.2057, simple_loss=0.2981, pruned_loss=0.05668, over 12434.00 frames. ], tot_loss[loss=0.1903, simple_loss=0.2781, pruned_loss=0.05124, over 3169236.79 frames. ], batch size: 246, lr: 4.54e-03, grad_scale: 8.0 2023-04-30 04:01:08,002 INFO [optim.py:368] (1/8) 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 04:02:12,002 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.43 vs. limit=2.0 2023-04-30 04:02:16,313 INFO [train.py:904] (1/8) Epoch 15, batch 4950, loss[loss=0.1961, simple_loss=0.2851, pruned_loss=0.05353, over 17141.00 frames. ], tot_loss[loss=0.19, simple_loss=0.2781, pruned_loss=0.05094, over 3175703.55 frames. ], batch size: 49, lr: 4.54e-03, grad_scale: 8.0 2023-04-30 04:03:28,736 INFO [train.py:904] (1/8) Epoch 15, batch 5000, loss[loss=0.1779, simple_loss=0.2775, pruned_loss=0.03913, over 16721.00 frames. ], tot_loss[loss=0.1912, simple_loss=0.2802, pruned_loss=0.05114, over 3181659.14 frames. ], batch size: 134, lr: 4.54e-03, grad_scale: 8.0 2023-04-30 04:03:31,911 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=4.96 vs. limit=5.0 2023-04-30 04:03:32,271 INFO [optim.py:368] (1/8) Clipping_scale=2.0, grad-norm quartiles 1.607e+02 2.177e+02 2.680e+02 3.002e+02 5.827e+02, threshold=5.360e+02, percent-clipped=3.0 2023-04-30 04:04:39,315 INFO [train.py:904] (1/8) Epoch 15, batch 5050, loss[loss=0.1639, simple_loss=0.2595, pruned_loss=0.03417, over 16431.00 frames. ], tot_loss[loss=0.1914, simple_loss=0.2807, pruned_loss=0.0511, over 3185795.45 frames. ], batch size: 68, lr: 4.54e-03, grad_scale: 8.0 2023-04-30 04:05:46,844 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 15, batch 5100, loss[loss=0.1831, simple_loss=0.2747, pruned_loss=0.04578, over 15327.00 frames. ], tot_loss[loss=0.19, simple_loss=0.2792, pruned_loss=0.05047, over 3182132.04 frames. ], batch size: 190, lr: 4.54e-03, grad_scale: 8.0 2023-04-30 04:05:52,960 INFO [optim.py:368] (1/8) 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 04:06:28,947 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.28 vs. limit=2.0 2023-04-30 04:06:45,595 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=147241.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 04:06:52,612 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=147247.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 04:06:59,912 INFO [train.py:904] (1/8) Epoch 15, batch 5150, loss[loss=0.2026, simple_loss=0.307, pruned_loss=0.04904, over 15389.00 frames. ], tot_loss[loss=0.189, simple_loss=0.2787, pruned_loss=0.04967, over 3186207.34 frames. ], batch size: 191, lr: 4.54e-03, grad_scale: 8.0 2023-04-30 04:07:09,857 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.3259, 2.1711, 2.2785, 4.0299, 2.0752, 2.5699, 2.2956, 2.4035], device='cuda:1'), covar=tensor([0.1094, 0.3507, 0.2578, 0.0398, 0.3853, 0.2443, 0.3448, 0.2753], device='cuda:1'), in_proj_covar=tensor([0.0375, 0.0414, 0.0343, 0.0318, 0.0418, 0.0475, 0.0379, 0.0482], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-30 04:07:28,270 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.7546, 3.7656, 2.2180, 4.3844, 2.7620, 4.2111, 2.3459, 2.8626], device='cuda:1'), covar=tensor([0.0228, 0.0357, 0.1635, 0.0138, 0.0867, 0.0486, 0.1501, 0.0772], device='cuda:1'), in_proj_covar=tensor([0.0159, 0.0169, 0.0189, 0.0142, 0.0167, 0.0208, 0.0196, 0.0172], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0003, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-04-30 04:07:40,359 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=4.82 vs. limit=5.0 2023-04-30 04:07:45,019 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=147282.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 04:07:59,967 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.69 vs. limit=2.0 2023-04-30 04:08:03,058 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 15, batch 5200, loss[loss=0.2, simple_loss=0.289, pruned_loss=0.05551, over 16727.00 frames. ], tot_loss[loss=0.1872, simple_loss=0.2767, pruned_loss=0.04885, over 3199308.77 frames. ], batch size: 134, lr: 4.54e-03, grad_scale: 8.0 2023-04-30 04:08:14,608 INFO [zipformer.py:625] (1/8) 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] (1/8) 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 04:08:25,213 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=147310.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 04:08:55,068 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=147330.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 04:09:27,035 INFO [train.py:904] (1/8) Epoch 15, batch 5250, loss[loss=0.1897, simple_loss=0.2766, pruned_loss=0.05138, over 16531.00 frames. ], tot_loss[loss=0.1856, simple_loss=0.2744, pruned_loss=0.04845, over 3210723.30 frames. ], batch size: 68, lr: 4.53e-03, grad_scale: 8.0 2023-04-30 04:09:54,282 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=147371.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 04:10:05,969 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.40 vs. limit=2.0 2023-04-30 04:10:37,526 INFO [train.py:904] (1/8) Epoch 15, batch 5300, loss[loss=0.1742, simple_loss=0.2497, pruned_loss=0.04932, over 16660.00 frames. ], tot_loss[loss=0.1831, simple_loss=0.2709, pruned_loss=0.0476, over 3210939.24 frames. ], batch size: 57, lr: 4.53e-03, grad_scale: 8.0 2023-04-30 04:10:40,967 INFO [optim.py:368] (1/8) 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 04:10:44,614 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=147407.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 04:11:00,917 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.2278, 3.4025, 3.6170, 1.6265, 3.7373, 3.8891, 2.9922, 2.9002], device='cuda:1'), covar=tensor([0.0984, 0.0197, 0.0170, 0.1475, 0.0083, 0.0102, 0.0390, 0.0495], device='cuda:1'), in_proj_covar=tensor([0.0145, 0.0104, 0.0091, 0.0137, 0.0073, 0.0113, 0.0122, 0.0128], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:1') 2023-04-30 04:11:23,868 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=147434.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 04:11:46,768 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=3.32 vs. limit=5.0 2023-04-30 04:11:49,962 INFO [train.py:904] (1/8) Epoch 15, batch 5350, loss[loss=0.2093, simple_loss=0.2934, pruned_loss=0.06259, over 15304.00 frames. ], tot_loss[loss=0.182, simple_loss=0.2698, pruned_loss=0.04711, over 3207233.50 frames. ], batch size: 190, lr: 4.53e-03, grad_scale: 8.0 2023-04-30 04:12:14,711 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=147468.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 04:12:28,160 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.2605, 3.2642, 3.6788, 1.5074, 3.8429, 3.9037, 2.8983, 2.7981], device='cuda:1'), covar=tensor([0.0886, 0.0253, 0.0178, 0.1385, 0.0057, 0.0109, 0.0402, 0.0489], device='cuda:1'), in_proj_covar=tensor([0.0146, 0.0105, 0.0091, 0.0138, 0.0073, 0.0114, 0.0123, 0.0129], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:1') 2023-04-30 04:12:53,035 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=147495.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 04:13:00,654 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=147500.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 04:13:03,289 INFO [train.py:904] (1/8) Epoch 15, batch 5400, loss[loss=0.1733, simple_loss=0.2604, pruned_loss=0.04312, over 17159.00 frames. ], tot_loss[loss=0.1836, simple_loss=0.2726, pruned_loss=0.04729, over 3220228.42 frames. ], batch size: 46, lr: 4.53e-03, grad_scale: 8.0 2023-04-30 04:13:07,672 INFO [optim.py:368] (1/8) 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 04:13:11,509 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.35 vs. limit=2.0 2023-04-30 04:13:22,312 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.0965, 3.6959, 3.6411, 2.2453, 3.2697, 3.6542, 3.3658, 1.8884], device='cuda:1'), covar=tensor([0.0501, 0.0037, 0.0038, 0.0388, 0.0097, 0.0078, 0.0078, 0.0455], device='cuda:1'), in_proj_covar=tensor([0.0131, 0.0073, 0.0074, 0.0130, 0.0089, 0.0097, 0.0085, 0.0123], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-04-30 04:14:13,937 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=147548.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 04:14:20,733 INFO [train.py:904] (1/8) Epoch 15, batch 5450, loss[loss=0.2116, simple_loss=0.2933, pruned_loss=0.06499, over 12062.00 frames. ], tot_loss[loss=0.187, simple_loss=0.2759, pruned_loss=0.04907, over 3199242.81 frames. ], batch size: 246, lr: 4.53e-03, grad_scale: 8.0 2023-04-30 04:15:32,460 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 15, batch 5500, loss[loss=0.2114, simple_loss=0.3059, pruned_loss=0.05844, over 17263.00 frames. ], tot_loss[loss=0.1949, simple_loss=0.2828, pruned_loss=0.05348, over 3169415.03 frames. ], batch size: 52, lr: 4.53e-03, grad_scale: 4.0 2023-04-30 04:15:45,682 INFO [optim.py:368] (1/8) 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 04:16:01,761 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.30 vs. limit=2.0 2023-04-30 04:16:58,368 INFO [train.py:904] (1/8) Epoch 15, batch 5550, loss[loss=0.2226, simple_loss=0.3046, pruned_loss=0.07026, over 16280.00 frames. ], tot_loss[loss=0.2036, simple_loss=0.2899, pruned_loss=0.05865, over 3157826.60 frames. ], batch size: 165, lr: 4.53e-03, grad_scale: 4.0 2023-04-30 04:17:10,637 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=147659.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 04:17:22,972 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=147666.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 04:17:46,822 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.2061, 3.2014, 3.6317, 1.4863, 3.7123, 3.8051, 2.8075, 2.7872], device='cuda:1'), covar=tensor([0.0895, 0.0247, 0.0201, 0.1485, 0.0082, 0.0175, 0.0434, 0.0496], device='cuda:1'), in_proj_covar=tensor([0.0147, 0.0105, 0.0092, 0.0139, 0.0074, 0.0115, 0.0124, 0.0129], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:1') 2023-04-30 04:18:21,668 INFO [train.py:904] (1/8) Epoch 15, batch 5600, loss[loss=0.2384, simple_loss=0.3148, pruned_loss=0.08098, over 15223.00 frames. ], tot_loss[loss=0.2112, simple_loss=0.2952, pruned_loss=0.06356, over 3121694.73 frames. ], batch size: 190, lr: 4.53e-03, grad_scale: 8.0 2023-04-30 04:18:28,273 INFO [optim.py:368] (1/8) 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 04:18:52,987 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=147720.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 04:19:39,346 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.71 vs. limit=2.0 2023-04-30 04:19:46,181 INFO [train.py:904] (1/8) Epoch 15, batch 5650, loss[loss=0.2034, simple_loss=0.2889, pruned_loss=0.05895, over 16666.00 frames. ], tot_loss[loss=0.2176, simple_loss=0.3005, pruned_loss=0.0674, over 3094933.23 frames. ], batch size: 57, lr: 4.53e-03, grad_scale: 8.0 2023-04-30 04:20:04,679 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=147763.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 04:20:47,700 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=147790.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 04:21:05,457 INFO [train.py:904] (1/8) Epoch 15, batch 5700, loss[loss=0.2237, simple_loss=0.3126, pruned_loss=0.06736, over 16880.00 frames. ], tot_loss[loss=0.2199, simple_loss=0.3019, pruned_loss=0.06891, over 3089418.07 frames. ], batch size: 116, lr: 4.53e-03, grad_scale: 8.0 2023-04-30 04:21:11,568 INFO [optim.py:368] (1/8) 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 04:21:26,358 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 15, batch 5750, loss[loss=0.2318, simple_loss=0.2949, pruned_loss=0.08434, over 11004.00 frames. ], tot_loss[loss=0.2228, simple_loss=0.3041, pruned_loss=0.07076, over 3057590.49 frames. ], batch size: 247, lr: 4.53e-03, grad_scale: 8.0 2023-04-30 04:22:32,744 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=147857.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 04:23:03,177 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=147876.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 04:23:36,736 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 15, batch 5800, loss[loss=0.2144, simple_loss=0.3015, pruned_loss=0.06365, over 17027.00 frames. ], tot_loss[loss=0.2219, simple_loss=0.3041, pruned_loss=0.06981, over 3047019.10 frames. ], batch size: 53, lr: 4.53e-03, grad_scale: 8.0 2023-04-30 04:23:51,419 INFO [optim.py:368] (1/8) 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 04:24:00,165 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=147911.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 04:24:12,306 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=147918.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 04:24:55,227 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 15, batch 5850, loss[loss=0.2086, simple_loss=0.2839, pruned_loss=0.06662, over 11614.00 frames. ], tot_loss[loss=0.2191, simple_loss=0.3017, pruned_loss=0.0683, over 3024093.45 frames. ], batch size: 246, lr: 4.53e-03, grad_scale: 8.0 2023-04-30 04:25:28,472 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=147966.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 04:25:38,362 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 15, batch 5900, loss[loss=0.2083, simple_loss=0.2925, pruned_loss=0.06207, over 16876.00 frames. ], tot_loss[loss=0.2186, simple_loss=0.3019, pruned_loss=0.06759, over 3063383.93 frames. ], batch size: 116, lr: 4.52e-03, grad_scale: 4.0 2023-04-30 04:26:38,846 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.19 vs. limit=2.0 2023-04-30 04:26:39,372 INFO [optim.py:368] (1/8) 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] (1/8) 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] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=148015.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 04:27:50,061 INFO [train.py:904] (1/8) Epoch 15, batch 5950, loss[loss=0.1927, simple_loss=0.2864, pruned_loss=0.04948, over 16513.00 frames. ], tot_loss[loss=0.2172, simple_loss=0.3024, pruned_loss=0.06594, over 3076991.86 frames. ], batch size: 68, lr: 4.52e-03, grad_scale: 4.0 2023-04-30 04:28:08,624 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=148063.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 04:28:48,992 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 15, batch 6000, loss[loss=0.225, simple_loss=0.3004, pruned_loss=0.07482, over 11394.00 frames. ], tot_loss[loss=0.2163, simple_loss=0.301, pruned_loss=0.06581, over 3062048.51 frames. ], batch size: 246, lr: 4.52e-03, grad_scale: 8.0 2023-04-30 04:29:08,854 INFO [train.py:929] (1/8) Computing validation loss 2023-04-30 04:29:19,436 INFO [train.py:938] (1/8) Epoch 15, validation: loss=0.1559, simple_loss=0.2691, pruned_loss=0.0214, over 944034.00 frames. 2023-04-30 04:29:19,436 INFO [train.py:939] (1/8) Maximum memory allocated so far is 17857MB 2023-04-30 04:29:21,823 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.1064, 2.0204, 2.1131, 3.6839, 1.9558, 2.3968, 2.1726, 2.1958], device='cuda:1'), covar=tensor([0.1242, 0.3483, 0.2662, 0.0541, 0.4167, 0.2461, 0.3305, 0.3363], device='cuda:1'), in_proj_covar=tensor([0.0374, 0.0413, 0.0341, 0.0316, 0.0420, 0.0473, 0.0378, 0.0481], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-30 04:29:26,128 INFO [optim.py:368] (1/8) 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 04:29:33,079 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=148111.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 04:30:15,355 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 15, batch 6050, loss[loss=0.2084, simple_loss=0.3032, pruned_loss=0.05678, over 16628.00 frames. ], tot_loss[loss=0.2149, simple_loss=0.2992, pruned_loss=0.06529, over 3071664.72 frames. ], batch size: 134, lr: 4.52e-03, grad_scale: 4.0 2023-04-30 04:31:06,631 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.19 vs. limit=2.0 2023-04-30 04:31:07,973 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=148171.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 04:31:59,321 INFO [train.py:904] (1/8) Epoch 15, batch 6100, loss[loss=0.1911, simple_loss=0.2809, pruned_loss=0.05068, over 17161.00 frames. ], tot_loss[loss=0.2135, simple_loss=0.2987, pruned_loss=0.06415, over 3103218.77 frames. ], batch size: 46, lr: 4.52e-03, grad_scale: 4.0 2023-04-30 04:32:08,603 INFO [optim.py:368] (1/8) 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 04:32:17,950 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=148213.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 04:33:09,679 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.41 vs. limit=2.0 2023-04-30 04:33:18,245 INFO [train.py:904] (1/8) Epoch 15, batch 6150, loss[loss=0.1913, simple_loss=0.2775, pruned_loss=0.05251, over 16767.00 frames. ], tot_loss[loss=0.2118, simple_loss=0.2967, pruned_loss=0.0634, over 3102908.69 frames. ], batch size: 124, lr: 4.52e-03, grad_scale: 4.0 2023-04-30 04:33:27,574 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.6873, 4.0351, 3.1454, 2.3369, 2.8082, 2.6043, 4.3398, 3.5898], device='cuda:1'), covar=tensor([0.2742, 0.0654, 0.1533, 0.2344, 0.2359, 0.1686, 0.0395, 0.1049], device='cuda:1'), in_proj_covar=tensor([0.0310, 0.0261, 0.0291, 0.0291, 0.0284, 0.0233, 0.0276, 0.0311], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-30 04:33:42,601 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=148267.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 04:34:08,931 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.1622, 4.0524, 4.2411, 4.4129, 4.5012, 4.0869, 4.4445, 4.5243], device='cuda:1'), covar=tensor([0.1647, 0.1185, 0.1441, 0.0665, 0.0618, 0.1194, 0.0725, 0.0633], device='cuda:1'), in_proj_covar=tensor([0.0570, 0.0705, 0.0846, 0.0715, 0.0540, 0.0564, 0.0572, 0.0666], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-30 04:34:15,478 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=148287.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 04:34:37,598 INFO [train.py:904] (1/8) Epoch 15, batch 6200, loss[loss=0.2112, simple_loss=0.2917, pruned_loss=0.06536, over 16720.00 frames. ], tot_loss[loss=0.21, simple_loss=0.2947, pruned_loss=0.06263, over 3104203.32 frames. ], batch size: 89, lr: 4.52e-03, grad_scale: 4.0 2023-04-30 04:34:46,166 INFO [optim.py:368] (1/8) 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 04:34:58,567 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=148315.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 04:35:48,396 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=148348.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 04:35:54,857 INFO [train.py:904] (1/8) Epoch 15, batch 6250, loss[loss=0.1709, simple_loss=0.2718, pruned_loss=0.03502, over 16687.00 frames. ], tot_loss[loss=0.2085, simple_loss=0.2938, pruned_loss=0.06167, over 3125281.79 frames. ], batch size: 89, lr: 4.52e-03, grad_scale: 4.0 2023-04-30 04:36:11,759 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=148363.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 04:37:11,930 INFO [train.py:904] (1/8) Epoch 15, batch 6300, loss[loss=0.2085, simple_loss=0.292, pruned_loss=0.06252, over 16976.00 frames. ], tot_loss[loss=0.2076, simple_loss=0.2935, pruned_loss=0.06088, over 3135524.67 frames. ], batch size: 55, lr: 4.52e-03, grad_scale: 4.0 2023-04-30 04:37:21,867 INFO [optim.py:368] (1/8) 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 04:37:57,239 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=148429.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 04:38:33,541 INFO [train.py:904] (1/8) Epoch 15, batch 6350, loss[loss=0.2479, simple_loss=0.3196, pruned_loss=0.0881, over 11911.00 frames. ], tot_loss[loss=0.2085, simple_loss=0.2938, pruned_loss=0.0616, over 3130808.94 frames. ], batch size: 246, lr: 4.52e-03, grad_scale: 4.0 2023-04-30 04:39:03,798 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=148471.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 04:39:33,512 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=148490.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 04:39:45,785 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 15, batch 6400, loss[loss=0.2057, simple_loss=0.2893, pruned_loss=0.06104, over 15318.00 frames. ], tot_loss[loss=0.2106, simple_loss=0.2946, pruned_loss=0.06332, over 3103253.20 frames. ], batch size: 190, lr: 4.52e-03, grad_scale: 8.0 2023-04-30 04:39:59,985 INFO [optim.py:368] (1/8) 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 04:40:08,875 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=148513.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 04:40:18,147 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=148519.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 04:41:07,997 INFO [train.py:904] (1/8) Epoch 15, batch 6450, loss[loss=0.1834, simple_loss=0.2799, pruned_loss=0.04345, over 16827.00 frames. ], tot_loss[loss=0.2101, simple_loss=0.2945, pruned_loss=0.06284, over 3094183.01 frames. ], batch size: 83, lr: 4.52e-03, grad_scale: 8.0 2023-04-30 04:41:19,834 INFO [zipformer.py:625] (1/8) 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] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=148561.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 04:41:30,707 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.9355, 5.3689, 5.5382, 5.2754, 5.2467, 5.8676, 5.4133, 5.1884], device='cuda:1'), covar=tensor([0.0954, 0.1802, 0.2166, 0.1880, 0.2620, 0.0982, 0.1472, 0.2219], device='cuda:1'), in_proj_covar=tensor([0.0383, 0.0543, 0.0589, 0.0460, 0.0610, 0.0619, 0.0468, 0.0614], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-04-30 04:41:31,916 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=148567.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 04:42:04,545 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.3376, 3.0162, 2.5179, 2.2492, 2.2533, 2.1096, 2.8989, 2.7685], device='cuda:1'), covar=tensor([0.2441, 0.0834, 0.1714, 0.2286, 0.2619, 0.2353, 0.0571, 0.1287], device='cuda:1'), in_proj_covar=tensor([0.0307, 0.0259, 0.0288, 0.0289, 0.0282, 0.0233, 0.0273, 0.0309], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-30 04:42:26,998 INFO [train.py:904] (1/8) Epoch 15, batch 6500, loss[loss=0.226, simple_loss=0.3007, pruned_loss=0.07571, over 15192.00 frames. ], tot_loss[loss=0.2093, simple_loss=0.2931, pruned_loss=0.06268, over 3097471.38 frames. ], batch size: 190, lr: 4.52e-03, grad_scale: 8.0 2023-04-30 04:42:36,999 INFO [optim.py:368] (1/8) 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 04:42:47,763 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=148615.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 04:43:33,197 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=148643.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 04:43:46,629 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 15, batch 6550, loss[loss=0.1987, simple_loss=0.3018, pruned_loss=0.04781, over 16186.00 frames. ], tot_loss[loss=0.2113, simple_loss=0.2958, pruned_loss=0.06344, over 3107340.00 frames. ], batch size: 165, lr: 4.51e-03, grad_scale: 8.0 2023-04-30 04:44:14,073 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.3362, 3.3974, 3.7795, 1.7543, 3.9452, 3.9651, 2.9779, 2.8961], device='cuda:1'), covar=tensor([0.0763, 0.0208, 0.0194, 0.1158, 0.0068, 0.0146, 0.0373, 0.0433], device='cuda:1'), in_proj_covar=tensor([0.0145, 0.0105, 0.0091, 0.0137, 0.0073, 0.0114, 0.0123, 0.0127], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:1') 2023-04-30 04:44:47,691 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.91 vs. limit=2.0 2023-04-30 04:45:05,098 INFO [train.py:904] (1/8) Epoch 15, batch 6600, loss[loss=0.2339, simple_loss=0.3145, pruned_loss=0.07667, over 15388.00 frames. ], tot_loss[loss=0.2135, simple_loss=0.2984, pruned_loss=0.06431, over 3104106.92 frames. ], batch size: 191, lr: 4.51e-03, grad_scale: 8.0 2023-04-30 04:45:13,945 INFO [optim.py:368] (1/8) 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 04:45:19,892 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=148712.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 04:45:26,255 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=148716.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 04:46:00,192 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.4554, 2.2820, 2.3740, 4.2360, 2.2897, 2.7095, 2.3880, 2.4619], device='cuda:1'), covar=tensor([0.1089, 0.3400, 0.2516, 0.0410, 0.3687, 0.2161, 0.3169, 0.3242], device='cuda:1'), in_proj_covar=tensor([0.0375, 0.0413, 0.0343, 0.0317, 0.0421, 0.0475, 0.0380, 0.0484], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-30 04:46:22,112 INFO [train.py:904] (1/8) Epoch 15, batch 6650, loss[loss=0.2749, simple_loss=0.3321, pruned_loss=0.1089, over 11114.00 frames. ], tot_loss[loss=0.2147, simple_loss=0.2989, pruned_loss=0.06522, over 3097468.82 frames. ], batch size: 247, lr: 4.51e-03, grad_scale: 8.0 2023-04-30 04:46:33,609 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.5819, 3.5714, 2.7643, 2.1233, 2.3535, 2.2804, 3.7754, 3.2413], device='cuda:1'), covar=tensor([0.2686, 0.0742, 0.1727, 0.2261, 0.2451, 0.1953, 0.0482, 0.1169], device='cuda:1'), in_proj_covar=tensor([0.0311, 0.0262, 0.0291, 0.0292, 0.0285, 0.0235, 0.0276, 0.0312], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:1') 2023-04-30 04:47:00,957 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=148777.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 04:47:13,451 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 15, batch 6700, loss[loss=0.1963, simple_loss=0.2808, pruned_loss=0.05591, over 16752.00 frames. ], tot_loss[loss=0.2142, simple_loss=0.2974, pruned_loss=0.06545, over 3093083.63 frames. ], batch size: 83, lr: 4.51e-03, grad_scale: 8.0 2023-04-30 04:47:47,145 INFO [optim.py:368] (1/8) 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 04:47:57,621 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.7759, 3.9371, 2.4494, 4.6871, 2.9836, 4.5606, 2.6654, 2.9858], device='cuda:1'), covar=tensor([0.0259, 0.0369, 0.1633, 0.0129, 0.0836, 0.0407, 0.1450, 0.0775], device='cuda:1'), in_proj_covar=tensor([0.0159, 0.0170, 0.0188, 0.0142, 0.0168, 0.0209, 0.0195, 0.0172], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0003, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-04-30 04:48:54,917 INFO [train.py:904] (1/8) Epoch 15, batch 6750, loss[loss=0.1796, simple_loss=0.2566, pruned_loss=0.05125, over 17128.00 frames. ], tot_loss[loss=0.2119, simple_loss=0.2954, pruned_loss=0.06421, over 3114453.22 frames. ], batch size: 48, lr: 4.51e-03, grad_scale: 8.0 2023-04-30 04:48:58,420 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=148854.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 04:49:01,582 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([5.4714, 5.4228, 5.2478, 4.9015, 4.8782, 5.2571, 5.3081, 4.9653], device='cuda:1'), covar=tensor([0.0543, 0.0333, 0.0279, 0.0259, 0.1048, 0.0404, 0.0222, 0.0637], device='cuda:1'), in_proj_covar=tensor([0.0262, 0.0363, 0.0310, 0.0293, 0.0325, 0.0340, 0.0209, 0.0368], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-30 04:49:08,853 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=4.76 vs. limit=5.0 2023-04-30 04:50:09,776 INFO [train.py:904] (1/8) Epoch 15, batch 6800, loss[loss=0.226, simple_loss=0.3043, pruned_loss=0.07379, over 15283.00 frames. ], tot_loss[loss=0.2119, simple_loss=0.2955, pruned_loss=0.06412, over 3119876.28 frames. ], batch size: 190, lr: 4.51e-03, grad_scale: 8.0 2023-04-30 04:50:21,236 INFO [optim.py:368] (1/8) 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 04:51:15,896 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=148943.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 04:51:28,452 INFO [train.py:904] (1/8) Epoch 15, batch 6850, loss[loss=0.2343, simple_loss=0.3042, pruned_loss=0.08222, over 11925.00 frames. ], tot_loss[loss=0.2134, simple_loss=0.2967, pruned_loss=0.065, over 3101644.94 frames. ], batch size: 246, lr: 4.51e-03, grad_scale: 8.0 2023-04-30 04:52:26,499 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=148991.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 04:52:32,764 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.0215, 2.3749, 2.3386, 2.8379, 2.0028, 3.2688, 1.7870, 2.6314], device='cuda:1'), covar=tensor([0.1158, 0.0553, 0.0967, 0.0168, 0.0121, 0.0362, 0.1365, 0.0695], device='cuda:1'), in_proj_covar=tensor([0.0159, 0.0165, 0.0186, 0.0167, 0.0202, 0.0210, 0.0191, 0.0187], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-04-30 04:52:43,265 INFO [train.py:904] (1/8) Epoch 15, batch 6900, loss[loss=0.2217, simple_loss=0.311, pruned_loss=0.06622, over 16653.00 frames. ], tot_loss[loss=0.2148, simple_loss=0.2994, pruned_loss=0.06511, over 3106469.84 frames. ], batch size: 89, lr: 4.51e-03, grad_scale: 8.0 2023-04-30 04:52:52,138 INFO [zipformer.py:625] (1/8) 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] (1/8) Clipping_scale=2.0, grad-norm quartiles 1.886e+02 2.687e+02 3.079e+02 3.976e+02 7.116e+02, threshold=6.157e+02, percent-clipped=1.0 2023-04-30 04:53:58,320 INFO [train.py:904] (1/8) Epoch 15, batch 6950, loss[loss=0.2622, simple_loss=0.3266, pruned_loss=0.09888, over 11342.00 frames. ], tot_loss[loss=0.217, simple_loss=0.301, pruned_loss=0.06646, over 3096345.85 frames. ], batch size: 248, lr: 4.51e-03, grad_scale: 8.0 2023-04-30 04:54:18,949 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.8543, 4.0573, 3.0049, 2.4467, 2.9972, 2.4931, 4.4418, 3.7168], device='cuda:1'), covar=tensor([0.2698, 0.0727, 0.1688, 0.2338, 0.2426, 0.1866, 0.0475, 0.1154], device='cuda:1'), in_proj_covar=tensor([0.0310, 0.0262, 0.0291, 0.0291, 0.0284, 0.0234, 0.0276, 0.0311], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-30 04:54:27,424 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=149072.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 04:54:44,357 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.32 vs. limit=2.0 2023-04-30 04:54:45,954 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=149085.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 04:55:10,868 INFO [train.py:904] (1/8) Epoch 15, batch 7000, loss[loss=0.2065, simple_loss=0.3053, pruned_loss=0.05386, over 16802.00 frames. ], tot_loss[loss=0.2162, simple_loss=0.3012, pruned_loss=0.06559, over 3102068.39 frames. ], batch size: 39, lr: 4.51e-03, grad_scale: 2.0 2023-04-30 04:55:23,348 INFO [optim.py:368] (1/8) 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 04:55:26,888 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.29 vs. limit=2.0 2023-04-30 04:55:57,574 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=149133.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 04:56:25,562 INFO [train.py:904] (1/8) Epoch 15, batch 7050, loss[loss=0.1971, simple_loss=0.2999, pruned_loss=0.04718, over 16823.00 frames. ], tot_loss[loss=0.216, simple_loss=0.3015, pruned_loss=0.06521, over 3083987.20 frames. ], batch size: 102, lr: 4.51e-03, grad_scale: 2.0 2023-04-30 04:56:28,533 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=149154.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 04:57:25,358 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=2.81 vs. limit=5.0 2023-04-30 04:57:40,343 INFO [train.py:904] (1/8) Epoch 15, batch 7100, loss[loss=0.2342, simple_loss=0.3006, pruned_loss=0.08386, over 11793.00 frames. ], tot_loss[loss=0.2147, simple_loss=0.3001, pruned_loss=0.0647, over 3094133.86 frames. ], batch size: 248, lr: 4.51e-03, grad_scale: 2.0 2023-04-30 04:57:40,683 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=149202.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 04:57:53,982 INFO [optim.py:368] (1/8) 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 04:57:55,156 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.8720, 5.3408, 5.5526, 5.2264, 5.3016, 5.8729, 5.3773, 5.1513], device='cuda:1'), covar=tensor([0.1035, 0.1791, 0.2093, 0.2023, 0.2428, 0.0929, 0.1602, 0.2496], device='cuda:1'), in_proj_covar=tensor([0.0380, 0.0533, 0.0581, 0.0451, 0.0600, 0.0610, 0.0463, 0.0603], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-04-30 04:58:55,136 INFO [train.py:904] (1/8) Epoch 15, batch 7150, loss[loss=0.2301, simple_loss=0.3069, pruned_loss=0.07666, over 16510.00 frames. ], tot_loss[loss=0.2129, simple_loss=0.2978, pruned_loss=0.064, over 3120521.14 frames. ], batch size: 68, lr: 4.51e-03, grad_scale: 2.0 2023-04-30 04:59:22,669 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.9356, 1.9827, 2.1262, 3.4864, 1.9962, 2.3284, 2.1515, 2.1440], device='cuda:1'), covar=tensor([0.1295, 0.3612, 0.2761, 0.0556, 0.4171, 0.2310, 0.3301, 0.3267], device='cuda:1'), in_proj_covar=tensor([0.0376, 0.0413, 0.0343, 0.0318, 0.0422, 0.0475, 0.0381, 0.0484], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-30 04:59:30,596 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=4.29 vs. limit=5.0 2023-04-30 05:00:05,782 INFO [train.py:904] (1/8) Epoch 15, batch 7200, loss[loss=0.1713, simple_loss=0.2721, pruned_loss=0.03524, over 16929.00 frames. ], tot_loss[loss=0.2097, simple_loss=0.2953, pruned_loss=0.06207, over 3114226.00 frames. ], batch size: 96, lr: 4.50e-03, grad_scale: 4.0 2023-04-30 05:00:13,324 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=149307.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 05:00:17,915 INFO [optim.py:368] (1/8) Clipping_scale=2.0, grad-norm quartiles 1.349e+02 2.693e+02 3.090e+02 3.791e+02 7.265e+02, threshold=6.181e+02, percent-clipped=1.0 2023-04-30 05:00:26,999 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.73 vs. limit=2.0 2023-04-30 05:00:55,168 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.9676, 2.0075, 2.2357, 3.4465, 2.0164, 2.3539, 2.1784, 2.1772], device='cuda:1'), covar=tensor([0.1262, 0.3456, 0.2607, 0.0596, 0.4069, 0.2328, 0.3161, 0.3259], device='cuda:1'), in_proj_covar=tensor([0.0374, 0.0411, 0.0341, 0.0317, 0.0420, 0.0472, 0.0379, 0.0481], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-30 05:01:26,135 INFO [train.py:904] (1/8) Epoch 15, batch 7250, loss[loss=0.2067, simple_loss=0.2841, pruned_loss=0.06462, over 16766.00 frames. ], tot_loss[loss=0.2077, simple_loss=0.2934, pruned_loss=0.06101, over 3105961.25 frames. ], batch size: 124, lr: 4.50e-03, grad_scale: 4.0 2023-04-30 05:01:31,288 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=149355.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 05:01:57,290 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=149372.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 05:02:42,286 INFO [train.py:904] (1/8) Epoch 15, batch 7300, loss[loss=0.1851, simple_loss=0.2833, pruned_loss=0.04346, over 16722.00 frames. ], tot_loss[loss=0.2075, simple_loss=0.2925, pruned_loss=0.0612, over 3092773.75 frames. ], batch size: 89, lr: 4.50e-03, grad_scale: 4.0 2023-04-30 05:02:55,636 INFO [optim.py:368] (1/8) 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 05:03:05,427 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.1580, 4.0518, 4.2126, 4.3491, 4.4454, 4.0730, 4.4048, 4.4374], device='cuda:1'), covar=tensor([0.1356, 0.0958, 0.1261, 0.0559, 0.0482, 0.1151, 0.0621, 0.0579], device='cuda:1'), in_proj_covar=tensor([0.0557, 0.0690, 0.0827, 0.0698, 0.0532, 0.0557, 0.0563, 0.0655], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-30 05:03:10,461 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=149420.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 05:03:58,890 INFO [train.py:904] (1/8) Epoch 15, batch 7350, loss[loss=0.215, simple_loss=0.2992, pruned_loss=0.06539, over 15316.00 frames. ], tot_loss[loss=0.2092, simple_loss=0.2938, pruned_loss=0.06235, over 3085774.00 frames. ], batch size: 190, lr: 4.50e-03, grad_scale: 4.0 2023-04-30 05:05:17,926 INFO [train.py:904] (1/8) Epoch 15, batch 7400, loss[loss=0.2148, simple_loss=0.2984, pruned_loss=0.06565, over 16279.00 frames. ], tot_loss[loss=0.2111, simple_loss=0.2956, pruned_loss=0.06329, over 3081999.48 frames. ], batch size: 165, lr: 4.50e-03, grad_scale: 4.0 2023-04-30 05:05:32,172 INFO [optim.py:368] (1/8) 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 05:05:37,339 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 15, batch 7450, loss[loss=0.1934, simple_loss=0.2798, pruned_loss=0.0535, over 16671.00 frames. ], tot_loss[loss=0.2121, simple_loss=0.2961, pruned_loss=0.06406, over 3080383.75 frames. ], batch size: 134, lr: 4.50e-03, grad_scale: 4.0 2023-04-30 05:07:00,484 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=149565.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 05:07:16,437 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=149575.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 05:07:18,113 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=3.04 vs. limit=5.0 2023-04-30 05:08:00,167 INFO [train.py:904] (1/8) Epoch 15, batch 7500, loss[loss=0.2039, simple_loss=0.293, pruned_loss=0.05734, over 16916.00 frames. ], tot_loss[loss=0.2114, simple_loss=0.2959, pruned_loss=0.06347, over 3067150.47 frames. ], batch size: 109, lr: 4.50e-03, grad_scale: 2.0 2023-04-30 05:08:16,081 INFO [optim.py:368] (1/8) 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 05:08:39,423 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=149626.0, num_to_drop=1, layers_to_drop={0} 2023-04-30 05:08:41,101 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.6274, 4.6514, 5.1127, 5.0897, 5.0714, 4.6674, 4.6987, 4.4170], device='cuda:1'), covar=tensor([0.0335, 0.0488, 0.0317, 0.0341, 0.0488, 0.0377, 0.0970, 0.0504], device='cuda:1'), in_proj_covar=tensor([0.0367, 0.0392, 0.0384, 0.0369, 0.0440, 0.0409, 0.0506, 0.0326], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:1') 2023-04-30 05:09:18,804 INFO [train.py:904] (1/8) Epoch 15, batch 7550, loss[loss=0.2057, simple_loss=0.2866, pruned_loss=0.06237, over 15361.00 frames. ], tot_loss[loss=0.2105, simple_loss=0.2946, pruned_loss=0.06319, over 3072440.70 frames. ], batch size: 191, lr: 4.50e-03, grad_scale: 2.0 2023-04-30 05:10:35,810 INFO [train.py:904] (1/8) Epoch 15, batch 7600, loss[loss=0.188, simple_loss=0.2747, pruned_loss=0.05068, over 17005.00 frames. ], tot_loss[loss=0.211, simple_loss=0.2944, pruned_loss=0.0638, over 3076650.84 frames. ], batch size: 55, lr: 4.50e-03, grad_scale: 4.0 2023-04-30 05:10:39,786 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=2.68 vs. limit=5.0 2023-04-30 05:10:50,664 INFO [optim.py:368] (1/8) Clipping_scale=2.0, grad-norm quartiles 1.987e+02 2.884e+02 3.508e+02 4.101e+02 6.446e+02, threshold=7.017e+02, percent-clipped=0.0 2023-04-30 05:11:00,650 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.1185, 3.1563, 1.8181, 3.4386, 2.3241, 3.4655, 1.9727, 2.5261], device='cuda:1'), covar=tensor([0.0299, 0.0388, 0.1760, 0.0184, 0.0869, 0.0593, 0.1579, 0.0763], device='cuda:1'), in_proj_covar=tensor([0.0160, 0.0170, 0.0190, 0.0143, 0.0169, 0.0210, 0.0198, 0.0174], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0003, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-04-30 05:11:07,350 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.5253, 4.4324, 4.5422, 4.7670, 4.9042, 4.5191, 4.8961, 4.9320], device='cuda:1'), covar=tensor([0.1775, 0.1200, 0.1717, 0.0737, 0.0614, 0.0911, 0.0701, 0.0652], device='cuda:1'), in_proj_covar=tensor([0.0553, 0.0686, 0.0822, 0.0696, 0.0530, 0.0552, 0.0561, 0.0650], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-30 05:11:41,155 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=4.69 vs. limit=5.0 2023-04-30 05:11:45,960 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 15, batch 7650, loss[loss=0.2632, simple_loss=0.3262, pruned_loss=0.1001, over 11398.00 frames. ], tot_loss[loss=0.2121, simple_loss=0.2955, pruned_loss=0.06436, over 3083592.57 frames. ], batch size: 247, lr: 4.50e-03, grad_scale: 4.0 2023-04-30 05:12:28,927 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.1421, 4.1363, 4.5780, 4.5355, 4.5357, 4.2125, 4.2385, 4.1198], device='cuda:1'), covar=tensor([0.0332, 0.0646, 0.0356, 0.0447, 0.0518, 0.0429, 0.0983, 0.0540], device='cuda:1'), in_proj_covar=tensor([0.0368, 0.0393, 0.0385, 0.0370, 0.0441, 0.0411, 0.0507, 0.0326], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:1') 2023-04-30 05:13:13,700 INFO [train.py:904] (1/8) Epoch 15, batch 7700, loss[loss=0.2025, simple_loss=0.2872, pruned_loss=0.05888, over 17238.00 frames. ], tot_loss[loss=0.2132, simple_loss=0.2961, pruned_loss=0.06512, over 3081020.68 frames. ], batch size: 52, lr: 4.50e-03, grad_scale: 4.0 2023-04-30 05:13:22,110 INFO [zipformer.py:625] (1/8) 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] (1/8) 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 05:14:05,303 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([1.8351, 1.3817, 1.6687, 1.6854, 1.7792, 1.9150, 1.5549, 1.8152], device='cuda:1'), covar=tensor([0.0180, 0.0311, 0.0163, 0.0222, 0.0220, 0.0144, 0.0344, 0.0104], device='cuda:1'), in_proj_covar=tensor([0.0170, 0.0181, 0.0165, 0.0170, 0.0179, 0.0137, 0.0184, 0.0128], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-30 05:14:33,190 INFO [train.py:904] (1/8) Epoch 15, batch 7750, loss[loss=0.2116, simple_loss=0.2977, pruned_loss=0.06273, over 16750.00 frames. ], tot_loss[loss=0.2133, simple_loss=0.2969, pruned_loss=0.06491, over 3089857.79 frames. ], batch size: 124, lr: 4.50e-03, grad_scale: 4.0 2023-04-30 05:15:02,327 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=149870.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 05:15:29,073 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.3422, 2.9888, 2.5629, 2.2365, 2.2817, 2.1717, 2.9775, 2.8325], device='cuda:1'), covar=tensor([0.2421, 0.0865, 0.1763, 0.2520, 0.2506, 0.2219, 0.0623, 0.1221], device='cuda:1'), in_proj_covar=tensor([0.0314, 0.0263, 0.0295, 0.0295, 0.0287, 0.0237, 0.0279, 0.0312], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:1') 2023-04-30 05:15:35,140 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([5.1643, 5.1511, 4.9972, 4.2743, 5.0181, 1.9292, 4.7472, 4.8305], device='cuda:1'), covar=tensor([0.0104, 0.0090, 0.0171, 0.0430, 0.0097, 0.2530, 0.0158, 0.0188], device='cuda:1'), in_proj_covar=tensor([0.0143, 0.0131, 0.0178, 0.0165, 0.0150, 0.0190, 0.0166, 0.0157], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-30 05:15:51,918 INFO [train.py:904] (1/8) Epoch 15, batch 7800, loss[loss=0.2074, simple_loss=0.2929, pruned_loss=0.06091, over 16719.00 frames. ], tot_loss[loss=0.2148, simple_loss=0.2979, pruned_loss=0.06586, over 3088300.04 frames. ], batch size: 134, lr: 4.50e-03, grad_scale: 4.0 2023-04-30 05:16:07,341 INFO [optim.py:368] (1/8) 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 05:16:20,889 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=149921.0, num_to_drop=1, layers_to_drop={1} 2023-04-30 05:17:06,632 INFO [train.py:904] (1/8) Epoch 15, batch 7850, loss[loss=0.2153, simple_loss=0.2879, pruned_loss=0.07136, over 11594.00 frames. ], tot_loss[loss=0.2147, simple_loss=0.2978, pruned_loss=0.06575, over 3050531.41 frames. ], batch size: 248, lr: 4.50e-03, grad_scale: 4.0 2023-04-30 05:17:41,547 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.8888, 4.8600, 4.7364, 4.4084, 4.4048, 4.7485, 4.7155, 4.4448], device='cuda:1'), covar=tensor([0.0574, 0.0516, 0.0284, 0.0295, 0.0970, 0.0465, 0.0378, 0.0781], device='cuda:1'), in_proj_covar=tensor([0.0257, 0.0354, 0.0302, 0.0287, 0.0317, 0.0334, 0.0208, 0.0361], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-30 05:18:25,132 INFO [train.py:904] (1/8) Epoch 15, batch 7900, loss[loss=0.2156, simple_loss=0.3004, pruned_loss=0.06541, over 16474.00 frames. ], tot_loss[loss=0.2131, simple_loss=0.2967, pruned_loss=0.06478, over 3062932.38 frames. ], batch size: 68, lr: 4.49e-03, grad_scale: 4.0 2023-04-30 05:18:40,305 INFO [optim.py:368] (1/8) Clipping_scale=2.0, grad-norm quartiles 2.219e+02 2.824e+02 3.475e+02 4.207e+02 7.504e+02, threshold=6.949e+02, percent-clipped=0.0 2023-04-30 05:19:24,986 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([5.5215, 5.4967, 5.2781, 4.6188, 5.3651, 2.0669, 5.1232, 5.1566], device='cuda:1'), covar=tensor([0.0067, 0.0054, 0.0156, 0.0385, 0.0078, 0.2502, 0.0107, 0.0163], device='cuda:1'), in_proj_covar=tensor([0.0144, 0.0131, 0.0178, 0.0165, 0.0151, 0.0191, 0.0166, 0.0158], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-30 05:19:43,691 INFO [train.py:904] (1/8) Epoch 15, batch 7950, loss[loss=0.2028, simple_loss=0.2848, pruned_loss=0.06044, over 15405.00 frames. ], tot_loss[loss=0.2133, simple_loss=0.2968, pruned_loss=0.06491, over 3077143.78 frames. ], batch size: 190, lr: 4.49e-03, grad_scale: 4.0 2023-04-30 05:20:07,263 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.9987, 4.0667, 3.8715, 3.6532, 3.6477, 3.9714, 3.6671, 3.7373], device='cuda:1'), covar=tensor([0.0607, 0.0577, 0.0272, 0.0272, 0.0730, 0.0510, 0.0983, 0.0594], device='cuda:1'), in_proj_covar=tensor([0.0257, 0.0354, 0.0301, 0.0287, 0.0317, 0.0334, 0.0208, 0.0360], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-30 05:20:20,280 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.27 vs. limit=2.0 2023-04-30 05:20:25,870 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=150079.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 05:20:58,677 INFO [train.py:904] (1/8) Epoch 15, batch 8000, loss[loss=0.2124, simple_loss=0.3086, pruned_loss=0.05808, over 16365.00 frames. ], tot_loss[loss=0.2141, simple_loss=0.2973, pruned_loss=0.06542, over 3080003.80 frames. ], batch size: 146, lr: 4.49e-03, grad_scale: 8.0 2023-04-30 05:20:59,764 INFO [zipformer.py:625] (1/8) 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] (1/8) 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 05:21:29,940 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.3846, 3.5491, 3.1982, 2.9599, 2.9469, 3.4048, 3.1996, 3.1671], device='cuda:1'), covar=tensor([0.0853, 0.0613, 0.0403, 0.0363, 0.0980, 0.0538, 0.1996, 0.0641], device='cuda:1'), in_proj_covar=tensor([0.0257, 0.0355, 0.0302, 0.0287, 0.0317, 0.0334, 0.0208, 0.0360], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-30 05:21:55,905 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 15, batch 8050, loss[loss=0.211, simple_loss=0.3, pruned_loss=0.06098, over 15333.00 frames. ], tot_loss[loss=0.2137, simple_loss=0.2974, pruned_loss=0.06505, over 3080736.28 frames. ], batch size: 191, lr: 4.49e-03, grad_scale: 8.0 2023-04-30 05:22:34,091 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.1242, 2.8855, 3.1442, 1.7691, 3.2696, 3.3551, 2.6930, 2.5707], device='cuda:1'), covar=tensor([0.0786, 0.0252, 0.0187, 0.1172, 0.0080, 0.0181, 0.0415, 0.0458], device='cuda:1'), in_proj_covar=tensor([0.0144, 0.0104, 0.0090, 0.0137, 0.0073, 0.0114, 0.0122, 0.0127], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:1') 2023-04-30 05:22:41,242 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=150170.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 05:23:19,942 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([5.4401, 5.8575, 5.4962, 5.5932, 5.2085, 5.1233, 5.1965, 5.9357], device='cuda:1'), covar=tensor([0.1119, 0.0826, 0.1009, 0.0824, 0.0923, 0.0681, 0.1116, 0.0883], device='cuda:1'), in_proj_covar=tensor([0.0597, 0.0734, 0.0613, 0.0531, 0.0462, 0.0475, 0.0613, 0.0563], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:1') 2023-04-30 05:23:20,068 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.5410, 2.5873, 1.9241, 2.4428, 3.0415, 2.7002, 3.2082, 3.2978], device='cuda:1'), covar=tensor([0.0095, 0.0372, 0.0598, 0.0404, 0.0220, 0.0370, 0.0245, 0.0201], device='cuda:1'), in_proj_covar=tensor([0.0170, 0.0215, 0.0210, 0.0210, 0.0214, 0.0216, 0.0217, 0.0209], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-30 05:23:30,576 INFO [train.py:904] (1/8) Epoch 15, batch 8100, loss[loss=0.2247, simple_loss=0.3028, pruned_loss=0.07326, over 15464.00 frames. ], tot_loss[loss=0.2128, simple_loss=0.2968, pruned_loss=0.06435, over 3092979.54 frames. ], batch size: 190, lr: 4.49e-03, grad_scale: 8.0 2023-04-30 05:23:45,513 INFO [optim.py:368] (1/8) 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 05:23:55,518 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=150218.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 05:23:59,919 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=150221.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 05:24:45,701 INFO [train.py:904] (1/8) Epoch 15, batch 8150, loss[loss=0.1841, simple_loss=0.2698, pruned_loss=0.04926, over 16833.00 frames. ], tot_loss[loss=0.2107, simple_loss=0.2945, pruned_loss=0.06338, over 3088504.75 frames. ], batch size: 90, lr: 4.49e-03, grad_scale: 8.0 2023-04-30 05:25:11,109 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=150269.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 05:25:18,980 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=150274.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 05:25:28,408 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.0082, 4.0501, 3.8726, 3.6383, 3.6341, 3.9729, 3.7002, 3.7485], device='cuda:1'), covar=tensor([0.0653, 0.0557, 0.0275, 0.0295, 0.0781, 0.0458, 0.0948, 0.0658], device='cuda:1'), in_proj_covar=tensor([0.0259, 0.0357, 0.0304, 0.0288, 0.0320, 0.0336, 0.0210, 0.0363], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-30 05:26:00,375 INFO [train.py:904] (1/8) Epoch 15, batch 8200, loss[loss=0.203, simple_loss=0.2928, pruned_loss=0.05662, over 16794.00 frames. ], tot_loss[loss=0.208, simple_loss=0.2918, pruned_loss=0.0621, over 3119948.03 frames. ], batch size: 102, lr: 4.49e-03, grad_scale: 8.0 2023-04-30 05:26:16,345 INFO [optim.py:368] (1/8) Clipping_scale=2.0, grad-norm quartiles 2.058e+02 2.882e+02 3.369e+02 3.879e+02 7.748e+02, threshold=6.737e+02, percent-clipped=3.0 2023-04-30 05:26:54,536 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=150335.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 05:27:18,686 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.79 vs. limit=2.0 2023-04-30 05:27:22,317 INFO [train.py:904] (1/8) Epoch 15, batch 8250, loss[loss=0.1722, simple_loss=0.2718, pruned_loss=0.03626, over 16470.00 frames. ], tot_loss[loss=0.205, simple_loss=0.2905, pruned_loss=0.05979, over 3093798.29 frames. ], batch size: 75, lr: 4.49e-03, grad_scale: 8.0 2023-04-30 05:27:38,602 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.45 vs. limit=2.0 2023-04-30 05:28:45,002 INFO [train.py:904] (1/8) Epoch 15, batch 8300, loss[loss=0.1663, simple_loss=0.2702, pruned_loss=0.03122, over 16856.00 frames. ], tot_loss[loss=0.2009, simple_loss=0.2877, pruned_loss=0.05704, over 3078381.49 frames. ], batch size: 96, lr: 4.49e-03, grad_scale: 8.0 2023-04-30 05:28:45,649 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=150402.0, num_to_drop=1, layers_to_drop={1} 2023-04-30 05:29:01,272 INFO [optim.py:368] (1/8) 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 05:29:31,548 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.64 vs. limit=2.0 2023-04-30 05:29:39,976 INFO [zipformer.py:625] (1/8) 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] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=150450.0, num_to_drop=1, layers_to_drop={0} 2023-04-30 05:30:07,755 INFO [train.py:904] (1/8) Epoch 15, batch 8350, loss[loss=0.2277, simple_loss=0.3034, pruned_loss=0.07603, over 12507.00 frames. ], tot_loss[loss=0.1989, simple_loss=0.2875, pruned_loss=0.05519, over 3073079.76 frames. ], batch size: 248, lr: 4.49e-03, grad_scale: 8.0 2023-04-30 05:31:29,423 INFO [train.py:904] (1/8) Epoch 15, batch 8400, loss[loss=0.1825, simple_loss=0.2652, pruned_loss=0.04992, over 12056.00 frames. ], tot_loss[loss=0.1954, simple_loss=0.2848, pruned_loss=0.053, over 3071247.14 frames. ], batch size: 248, lr: 4.49e-03, grad_scale: 8.0 2023-04-30 05:31:46,287 INFO [optim.py:368] (1/8) 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 05:32:19,288 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.55 vs. limit=2.0 2023-04-30 05:32:27,442 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 15, batch 8450, loss[loss=0.178, simple_loss=0.273, pruned_loss=0.0415, over 15386.00 frames. ], tot_loss[loss=0.1928, simple_loss=0.2831, pruned_loss=0.05131, over 3081343.03 frames. ], batch size: 191, lr: 4.49e-03, grad_scale: 8.0 2023-04-30 05:34:02,215 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=150598.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 05:34:08,631 INFO [train.py:904] (1/8) Epoch 15, batch 8500, loss[loss=0.1845, simple_loss=0.2678, pruned_loss=0.05063, over 15380.00 frames. ], tot_loss[loss=0.1888, simple_loss=0.2794, pruned_loss=0.04914, over 3073356.42 frames. ], batch size: 190, lr: 4.49e-03, grad_scale: 8.0 2023-04-30 05:34:25,005 INFO [optim.py:368] (1/8) 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 05:34:54,268 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=150630.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 05:35:23,706 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([1.8999, 2.7634, 2.6208, 2.0955, 2.5027, 2.7491, 2.6496, 1.8788], device='cuda:1'), covar=tensor([0.0371, 0.0063, 0.0055, 0.0285, 0.0092, 0.0082, 0.0087, 0.0391], device='cuda:1'), in_proj_covar=tensor([0.0129, 0.0073, 0.0072, 0.0131, 0.0086, 0.0095, 0.0084, 0.0122], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:1') 2023-04-30 05:35:30,232 INFO [train.py:904] (1/8) Epoch 15, batch 8550, loss[loss=0.205, simple_loss=0.2948, pruned_loss=0.05756, over 16967.00 frames. ], tot_loss[loss=0.1862, simple_loss=0.2766, pruned_loss=0.04792, over 3066327.89 frames. ], batch size: 109, lr: 4.48e-03, grad_scale: 8.0 2023-04-30 05:36:46,679 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.47 vs. limit=2.0 2023-04-30 05:37:12,303 INFO [train.py:904] (1/8) Epoch 15, batch 8600, loss[loss=0.1761, simple_loss=0.2768, pruned_loss=0.03767, over 16604.00 frames. ], tot_loss[loss=0.186, simple_loss=0.277, pruned_loss=0.04749, over 3041280.51 frames. ], batch size: 76, lr: 4.48e-03, grad_scale: 8.0 2023-04-30 05:37:32,421 INFO [optim.py:368] (1/8) 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 05:38:14,926 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=150732.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 05:38:20,333 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=150735.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 05:38:42,148 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.71 vs. limit=2.0 2023-04-30 05:38:51,976 INFO [train.py:904] (1/8) Epoch 15, batch 8650, loss[loss=0.1783, simple_loss=0.2787, pruned_loss=0.03897, over 16947.00 frames. ], tot_loss[loss=0.1834, simple_loss=0.2749, pruned_loss=0.04596, over 3032955.46 frames. ], batch size: 116, lr: 4.48e-03, grad_scale: 8.0 2023-04-30 05:40:03,477 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=150783.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 05:40:23,193 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=150793.0, num_to_drop=1, layers_to_drop={2} 2023-04-30 05:40:39,985 INFO [train.py:904] (1/8) Epoch 15, batch 8700, loss[loss=0.1535, simple_loss=0.2409, pruned_loss=0.03309, over 16648.00 frames. ], tot_loss[loss=0.1814, simple_loss=0.2724, pruned_loss=0.04524, over 3034072.03 frames. ], batch size: 62, lr: 4.48e-03, grad_scale: 4.0 2023-04-30 05:40:48,632 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.1809, 3.2483, 1.9027, 3.4950, 2.3901, 3.4990, 2.1505, 2.6981], device='cuda:1'), covar=tensor([0.0275, 0.0382, 0.1567, 0.0237, 0.0831, 0.0546, 0.1506, 0.0689], device='cuda:1'), in_proj_covar=tensor([0.0153, 0.0164, 0.0184, 0.0137, 0.0164, 0.0201, 0.0193, 0.0170], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0003, 0.0004, 0.0004, 0.0003], device='cuda:1') 2023-04-30 05:41:01,899 INFO [optim.py:368] (1/8) 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 05:41:55,358 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.6247, 3.8680, 4.1522, 1.9688, 4.2784, 4.4878, 3.2514, 3.2782], device='cuda:1'), covar=tensor([0.0883, 0.0151, 0.0149, 0.1216, 0.0043, 0.0078, 0.0320, 0.0445], device='cuda:1'), in_proj_covar=tensor([0.0140, 0.0101, 0.0087, 0.0134, 0.0070, 0.0110, 0.0119, 0.0123], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:1') 2023-04-30 05:42:16,603 INFO [train.py:904] (1/8) Epoch 15, batch 8750, loss[loss=0.1911, simple_loss=0.2847, pruned_loss=0.04876, over 16173.00 frames. ], tot_loss[loss=0.1809, simple_loss=0.2723, pruned_loss=0.04477, over 3036369.36 frames. ], batch size: 165, lr: 4.48e-03, grad_scale: 4.0 2023-04-30 05:42:46,755 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.44 vs. limit=2.0 2023-04-30 05:43:41,300 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.7667, 3.2510, 3.3345, 2.0377, 2.8393, 2.2548, 3.2089, 3.3514], device='cuda:1'), covar=tensor([0.0288, 0.0713, 0.0483, 0.1794, 0.0749, 0.0895, 0.0636, 0.0748], device='cuda:1'), in_proj_covar=tensor([0.0145, 0.0148, 0.0157, 0.0144, 0.0136, 0.0123, 0.0136, 0.0156], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:1') 2023-04-30 05:43:52,657 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=150893.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 05:44:06,437 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 15, batch 8800, loss[loss=0.1679, simple_loss=0.2621, pruned_loss=0.0369, over 15492.00 frames. ], tot_loss[loss=0.1786, simple_loss=0.2708, pruned_loss=0.04323, over 3058415.64 frames. ], batch size: 191, lr: 4.48e-03, grad_scale: 8.0 2023-04-30 05:44:28,830 INFO [optim.py:368] (1/8) 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 05:45:06,329 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=150930.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 05:45:50,511 INFO [train.py:904] (1/8) Epoch 15, batch 8850, loss[loss=0.1774, simple_loss=0.2612, pruned_loss=0.04677, over 12512.00 frames. ], tot_loss[loss=0.1791, simple_loss=0.2732, pruned_loss=0.04251, over 3067422.61 frames. ], batch size: 250, lr: 4.48e-03, grad_scale: 8.0 2023-04-30 05:46:10,231 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=150961.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 05:46:17,299 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.98 vs. limit=2.0 2023-04-30 05:46:46,307 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=150978.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 05:47:36,865 INFO [train.py:904] (1/8) Epoch 15, batch 8900, loss[loss=0.1799, simple_loss=0.2854, pruned_loss=0.03718, over 16909.00 frames. ], tot_loss[loss=0.178, simple_loss=0.2732, pruned_loss=0.04145, over 3075467.65 frames. ], batch size: 96, lr: 4.48e-03, grad_scale: 8.0 2023-04-30 05:47:59,489 INFO [optim.py:368] (1/8) 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 05:49:42,111 INFO [train.py:904] (1/8) Epoch 15, batch 8950, loss[loss=0.1631, simple_loss=0.2574, pruned_loss=0.03437, over 16280.00 frames. ], tot_loss[loss=0.1781, simple_loss=0.2728, pruned_loss=0.04173, over 3090578.35 frames. ], batch size: 165, lr: 4.48e-03, grad_scale: 8.0 2023-04-30 05:51:01,605 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=151088.0, num_to_drop=1, layers_to_drop={3} 2023-04-30 05:51:16,281 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.47 vs. limit=2.0 2023-04-30 05:51:30,995 INFO [train.py:904] (1/8) Epoch 15, batch 9000, loss[loss=0.1808, simple_loss=0.2699, pruned_loss=0.04582, over 16826.00 frames. ], tot_loss[loss=0.1759, simple_loss=0.2699, pruned_loss=0.04091, over 3063906.19 frames. ], batch size: 124, lr: 4.48e-03, grad_scale: 8.0 2023-04-30 05:51:30,995 INFO [train.py:929] (1/8) Computing validation loss 2023-04-30 05:51:40,827 INFO [train.py:938] (1/8) Epoch 15, validation: loss=0.15, simple_loss=0.2539, pruned_loss=0.02307, over 944034.00 frames. 2023-04-30 05:51:40,827 INFO [train.py:939] (1/8) Maximum memory allocated so far is 17857MB 2023-04-30 05:52:03,718 INFO [optim.py:368] (1/8) 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 05:52:34,130 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.0355, 3.1071, 1.8687, 3.3224, 2.3088, 3.3090, 2.0479, 2.6274], device='cuda:1'), covar=tensor([0.0279, 0.0356, 0.1588, 0.0206, 0.0802, 0.0538, 0.1454, 0.0647], device='cuda:1'), in_proj_covar=tensor([0.0152, 0.0162, 0.0182, 0.0135, 0.0163, 0.0198, 0.0191, 0.0167], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0003, 0.0004, 0.0004, 0.0003], device='cuda:1') 2023-04-30 05:53:23,275 INFO [train.py:904] (1/8) Epoch 15, batch 9050, loss[loss=0.1779, simple_loss=0.2634, pruned_loss=0.0462, over 16594.00 frames. ], tot_loss[loss=0.1772, simple_loss=0.271, pruned_loss=0.04167, over 3063696.86 frames. ], batch size: 62, lr: 4.48e-03, grad_scale: 8.0 2023-04-30 05:54:27,752 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=151184.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 05:54:51,354 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 15, batch 9100, loss[loss=0.1685, simple_loss=0.2718, pruned_loss=0.0326, over 16774.00 frames. ], tot_loss[loss=0.1773, simple_loss=0.2704, pruned_loss=0.04214, over 3051418.96 frames. ], batch size: 83, lr: 4.48e-03, grad_scale: 8.0 2023-04-30 05:55:31,069 INFO [optim.py:368] (1/8) Clipping_scale=2.0, grad-norm quartiles 1.671e+02 2.276e+02 2.684e+02 3.281e+02 5.650e+02, threshold=5.369e+02, percent-clipped=3.0 2023-04-30 05:56:08,303 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.21 vs. limit=2.0 2023-04-30 05:56:40,113 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.56 vs. limit=2.0 2023-04-30 05:56:44,482 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=151241.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 05:56:54,388 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=151245.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 05:57:08,960 INFO [train.py:904] (1/8) Epoch 15, batch 9150, loss[loss=0.1658, simple_loss=0.2621, pruned_loss=0.03477, over 15435.00 frames. ], tot_loss[loss=0.1771, simple_loss=0.2709, pruned_loss=0.04167, over 3055578.96 frames. ], batch size: 191, lr: 4.48e-03, grad_scale: 8.0 2023-04-30 05:57:18,558 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=151256.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 05:58:12,654 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=151282.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 05:58:46,465 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=151298.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 05:58:51,676 INFO [train.py:904] (1/8) Epoch 15, batch 9200, loss[loss=0.1655, simple_loss=0.2577, pruned_loss=0.0367, over 16831.00 frames. ], tot_loss[loss=0.1734, simple_loss=0.2661, pruned_loss=0.04036, over 3074531.77 frames. ], batch size: 124, lr: 4.48e-03, grad_scale: 8.0 2023-04-30 05:59:01,635 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.44 vs. limit=2.0 2023-04-30 05:59:12,237 INFO [optim.py:368] (1/8) Clipping_scale=2.0, grad-norm quartiles 1.606e+02 2.499e+02 3.414e+02 4.138e+02 7.678e+02, threshold=6.827e+02, percent-clipped=7.0 2023-04-30 06:00:09,401 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 15, batch 9250, loss[loss=0.1687, simple_loss=0.2632, pruned_loss=0.0371, over 16170.00 frames. ], tot_loss[loss=0.1741, simple_loss=0.2665, pruned_loss=0.04087, over 3080520.46 frames. ], batch size: 165, lr: 4.47e-03, grad_scale: 8.0 2023-04-30 06:00:40,461 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=151359.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 06:01:07,615 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.5375, 3.4890, 3.5103, 2.6150, 3.4237, 1.9509, 3.2182, 2.8153], device='cuda:1'), covar=tensor([0.0156, 0.0141, 0.0182, 0.0306, 0.0120, 0.2809, 0.0170, 0.0299], device='cuda:1'), in_proj_covar=tensor([0.0141, 0.0129, 0.0175, 0.0158, 0.0149, 0.0189, 0.0163, 0.0153], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-30 06:01:43,154 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=151388.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 06:02:14,983 INFO [train.py:904] (1/8) Epoch 15, batch 9300, loss[loss=0.1664, simple_loss=0.2488, pruned_loss=0.04197, over 12263.00 frames. ], tot_loss[loss=0.1725, simple_loss=0.2648, pruned_loss=0.04005, over 3064545.70 frames. ], batch size: 248, lr: 4.47e-03, grad_scale: 8.0 2023-04-30 06:02:37,909 INFO [optim.py:368] (1/8) Clipping_scale=2.0, grad-norm quartiles 1.525e+02 2.225e+02 2.604e+02 3.286e+02 5.862e+02, threshold=5.207e+02, percent-clipped=0.0 2023-04-30 06:03:16,621 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.2439, 2.0815, 2.2306, 3.8063, 2.0494, 2.4456, 2.2506, 2.2437], device='cuda:1'), covar=tensor([0.1043, 0.3631, 0.2762, 0.0472, 0.4166, 0.2589, 0.3474, 0.3579], device='cuda:1'), in_proj_covar=tensor([0.0369, 0.0406, 0.0343, 0.0312, 0.0419, 0.0464, 0.0374, 0.0475], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-30 06:03:21,554 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.21 vs. limit=2.0 2023-04-30 06:03:29,231 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=151436.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 06:03:59,109 INFO [train.py:904] (1/8) Epoch 15, batch 9350, loss[loss=0.1979, simple_loss=0.2852, pruned_loss=0.05531, over 16938.00 frames. ], tot_loss[loss=0.1724, simple_loss=0.2649, pruned_loss=0.03991, over 3089606.77 frames. ], batch size: 109, lr: 4.47e-03, grad_scale: 8.0 2023-04-30 06:04:20,054 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=4.72 vs. limit=5.0 2023-04-30 06:04:45,718 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.3239, 3.0796, 2.6639, 2.2466, 2.2409, 2.0956, 2.9615, 2.7685], device='cuda:1'), covar=tensor([0.2476, 0.0737, 0.1654, 0.2822, 0.2681, 0.2418, 0.0584, 0.1368], device='cuda:1'), in_proj_covar=tensor([0.0304, 0.0254, 0.0285, 0.0286, 0.0270, 0.0230, 0.0268, 0.0300], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-30 06:05:36,958 INFO [train.py:904] (1/8) Epoch 15, batch 9400, loss[loss=0.1771, simple_loss=0.275, pruned_loss=0.03964, over 15290.00 frames. ], tot_loss[loss=0.1721, simple_loss=0.2649, pruned_loss=0.03965, over 3076564.47 frames. ], batch size: 190, lr: 4.47e-03, grad_scale: 8.0 2023-04-30 06:05:59,180 INFO [optim.py:368] (1/8) 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 06:06:08,393 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.39 vs. limit=2.0 2023-04-30 06:06:55,071 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=151540.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 06:07:17,456 INFO [train.py:904] (1/8) Epoch 15, batch 9450, loss[loss=0.1743, simple_loss=0.2613, pruned_loss=0.0436, over 12622.00 frames. ], tot_loss[loss=0.1729, simple_loss=0.2658, pruned_loss=0.03994, over 3041202.43 frames. ], batch size: 248, lr: 4.47e-03, grad_scale: 8.0 2023-04-30 06:07:24,108 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=151556.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 06:08:58,267 INFO [train.py:904] (1/8) Epoch 15, batch 9500, loss[loss=0.1677, simple_loss=0.2587, pruned_loss=0.03834, over 17055.00 frames. ], tot_loss[loss=0.1722, simple_loss=0.2652, pruned_loss=0.0396, over 3063708.41 frames. ], batch size: 55, lr: 4.47e-03, grad_scale: 8.0 2023-04-30 06:09:04,123 INFO [zipformer.py:625] (1/8) 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] (1/8) 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 06:09:47,490 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([1.9329, 2.6490, 2.4037, 4.2256, 2.7778, 4.0675, 1.3361, 2.8188], device='cuda:1'), covar=tensor([0.1147, 0.0701, 0.1164, 0.0156, 0.0112, 0.0419, 0.1524, 0.0780], device='cuda:1'), in_proj_covar=tensor([0.0156, 0.0160, 0.0182, 0.0158, 0.0188, 0.0202, 0.0186, 0.0179], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0003, 0.0003], device='cuda:1') 2023-04-30 06:10:11,839 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=151638.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 06:10:44,451 INFO [train.py:904] (1/8) Epoch 15, batch 9550, loss[loss=0.1854, simple_loss=0.2791, pruned_loss=0.04583, over 16845.00 frames. ], tot_loss[loss=0.1723, simple_loss=0.2655, pruned_loss=0.03962, over 3087975.31 frames. ], batch size: 116, lr: 4.47e-03, grad_scale: 8.0 2023-04-30 06:10:49,179 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=151654.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 06:10:52,351 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.4861, 3.5174, 2.8108, 2.0477, 2.2805, 2.2603, 3.6909, 3.1392], device='cuda:1'), covar=tensor([0.2826, 0.0633, 0.1580, 0.2788, 0.2609, 0.1990, 0.0396, 0.1194], device='cuda:1'), in_proj_covar=tensor([0.0303, 0.0253, 0.0284, 0.0283, 0.0268, 0.0229, 0.0267, 0.0298], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-30 06:11:58,523 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=151688.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 06:12:03,756 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.0352, 4.0174, 3.9710, 3.4384, 3.9841, 1.8038, 3.7951, 3.6955], device='cuda:1'), covar=tensor([0.0097, 0.0087, 0.0143, 0.0237, 0.0096, 0.2439, 0.0126, 0.0232], device='cuda:1'), in_proj_covar=tensor([0.0138, 0.0126, 0.0171, 0.0155, 0.0146, 0.0187, 0.0161, 0.0151], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-30 06:12:19,304 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=2.63 vs. limit=5.0 2023-04-30 06:12:24,738 INFO [train.py:904] (1/8) Epoch 15, batch 9600, loss[loss=0.1603, simple_loss=0.2472, pruned_loss=0.03675, over 12539.00 frames. ], tot_loss[loss=0.1743, simple_loss=0.2677, pruned_loss=0.04046, over 3088469.77 frames. ], batch size: 248, lr: 4.47e-03, grad_scale: 8.0 2023-04-30 06:12:25,693 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.0635, 2.0901, 2.1688, 3.5625, 2.0108, 2.4086, 2.2172, 2.2091], device='cuda:1'), covar=tensor([0.1127, 0.3377, 0.2652, 0.0510, 0.4084, 0.2388, 0.3214, 0.3195], device='cuda:1'), in_proj_covar=tensor([0.0366, 0.0403, 0.0339, 0.0309, 0.0416, 0.0459, 0.0372, 0.0470], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-30 06:12:44,296 INFO [optim.py:368] (1/8) 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 06:14:04,284 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=151749.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 06:14:11,687 INFO [train.py:904] (1/8) Epoch 15, batch 9650, loss[loss=0.1707, simple_loss=0.2704, pruned_loss=0.0355, over 16378.00 frames. ], tot_loss[loss=0.1761, simple_loss=0.2699, pruned_loss=0.04113, over 3072619.82 frames. ], batch size: 146, lr: 4.47e-03, grad_scale: 8.0 2023-04-30 06:15:01,167 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.22 vs. limit=2.0 2023-04-30 06:15:57,513 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=151801.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 06:15:58,768 INFO [train.py:904] (1/8) Epoch 15, batch 9700, loss[loss=0.1701, simple_loss=0.2661, pruned_loss=0.03707, over 12135.00 frames. ], tot_loss[loss=0.1753, simple_loss=0.2688, pruned_loss=0.04092, over 3065463.99 frames. ], batch size: 248, lr: 4.47e-03, grad_scale: 8.0 2023-04-30 06:16:07,776 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.28 vs. limit=2.0 2023-04-30 06:16:19,905 INFO [optim.py:368] (1/8) Clipping_scale=2.0, grad-norm quartiles 1.752e+02 2.225e+02 2.684e+02 3.461e+02 6.863e+02, threshold=5.368e+02, percent-clipped=3.0 2023-04-30 06:16:55,201 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.8794, 4.2106, 4.0709, 4.0812, 3.7526, 3.8263, 3.8614, 4.2221], device='cuda:1'), covar=tensor([0.1183, 0.0915, 0.0897, 0.0742, 0.0764, 0.1436, 0.0944, 0.0910], device='cuda:1'), in_proj_covar=tensor([0.0570, 0.0707, 0.0574, 0.0508, 0.0445, 0.0458, 0.0585, 0.0538], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:1') 2023-04-30 06:17:18,824 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 15, batch 9750, loss[loss=0.1815, simple_loss=0.2759, pruned_loss=0.04355, over 16332.00 frames. ], tot_loss[loss=0.1747, simple_loss=0.2673, pruned_loss=0.04107, over 3057194.71 frames. ], batch size: 146, lr: 4.47e-03, grad_scale: 8.0 2023-04-30 06:18:01,494 INFO [zipformer.py:625] (1/8) 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] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=151888.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 06:19:12,921 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.2454, 4.4193, 2.8749, 4.9288, 3.2825, 4.8015, 3.1038, 3.5060], device='cuda:1'), covar=tensor([0.0186, 0.0259, 0.1307, 0.0149, 0.0716, 0.0393, 0.1161, 0.0574], device='cuda:1'), in_proj_covar=tensor([0.0152, 0.0162, 0.0183, 0.0136, 0.0165, 0.0198, 0.0193, 0.0168], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0003, 0.0004, 0.0004, 0.0003], device='cuda:1') 2023-04-30 06:19:18,823 INFO [train.py:904] (1/8) Epoch 15, batch 9800, loss[loss=0.1748, simple_loss=0.2739, pruned_loss=0.03787, over 16669.00 frames. ], tot_loss[loss=0.1733, simple_loss=0.267, pruned_loss=0.03982, over 3084916.64 frames. ], batch size: 134, lr: 4.47e-03, grad_scale: 8.0 2023-04-30 06:19:40,659 INFO [optim.py:368] (1/8) 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 06:20:30,613 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=151938.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 06:21:03,082 INFO [train.py:904] (1/8) Epoch 15, batch 9850, loss[loss=0.1899, simple_loss=0.2695, pruned_loss=0.05519, over 12682.00 frames. ], tot_loss[loss=0.173, simple_loss=0.2674, pruned_loss=0.03924, over 3078666.01 frames. ], batch size: 248, lr: 4.47e-03, grad_scale: 8.0 2023-04-30 06:21:08,480 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=151954.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 06:21:16,516 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=4.92 vs. limit=5.0 2023-04-30 06:21:24,062 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.5786, 3.6741, 3.4356, 3.1337, 3.2656, 3.5955, 3.3078, 3.3918], device='cuda:1'), covar=tensor([0.0578, 0.0534, 0.0268, 0.0241, 0.0505, 0.0386, 0.1255, 0.0436], device='cuda:1'), in_proj_covar=tensor([0.0249, 0.0339, 0.0292, 0.0276, 0.0302, 0.0320, 0.0201, 0.0344], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002], device='cuda:1') 2023-04-30 06:22:17,651 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 15, batch 9900, loss[loss=0.1725, simple_loss=0.2779, pruned_loss=0.03354, over 16781.00 frames. ], tot_loss[loss=0.1726, simple_loss=0.2673, pruned_loss=0.03895, over 3067027.67 frames. ], batch size: 76, lr: 4.46e-03, grad_scale: 8.0 2023-04-30 06:22:58,683 INFO [zipformer.py:625] (1/8) 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] (1/8) 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 06:23:50,280 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.85 vs. limit=2.0 2023-04-30 06:24:37,869 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=152044.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 06:24:55,577 INFO [train.py:904] (1/8) Epoch 15, batch 9950, loss[loss=0.1839, simple_loss=0.273, pruned_loss=0.04745, over 12282.00 frames. ], tot_loss[loss=0.1749, simple_loss=0.2702, pruned_loss=0.03977, over 3067537.75 frames. ], batch size: 247, lr: 4.46e-03, grad_scale: 8.0 2023-04-30 06:26:56,976 INFO [train.py:904] (1/8) Epoch 15, batch 10000, loss[loss=0.1518, simple_loss=0.2571, pruned_loss=0.02331, over 16898.00 frames. ], tot_loss[loss=0.1734, simple_loss=0.2687, pruned_loss=0.03907, over 3102185.34 frames. ], batch size: 96, lr: 4.46e-03, grad_scale: 8.0 2023-04-30 06:27:06,629 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.1802, 3.1134, 3.3615, 1.6957, 3.5044, 3.6108, 2.8526, 2.7753], device='cuda:1'), covar=tensor([0.0779, 0.0239, 0.0193, 0.1206, 0.0074, 0.0134, 0.0378, 0.0410], device='cuda:1'), in_proj_covar=tensor([0.0138, 0.0100, 0.0085, 0.0133, 0.0069, 0.0108, 0.0118, 0.0122], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0003, 0.0002, 0.0004, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:1') 2023-04-30 06:27:18,765 INFO [optim.py:368] (1/8) Clipping_scale=2.0, grad-norm quartiles 1.352e+02 2.235e+02 2.836e+02 3.494e+02 9.282e+02, threshold=5.672e+02, percent-clipped=5.0 2023-04-30 06:28:16,511 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([1.8943, 2.7678, 2.6850, 2.0281, 2.5333, 2.8235, 2.7331, 1.8771], device='cuda:1'), covar=tensor([0.0395, 0.0051, 0.0049, 0.0314, 0.0109, 0.0068, 0.0073, 0.0403], device='cuda:1'), in_proj_covar=tensor([0.0129, 0.0071, 0.0071, 0.0128, 0.0086, 0.0093, 0.0082, 0.0121], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:1') 2023-04-30 06:28:35,911 INFO [train.py:904] (1/8) Epoch 15, batch 10050, loss[loss=0.1787, simple_loss=0.2663, pruned_loss=0.0456, over 12271.00 frames. ], tot_loss[loss=0.1731, simple_loss=0.2687, pruned_loss=0.03876, over 3102925.21 frames. ], batch size: 246, lr: 4.46e-03, grad_scale: 4.0 2023-04-30 06:28:45,745 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=152157.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 06:29:12,566 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.35 vs. limit=2.0 2023-04-30 06:29:37,276 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.5854, 2.0530, 1.7063, 1.9094, 2.4074, 2.0825, 2.0932, 2.4964], device='cuda:1'), covar=tensor([0.0159, 0.0414, 0.0504, 0.0445, 0.0254, 0.0375, 0.0196, 0.0253], device='cuda:1'), in_proj_covar=tensor([0.0166, 0.0214, 0.0208, 0.0208, 0.0212, 0.0214, 0.0210, 0.0202], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-30 06:30:06,249 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.3643, 2.1950, 2.2648, 4.2367, 2.1203, 2.5955, 2.3054, 2.3928], device='cuda:1'), covar=tensor([0.1063, 0.3571, 0.2681, 0.0344, 0.4028, 0.2356, 0.3278, 0.3218], device='cuda:1'), in_proj_covar=tensor([0.0362, 0.0398, 0.0337, 0.0305, 0.0410, 0.0453, 0.0367, 0.0464], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-30 06:30:08,500 INFO [train.py:904] (1/8) Epoch 15, batch 10100, loss[loss=0.1994, simple_loss=0.2847, pruned_loss=0.05706, over 16927.00 frames. ], tot_loss[loss=0.1741, simple_loss=0.2694, pruned_loss=0.03937, over 3088552.48 frames. ], batch size: 109, lr: 4.46e-03, grad_scale: 4.0 2023-04-30 06:30:17,241 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.68 vs. limit=2.0 2023-04-30 06:30:28,184 INFO [optim.py:368] (1/8) 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 06:30:31,325 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.0837, 2.9955, 3.1730, 1.8128, 3.3419, 3.4183, 2.7572, 2.6637], device='cuda:1'), covar=tensor([0.0834, 0.0229, 0.0164, 0.1132, 0.0066, 0.0158, 0.0416, 0.0427], device='cuda:1'), in_proj_covar=tensor([0.0138, 0.0100, 0.0084, 0.0132, 0.0068, 0.0108, 0.0118, 0.0121], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0003, 0.0002, 0.0004, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:1') 2023-04-30 06:30:55,330 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=152225.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 06:30:56,874 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.4377, 3.3964, 3.4803, 3.5434, 3.5732, 3.2832, 3.5662, 3.6273], device='cuda:1'), covar=tensor([0.1207, 0.0943, 0.1000, 0.0645, 0.0621, 0.2078, 0.0825, 0.0752], device='cuda:1'), in_proj_covar=tensor([0.0529, 0.0660, 0.0776, 0.0673, 0.0507, 0.0526, 0.0539, 0.0623], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:1') 2023-04-30 06:31:53,404 INFO [train.py:904] (1/8) Epoch 16, batch 0, loss[loss=0.2639, simple_loss=0.341, pruned_loss=0.0934, over 12167.00 frames. ], tot_loss[loss=0.2639, simple_loss=0.341, pruned_loss=0.0934, over 12167.00 frames. ], batch size: 246, lr: 4.32e-03, grad_scale: 8.0 2023-04-30 06:31:53,405 INFO [train.py:929] (1/8) Computing validation loss 2023-04-30 06:32:00,887 INFO [train.py:938] (1/8) Epoch 16, validation: loss=0.1492, simple_loss=0.2525, pruned_loss=0.02288, over 944034.00 frames. 2023-04-30 06:32:00,888 INFO [train.py:939] (1/8) Maximum memory allocated so far is 17857MB 2023-04-30 06:32:10,191 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.12 vs. limit=2.0 2023-04-30 06:32:29,734 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=152273.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 06:32:29,861 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.2278, 3.4591, 3.5117, 2.2972, 2.8265, 2.4513, 3.6603, 3.6553], device='cuda:1'), covar=tensor([0.0262, 0.0758, 0.0618, 0.1689, 0.0922, 0.0919, 0.0567, 0.0829], device='cuda:1'), in_proj_covar=tensor([0.0147, 0.0145, 0.0157, 0.0145, 0.0137, 0.0124, 0.0137, 0.0156], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:1') 2023-04-30 06:32:48,532 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=152286.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 06:33:09,962 INFO [train.py:904] (1/8) Epoch 16, batch 50, loss[loss=0.1721, simple_loss=0.271, pruned_loss=0.03658, over 17115.00 frames. ], tot_loss[loss=0.1901, simple_loss=0.2754, pruned_loss=0.05239, over 754910.67 frames. ], batch size: 47, lr: 4.32e-03, grad_scale: 1.0 2023-04-30 06:33:29,876 INFO [optim.py:368] (1/8) 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 06:33:30,349 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.7959, 3.9218, 2.4688, 4.3583, 2.8777, 4.3581, 2.3843, 3.1060], device='cuda:1'), covar=tensor([0.0285, 0.0388, 0.1504, 0.0275, 0.0867, 0.0508, 0.1536, 0.0722], device='cuda:1'), in_proj_covar=tensor([0.0154, 0.0164, 0.0186, 0.0139, 0.0167, 0.0201, 0.0197, 0.0170], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0003, 0.0004, 0.0004, 0.0003], device='cuda:1') 2023-04-30 06:33:53,545 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=152334.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 06:34:07,525 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=152344.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 06:34:18,730 INFO [train.py:904] (1/8) Epoch 16, batch 100, loss[loss=0.2059, simple_loss=0.2844, pruned_loss=0.06369, over 15403.00 frames. ], tot_loss[loss=0.185, simple_loss=0.2709, pruned_loss=0.04955, over 1331464.82 frames. ], batch size: 190, lr: 4.32e-03, grad_scale: 1.0 2023-04-30 06:34:41,210 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=3.47 vs. limit=5.0 2023-04-30 06:35:14,869 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=152392.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 06:35:26,678 INFO [train.py:904] (1/8) Epoch 16, batch 150, loss[loss=0.1754, simple_loss=0.2546, pruned_loss=0.04812, over 15754.00 frames. ], tot_loss[loss=0.1828, simple_loss=0.2685, pruned_loss=0.04852, over 1766239.06 frames. ], batch size: 35, lr: 4.32e-03, grad_scale: 1.0 2023-04-30 06:35:48,057 INFO [optim.py:368] (1/8) Clipping_scale=2.0, grad-norm quartiles 1.712e+02 2.454e+02 2.817e+02 3.397e+02 1.160e+03, threshold=5.634e+02, percent-clipped=3.0 2023-04-30 06:36:24,982 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=152445.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 06:36:30,991 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([5.2372, 4.9205, 5.1742, 5.3459, 5.5608, 4.8479, 5.4758, 5.4957], device='cuda:1'), covar=tensor([0.1516, 0.1201, 0.1669, 0.0791, 0.0518, 0.0828, 0.0645, 0.0568], device='cuda:1'), in_proj_covar=tensor([0.0554, 0.0693, 0.0810, 0.0705, 0.0530, 0.0550, 0.0567, 0.0651], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-30 06:36:35,177 INFO [train.py:904] (1/8) Epoch 16, batch 200, loss[loss=0.1741, simple_loss=0.2621, pruned_loss=0.04303, over 17213.00 frames. ], tot_loss[loss=0.1814, simple_loss=0.2669, pruned_loss=0.04792, over 2120090.92 frames. ], batch size: 44, lr: 4.32e-03, grad_scale: 1.0 2023-04-30 06:36:39,817 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.3467, 3.3313, 2.0741, 3.4918, 2.6120, 3.5097, 2.1211, 2.7305], device='cuda:1'), covar=tensor([0.0274, 0.0472, 0.1565, 0.0336, 0.0811, 0.0807, 0.1441, 0.0710], device='cuda:1'), in_proj_covar=tensor([0.0156, 0.0167, 0.0188, 0.0142, 0.0170, 0.0205, 0.0199, 0.0174], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-04-30 06:36:42,909 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=152457.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 06:37:01,583 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.4686, 3.4659, 3.4707, 2.9314, 3.3677, 2.0466, 3.1454, 2.8997], device='cuda:1'), covar=tensor([0.0127, 0.0103, 0.0142, 0.0204, 0.0080, 0.2145, 0.0124, 0.0219], device='cuda:1'), in_proj_covar=tensor([0.0143, 0.0131, 0.0176, 0.0159, 0.0151, 0.0192, 0.0165, 0.0157], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-30 06:37:44,241 INFO [train.py:904] (1/8) Epoch 16, batch 250, loss[loss=0.184, simple_loss=0.2583, pruned_loss=0.05485, over 16534.00 frames. ], tot_loss[loss=0.1809, simple_loss=0.2659, pruned_loss=0.04801, over 2374735.78 frames. ], batch size: 75, lr: 4.31e-03, grad_scale: 1.0 2023-04-30 06:37:48,058 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=152505.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 06:37:49,466 INFO [zipformer.py:625] (1/8) 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] (1/8) 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] (1/8) Epoch 16, batch 300, loss[loss=0.179, simple_loss=0.2582, pruned_loss=0.04986, over 16467.00 frames. ], tot_loss[loss=0.1795, simple_loss=0.2638, pruned_loss=0.04754, over 2570694.32 frames. ], batch size: 75, lr: 4.31e-03, grad_scale: 1.0 2023-04-30 06:39:33,772 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=152581.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 06:39:35,000 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=152582.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 06:39:43,031 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=152589.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 06:39:56,500 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=4.93 vs. limit=5.0 2023-04-30 06:40:01,537 INFO [train.py:904] (1/8) Epoch 16, batch 350, loss[loss=0.1635, simple_loss=0.2451, pruned_loss=0.04099, over 17053.00 frames. ], tot_loss[loss=0.1768, simple_loss=0.2613, pruned_loss=0.04618, over 2743514.82 frames. ], batch size: 41, lr: 4.31e-03, grad_scale: 1.0 2023-04-30 06:40:12,477 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.7821, 5.0429, 5.2291, 5.0452, 5.0828, 5.6742, 5.1676, 4.8577], device='cuda:1'), covar=tensor([0.1251, 0.1894, 0.2100, 0.1938, 0.2707, 0.1091, 0.1496, 0.2387], device='cuda:1'), in_proj_covar=tensor([0.0381, 0.0540, 0.0589, 0.0455, 0.0603, 0.0627, 0.0463, 0.0602], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-04-30 06:40:20,730 INFO [optim.py:368] (1/8) Clipping_scale=2.0, grad-norm quartiles 1.536e+02 2.239e+02 2.608e+02 3.067e+02 7.666e+02, threshold=5.216e+02, percent-clipped=3.0 2023-04-30 06:40:38,379 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=152629.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 06:40:58,676 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=152643.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 06:41:06,965 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=152650.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 06:41:08,823 INFO [train.py:904] (1/8) Epoch 16, batch 400, loss[loss=0.1654, simple_loss=0.2603, pruned_loss=0.03519, over 17165.00 frames. ], tot_loss[loss=0.176, simple_loss=0.2598, pruned_loss=0.04613, over 2872866.11 frames. ], batch size: 46, lr: 4.31e-03, grad_scale: 2.0 2023-04-30 06:41:13,646 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.7538, 2.3032, 2.3073, 4.6306, 2.3358, 2.7805, 2.4294, 2.4903], device='cuda:1'), covar=tensor([0.1040, 0.3629, 0.2974, 0.0398, 0.4130, 0.2542, 0.3402, 0.3670], device='cuda:1'), in_proj_covar=tensor([0.0378, 0.0416, 0.0351, 0.0321, 0.0426, 0.0476, 0.0384, 0.0487], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-30 06:42:02,957 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.8056, 4.7971, 4.7034, 4.1320, 4.7156, 1.9522, 4.5055, 4.4220], device='cuda:1'), covar=tensor([0.0117, 0.0094, 0.0175, 0.0366, 0.0105, 0.2515, 0.0149, 0.0217], device='cuda:1'), in_proj_covar=tensor([0.0147, 0.0135, 0.0181, 0.0164, 0.0155, 0.0196, 0.0170, 0.0162], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-30 06:42:16,089 INFO [train.py:904] (1/8) Epoch 16, batch 450, loss[loss=0.1541, simple_loss=0.2348, pruned_loss=0.03666, over 16868.00 frames. ], tot_loss[loss=0.1751, simple_loss=0.2588, pruned_loss=0.04568, over 2974231.25 frames. ], batch size: 102, lr: 4.31e-03, grad_scale: 2.0 2023-04-30 06:42:36,231 INFO [optim.py:368] (1/8) 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 06:42:50,497 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=152726.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 06:43:25,828 INFO [train.py:904] (1/8) Epoch 16, batch 500, loss[loss=0.1532, simple_loss=0.2366, pruned_loss=0.03493, over 16804.00 frames. ], tot_loss[loss=0.1745, simple_loss=0.2583, pruned_loss=0.04538, over 3055626.15 frames. ], batch size: 39, lr: 4.31e-03, grad_scale: 2.0 2023-04-30 06:43:43,758 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.7013, 4.7053, 5.0868, 5.0925, 5.1231, 4.8047, 4.7315, 4.6144], device='cuda:1'), covar=tensor([0.0332, 0.0584, 0.0529, 0.0513, 0.0448, 0.0423, 0.0911, 0.0483], device='cuda:1'), in_proj_covar=tensor([0.0364, 0.0393, 0.0383, 0.0367, 0.0433, 0.0407, 0.0497, 0.0326], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:1') 2023-04-30 06:44:12,176 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.4710, 3.7025, 3.9075, 2.7414, 3.5594, 3.8872, 3.7329, 2.3387], device='cuda:1'), covar=tensor([0.0434, 0.0227, 0.0044, 0.0326, 0.0096, 0.0106, 0.0080, 0.0387], device='cuda:1'), in_proj_covar=tensor([0.0131, 0.0074, 0.0073, 0.0130, 0.0087, 0.0096, 0.0085, 0.0123], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-04-30 06:44:14,761 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=152787.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 06:44:22,456 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.0961, 3.3015, 2.8845, 5.1849, 4.3266, 4.5154, 1.7724, 3.3521], device='cuda:1'), covar=tensor([0.1197, 0.0633, 0.1111, 0.0165, 0.0219, 0.0382, 0.1507, 0.0678], device='cuda:1'), in_proj_covar=tensor([0.0159, 0.0164, 0.0186, 0.0165, 0.0194, 0.0209, 0.0189, 0.0184], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-04-30 06:44:24,661 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.7995, 3.9741, 2.2126, 4.4613, 3.0674, 4.3126, 2.1047, 3.1423], device='cuda:1'), covar=tensor([0.0268, 0.0325, 0.1814, 0.0270, 0.0721, 0.0425, 0.1919, 0.0684], device='cuda:1'), in_proj_covar=tensor([0.0158, 0.0169, 0.0190, 0.0147, 0.0171, 0.0209, 0.0200, 0.0175], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-04-30 06:44:27,213 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=3.14 vs. limit=5.0 2023-04-30 06:44:33,454 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=152801.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 06:44:34,215 INFO [train.py:904] (1/8) Epoch 16, batch 550, loss[loss=0.1867, simple_loss=0.2668, pruned_loss=0.05328, over 16507.00 frames. ], tot_loss[loss=0.1741, simple_loss=0.2576, pruned_loss=0.04534, over 3120525.96 frames. ], batch size: 68, lr: 4.31e-03, grad_scale: 2.0 2023-04-30 06:44:55,110 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.20 vs. limit=2.0 2023-04-30 06:44:55,529 INFO [optim.py:368] (1/8) 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 06:45:20,083 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.56 vs. limit=2.0 2023-04-30 06:45:46,473 INFO [train.py:904] (1/8) Epoch 16, batch 600, loss[loss=0.1511, simple_loss=0.2233, pruned_loss=0.03948, over 16532.00 frames. ], tot_loss[loss=0.1735, simple_loss=0.2565, pruned_loss=0.04529, over 3166922.46 frames. ], batch size: 75, lr: 4.31e-03, grad_scale: 2.0 2023-04-30 06:46:24,472 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=152881.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 06:46:41,210 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.2551, 3.3526, 3.7813, 2.0585, 3.0389, 2.4325, 3.7340, 3.5790], device='cuda:1'), covar=tensor([0.0268, 0.0997, 0.0504, 0.1962, 0.0861, 0.0935, 0.0586, 0.1044], device='cuda:1'), in_proj_covar=tensor([0.0148, 0.0152, 0.0160, 0.0147, 0.0139, 0.0125, 0.0139, 0.0162], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:1') 2023-04-30 06:46:53,568 INFO [train.py:904] (1/8) Epoch 16, batch 650, loss[loss=0.1504, simple_loss=0.2309, pruned_loss=0.03495, over 16572.00 frames. ], tot_loss[loss=0.1726, simple_loss=0.2552, pruned_loss=0.04501, over 3205957.05 frames. ], batch size: 75, lr: 4.31e-03, grad_scale: 2.0 2023-04-30 06:47:14,336 INFO [optim.py:368] (1/8) Clipping_scale=2.0, grad-norm quartiles 1.748e+02 2.133e+02 2.528e+02 3.105e+02 5.746e+02, threshold=5.056e+02, percent-clipped=1.0 2023-04-30 06:47:30,635 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=152929.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 06:47:30,681 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=152929.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 06:47:43,475 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=152938.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 06:47:52,080 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=152945.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 06:48:00,886 INFO [train.py:904] (1/8) Epoch 16, batch 700, loss[loss=0.1501, simple_loss=0.2352, pruned_loss=0.03253, over 15869.00 frames. ], tot_loss[loss=0.1718, simple_loss=0.2547, pruned_loss=0.04446, over 3232012.93 frames. ], batch size: 35, lr: 4.31e-03, grad_scale: 2.0 2023-04-30 06:48:37,711 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=152977.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 06:49:02,418 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.39 vs. limit=2.0 2023-04-30 06:49:10,743 INFO [train.py:904] (1/8) Epoch 16, batch 750, loss[loss=0.1669, simple_loss=0.2475, pruned_loss=0.04314, over 16237.00 frames. ], tot_loss[loss=0.1729, simple_loss=0.2562, pruned_loss=0.04483, over 3259190.23 frames. ], batch size: 165, lr: 4.31e-03, grad_scale: 2.0 2023-04-30 06:49:31,100 INFO [optim.py:368] (1/8) 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 06:50:17,790 INFO [train.py:904] (1/8) Epoch 16, batch 800, loss[loss=0.1845, simple_loss=0.258, pruned_loss=0.05547, over 16856.00 frames. ], tot_loss[loss=0.1723, simple_loss=0.2555, pruned_loss=0.04456, over 3269867.75 frames. ], batch size: 116, lr: 4.31e-03, grad_scale: 4.0 2023-04-30 06:50:18,567 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.31 vs. limit=2.0 2023-04-30 06:51:00,654 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=153082.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 06:51:24,848 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=153101.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 06:51:26,408 INFO [train.py:904] (1/8) Epoch 16, batch 850, loss[loss=0.1718, simple_loss=0.2646, pruned_loss=0.03954, over 17052.00 frames. ], tot_loss[loss=0.172, simple_loss=0.2553, pruned_loss=0.04438, over 3274161.31 frames. ], batch size: 55, lr: 4.31e-03, grad_scale: 4.0 2023-04-30 06:51:43,586 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.3636, 3.5072, 3.7773, 1.9583, 3.0867, 2.4935, 3.7954, 3.7585], device='cuda:1'), covar=tensor([0.0289, 0.0966, 0.0532, 0.2009, 0.0815, 0.0960, 0.0614, 0.0966], device='cuda:1'), in_proj_covar=tensor([0.0148, 0.0153, 0.0160, 0.0147, 0.0139, 0.0125, 0.0140, 0.0163], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:1') 2023-04-30 06:51:46,528 INFO [optim.py:368] (1/8) 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 06:51:47,001 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=153117.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 06:52:16,946 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=4.25 vs. limit=5.0 2023-04-30 06:52:20,151 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.44 vs. limit=2.0 2023-04-30 06:52:32,207 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=153149.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 06:52:35,389 INFO [train.py:904] (1/8) Epoch 16, batch 900, loss[loss=0.1765, simple_loss=0.2661, pruned_loss=0.0435, over 16774.00 frames. ], tot_loss[loss=0.1713, simple_loss=0.255, pruned_loss=0.04377, over 3273014.62 frames. ], batch size: 57, lr: 4.31e-03, grad_scale: 4.0 2023-04-30 06:52:56,732 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.28 vs. limit=2.0 2023-04-30 06:53:11,033 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=153178.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 06:53:36,537 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([5.1354, 5.0421, 4.9359, 4.4919, 4.5919, 5.0182, 4.9498, 4.6428], device='cuda:1'), covar=tensor([0.0598, 0.0571, 0.0317, 0.0344, 0.1121, 0.0452, 0.0348, 0.0808], device='cuda:1'), in_proj_covar=tensor([0.0280, 0.0383, 0.0327, 0.0314, 0.0344, 0.0362, 0.0223, 0.0392], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002], device='cuda:1') 2023-04-30 06:53:43,174 INFO [train.py:904] (1/8) Epoch 16, batch 950, loss[loss=0.167, simple_loss=0.2591, pruned_loss=0.0374, over 17245.00 frames. ], tot_loss[loss=0.172, simple_loss=0.2558, pruned_loss=0.04408, over 3261772.98 frames. ], batch size: 44, lr: 4.30e-03, grad_scale: 4.0 2023-04-30 06:54:04,604 INFO [optim.py:368] (1/8) 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 06:54:33,216 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=153238.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 06:54:34,322 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.9944, 5.0549, 5.4252, 5.4006, 5.4172, 5.1199, 5.0207, 4.8598], device='cuda:1'), covar=tensor([0.0295, 0.0483, 0.0371, 0.0445, 0.0456, 0.0346, 0.0842, 0.0406], device='cuda:1'), in_proj_covar=tensor([0.0375, 0.0406, 0.0394, 0.0378, 0.0446, 0.0419, 0.0511, 0.0333], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:1') 2023-04-30 06:54:42,484 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 16, batch 1000, loss[loss=0.1876, simple_loss=0.2567, pruned_loss=0.05924, over 16710.00 frames. ], tot_loss[loss=0.1714, simple_loss=0.2546, pruned_loss=0.04407, over 3266964.42 frames. ], batch size: 83, lr: 4.30e-03, grad_scale: 4.0 2023-04-30 06:55:29,854 INFO [zipformer.py:625] (1/8) 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] (1/8) 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] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=153293.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 06:56:02,850 INFO [train.py:904] (1/8) Epoch 16, batch 1050, loss[loss=0.1824, simple_loss=0.2655, pruned_loss=0.04963, over 16451.00 frames. ], tot_loss[loss=0.1707, simple_loss=0.2539, pruned_loss=0.04381, over 3285196.35 frames. ], batch size: 68, lr: 4.30e-03, grad_scale: 4.0 2023-04-30 06:56:24,069 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.9106, 2.0263, 2.4616, 2.8293, 2.6893, 3.2023, 2.0259, 3.2021], device='cuda:1'), covar=tensor([0.0197, 0.0407, 0.0268, 0.0289, 0.0271, 0.0186, 0.0449, 0.0140], device='cuda:1'), in_proj_covar=tensor([0.0175, 0.0184, 0.0170, 0.0173, 0.0182, 0.0140, 0.0185, 0.0130], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-30 06:56:24,685 INFO [optim.py:368] (1/8) 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 06:56:55,498 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=153339.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 06:56:55,585 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.9753, 4.5409, 3.1948, 2.4297, 2.8607, 2.6242, 4.8666, 3.8029], device='cuda:1'), covar=tensor([0.2644, 0.0530, 0.1683, 0.2562, 0.2773, 0.1949, 0.0320, 0.1230], device='cuda:1'), in_proj_covar=tensor([0.0313, 0.0263, 0.0294, 0.0291, 0.0282, 0.0237, 0.0278, 0.0315], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:1') 2023-04-30 06:57:12,821 INFO [train.py:904] (1/8) Epoch 16, batch 1100, loss[loss=0.1582, simple_loss=0.239, pruned_loss=0.03867, over 16445.00 frames. ], tot_loss[loss=0.1691, simple_loss=0.2526, pruned_loss=0.04281, over 3299455.25 frames. ], batch size: 75, lr: 4.30e-03, grad_scale: 4.0 2023-04-30 06:57:53,721 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 16, batch 1150, loss[loss=0.1751, simple_loss=0.2577, pruned_loss=0.04622, over 16393.00 frames. ], tot_loss[loss=0.1687, simple_loss=0.2526, pruned_loss=0.04246, over 3309579.04 frames. ], batch size: 75, lr: 4.30e-03, grad_scale: 4.0 2023-04-30 06:58:42,014 INFO [optim.py:368] (1/8) 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 06:58:46,830 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=153421.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 06:58:58,553 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 16, batch 1200, loss[loss=0.196, simple_loss=0.2595, pruned_loss=0.06618, over 16922.00 frames. ], tot_loss[loss=0.1685, simple_loss=0.252, pruned_loss=0.04254, over 3313714.31 frames. ], batch size: 96, lr: 4.30e-03, grad_scale: 8.0 2023-04-30 06:59:57,451 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=153473.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 07:00:10,716 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=153482.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 07:00:34,382 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.4066, 4.1506, 4.2673, 4.5938, 4.6396, 4.3094, 4.6394, 4.6880], device='cuda:1'), covar=tensor([0.1611, 0.1456, 0.1953, 0.0949, 0.0855, 0.1119, 0.1673, 0.0929], device='cuda:1'), in_proj_covar=tensor([0.0611, 0.0759, 0.0899, 0.0770, 0.0578, 0.0603, 0.0618, 0.0712], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-30 07:00:37,161 INFO [train.py:904] (1/8) Epoch 16, batch 1250, loss[loss=0.1695, simple_loss=0.2599, pruned_loss=0.0395, over 17135.00 frames. ], tot_loss[loss=0.17, simple_loss=0.2529, pruned_loss=0.04352, over 3319006.91 frames. ], batch size: 49, lr: 4.30e-03, grad_scale: 8.0 2023-04-30 07:00:57,395 INFO [optim.py:368] (1/8) 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 07:01:43,331 INFO [train.py:904] (1/8) Epoch 16, batch 1300, loss[loss=0.1749, simple_loss=0.2697, pruned_loss=0.04002, over 17052.00 frames. ], tot_loss[loss=0.1689, simple_loss=0.2523, pruned_loss=0.04278, over 3329545.95 frames. ], batch size: 53, lr: 4.30e-03, grad_scale: 8.0 2023-04-30 07:02:24,850 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.37 vs. limit=2.0 2023-04-30 07:02:44,872 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.7836, 3.7569, 4.2853, 1.8982, 4.3588, 4.5748, 3.2299, 3.5075], device='cuda:1'), covar=tensor([0.0755, 0.0245, 0.0221, 0.1287, 0.0088, 0.0142, 0.0426, 0.0384], device='cuda:1'), in_proj_covar=tensor([0.0146, 0.0106, 0.0093, 0.0139, 0.0074, 0.0118, 0.0125, 0.0129], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:1') 2023-04-30 07:02:52,311 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.83 vs. limit=2.0 2023-04-30 07:02:52,642 INFO [train.py:904] (1/8) Epoch 16, batch 1350, loss[loss=0.1773, simple_loss=0.2733, pruned_loss=0.04063, over 17076.00 frames. ], tot_loss[loss=0.1682, simple_loss=0.2522, pruned_loss=0.0421, over 3326762.03 frames. ], batch size: 53, lr: 4.30e-03, grad_scale: 8.0 2023-04-30 07:03:12,846 INFO [optim.py:368] (1/8) 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 07:03:37,023 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=153634.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 07:04:02,090 INFO [train.py:904] (1/8) Epoch 16, batch 1400, loss[loss=0.1348, simple_loss=0.2195, pruned_loss=0.02498, over 16758.00 frames. ], tot_loss[loss=0.1686, simple_loss=0.2523, pruned_loss=0.04243, over 3324733.21 frames. ], batch size: 39, lr: 4.30e-03, grad_scale: 8.0 2023-04-30 07:05:12,109 INFO [train.py:904] (1/8) Epoch 16, batch 1450, loss[loss=0.1578, simple_loss=0.2358, pruned_loss=0.0399, over 15507.00 frames. ], tot_loss[loss=0.1685, simple_loss=0.2522, pruned_loss=0.04239, over 3317895.49 frames. ], batch size: 190, lr: 4.30e-03, grad_scale: 4.0 2023-04-30 07:05:34,100 INFO [optim.py:368] (1/8) 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 07:06:11,408 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.0945, 4.2308, 2.8819, 4.7648, 3.3923, 4.7737, 2.9794, 3.6356], device='cuda:1'), covar=tensor([0.0243, 0.0284, 0.1290, 0.0220, 0.0688, 0.0352, 0.1147, 0.0543], device='cuda:1'), in_proj_covar=tensor([0.0162, 0.0172, 0.0193, 0.0152, 0.0174, 0.0214, 0.0202, 0.0178], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-04-30 07:06:22,460 INFO [train.py:904] (1/8) Epoch 16, batch 1500, loss[loss=0.1863, simple_loss=0.2574, pruned_loss=0.05765, over 16677.00 frames. ], tot_loss[loss=0.1685, simple_loss=0.252, pruned_loss=0.04252, over 3312420.41 frames. ], batch size: 89, lr: 4.30e-03, grad_scale: 4.0 2023-04-30 07:06:50,893 INFO [zipformer.py:625] (1/8) 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] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=153777.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 07:06:59,039 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.8885, 4.9136, 5.4707, 5.4057, 5.4183, 5.0165, 4.9641, 4.8363], device='cuda:1'), covar=tensor([0.0343, 0.0558, 0.0384, 0.0491, 0.0493, 0.0401, 0.1038, 0.0440], device='cuda:1'), in_proj_covar=tensor([0.0384, 0.0415, 0.0403, 0.0385, 0.0455, 0.0430, 0.0524, 0.0341], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:1') 2023-04-30 07:07:01,432 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.0485, 4.3813, 3.2820, 2.3029, 2.8705, 2.6042, 4.7015, 3.8327], device='cuda:1'), covar=tensor([0.2476, 0.0590, 0.1555, 0.2712, 0.2645, 0.1879, 0.0345, 0.1122], device='cuda:1'), in_proj_covar=tensor([0.0312, 0.0264, 0.0293, 0.0292, 0.0283, 0.0237, 0.0279, 0.0316], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:1') 2023-04-30 07:07:09,835 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.3573, 3.3690, 2.2036, 3.5863, 2.6781, 3.5489, 2.1858, 2.7328], device='cuda:1'), covar=tensor([0.0227, 0.0367, 0.1322, 0.0225, 0.0665, 0.0602, 0.1276, 0.0638], device='cuda:1'), in_proj_covar=tensor([0.0161, 0.0172, 0.0192, 0.0152, 0.0173, 0.0214, 0.0201, 0.0177], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-04-30 07:07:24,805 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.20 vs. limit=2.0 2023-04-30 07:07:30,670 INFO [train.py:904] (1/8) Epoch 16, batch 1550, loss[loss=0.1767, simple_loss=0.2403, pruned_loss=0.05652, over 16796.00 frames. ], tot_loss[loss=0.1699, simple_loss=0.2531, pruned_loss=0.04335, over 3314275.79 frames. ], batch size: 96, lr: 4.30e-03, grad_scale: 4.0 2023-04-30 07:07:52,793 INFO [scaling.py:679] (1/8) 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] (1/8) 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] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=153821.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 07:08:39,725 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.4850, 3.2251, 2.6429, 2.0648, 2.2273, 2.2121, 3.3394, 3.0192], device='cuda:1'), covar=tensor([0.2627, 0.0775, 0.1628, 0.2554, 0.2346, 0.1966, 0.0563, 0.1271], device='cuda:1'), in_proj_covar=tensor([0.0312, 0.0263, 0.0293, 0.0291, 0.0282, 0.0237, 0.0279, 0.0315], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:1') 2023-04-30 07:08:40,378 INFO [train.py:904] (1/8) Epoch 16, batch 1600, loss[loss=0.2027, simple_loss=0.2868, pruned_loss=0.0593, over 15620.00 frames. ], tot_loss[loss=0.172, simple_loss=0.2555, pruned_loss=0.04422, over 3308363.29 frames. ], batch size: 191, lr: 4.30e-03, grad_scale: 8.0 2023-04-30 07:09:07,633 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.2249, 4.1868, 4.6336, 2.4738, 4.8008, 4.8706, 3.5697, 3.9017], device='cuda:1'), covar=tensor([0.0614, 0.0200, 0.0191, 0.1038, 0.0060, 0.0121, 0.0367, 0.0306], device='cuda:1'), in_proj_covar=tensor([0.0144, 0.0105, 0.0092, 0.0137, 0.0073, 0.0117, 0.0124, 0.0127], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:1') 2023-04-30 07:09:47,765 INFO [train.py:904] (1/8) Epoch 16, batch 1650, loss[loss=0.2099, simple_loss=0.3031, pruned_loss=0.05833, over 15515.00 frames. ], tot_loss[loss=0.1741, simple_loss=0.2576, pruned_loss=0.04523, over 3306539.58 frames. ], batch size: 190, lr: 4.29e-03, grad_scale: 8.0 2023-04-30 07:10:09,060 INFO [optim.py:368] (1/8) 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 07:10:15,992 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=153923.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 07:10:17,456 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.71 vs. limit=2.0 2023-04-30 07:10:32,363 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=153934.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 07:10:46,556 INFO [scaling.py:679] (1/8) 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] (1/8) Epoch 16, batch 1700, loss[loss=0.1737, simple_loss=0.2656, pruned_loss=0.04088, over 16737.00 frames. ], tot_loss[loss=0.1747, simple_loss=0.2584, pruned_loss=0.04554, over 3310070.24 frames. ], batch size: 62, lr: 4.29e-03, grad_scale: 8.0 2023-04-30 07:11:38,410 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=153982.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 07:11:40,889 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=153984.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 07:12:09,296 INFO [train.py:904] (1/8) Epoch 16, batch 1750, loss[loss=0.1659, simple_loss=0.2469, pruned_loss=0.04245, over 16857.00 frames. ], tot_loss[loss=0.1756, simple_loss=0.2595, pruned_loss=0.04578, over 3309956.97 frames. ], batch size: 102, lr: 4.29e-03, grad_scale: 8.0 2023-04-30 07:12:33,137 INFO [optim.py:368] (1/8) 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 07:13:15,095 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=154049.0, num_to_drop=1, layers_to_drop={1} 2023-04-30 07:13:18,924 INFO [train.py:904] (1/8) Epoch 16, batch 1800, loss[loss=0.1883, simple_loss=0.2645, pruned_loss=0.05601, over 16778.00 frames. ], tot_loss[loss=0.1765, simple_loss=0.2605, pruned_loss=0.04629, over 3304569.57 frames. ], batch size: 83, lr: 4.29e-03, grad_scale: 8.0 2023-04-30 07:13:52,623 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=154076.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 07:13:54,276 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=154077.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 07:14:03,066 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=154084.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 07:14:28,054 INFO [train.py:904] (1/8) Epoch 16, batch 1850, loss[loss=0.1395, simple_loss=0.2312, pruned_loss=0.02384, over 15973.00 frames. ], tot_loss[loss=0.1772, simple_loss=0.2615, pruned_loss=0.04644, over 3306003.46 frames. ], batch size: 35, lr: 4.29e-03, grad_scale: 8.0 2023-04-30 07:14:38,641 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=154110.0, num_to_drop=1, layers_to_drop={0} 2023-04-30 07:14:50,217 INFO [optim.py:368] (1/8) 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] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=154125.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 07:15:15,165 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=154137.0, num_to_drop=1, layers_to_drop={2} 2023-04-30 07:15:25,835 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 16, batch 1900, loss[loss=0.1751, simple_loss=0.254, pruned_loss=0.04817, over 16736.00 frames. ], tot_loss[loss=0.1759, simple_loss=0.2603, pruned_loss=0.04574, over 3315195.73 frames. ], batch size: 124, lr: 4.29e-03, grad_scale: 8.0 2023-04-30 07:15:55,671 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([5.0857, 5.6859, 5.7870, 5.5441, 5.5991, 6.1084, 5.6734, 5.4003], device='cuda:1'), covar=tensor([0.0932, 0.1714, 0.2292, 0.2062, 0.2867, 0.1029, 0.1376, 0.2472], device='cuda:1'), in_proj_covar=tensor([0.0394, 0.0564, 0.0611, 0.0472, 0.0632, 0.0645, 0.0479, 0.0626], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-04-30 07:16:41,950 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.7591, 3.6047, 3.6427, 3.9332, 3.9624, 3.6287, 3.8520, 4.0130], device='cuda:1'), covar=tensor([0.1572, 0.1374, 0.1876, 0.0857, 0.0823, 0.1972, 0.2056, 0.0906], device='cuda:1'), in_proj_covar=tensor([0.0616, 0.0763, 0.0906, 0.0772, 0.0579, 0.0603, 0.0615, 0.0714], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-30 07:16:44,644 INFO [train.py:904] (1/8) Epoch 16, batch 1950, loss[loss=0.1814, simple_loss=0.2793, pruned_loss=0.04177, over 16730.00 frames. ], tot_loss[loss=0.176, simple_loss=0.2612, pruned_loss=0.04539, over 3304618.64 frames. ], batch size: 62, lr: 4.29e-03, grad_scale: 8.0 2023-04-30 07:17:04,523 INFO [optim.py:368] (1/8) 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 07:17:09,830 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=154221.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 07:17:51,474 INFO [train.py:904] (1/8) Epoch 16, batch 2000, loss[loss=0.1554, simple_loss=0.2398, pruned_loss=0.03548, over 16856.00 frames. ], tot_loss[loss=0.1748, simple_loss=0.2604, pruned_loss=0.04465, over 3314678.95 frames. ], batch size: 42, lr: 4.29e-03, grad_scale: 8.0 2023-04-30 07:18:27,772 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=154279.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 07:18:32,725 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=154282.0, num_to_drop=1, layers_to_drop={0} 2023-04-30 07:18:38,743 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=154286.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 07:18:44,477 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 16, batch 2050, loss[loss=0.192, simple_loss=0.2749, pruned_loss=0.05453, over 16835.00 frames. ], tot_loss[loss=0.1759, simple_loss=0.2612, pruned_loss=0.04532, over 3308068.77 frames. ], batch size: 96, lr: 4.29e-03, grad_scale: 8.0 2023-04-30 07:19:17,353 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.11 vs. limit=2.0 2023-04-30 07:19:19,609 INFO [optim.py:368] (1/8) 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 07:19:31,182 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.35 vs. limit=2.0 2023-04-30 07:19:46,259 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=154337.0, num_to_drop=1, layers_to_drop={1} 2023-04-30 07:20:01,016 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=154347.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 07:20:06,439 INFO [train.py:904] (1/8) Epoch 16, batch 2100, loss[loss=0.1886, simple_loss=0.2716, pruned_loss=0.05279, over 16753.00 frames. ], tot_loss[loss=0.1777, simple_loss=0.263, pruned_loss=0.0462, over 3308424.75 frames. ], batch size: 124, lr: 4.29e-03, grad_scale: 8.0 2023-04-30 07:20:06,876 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=154352.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 07:20:08,284 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.5602, 2.5697, 2.2787, 2.3961, 2.8599, 2.6695, 3.2257, 3.1429], device='cuda:1'), covar=tensor([0.0124, 0.0352, 0.0432, 0.0419, 0.0258, 0.0340, 0.0226, 0.0225], device='cuda:1'), in_proj_covar=tensor([0.0187, 0.0225, 0.0218, 0.0218, 0.0227, 0.0229, 0.0232, 0.0223], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-30 07:20:46,374 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.25 vs. limit=2.0 2023-04-30 07:21:09,659 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=154398.0, num_to_drop=1, layers_to_drop={0} 2023-04-30 07:21:14,694 INFO [train.py:904] (1/8) Epoch 16, batch 2150, loss[loss=0.1682, simple_loss=0.2643, pruned_loss=0.03607, over 17252.00 frames. ], tot_loss[loss=0.1784, simple_loss=0.263, pruned_loss=0.04686, over 3305348.27 frames. ], batch size: 52, lr: 4.29e-03, grad_scale: 8.0 2023-04-30 07:21:18,544 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=154405.0, num_to_drop=1, layers_to_drop={1} 2023-04-30 07:21:36,517 INFO [optim.py:368] (1/8) 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] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=154432.0, num_to_drop=1, layers_to_drop={1} 2023-04-30 07:21:58,105 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.5360, 2.2709, 2.2876, 4.3431, 2.1567, 2.6844, 2.3078, 2.4621], device='cuda:1'), covar=tensor([0.1156, 0.3591, 0.2849, 0.0478, 0.4226, 0.2615, 0.3438, 0.3623], device='cuda:1'), in_proj_covar=tensor([0.0384, 0.0421, 0.0352, 0.0324, 0.0425, 0.0484, 0.0390, 0.0493], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-30 07:22:07,247 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=154440.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 07:22:25,080 INFO [train.py:904] (1/8) Epoch 16, batch 2200, loss[loss=0.1494, simple_loss=0.2373, pruned_loss=0.03072, over 17210.00 frames. ], tot_loss[loss=0.1792, simple_loss=0.2639, pruned_loss=0.04731, over 3307950.78 frames. ], batch size: 44, lr: 4.29e-03, grad_scale: 8.0 2023-04-30 07:22:55,729 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=154474.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 07:23:32,242 INFO [train.py:904] (1/8) Epoch 16, batch 2250, loss[loss=0.1795, simple_loss=0.254, pruned_loss=0.05245, over 16810.00 frames. ], tot_loss[loss=0.1802, simple_loss=0.2645, pruned_loss=0.04794, over 3303274.76 frames. ], batch size: 96, lr: 4.29e-03, grad_scale: 8.0 2023-04-30 07:23:55,133 INFO [optim.py:368] (1/8) 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 07:24:18,258 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=154535.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 07:24:25,942 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.16 vs. limit=2.0 2023-04-30 07:24:31,554 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([5.0626, 4.8114, 5.0559, 5.2504, 5.5065, 4.7654, 5.4512, 5.4739], device='cuda:1'), covar=tensor([0.1745, 0.1446, 0.1693, 0.0808, 0.0472, 0.0800, 0.0538, 0.0530], device='cuda:1'), in_proj_covar=tensor([0.0613, 0.0762, 0.0904, 0.0769, 0.0577, 0.0603, 0.0614, 0.0714], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-30 07:24:40,019 INFO [train.py:904] (1/8) Epoch 16, batch 2300, loss[loss=0.1608, simple_loss=0.2533, pruned_loss=0.03417, over 17174.00 frames. ], tot_loss[loss=0.1792, simple_loss=0.2638, pruned_loss=0.04733, over 3304063.45 frames. ], batch size: 46, lr: 4.29e-03, grad_scale: 8.0 2023-04-30 07:25:16,140 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=154577.0, num_to_drop=1, layers_to_drop={0} 2023-04-30 07:25:20,042 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=154579.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 07:25:51,249 INFO [train.py:904] (1/8) Epoch 16, batch 2350, loss[loss=0.2212, simple_loss=0.3081, pruned_loss=0.06716, over 11983.00 frames. ], tot_loss[loss=0.1789, simple_loss=0.2639, pruned_loss=0.04694, over 3309848.91 frames. ], batch size: 246, lr: 4.29e-03, grad_scale: 8.0 2023-04-30 07:26:11,866 INFO [optim.py:368] (1/8) 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] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=154627.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 07:26:26,627 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=154628.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 07:26:45,806 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=154642.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 07:26:52,305 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=154647.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 07:26:57,937 INFO [train.py:904] (1/8) Epoch 16, batch 2400, loss[loss=0.1611, simple_loss=0.2425, pruned_loss=0.03986, over 16877.00 frames. ], tot_loss[loss=0.1797, simple_loss=0.2647, pruned_loss=0.04738, over 3307721.18 frames. ], batch size: 102, lr: 4.28e-03, grad_scale: 8.0 2023-04-30 07:27:50,357 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=154689.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 07:27:55,501 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=154693.0, num_to_drop=1, layers_to_drop={1} 2023-04-30 07:28:06,455 INFO [train.py:904] (1/8) Epoch 16, batch 2450, loss[loss=0.1563, simple_loss=0.2368, pruned_loss=0.03791, over 16868.00 frames. ], tot_loss[loss=0.1784, simple_loss=0.264, pruned_loss=0.04636, over 3311225.59 frames. ], batch size: 96, lr: 4.28e-03, grad_scale: 8.0 2023-04-30 07:28:11,474 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=154705.0, num_to_drop=1, layers_to_drop={1} 2023-04-30 07:28:28,918 INFO [optim.py:368] (1/8) 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 07:28:31,781 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=154720.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 07:28:49,657 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=154732.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 07:28:59,814 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=154740.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 07:29:16,072 INFO [train.py:904] (1/8) Epoch 16, batch 2500, loss[loss=0.1973, simple_loss=0.2803, pruned_loss=0.05712, over 16822.00 frames. ], tot_loss[loss=0.1788, simple_loss=0.2642, pruned_loss=0.04667, over 3293103.46 frames. ], batch size: 83, lr: 4.28e-03, grad_scale: 8.0 2023-04-30 07:29:18,086 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=154753.0, num_to_drop=1, layers_to_drop={0} 2023-04-30 07:29:53,783 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=154780.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 07:29:54,291 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.37 vs. limit=2.0 2023-04-30 07:29:55,099 INFO [zipformer.py:625] (1/8) 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] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=154788.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 07:30:24,149 INFO [train.py:904] (1/8) Epoch 16, batch 2550, loss[loss=0.1907, simple_loss=0.2796, pruned_loss=0.05085, over 16418.00 frames. ], tot_loss[loss=0.1786, simple_loss=0.264, pruned_loss=0.04659, over 3300240.71 frames. ], batch size: 68, lr: 4.28e-03, grad_scale: 8.0 2023-04-30 07:30:47,016 INFO [optim.py:368] (1/8) 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 07:31:02,556 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=154830.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 07:31:32,684 INFO [train.py:904] (1/8) Epoch 16, batch 2600, loss[loss=0.1967, simple_loss=0.2917, pruned_loss=0.05089, over 16717.00 frames. ], tot_loss[loss=0.1776, simple_loss=0.2634, pruned_loss=0.04594, over 3312133.57 frames. ], batch size: 57, lr: 4.28e-03, grad_scale: 8.0 2023-04-30 07:32:07,857 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=154877.0, num_to_drop=1, layers_to_drop={0} 2023-04-30 07:32:28,400 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.6342, 2.4678, 1.9194, 2.1374, 2.8837, 2.6244, 3.3947, 3.1876], device='cuda:1'), covar=tensor([0.0164, 0.0550, 0.0658, 0.0544, 0.0293, 0.0414, 0.0265, 0.0277], device='cuda:1'), in_proj_covar=tensor([0.0188, 0.0225, 0.0218, 0.0218, 0.0227, 0.0228, 0.0233, 0.0223], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-30 07:32:31,481 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.59 vs. limit=2.0 2023-04-30 07:32:40,429 INFO [train.py:904] (1/8) Epoch 16, batch 2650, loss[loss=0.1933, simple_loss=0.2744, pruned_loss=0.05611, over 16544.00 frames. ], tot_loss[loss=0.1773, simple_loss=0.2638, pruned_loss=0.0454, over 3319261.97 frames. ], batch size: 146, lr: 4.28e-03, grad_scale: 8.0 2023-04-30 07:32:48,568 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.1682, 4.0008, 4.2213, 4.3501, 4.4591, 4.0283, 4.2392, 4.4559], device='cuda:1'), covar=tensor([0.1571, 0.1140, 0.1295, 0.0700, 0.0669, 0.1328, 0.1903, 0.0814], device='cuda:1'), in_proj_covar=tensor([0.0618, 0.0769, 0.0913, 0.0777, 0.0584, 0.0615, 0.0619, 0.0723], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-30 07:33:01,394 INFO [optim.py:368] (1/8) 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 07:33:12,798 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=154925.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 07:33:32,274 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.29 vs. limit=2.0 2023-04-30 07:33:35,203 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=154942.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 07:33:37,478 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.34 vs. limit=2.0 2023-04-30 07:33:43,614 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 16, batch 2700, loss[loss=0.153, simple_loss=0.2466, pruned_loss=0.02968, over 17240.00 frames. ], tot_loss[loss=0.1775, simple_loss=0.2646, pruned_loss=0.04518, over 3324754.34 frames. ], batch size: 45, lr: 4.28e-03, grad_scale: 8.0 2023-04-30 07:34:21,645 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=154976.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 07:34:31,999 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=154984.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 07:34:34,846 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=4.00 vs. limit=5.0 2023-04-30 07:34:41,399 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=154990.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 07:34:45,802 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=154993.0, num_to_drop=1, layers_to_drop={1} 2023-04-30 07:34:47,946 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=154995.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 07:34:56,796 INFO [train.py:904] (1/8) Epoch 16, batch 2750, loss[loss=0.1964, simple_loss=0.2851, pruned_loss=0.05382, over 17239.00 frames. ], tot_loss[loss=0.177, simple_loss=0.2645, pruned_loss=0.04477, over 3335776.39 frames. ], batch size: 52, lr: 4.28e-03, grad_scale: 8.0 2023-04-30 07:35:18,259 INFO [optim.py:368] (1/8) 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 07:35:46,119 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=155037.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 07:35:47,457 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=3.52 vs. limit=5.0 2023-04-30 07:35:50,509 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=155041.0, num_to_drop=1, layers_to_drop={1} 2023-04-30 07:36:02,155 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.7398, 1.8366, 2.2380, 2.6679, 2.7011, 2.6915, 1.9411, 2.9024], device='cuda:1'), covar=tensor([0.0144, 0.0432, 0.0309, 0.0243, 0.0231, 0.0223, 0.0427, 0.0116], device='cuda:1'), in_proj_covar=tensor([0.0179, 0.0186, 0.0172, 0.0176, 0.0186, 0.0142, 0.0185, 0.0135], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-30 07:36:03,284 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.4207, 3.3842, 3.6479, 2.6936, 3.3469, 3.7578, 3.5139, 2.2326], device='cuda:1'), covar=tensor([0.0439, 0.0165, 0.0048, 0.0314, 0.0090, 0.0082, 0.0083, 0.0404], device='cuda:1'), in_proj_covar=tensor([0.0134, 0.0078, 0.0077, 0.0133, 0.0090, 0.0100, 0.0089, 0.0127], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-04-30 07:36:04,755 INFO [train.py:904] (1/8) Epoch 16, batch 2800, loss[loss=0.1874, simple_loss=0.2642, pruned_loss=0.05531, over 16844.00 frames. ], tot_loss[loss=0.1767, simple_loss=0.2641, pruned_loss=0.0446, over 3330297.03 frames. ], batch size: 116, lr: 4.28e-03, grad_scale: 8.0 2023-04-30 07:36:28,414 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=155069.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 07:36:38,303 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=155076.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 07:36:59,227 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.20 vs. limit=2.0 2023-04-30 07:37:15,726 INFO [train.py:904] (1/8) Epoch 16, batch 2850, loss[loss=0.1551, simple_loss=0.2441, pruned_loss=0.03303, over 15893.00 frames. ], tot_loss[loss=0.1765, simple_loss=0.2635, pruned_loss=0.04473, over 3319609.88 frames. ], batch size: 35, lr: 4.28e-03, grad_scale: 8.0 2023-04-30 07:37:36,420 INFO [optim.py:368] (1/8) Clipping_scale=2.0, grad-norm quartiles 1.487e+02 2.238e+02 2.628e+02 3.275e+02 9.061e+02, threshold=5.256e+02, percent-clipped=3.0 2023-04-30 07:37:43,264 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.41 vs. limit=2.0 2023-04-30 07:37:54,417 INFO [zipformer.py:625] (1/8) 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] (1/8) 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] (1/8) Epoch 16, batch 2900, loss[loss=0.1599, simple_loss=0.2398, pruned_loss=0.04001, over 16718.00 frames. ], tot_loss[loss=0.1768, simple_loss=0.2629, pruned_loss=0.04539, over 3318824.75 frames. ], batch size: 89, lr: 4.28e-03, grad_scale: 8.0 2023-04-30 07:38:55,026 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.7961, 3.8941, 2.7087, 4.4983, 3.0723, 4.4509, 2.4992, 3.1489], device='cuda:1'), covar=tensor([0.0278, 0.0371, 0.1296, 0.0219, 0.0761, 0.0461, 0.1468, 0.0718], device='cuda:1'), in_proj_covar=tensor([0.0164, 0.0177, 0.0194, 0.0156, 0.0175, 0.0219, 0.0204, 0.0180], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-04-30 07:39:00,496 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=155178.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 07:39:33,218 INFO [train.py:904] (1/8) Epoch 16, batch 2950, loss[loss=0.207, simple_loss=0.2866, pruned_loss=0.06367, over 15352.00 frames. ], tot_loss[loss=0.1759, simple_loss=0.2612, pruned_loss=0.04531, over 3323419.66 frames. ], batch size: 190, lr: 4.28e-03, grad_scale: 8.0 2023-04-30 07:39:54,114 INFO [optim.py:368] (1/8) 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 07:40:40,759 INFO [train.py:904] (1/8) Epoch 16, batch 3000, loss[loss=0.1815, simple_loss=0.2683, pruned_loss=0.04739, over 17043.00 frames. ], tot_loss[loss=0.1768, simple_loss=0.2619, pruned_loss=0.04584, over 3330776.44 frames. ], batch size: 50, lr: 4.28e-03, grad_scale: 8.0 2023-04-30 07:40:40,759 INFO [train.py:929] (1/8) Computing validation loss 2023-04-30 07:40:47,342 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.8897, 4.9147, 5.2567, 5.2290, 5.2049, 4.9385, 4.8525, 4.7306], device='cuda:1'), covar=tensor([0.0294, 0.0390, 0.0360, 0.0360, 0.0336, 0.0310, 0.0830, 0.0358], device='cuda:1'), in_proj_covar=tensor([0.0389, 0.0421, 0.0409, 0.0387, 0.0460, 0.0436, 0.0532, 0.0343], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:1') 2023-04-30 07:40:49,853 INFO [train.py:938] (1/8) Epoch 16, validation: loss=0.1368, simple_loss=0.2429, pruned_loss=0.01541, over 944034.00 frames. 2023-04-30 07:40:49,854 INFO [train.py:939] (1/8) Maximum memory allocated so far is 17857MB 2023-04-30 07:41:03,223 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.7577, 5.0678, 4.8117, 4.8610, 4.5974, 4.5753, 4.4698, 5.1621], device='cuda:1'), covar=tensor([0.1070, 0.0833, 0.1005, 0.0765, 0.0903, 0.1100, 0.1071, 0.0838], device='cuda:1'), in_proj_covar=tensor([0.0641, 0.0791, 0.0644, 0.0570, 0.0499, 0.0505, 0.0658, 0.0606], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-30 07:41:34,226 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 16, batch 3050, loss[loss=0.1767, simple_loss=0.2607, pruned_loss=0.04635, over 16451.00 frames. ], tot_loss[loss=0.1772, simple_loss=0.2624, pruned_loss=0.04602, over 3324538.63 frames. ], batch size: 68, lr: 4.28e-03, grad_scale: 8.0 2023-04-30 07:42:21,041 INFO [optim.py:368] (1/8) 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 07:42:23,347 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.4608, 3.5831, 3.8656, 2.3791, 3.1977, 2.2516, 3.7887, 3.7719], device='cuda:1'), covar=tensor([0.0252, 0.0878, 0.0528, 0.1766, 0.0810, 0.1080, 0.0634, 0.1026], device='cuda:1'), in_proj_covar=tensor([0.0150, 0.0156, 0.0161, 0.0148, 0.0139, 0.0126, 0.0141, 0.0168], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:1') 2023-04-30 07:42:42,355 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=155332.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 07:42:42,368 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 16, batch 3100, loss[loss=0.1595, simple_loss=0.2514, pruned_loss=0.03383, over 17107.00 frames. ], tot_loss[loss=0.1782, simple_loss=0.2632, pruned_loss=0.04662, over 3324994.28 frames. ], batch size: 48, lr: 4.27e-03, grad_scale: 8.0 2023-04-30 07:43:42,801 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=155376.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 07:43:59,308 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.8557, 3.8406, 3.0463, 2.2625, 2.5458, 2.3704, 3.9844, 3.4711], device='cuda:1'), covar=tensor([0.2232, 0.0555, 0.1448, 0.2708, 0.2564, 0.1916, 0.0431, 0.1165], device='cuda:1'), in_proj_covar=tensor([0.0312, 0.0264, 0.0293, 0.0293, 0.0287, 0.0239, 0.0279, 0.0320], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:1') 2023-04-30 07:44:17,693 INFO [train.py:904] (1/8) Epoch 16, batch 3150, loss[loss=0.1929, simple_loss=0.2828, pruned_loss=0.05153, over 16628.00 frames. ], tot_loss[loss=0.1779, simple_loss=0.2627, pruned_loss=0.04656, over 3328165.41 frames. ], batch size: 62, lr: 4.27e-03, grad_scale: 8.0 2023-04-30 07:44:29,344 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.9614, 3.8600, 4.3123, 2.1935, 4.4863, 4.5163, 3.2599, 3.4942], device='cuda:1'), covar=tensor([0.0665, 0.0194, 0.0176, 0.1054, 0.0054, 0.0159, 0.0382, 0.0341], device='cuda:1'), in_proj_covar=tensor([0.0146, 0.0105, 0.0093, 0.0138, 0.0074, 0.0119, 0.0125, 0.0128], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:1') 2023-04-30 07:44:39,888 INFO [optim.py:368] (1/8) 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 07:44:43,585 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.8517, 2.3504, 2.5046, 4.6381, 2.4010, 2.7985, 2.5683, 2.6221], device='cuda:1'), covar=tensor([0.0988, 0.3551, 0.2618, 0.0407, 0.3758, 0.2410, 0.3148, 0.3508], device='cuda:1'), in_proj_covar=tensor([0.0386, 0.0423, 0.0352, 0.0325, 0.0425, 0.0487, 0.0391, 0.0495], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-30 07:44:47,893 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=155424.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 07:44:49,176 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([5.6148, 6.0231, 5.7544, 5.8521, 5.4618, 5.3428, 5.4359, 6.1838], device='cuda:1'), covar=tensor([0.1165, 0.0808, 0.0926, 0.0739, 0.0795, 0.0731, 0.1046, 0.0905], device='cuda:1'), in_proj_covar=tensor([0.0646, 0.0799, 0.0649, 0.0576, 0.0502, 0.0510, 0.0666, 0.0613], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-30 07:44:49,180 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=155425.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 07:45:00,337 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.32 vs. limit=2.0 2023-04-30 07:45:27,253 INFO [train.py:904] (1/8) Epoch 16, batch 3200, loss[loss=0.1893, simple_loss=0.2799, pruned_loss=0.04937, over 17089.00 frames. ], tot_loss[loss=0.1773, simple_loss=0.2623, pruned_loss=0.04612, over 3325584.09 frames. ], batch size: 53, lr: 4.27e-03, grad_scale: 8.0 2023-04-30 07:45:33,540 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=155457.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 07:46:31,579 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.5544, 4.9803, 4.4729, 4.8174, 4.5549, 4.4899, 4.5087, 5.0492], device='cuda:1'), covar=tensor([0.2436, 0.1713, 0.2747, 0.1477, 0.1703, 0.2053, 0.2241, 0.2031], device='cuda:1'), in_proj_covar=tensor([0.0649, 0.0803, 0.0650, 0.0579, 0.0504, 0.0512, 0.0668, 0.0616], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-30 07:46:36,135 INFO [train.py:904] (1/8) Epoch 16, batch 3250, loss[loss=0.2168, simple_loss=0.2961, pruned_loss=0.0688, over 12181.00 frames. ], tot_loss[loss=0.1762, simple_loss=0.2615, pruned_loss=0.04549, over 3325022.65 frames. ], batch size: 246, lr: 4.27e-03, grad_scale: 8.0 2023-04-30 07:46:58,461 INFO [optim.py:368] (1/8) 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 07:46:58,913 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=155518.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 07:47:45,922 INFO [train.py:904] (1/8) Epoch 16, batch 3300, loss[loss=0.1921, simple_loss=0.2705, pruned_loss=0.05682, over 16185.00 frames. ], tot_loss[loss=0.1761, simple_loss=0.2614, pruned_loss=0.04539, over 3314136.48 frames. ], batch size: 164, lr: 4.27e-03, grad_scale: 8.0 2023-04-30 07:48:21,132 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([1.8455, 2.9737, 2.6364, 4.4595, 3.6661, 4.2653, 1.8057, 3.0005], device='cuda:1'), covar=tensor([0.1334, 0.0692, 0.1117, 0.0169, 0.0267, 0.0379, 0.1425, 0.0813], device='cuda:1'), in_proj_covar=tensor([0.0159, 0.0167, 0.0188, 0.0175, 0.0202, 0.0214, 0.0189, 0.0186], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-04-30 07:48:56,952 INFO [train.py:904] (1/8) Epoch 16, batch 3350, loss[loss=0.1622, simple_loss=0.244, pruned_loss=0.04016, over 16830.00 frames. ], tot_loss[loss=0.1753, simple_loss=0.2611, pruned_loss=0.04481, over 3319493.47 frames. ], batch size: 102, lr: 4.27e-03, grad_scale: 8.0 2023-04-30 07:49:01,204 INFO [zipformer.py:625] (1/8) 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] (1/8) 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 07:49:39,759 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=155632.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 07:49:58,343 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.6720, 3.7952, 2.3007, 4.0896, 2.8911, 4.0881, 2.4247, 2.9479], device='cuda:1'), covar=tensor([0.0255, 0.0337, 0.1511, 0.0275, 0.0759, 0.0557, 0.1318, 0.0719], device='cuda:1'), in_proj_covar=tensor([0.0164, 0.0175, 0.0194, 0.0156, 0.0174, 0.0218, 0.0202, 0.0179], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-04-30 07:50:08,456 INFO [train.py:904] (1/8) Epoch 16, batch 3400, loss[loss=0.1855, simple_loss=0.2576, pruned_loss=0.0567, over 16871.00 frames. ], tot_loss[loss=0.1763, simple_loss=0.2617, pruned_loss=0.04543, over 3321177.63 frames. ], batch size: 116, lr: 4.27e-03, grad_scale: 8.0 2023-04-30 07:50:16,696 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=155657.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 07:50:27,012 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=2.64 vs. limit=5.0 2023-04-30 07:50:28,028 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=155666.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 07:50:45,728 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.95 vs. limit=2.0 2023-04-30 07:50:49,343 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=155680.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 07:50:50,791 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.25 vs. limit=2.0 2023-04-30 07:51:10,373 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([5.4062, 5.8463, 5.6112, 5.6760, 5.2432, 5.1917, 5.3048, 6.0238], device='cuda:1'), covar=tensor([0.1415, 0.0900, 0.1010, 0.0792, 0.0958, 0.0744, 0.1067, 0.0849], device='cuda:1'), in_proj_covar=tensor([0.0643, 0.0798, 0.0644, 0.0575, 0.0499, 0.0508, 0.0661, 0.0610], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-30 07:51:19,277 INFO [train.py:904] (1/8) Epoch 16, batch 3450, loss[loss=0.1896, simple_loss=0.2762, pruned_loss=0.05151, over 16568.00 frames. ], tot_loss[loss=0.1746, simple_loss=0.2596, pruned_loss=0.04483, over 3329769.62 frames. ], batch size: 68, lr: 4.27e-03, grad_scale: 16.0 2023-04-30 07:51:41,375 INFO [optim.py:368] (1/8) 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 07:51:41,686 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([5.2144, 5.7199, 5.8339, 5.5793, 5.7012, 6.2242, 5.7937, 5.4818], device='cuda:1'), covar=tensor([0.0887, 0.2049, 0.2349, 0.1960, 0.2604, 0.0931, 0.1455, 0.2212], device='cuda:1'), in_proj_covar=tensor([0.0401, 0.0575, 0.0620, 0.0483, 0.0645, 0.0650, 0.0490, 0.0641], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-04-30 07:51:41,856 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=155718.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 07:51:52,545 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=155725.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 07:52:22,853 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.21 vs. limit=2.0 2023-04-30 07:52:23,604 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.9533, 5.2431, 5.4245, 5.2169, 5.2151, 5.8435, 5.3408, 5.0244], device='cuda:1'), covar=tensor([0.1217, 0.2130, 0.1970, 0.2034, 0.2755, 0.1028, 0.1611, 0.2717], device='cuda:1'), in_proj_covar=tensor([0.0400, 0.0574, 0.0619, 0.0483, 0.0644, 0.0649, 0.0489, 0.0640], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-04-30 07:52:28,776 INFO [train.py:904] (1/8) Epoch 16, batch 3500, loss[loss=0.1676, simple_loss=0.2602, pruned_loss=0.03751, over 16727.00 frames. ], tot_loss[loss=0.1738, simple_loss=0.2585, pruned_loss=0.04451, over 3322408.63 frames. ], batch size: 57, lr: 4.27e-03, grad_scale: 16.0 2023-04-30 07:52:52,025 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.89 vs. limit=2.0 2023-04-30 07:52:58,253 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=155773.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 07:53:14,816 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=155784.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 07:53:19,024 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.0152, 4.4002, 3.1629, 2.4402, 2.8126, 2.6434, 4.7593, 3.7736], device='cuda:1'), covar=tensor([0.2425, 0.0553, 0.1610, 0.2492, 0.2632, 0.1823, 0.0332, 0.1120], device='cuda:1'), in_proj_covar=tensor([0.0313, 0.0266, 0.0293, 0.0294, 0.0288, 0.0240, 0.0281, 0.0321], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:1') 2023-04-30 07:53:40,574 INFO [train.py:904] (1/8) Epoch 16, batch 3550, loss[loss=0.1664, simple_loss=0.2634, pruned_loss=0.03466, over 17062.00 frames. ], tot_loss[loss=0.1733, simple_loss=0.2581, pruned_loss=0.04425, over 3326352.57 frames. ], batch size: 50, lr: 4.27e-03, grad_scale: 8.0 2023-04-30 07:53:56,147 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=155813.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 07:53:56,303 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([5.2687, 5.2269, 4.9945, 4.4840, 5.1309, 1.8988, 4.8285, 5.0363], device='cuda:1'), covar=tensor([0.0069, 0.0066, 0.0170, 0.0381, 0.0081, 0.2678, 0.0128, 0.0152], device='cuda:1'), in_proj_covar=tensor([0.0156, 0.0144, 0.0192, 0.0177, 0.0165, 0.0203, 0.0181, 0.0173], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-30 07:54:04,791 INFO [optim.py:368] (1/8) Clipping_scale=2.0, grad-norm quartiles 1.461e+02 2.310e+02 2.739e+02 3.224e+02 5.856e+02, threshold=5.477e+02, percent-clipped=2.0 2023-04-30 07:54:42,479 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=155845.0, num_to_drop=1, layers_to_drop={1} 2023-04-30 07:54:52,023 INFO [train.py:904] (1/8) Epoch 16, batch 3600, loss[loss=0.1519, simple_loss=0.2309, pruned_loss=0.03641, over 16838.00 frames. ], tot_loss[loss=0.1727, simple_loss=0.2574, pruned_loss=0.04399, over 3329238.68 frames. ], batch size: 39, lr: 4.27e-03, grad_scale: 8.0 2023-04-30 07:55:12,371 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([5.6042, 6.0107, 5.7494, 5.8091, 5.4213, 5.3827, 5.4170, 6.1328], device='cuda:1'), covar=tensor([0.1332, 0.0893, 0.1011, 0.0814, 0.0852, 0.0686, 0.1134, 0.0965], device='cuda:1'), in_proj_covar=tensor([0.0642, 0.0799, 0.0644, 0.0575, 0.0499, 0.0508, 0.0662, 0.0611], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-30 07:56:03,880 INFO [train.py:904] (1/8) Epoch 16, batch 3650, loss[loss=0.1683, simple_loss=0.2456, pruned_loss=0.04549, over 16822.00 frames. ], tot_loss[loss=0.1729, simple_loss=0.2566, pruned_loss=0.04453, over 3320436.53 frames. ], batch size: 102, lr: 4.27e-03, grad_scale: 8.0 2023-04-30 07:56:28,564 INFO [optim.py:368] (1/8) Clipping_scale=2.0, grad-norm quartiles 1.735e+02 2.258e+02 2.651e+02 3.350e+02 6.145e+02, threshold=5.302e+02, percent-clipped=1.0 2023-04-30 07:56:43,945 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([5.2524, 5.4114, 5.1084, 4.8132, 4.3683, 5.3426, 5.3157, 4.8703], device='cuda:1'), covar=tensor([0.0820, 0.0490, 0.0449, 0.0392, 0.1876, 0.0402, 0.0330, 0.0731], device='cuda:1'), in_proj_covar=tensor([0.0287, 0.0397, 0.0337, 0.0326, 0.0353, 0.0377, 0.0230, 0.0406], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-04-30 07:57:17,739 INFO [train.py:904] (1/8) Epoch 16, batch 3700, loss[loss=0.1797, simple_loss=0.2608, pruned_loss=0.04932, over 16673.00 frames. ], tot_loss[loss=0.1733, simple_loss=0.255, pruned_loss=0.04575, over 3309550.46 frames. ], batch size: 134, lr: 4.27e-03, grad_scale: 8.0 2023-04-30 07:57:31,974 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=155961.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 07:57:43,294 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.37 vs. limit=2.0 2023-04-30 07:58:07,558 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.6415, 1.7767, 2.2915, 2.5294, 2.6375, 2.6110, 1.8803, 2.7374], device='cuda:1'), covar=tensor([0.0137, 0.0392, 0.0227, 0.0198, 0.0213, 0.0206, 0.0402, 0.0129], device='cuda:1'), in_proj_covar=tensor([0.0179, 0.0186, 0.0172, 0.0176, 0.0187, 0.0143, 0.0185, 0.0136], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-30 07:58:36,928 INFO [train.py:904] (1/8) Epoch 16, batch 3750, loss[loss=0.1961, simple_loss=0.2738, pruned_loss=0.05918, over 11838.00 frames. ], tot_loss[loss=0.1751, simple_loss=0.2556, pruned_loss=0.04735, over 3282529.94 frames. ], batch size: 248, lr: 4.27e-03, grad_scale: 8.0 2023-04-30 07:58:53,127 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=156013.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 07:59:02,684 INFO [optim.py:368] (1/8) 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 07:59:51,246 INFO [train.py:904] (1/8) Epoch 16, batch 3800, loss[loss=0.172, simple_loss=0.2513, pruned_loss=0.0464, over 16448.00 frames. ], tot_loss[loss=0.1774, simple_loss=0.2571, pruned_loss=0.04882, over 3265663.44 frames. ], batch size: 68, lr: 4.27e-03, grad_scale: 8.0 2023-04-30 08:00:17,810 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=156070.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 08:01:03,106 INFO [train.py:904] (1/8) Epoch 16, batch 3850, loss[loss=0.1967, simple_loss=0.2736, pruned_loss=0.05993, over 16311.00 frames. ], tot_loss[loss=0.1787, simple_loss=0.258, pruned_loss=0.0497, over 3262121.31 frames. ], batch size: 35, lr: 4.26e-03, grad_scale: 8.0 2023-04-30 08:01:20,097 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=156113.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 08:01:28,704 INFO [optim.py:368] (1/8) 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 08:01:47,452 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=156131.0, num_to_drop=1, layers_to_drop={0} 2023-04-30 08:01:59,096 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=156140.0, num_to_drop=1, layers_to_drop={2} 2023-04-30 08:02:16,231 INFO [train.py:904] (1/8) Epoch 16, batch 3900, loss[loss=0.1718, simple_loss=0.2536, pruned_loss=0.04498, over 16586.00 frames. ], tot_loss[loss=0.1786, simple_loss=0.2571, pruned_loss=0.05002, over 3263934.47 frames. ], batch size: 57, lr: 4.26e-03, grad_scale: 8.0 2023-04-30 08:02:29,402 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=156161.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 08:03:26,081 INFO [train.py:904] (1/8) Epoch 16, batch 3950, loss[loss=0.1898, simple_loss=0.2582, pruned_loss=0.0607, over 16892.00 frames. ], tot_loss[loss=0.1794, simple_loss=0.2571, pruned_loss=0.05082, over 3271544.03 frames. ], batch size: 109, lr: 4.26e-03, grad_scale: 8.0 2023-04-30 08:03:47,380 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=156216.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 08:03:50,340 INFO [optim.py:368] (1/8) 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 08:04:37,103 INFO [train.py:904] (1/8) Epoch 16, batch 4000, loss[loss=0.1875, simple_loss=0.2638, pruned_loss=0.05562, over 17053.00 frames. ], tot_loss[loss=0.1801, simple_loss=0.2576, pruned_loss=0.05129, over 3279509.47 frames. ], batch size: 55, lr: 4.26e-03, grad_scale: 8.0 2023-04-30 08:04:49,160 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=156261.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 08:05:12,197 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=156277.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 08:05:12,576 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.75 vs. limit=2.0 2023-04-30 08:05:31,050 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=156290.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 08:05:36,935 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([1.8097, 2.7117, 2.5620, 4.4158, 3.3319, 4.1273, 1.5355, 3.0616], device='cuda:1'), covar=tensor([0.1319, 0.0803, 0.1188, 0.0147, 0.0303, 0.0336, 0.1636, 0.0803], device='cuda:1'), in_proj_covar=tensor([0.0160, 0.0169, 0.0188, 0.0175, 0.0203, 0.0214, 0.0190, 0.0187], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-04-30 08:05:46,750 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=3.06 vs. limit=5.0 2023-04-30 08:05:49,241 INFO [train.py:904] (1/8) Epoch 16, batch 4050, loss[loss=0.1746, simple_loss=0.2609, pruned_loss=0.04418, over 16816.00 frames. ], tot_loss[loss=0.1786, simple_loss=0.2574, pruned_loss=0.04996, over 3278630.23 frames. ], batch size: 102, lr: 4.26e-03, grad_scale: 8.0 2023-04-30 08:05:59,640 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=156309.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 08:06:04,439 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=156313.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 08:06:14,226 INFO [optim.py:368] (1/8) 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 08:07:01,378 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=156351.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 08:07:01,986 INFO [train.py:904] (1/8) Epoch 16, batch 4100, loss[loss=0.1707, simple_loss=0.2568, pruned_loss=0.04228, over 16482.00 frames. ], tot_loss[loss=0.1791, simple_loss=0.2589, pruned_loss=0.04962, over 3265877.16 frames. ], batch size: 75, lr: 4.26e-03, grad_scale: 8.0 2023-04-30 08:07:15,155 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=156361.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 08:08:15,973 INFO [train.py:904] (1/8) Epoch 16, batch 4150, loss[loss=0.1962, simple_loss=0.2765, pruned_loss=0.05797, over 16828.00 frames. ], tot_loss[loss=0.1853, simple_loss=0.2665, pruned_loss=0.05201, over 3239522.00 frames. ], batch size: 39, lr: 4.26e-03, grad_scale: 8.0 2023-04-30 08:08:40,340 INFO [optim.py:368] (1/8) 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 08:08:52,134 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=156426.0, num_to_drop=1, layers_to_drop={1} 2023-04-30 08:09:13,246 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=156440.0, num_to_drop=1, layers_to_drop={0} 2023-04-30 08:09:30,675 INFO [train.py:904] (1/8) Epoch 16, batch 4200, loss[loss=0.2474, simple_loss=0.3187, pruned_loss=0.08802, over 11754.00 frames. ], tot_loss[loss=0.1905, simple_loss=0.2735, pruned_loss=0.05378, over 3192605.91 frames. ], batch size: 247, lr: 4.26e-03, grad_scale: 8.0 2023-04-30 08:10:24,400 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=156488.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 08:10:35,401 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.4084, 2.3025, 2.3756, 4.2009, 2.2247, 2.6365, 2.3382, 2.4961], device='cuda:1'), covar=tensor([0.1111, 0.3432, 0.2485, 0.0411, 0.3842, 0.2213, 0.3300, 0.3066], device='cuda:1'), in_proj_covar=tensor([0.0383, 0.0423, 0.0352, 0.0324, 0.0423, 0.0488, 0.0390, 0.0494], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-30 08:10:37,777 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.6672, 3.7759, 2.0887, 4.2717, 2.8709, 4.2869, 2.5782, 3.0645], device='cuda:1'), covar=tensor([0.0242, 0.0331, 0.1756, 0.0233, 0.0797, 0.0356, 0.1303, 0.0683], device='cuda:1'), in_proj_covar=tensor([0.0163, 0.0173, 0.0191, 0.0151, 0.0173, 0.0213, 0.0200, 0.0176], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-04-30 08:10:43,947 INFO [train.py:904] (1/8) Epoch 16, batch 4250, loss[loss=0.167, simple_loss=0.2653, pruned_loss=0.03431, over 16824.00 frames. ], tot_loss[loss=0.1914, simple_loss=0.2764, pruned_loss=0.05323, over 3194446.56 frames. ], batch size: 83, lr: 4.26e-03, grad_scale: 8.0 2023-04-30 08:11:09,162 INFO [optim.py:368] (1/8) 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 08:11:35,500 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.7185, 3.6469, 4.0604, 1.9270, 4.2963, 4.3442, 3.0789, 3.0980], device='cuda:1'), covar=tensor([0.0715, 0.0228, 0.0200, 0.1193, 0.0049, 0.0090, 0.0399, 0.0432], device='cuda:1'), in_proj_covar=tensor([0.0145, 0.0105, 0.0092, 0.0138, 0.0074, 0.0118, 0.0124, 0.0128], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:1') 2023-04-30 08:11:56,452 INFO [train.py:904] (1/8) Epoch 16, batch 4300, loss[loss=0.2165, simple_loss=0.3051, pruned_loss=0.06399, over 16787.00 frames. ], tot_loss[loss=0.1914, simple_loss=0.2779, pruned_loss=0.05238, over 3204785.28 frames. ], batch size: 83, lr: 4.26e-03, grad_scale: 8.0 2023-04-30 08:12:16,018 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.4307, 2.9154, 2.8601, 1.6150, 2.6056, 1.8307, 2.9974, 3.1274], device='cuda:1'), covar=tensor([0.0281, 0.0690, 0.0634, 0.2175, 0.0991, 0.1105, 0.0632, 0.0706], device='cuda:1'), in_proj_covar=tensor([0.0148, 0.0156, 0.0161, 0.0148, 0.0139, 0.0126, 0.0140, 0.0167], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:1') 2023-04-30 08:12:19,426 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([1.9489, 3.8530, 3.8699, 2.3192, 3.4250, 3.8435, 3.5369, 2.1406], device='cuda:1'), covar=tensor([0.0548, 0.0032, 0.0037, 0.0406, 0.0088, 0.0074, 0.0081, 0.0426], device='cuda:1'), in_proj_covar=tensor([0.0132, 0.0076, 0.0077, 0.0131, 0.0090, 0.0099, 0.0088, 0.0125], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-04-30 08:12:27,167 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=156572.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 08:13:09,396 INFO [train.py:904] (1/8) Epoch 16, batch 4350, loss[loss=0.2002, simple_loss=0.2781, pruned_loss=0.06121, over 11617.00 frames. ], tot_loss[loss=0.1935, simple_loss=0.2807, pruned_loss=0.05315, over 3201634.38 frames. ], batch size: 248, lr: 4.26e-03, grad_scale: 8.0 2023-04-30 08:13:34,597 INFO [optim.py:368] (1/8) 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] (1/8) 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] (1/8) Epoch 16, batch 4400, loss[loss=0.2046, simple_loss=0.2881, pruned_loss=0.06054, over 16440.00 frames. ], tot_loss[loss=0.1962, simple_loss=0.2831, pruned_loss=0.05466, over 3191966.53 frames. ], batch size: 68, lr: 4.26e-03, grad_scale: 8.0 2023-04-30 08:15:06,774 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.8083, 1.8259, 2.3298, 2.7180, 2.6077, 3.1470, 1.8381, 3.0708], device='cuda:1'), covar=tensor([0.0159, 0.0440, 0.0276, 0.0251, 0.0240, 0.0123, 0.0483, 0.0104], device='cuda:1'), in_proj_covar=tensor([0.0175, 0.0184, 0.0170, 0.0174, 0.0185, 0.0140, 0.0183, 0.0134], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-30 08:15:32,099 INFO [train.py:904] (1/8) Epoch 16, batch 4450, loss[loss=0.1984, simple_loss=0.287, pruned_loss=0.05496, over 16505.00 frames. ], tot_loss[loss=0.1989, simple_loss=0.2865, pruned_loss=0.05559, over 3212706.32 frames. ], batch size: 75, lr: 4.26e-03, grad_scale: 8.0 2023-04-30 08:15:57,563 INFO [optim.py:368] (1/8) 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 08:16:04,234 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.1607, 3.1779, 1.8975, 3.4359, 2.3640, 3.4902, 2.1056, 2.5070], device='cuda:1'), covar=tensor([0.0290, 0.0407, 0.1669, 0.0152, 0.0870, 0.0471, 0.1442, 0.0789], device='cuda:1'), in_proj_covar=tensor([0.0162, 0.0172, 0.0190, 0.0149, 0.0172, 0.0212, 0.0199, 0.0175], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-04-30 08:16:09,706 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=156726.0, num_to_drop=1, layers_to_drop={1} 2023-04-30 08:16:41,456 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=156749.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 08:16:45,641 INFO [train.py:904] (1/8) Epoch 16, batch 4500, loss[loss=0.1815, simple_loss=0.2676, pruned_loss=0.0477, over 17221.00 frames. ], tot_loss[loss=0.1995, simple_loss=0.2866, pruned_loss=0.05622, over 3208109.68 frames. ], batch size: 44, lr: 4.26e-03, grad_scale: 8.0 2023-04-30 08:16:47,472 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.6279, 2.6966, 2.2423, 2.4432, 3.0920, 2.6569, 3.3227, 3.2687], device='cuda:1'), covar=tensor([0.0063, 0.0306, 0.0406, 0.0335, 0.0180, 0.0308, 0.0154, 0.0189], device='cuda:1'), in_proj_covar=tensor([0.0181, 0.0217, 0.0210, 0.0211, 0.0220, 0.0221, 0.0224, 0.0217], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-30 08:16:49,095 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=4.42 vs. limit=5.0 2023-04-30 08:17:16,165 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=156774.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 08:17:19,685 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.25 vs. limit=2.0 2023-04-30 08:17:54,762 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([1.7258, 3.0546, 2.7212, 4.8619, 3.8382, 4.1343, 1.6138, 3.1254], device='cuda:1'), covar=tensor([0.1310, 0.0628, 0.1105, 0.0108, 0.0292, 0.0352, 0.1533, 0.0768], device='cuda:1'), in_proj_covar=tensor([0.0159, 0.0168, 0.0187, 0.0172, 0.0203, 0.0213, 0.0190, 0.0186], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-04-30 08:17:56,547 INFO [train.py:904] (1/8) Epoch 16, batch 4550, loss[loss=0.226, simple_loss=0.2903, pruned_loss=0.08089, over 11870.00 frames. ], tot_loss[loss=0.2004, simple_loss=0.2869, pruned_loss=0.05696, over 3205205.01 frames. ], batch size: 246, lr: 4.26e-03, grad_scale: 8.0 2023-04-30 08:18:07,326 INFO [zipformer.py:625] (1/8) 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] (1/8) attn_weights_entropy = tensor([2.3423, 3.2548, 2.6081, 2.0464, 2.1657, 2.0437, 3.3131, 2.9788], device='cuda:1'), covar=tensor([0.2877, 0.0674, 0.1691, 0.2537, 0.2515, 0.2195, 0.0494, 0.1166], device='cuda:1'), in_proj_covar=tensor([0.0315, 0.0264, 0.0295, 0.0295, 0.0291, 0.0240, 0.0282, 0.0320], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:1') 2023-04-30 08:18:19,417 INFO [optim.py:368] (1/8) 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] (1/8) Epoch 16, batch 4600, loss[loss=0.2001, simple_loss=0.286, pruned_loss=0.05706, over 16223.00 frames. ], tot_loss[loss=0.2013, simple_loss=0.2881, pruned_loss=0.05719, over 3220016.39 frames. ], batch size: 35, lr: 4.25e-03, grad_scale: 8.0 2023-04-30 08:19:36,305 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=156872.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 08:19:55,729 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 16, batch 4650, loss[loss=0.2549, simple_loss=0.3123, pruned_loss=0.09872, over 11787.00 frames. ], tot_loss[loss=0.2012, simple_loss=0.2874, pruned_loss=0.05749, over 3216415.30 frames. ], batch size: 247, lr: 4.25e-03, grad_scale: 8.0 2023-04-30 08:20:37,619 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.74 vs. limit=2.0 2023-04-30 08:20:39,057 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.71 vs. limit=2.0 2023-04-30 08:20:45,036 INFO [optim.py:368] (1/8) 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] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=156920.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 08:20:48,376 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.0558, 4.0182, 4.5255, 2.3995, 4.7944, 4.8054, 3.3417, 3.5343], device='cuda:1'), covar=tensor([0.0771, 0.0233, 0.0205, 0.1102, 0.0073, 0.0085, 0.0404, 0.0429], device='cuda:1'), in_proj_covar=tensor([0.0147, 0.0106, 0.0093, 0.0139, 0.0074, 0.0119, 0.0126, 0.0129], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0004, 0.0003], device='cuda:1') 2023-04-30 08:21:12,272 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.62 vs. limit=2.0 2023-04-30 08:21:25,715 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=156946.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 08:21:25,800 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 16, batch 4700, loss[loss=0.1774, simple_loss=0.2645, pruned_loss=0.0451, over 16885.00 frames. ], tot_loss[loss=0.1982, simple_loss=0.2844, pruned_loss=0.05605, over 3238508.07 frames. ], batch size: 109, lr: 4.25e-03, grad_scale: 8.0 2023-04-30 08:22:36,260 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 16, batch 4750, loss[loss=0.1577, simple_loss=0.2494, pruned_loss=0.03296, over 16731.00 frames. ], tot_loss[loss=0.1954, simple_loss=0.2815, pruned_loss=0.05463, over 3222509.77 frames. ], batch size: 89, lr: 4.25e-03, grad_scale: 8.0 2023-04-30 08:22:59,368 INFO [zipformer.py:625] (1/8) 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] (1/8) 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 08:23:59,603 INFO [train.py:904] (1/8) Epoch 16, batch 4800, loss[loss=0.1977, simple_loss=0.2762, pruned_loss=0.05965, over 11931.00 frames. ], tot_loss[loss=0.1917, simple_loss=0.2777, pruned_loss=0.05285, over 3205074.01 frames. ], batch size: 248, lr: 4.25e-03, grad_scale: 8.0 2023-04-30 08:24:28,410 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=157071.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 08:25:14,776 INFO [train.py:904] (1/8) Epoch 16, batch 4850, loss[loss=0.1656, simple_loss=0.2605, pruned_loss=0.03533, over 16465.00 frames. ], tot_loss[loss=0.1918, simple_loss=0.2786, pruned_loss=0.0525, over 3172216.87 frames. ], batch size: 75, lr: 4.25e-03, grad_scale: 8.0 2023-04-30 08:25:20,104 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=157105.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 08:25:39,363 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.7912, 5.0255, 5.2447, 5.0196, 5.0622, 5.6316, 5.1081, 4.8755], device='cuda:1'), covar=tensor([0.0920, 0.1662, 0.1740, 0.1713, 0.2287, 0.0808, 0.1227, 0.2143], device='cuda:1'), in_proj_covar=tensor([0.0384, 0.0544, 0.0586, 0.0460, 0.0613, 0.0622, 0.0464, 0.0614], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-04-30 08:25:41,920 INFO [optim.py:368] (1/8) 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] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=157142.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 08:26:30,907 INFO [train.py:904] (1/8) Epoch 16, batch 4900, loss[loss=0.1823, simple_loss=0.2654, pruned_loss=0.04956, over 11951.00 frames. ], tot_loss[loss=0.1898, simple_loss=0.2778, pruned_loss=0.05095, over 3167191.00 frames. ], batch size: 247, lr: 4.25e-03, grad_scale: 8.0 2023-04-30 08:27:20,680 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=4.65 vs. limit=5.0 2023-04-30 08:27:45,015 INFO [train.py:904] (1/8) Epoch 16, batch 4950, loss[loss=0.1919, simple_loss=0.2888, pruned_loss=0.04752, over 16645.00 frames. ], tot_loss[loss=0.1893, simple_loss=0.2777, pruned_loss=0.05039, over 3184796.24 frames. ], batch size: 134, lr: 4.25e-03, grad_scale: 8.0 2023-04-30 08:27:46,688 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=157203.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 08:28:08,808 INFO [optim.py:368] (1/8) 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 08:28:30,263 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.35 vs. limit=2.0 2023-04-30 08:28:40,328 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=157241.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 08:28:52,533 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=4.88 vs. limit=5.0 2023-04-30 08:28:57,707 INFO [train.py:904] (1/8) Epoch 16, batch 5000, loss[loss=0.2041, simple_loss=0.2939, pruned_loss=0.0572, over 16347.00 frames. ], tot_loss[loss=0.19, simple_loss=0.2788, pruned_loss=0.05057, over 3183624.67 frames. ], batch size: 165, lr: 4.25e-03, grad_scale: 8.0 2023-04-30 08:29:13,975 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.32 vs. limit=2.0 2023-04-30 08:29:50,672 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.35 vs. limit=2.0 2023-04-30 08:30:10,076 INFO [train.py:904] (1/8) Epoch 16, batch 5050, loss[loss=0.2064, simple_loss=0.2836, pruned_loss=0.06462, over 16582.00 frames. ], tot_loss[loss=0.19, simple_loss=0.2793, pruned_loss=0.05029, over 3210034.71 frames. ], batch size: 57, lr: 4.25e-03, grad_scale: 8.0 2023-04-30 08:30:33,953 INFO [optim.py:368] (1/8) 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] (1/8) Epoch 16, batch 5100, loss[loss=0.1676, simple_loss=0.2539, pruned_loss=0.0407, over 16766.00 frames. ], tot_loss[loss=0.1877, simple_loss=0.277, pruned_loss=0.04924, over 3228463.87 frames. ], batch size: 83, lr: 4.25e-03, grad_scale: 8.0 2023-04-30 08:31:23,023 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([5.0439, 5.0131, 4.8715, 4.5283, 4.4873, 4.9329, 4.8627, 4.6486], device='cuda:1'), covar=tensor([0.0549, 0.0370, 0.0276, 0.0278, 0.1168, 0.0485, 0.0304, 0.0621], device='cuda:1'), in_proj_covar=tensor([0.0266, 0.0372, 0.0315, 0.0304, 0.0331, 0.0353, 0.0214, 0.0377], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-30 08:31:28,535 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=157356.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 08:31:44,277 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=157366.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 08:32:07,542 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=157382.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 08:32:18,411 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([5.4125, 5.7848, 5.4121, 5.5521, 5.1903, 5.1395, 5.1345, 5.8763], device='cuda:1'), covar=tensor([0.1258, 0.0848, 0.1016, 0.0726, 0.0822, 0.0615, 0.1064, 0.0838], device='cuda:1'), in_proj_covar=tensor([0.0606, 0.0752, 0.0613, 0.0544, 0.0472, 0.0481, 0.0622, 0.0577], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:1') 2023-04-30 08:32:38,193 INFO [train.py:904] (1/8) Epoch 16, batch 5150, loss[loss=0.1783, simple_loss=0.2757, pruned_loss=0.04045, over 16858.00 frames. ], tot_loss[loss=0.1868, simple_loss=0.2766, pruned_loss=0.04853, over 3216400.83 frames. ], batch size: 102, lr: 4.25e-03, grad_scale: 8.0 2023-04-30 08:32:42,246 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=157404.0, num_to_drop=1, layers_to_drop={0} 2023-04-30 08:32:43,987 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=157405.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 08:33:02,246 INFO [zipformer.py:625] (1/8) 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] (1/8) Clipping_scale=2.0, grad-norm quartiles 1.305e+02 1.981e+02 2.319e+02 2.667e+02 4.266e+02, threshold=4.638e+02, percent-clipped=0.0 2023-04-30 08:33:05,818 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.0130, 3.4924, 3.3320, 2.0214, 2.9035, 2.4958, 3.5330, 3.5762], device='cuda:1'), covar=tensor([0.0279, 0.0666, 0.0718, 0.1866, 0.0889, 0.0887, 0.0666, 0.0821], device='cuda:1'), in_proj_covar=tensor([0.0148, 0.0155, 0.0161, 0.0148, 0.0139, 0.0125, 0.0140, 0.0165], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:1') 2023-04-30 08:33:37,805 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.6570, 1.7910, 2.1924, 2.6581, 2.6230, 2.9593, 1.8006, 2.9957], device='cuda:1'), covar=tensor([0.0188, 0.0478, 0.0316, 0.0268, 0.0282, 0.0151, 0.0532, 0.0109], device='cuda:1'), in_proj_covar=tensor([0.0177, 0.0185, 0.0172, 0.0176, 0.0186, 0.0141, 0.0186, 0.0135], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-30 08:33:40,840 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=157443.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 08:33:53,087 INFO [train.py:904] (1/8) Epoch 16, batch 5200, loss[loss=0.1841, simple_loss=0.2658, pruned_loss=0.0512, over 16556.00 frames. ], tot_loss[loss=0.1854, simple_loss=0.275, pruned_loss=0.04791, over 3218497.78 frames. ], batch size: 68, lr: 4.25e-03, grad_scale: 8.0 2023-04-30 08:33:54,777 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=157453.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 08:34:12,577 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.7813, 4.8354, 5.2207, 5.1709, 5.1883, 4.8451, 4.8112, 4.6437], device='cuda:1'), covar=tensor([0.0252, 0.0450, 0.0294, 0.0355, 0.0446, 0.0316, 0.0868, 0.0401], device='cuda:1'), in_proj_covar=tensor([0.0373, 0.0400, 0.0389, 0.0369, 0.0442, 0.0415, 0.0511, 0.0330], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:1') 2023-04-30 08:34:12,645 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=157465.0, num_to_drop=1, layers_to_drop={1} 2023-04-30 08:35:01,963 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 16, batch 5250, loss[loss=0.1675, simple_loss=0.2462, pruned_loss=0.0444, over 16972.00 frames. ], tot_loss[loss=0.1838, simple_loss=0.2725, pruned_loss=0.04753, over 3224702.81 frames. ], batch size: 55, lr: 4.25e-03, grad_scale: 8.0 2023-04-30 08:35:13,327 INFO [zipformer.py:625] (1/8) 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] (1/8) 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 08:36:05,737 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=157541.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 08:36:21,503 INFO [train.py:904] (1/8) Epoch 16, batch 5300, loss[loss=0.1662, simple_loss=0.2554, pruned_loss=0.03852, over 16595.00 frames. ], tot_loss[loss=0.181, simple_loss=0.2692, pruned_loss=0.04641, over 3226320.33 frames. ], batch size: 62, lr: 4.24e-03, grad_scale: 8.0 2023-04-30 08:36:45,080 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=157567.0, num_to_drop=1, layers_to_drop={3} 2023-04-30 08:36:46,079 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=157568.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 08:37:17,189 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=157589.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 08:37:35,558 INFO [train.py:904] (1/8) Epoch 16, batch 5350, loss[loss=0.1826, simple_loss=0.2705, pruned_loss=0.04738, over 16626.00 frames. ], tot_loss[loss=0.1795, simple_loss=0.2676, pruned_loss=0.0457, over 3220410.58 frames. ], batch size: 62, lr: 4.24e-03, grad_scale: 8.0 2023-04-30 08:38:00,208 INFO [optim.py:368] (1/8) 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 08:38:16,641 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=157629.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 08:38:42,636 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.0110, 3.3400, 3.4137, 1.9726, 2.9550, 2.3393, 3.5388, 3.5162], device='cuda:1'), covar=tensor([0.0283, 0.0713, 0.0582, 0.1900, 0.0823, 0.0900, 0.0588, 0.0891], device='cuda:1'), in_proj_covar=tensor([0.0150, 0.0156, 0.0163, 0.0149, 0.0140, 0.0126, 0.0141, 0.0167], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:1') 2023-04-30 08:38:49,047 INFO [train.py:904] (1/8) Epoch 16, batch 5400, loss[loss=0.1796, simple_loss=0.2726, pruned_loss=0.04335, over 16821.00 frames. ], tot_loss[loss=0.1817, simple_loss=0.2705, pruned_loss=0.04646, over 3205306.51 frames. ], batch size: 83, lr: 4.24e-03, grad_scale: 8.0 2023-04-30 08:39:09,646 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=157666.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 08:39:24,994 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.3121, 4.2792, 4.1834, 3.4825, 4.2205, 1.6508, 3.9550, 3.8938], device='cuda:1'), covar=tensor([0.0087, 0.0086, 0.0141, 0.0340, 0.0083, 0.2678, 0.0122, 0.0201], device='cuda:1'), in_proj_covar=tensor([0.0147, 0.0136, 0.0183, 0.0171, 0.0155, 0.0194, 0.0171, 0.0164], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-30 08:40:04,938 INFO [train.py:904] (1/8) Epoch 16, batch 5450, loss[loss=0.2276, simple_loss=0.3154, pruned_loss=0.06989, over 15284.00 frames. ], tot_loss[loss=0.1855, simple_loss=0.2744, pruned_loss=0.04829, over 3200637.02 frames. ], batch size: 190, lr: 4.24e-03, grad_scale: 8.0 2023-04-30 08:40:07,794 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=2.55 vs. limit=5.0 2023-04-30 08:40:08,083 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=4.66 vs. limit=5.0 2023-04-30 08:40:12,603 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.1937, 3.4280, 3.5696, 3.5472, 3.5576, 3.3910, 3.3923, 3.4690], device='cuda:1'), covar=tensor([0.0442, 0.0762, 0.0508, 0.0471, 0.0552, 0.0639, 0.1024, 0.0588], device='cuda:1'), in_proj_covar=tensor([0.0374, 0.0403, 0.0392, 0.0372, 0.0442, 0.0416, 0.0515, 0.0331], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:1') 2023-04-30 08:40:19,856 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=157712.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 08:40:23,495 INFO [zipformer.py:625] (1/8) 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] (1/8) 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 08:41:00,185 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=157738.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 08:41:20,975 INFO [train.py:904] (1/8) Epoch 16, batch 5500, loss[loss=0.2255, simple_loss=0.3215, pruned_loss=0.06477, over 17272.00 frames. ], tot_loss[loss=0.1938, simple_loss=0.2819, pruned_loss=0.0528, over 3180882.09 frames. ], batch size: 52, lr: 4.24e-03, grad_scale: 8.0 2023-04-30 08:41:33,730 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=157760.0, num_to_drop=1, layers_to_drop={2} 2023-04-30 08:42:25,161 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.77 vs. limit=2.0 2023-04-30 08:42:30,998 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=157798.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 08:42:36,591 INFO [train.py:904] (1/8) Epoch 16, batch 5550, loss[loss=0.2828, simple_loss=0.3404, pruned_loss=0.1126, over 11542.00 frames. ], tot_loss[loss=0.2047, simple_loss=0.2904, pruned_loss=0.0595, over 3119729.76 frames. ], batch size: 248, lr: 4.24e-03, grad_scale: 16.0 2023-04-30 08:43:04,447 INFO [optim.py:368] (1/8) 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 08:43:47,867 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=157846.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 08:43:57,056 INFO [train.py:904] (1/8) Epoch 16, batch 5600, loss[loss=0.2781, simple_loss=0.3303, pruned_loss=0.113, over 11068.00 frames. ], tot_loss[loss=0.21, simple_loss=0.2945, pruned_loss=0.06279, over 3101733.02 frames. ], batch size: 248, lr: 4.24e-03, grad_scale: 16.0 2023-04-30 08:43:59,494 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=157853.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 08:44:12,211 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.6841, 4.7076, 4.5699, 4.2862, 4.2282, 4.6366, 4.4662, 4.3639], device='cuda:1'), covar=tensor([0.0624, 0.0595, 0.0267, 0.0299, 0.0923, 0.0470, 0.0465, 0.0659], device='cuda:1'), in_proj_covar=tensor([0.0271, 0.0378, 0.0318, 0.0308, 0.0333, 0.0358, 0.0218, 0.0381], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002], device='cuda:1') 2023-04-30 08:44:13,639 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=157862.0, num_to_drop=1, layers_to_drop={3} 2023-04-30 08:44:13,859 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.0077, 2.8376, 2.7729, 2.1273, 2.6109, 2.1793, 2.7222, 2.9715], device='cuda:1'), covar=tensor([0.0256, 0.0698, 0.0520, 0.1654, 0.0778, 0.0963, 0.0548, 0.0730], device='cuda:1'), in_proj_covar=tensor([0.0148, 0.0155, 0.0161, 0.0148, 0.0139, 0.0126, 0.0140, 0.0165], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:1') 2023-04-30 08:45:21,493 INFO [train.py:904] (1/8) Epoch 16, batch 5650, loss[loss=0.2065, simple_loss=0.2874, pruned_loss=0.06274, over 16475.00 frames. ], tot_loss[loss=0.2164, simple_loss=0.2997, pruned_loss=0.06652, over 3079611.69 frames. ], batch size: 68, lr: 4.24e-03, grad_scale: 8.0 2023-04-30 08:45:35,309 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.7148, 3.7920, 2.9143, 2.2561, 2.6670, 2.4406, 4.1478, 3.4993], device='cuda:1'), covar=tensor([0.2622, 0.0686, 0.1615, 0.2376, 0.2358, 0.1853, 0.0422, 0.1162], device='cuda:1'), in_proj_covar=tensor([0.0313, 0.0261, 0.0294, 0.0294, 0.0287, 0.0238, 0.0279, 0.0315], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:1') 2023-04-30 08:45:40,963 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=157914.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 08:45:50,669 INFO [optim.py:368] (1/8) Clipping_scale=2.0, grad-norm quartiles 2.188e+02 3.549e+02 4.142e+02 4.836e+02 1.033e+03, threshold=8.283e+02, percent-clipped=2.0 2023-04-30 08:45:56,909 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=157924.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 08:46:08,864 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.19 vs. limit=2.0 2023-04-30 08:46:12,169 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([5.3294, 5.3758, 5.2319, 4.9079, 4.8773, 5.3023, 5.1974, 4.9949], device='cuda:1'), covar=tensor([0.0624, 0.0337, 0.0257, 0.0285, 0.0921, 0.0384, 0.0253, 0.0635], device='cuda:1'), in_proj_covar=tensor([0.0269, 0.0376, 0.0316, 0.0306, 0.0331, 0.0356, 0.0217, 0.0378], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-30 08:46:39,301 INFO [train.py:904] (1/8) Epoch 16, batch 5700, loss[loss=0.2105, simple_loss=0.3016, pruned_loss=0.05966, over 16819.00 frames. ], tot_loss[loss=0.2187, simple_loss=0.301, pruned_loss=0.06815, over 3078652.08 frames. ], batch size: 39, lr: 4.24e-03, grad_scale: 8.0 2023-04-30 08:47:44,930 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.38 vs. limit=2.0 2023-04-30 08:48:01,931 INFO [train.py:904] (1/8) Epoch 16, batch 5750, loss[loss=0.2419, simple_loss=0.3228, pruned_loss=0.0805, over 16803.00 frames. ], tot_loss[loss=0.2225, simple_loss=0.3044, pruned_loss=0.0703, over 3045457.52 frames. ], batch size: 116, lr: 4.24e-03, grad_scale: 8.0 2023-04-30 08:48:16,706 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=158012.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 08:48:30,519 INFO [optim.py:368] (1/8) 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 08:49:00,679 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=158038.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 08:49:04,520 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.8861, 4.7978, 4.6643, 3.9770, 4.7381, 1.6809, 4.4687, 4.4422], device='cuda:1'), covar=tensor([0.0099, 0.0086, 0.0159, 0.0365, 0.0092, 0.2678, 0.0145, 0.0197], device='cuda:1'), in_proj_covar=tensor([0.0146, 0.0136, 0.0182, 0.0170, 0.0155, 0.0193, 0.0169, 0.0163], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-30 08:49:20,318 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.29 vs. limit=2.0 2023-04-30 08:49:22,774 INFO [train.py:904] (1/8) Epoch 16, batch 5800, loss[loss=0.2085, simple_loss=0.2945, pruned_loss=0.06121, over 16483.00 frames. ], tot_loss[loss=0.22, simple_loss=0.3034, pruned_loss=0.06829, over 3070827.01 frames. ], batch size: 68, lr: 4.24e-03, grad_scale: 8.0 2023-04-30 08:49:32,302 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.1795, 4.2906, 4.4353, 4.2820, 4.3012, 4.8264, 4.4253, 4.1667], device='cuda:1'), covar=tensor([0.1768, 0.1811, 0.2145, 0.1767, 0.2430, 0.0982, 0.1321, 0.2354], device='cuda:1'), in_proj_covar=tensor([0.0385, 0.0544, 0.0593, 0.0458, 0.0617, 0.0622, 0.0465, 0.0619], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-04-30 08:49:35,950 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=158060.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 08:49:36,018 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=158060.0, num_to_drop=1, layers_to_drop={1} 2023-04-30 08:50:16,570 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=158086.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 08:50:41,299 INFO [train.py:904] (1/8) Epoch 16, batch 5850, loss[loss=0.2265, simple_loss=0.3107, pruned_loss=0.07113, over 16662.00 frames. ], tot_loss[loss=0.2167, simple_loss=0.3008, pruned_loss=0.06635, over 3073610.89 frames. ], batch size: 134, lr: 4.24e-03, grad_scale: 8.0 2023-04-30 08:50:51,620 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=158108.0, num_to_drop=1, layers_to_drop={0} 2023-04-30 08:50:54,109 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.8156, 4.7638, 4.5717, 3.6555, 4.7153, 1.5491, 4.4267, 4.3311], device='cuda:1'), covar=tensor([0.0100, 0.0105, 0.0206, 0.0526, 0.0102, 0.3189, 0.0147, 0.0267], device='cuda:1'), in_proj_covar=tensor([0.0146, 0.0136, 0.0182, 0.0170, 0.0154, 0.0192, 0.0169, 0.0162], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-30 08:51:08,529 INFO [optim.py:368] (1/8) 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 08:51:12,986 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 16, batch 5900, loss[loss=0.2069, simple_loss=0.2942, pruned_loss=0.05973, over 16607.00 frames. ], tot_loss[loss=0.2165, simple_loss=0.3003, pruned_loss=0.06635, over 3062081.00 frames. ], batch size: 57, lr: 4.24e-03, grad_scale: 8.0 2023-04-30 08:52:23,700 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=158162.0, num_to_drop=1, layers_to_drop={1} 2023-04-30 08:52:56,025 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=158183.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 08:52:59,477 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.22 vs. limit=2.0 2023-04-30 08:53:25,266 INFO [train.py:904] (1/8) Epoch 16, batch 5950, loss[loss=0.1993, simple_loss=0.2926, pruned_loss=0.053, over 16744.00 frames. ], tot_loss[loss=0.2154, simple_loss=0.3009, pruned_loss=0.06493, over 3094645.67 frames. ], batch size: 89, lr: 4.24e-03, grad_scale: 8.0 2023-04-30 08:53:36,984 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=158209.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 08:53:38,262 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=158210.0, num_to_drop=1, layers_to_drop={1} 2023-04-30 08:53:55,265 INFO [optim.py:368] (1/8) 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 08:54:01,920 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=158224.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 08:54:45,445 INFO [train.py:904] (1/8) Epoch 16, batch 6000, loss[loss=0.2036, simple_loss=0.2877, pruned_loss=0.05979, over 16327.00 frames. ], tot_loss[loss=0.2144, simple_loss=0.2998, pruned_loss=0.06448, over 3112971.05 frames. ], batch size: 146, lr: 4.24e-03, grad_scale: 8.0 2023-04-30 08:54:45,445 INFO [train.py:929] (1/8) Computing validation loss 2023-04-30 08:54:56,480 INFO [train.py:938] (1/8) Epoch 16, validation: loss=0.1553, simple_loss=0.2682, pruned_loss=0.0212, over 944034.00 frames. 2023-04-30 08:54:56,481 INFO [train.py:939] (1/8) Maximum memory allocated so far is 17857MB 2023-04-30 08:55:03,830 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([1.6975, 1.8266, 1.5967, 1.4938, 1.9326, 1.6145, 1.7141, 1.9050], device='cuda:1'), covar=tensor([0.0168, 0.0235, 0.0368, 0.0308, 0.0178, 0.0239, 0.0184, 0.0182], device='cuda:1'), in_proj_covar=tensor([0.0179, 0.0219, 0.0213, 0.0212, 0.0219, 0.0222, 0.0224, 0.0215], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-30 08:55:27,052 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=158272.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 08:56:02,728 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.6270, 1.7176, 2.2202, 2.5712, 2.6154, 2.9240, 1.8681, 2.9218], device='cuda:1'), covar=tensor([0.0172, 0.0471, 0.0301, 0.0291, 0.0262, 0.0172, 0.0462, 0.0121], device='cuda:1'), in_proj_covar=tensor([0.0174, 0.0183, 0.0169, 0.0172, 0.0183, 0.0140, 0.0183, 0.0133], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-30 08:56:13,210 INFO [train.py:904] (1/8) Epoch 16, batch 6050, loss[loss=0.2222, simple_loss=0.3347, pruned_loss=0.05479, over 16556.00 frames. ], tot_loss[loss=0.2128, simple_loss=0.2983, pruned_loss=0.06366, over 3127819.75 frames. ], batch size: 75, lr: 4.23e-03, grad_scale: 8.0 2023-04-30 08:56:40,245 INFO [optim.py:368] (1/8) 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 08:57:31,986 INFO [train.py:904] (1/8) Epoch 16, batch 6100, loss[loss=0.1824, simple_loss=0.272, pruned_loss=0.0464, over 16843.00 frames. ], tot_loss[loss=0.2112, simple_loss=0.2971, pruned_loss=0.0627, over 3117243.53 frames. ], batch size: 42, lr: 4.23e-03, grad_scale: 8.0 2023-04-30 08:57:46,551 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.9298, 4.1967, 3.9787, 4.0284, 3.7496, 3.8237, 3.8565, 4.1819], device='cuda:1'), covar=tensor([0.1061, 0.0810, 0.0958, 0.0786, 0.0776, 0.1476, 0.0904, 0.0994], device='cuda:1'), in_proj_covar=tensor([0.0603, 0.0745, 0.0610, 0.0544, 0.0466, 0.0476, 0.0618, 0.0570], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:1') 2023-04-30 08:58:23,957 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=158384.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 08:58:28,974 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.6930, 3.7877, 2.3061, 4.4793, 2.9241, 4.3821, 2.4926, 2.9720], device='cuda:1'), covar=tensor([0.0280, 0.0345, 0.1669, 0.0131, 0.0815, 0.0411, 0.1446, 0.0778], device='cuda:1'), in_proj_covar=tensor([0.0164, 0.0172, 0.0193, 0.0148, 0.0174, 0.0211, 0.0201, 0.0177], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-04-30 08:58:49,711 INFO [train.py:904] (1/8) Epoch 16, batch 6150, loss[loss=0.1825, simple_loss=0.2671, pruned_loss=0.04898, over 16504.00 frames. ], tot_loss[loss=0.211, simple_loss=0.296, pruned_loss=0.06295, over 3078256.61 frames. ], batch size: 68, lr: 4.23e-03, grad_scale: 8.0 2023-04-30 08:59:18,314 INFO [optim.py:368] (1/8) 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 08:59:55,869 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.9991, 2.0794, 2.3004, 3.5184, 2.0470, 2.4080, 2.2143, 2.2072], device='cuda:1'), covar=tensor([0.1178, 0.3336, 0.2477, 0.0546, 0.3960, 0.2334, 0.3269, 0.3117], device='cuda:1'), in_proj_covar=tensor([0.0378, 0.0419, 0.0348, 0.0319, 0.0423, 0.0483, 0.0387, 0.0489], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-30 08:59:57,077 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=158445.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 09:00:07,998 INFO [train.py:904] (1/8) Epoch 16, batch 6200, loss[loss=0.1908, simple_loss=0.2814, pruned_loss=0.05009, over 16364.00 frames. ], tot_loss[loss=0.2082, simple_loss=0.293, pruned_loss=0.06169, over 3099325.75 frames. ], batch size: 35, lr: 4.23e-03, grad_scale: 8.0 2023-04-30 09:00:33,401 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=2.68 vs. limit=5.0 2023-04-30 09:00:47,989 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=158478.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 09:01:21,942 INFO [train.py:904] (1/8) Epoch 16, batch 6250, loss[loss=0.2015, simple_loss=0.2838, pruned_loss=0.05955, over 17009.00 frames. ], tot_loss[loss=0.2087, simple_loss=0.2936, pruned_loss=0.06189, over 3105924.71 frames. ], batch size: 55, lr: 4.23e-03, grad_scale: 4.0 2023-04-30 09:01:34,522 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=158509.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 09:01:46,293 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.72 vs. limit=2.0 2023-04-30 09:01:50,901 INFO [optim.py:368] (1/8) 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 09:02:38,738 INFO [train.py:904] (1/8) Epoch 16, batch 6300, loss[loss=0.2439, simple_loss=0.311, pruned_loss=0.08841, over 11917.00 frames. ], tot_loss[loss=0.2076, simple_loss=0.293, pruned_loss=0.06108, over 3111378.99 frames. ], batch size: 248, lr: 4.23e-03, grad_scale: 4.0 2023-04-30 09:02:39,326 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([1.8205, 1.3814, 1.7432, 1.6565, 1.7840, 1.9430, 1.5837, 1.8305], device='cuda:1'), covar=tensor([0.0205, 0.0300, 0.0172, 0.0250, 0.0223, 0.0153, 0.0331, 0.0095], device='cuda:1'), in_proj_covar=tensor([0.0174, 0.0184, 0.0169, 0.0173, 0.0184, 0.0141, 0.0183, 0.0134], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-30 09:02:45,559 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=158557.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 09:03:00,509 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.2653, 4.3046, 4.1499, 3.9083, 3.8122, 4.2719, 3.9873, 3.9305], device='cuda:1'), covar=tensor([0.0673, 0.0610, 0.0329, 0.0323, 0.0955, 0.0487, 0.0698, 0.0710], device='cuda:1'), in_proj_covar=tensor([0.0266, 0.0370, 0.0312, 0.0300, 0.0325, 0.0349, 0.0215, 0.0371], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-30 09:03:55,407 INFO [train.py:904] (1/8) Epoch 16, batch 6350, loss[loss=0.181, simple_loss=0.2772, pruned_loss=0.04239, over 16860.00 frames. ], tot_loss[loss=0.2087, simple_loss=0.2936, pruned_loss=0.0619, over 3100969.27 frames. ], batch size: 96, lr: 4.23e-03, grad_scale: 4.0 2023-04-30 09:04:24,044 INFO [optim.py:368] (1/8) 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 09:05:11,913 INFO [train.py:904] (1/8) Epoch 16, batch 6400, loss[loss=0.1948, simple_loss=0.2761, pruned_loss=0.05673, over 16667.00 frames. ], tot_loss[loss=0.2094, simple_loss=0.2935, pruned_loss=0.06267, over 3102367.80 frames. ], batch size: 62, lr: 4.23e-03, grad_scale: 8.0 2023-04-30 09:06:09,485 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.1534, 4.1691, 4.0794, 3.2220, 4.1303, 1.5602, 3.8700, 3.6518], device='cuda:1'), covar=tensor([0.0111, 0.0093, 0.0180, 0.0384, 0.0089, 0.3024, 0.0166, 0.0276], device='cuda:1'), in_proj_covar=tensor([0.0148, 0.0137, 0.0184, 0.0171, 0.0156, 0.0195, 0.0171, 0.0164], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-30 09:06:17,609 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.85 vs. limit=2.0 2023-04-30 09:06:27,727 INFO [train.py:904] (1/8) Epoch 16, batch 6450, loss[loss=0.1999, simple_loss=0.2879, pruned_loss=0.05597, over 15313.00 frames. ], tot_loss[loss=0.21, simple_loss=0.2944, pruned_loss=0.06286, over 3093657.29 frames. ], batch size: 191, lr: 4.23e-03, grad_scale: 4.0 2023-04-30 09:06:31,366 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.65 vs. limit=2.0 2023-04-30 09:06:56,386 INFO [optim.py:368] (1/8) 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 09:07:25,933 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 16, batch 6500, loss[loss=0.2288, simple_loss=0.295, pruned_loss=0.08123, over 11788.00 frames. ], tot_loss[loss=0.2085, simple_loss=0.2924, pruned_loss=0.06226, over 3078406.79 frames. ], batch size: 246, lr: 4.23e-03, grad_scale: 4.0 2023-04-30 09:08:21,060 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=158777.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 09:08:22,835 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=158778.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 09:08:31,611 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=158783.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 09:08:54,727 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=5.02 vs. limit=5.0 2023-04-30 09:09:01,976 INFO [train.py:904] (1/8) Epoch 16, batch 6550, loss[loss=0.1964, simple_loss=0.2943, pruned_loss=0.04928, over 15440.00 frames. ], tot_loss[loss=0.2101, simple_loss=0.2954, pruned_loss=0.06241, over 3113246.09 frames. ], batch size: 191, lr: 4.23e-03, grad_scale: 4.0 2023-04-30 09:09:33,175 INFO [optim.py:368] (1/8) Clipping_scale=2.0, grad-norm quartiles 2.010e+02 2.836e+02 3.301e+02 3.944e+02 7.483e+02, threshold=6.603e+02, percent-clipped=1.0 2023-04-30 09:09:39,591 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=158826.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 09:09:52,653 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=158835.0, num_to_drop=1, layers_to_drop={0} 2023-04-30 09:09:58,232 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=158838.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 09:10:06,322 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=158844.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 09:10:18,011 INFO [train.py:904] (1/8) Epoch 16, batch 6600, loss[loss=0.2352, simple_loss=0.3201, pruned_loss=0.07519, over 15246.00 frames. ], tot_loss[loss=0.2112, simple_loss=0.2971, pruned_loss=0.06267, over 3110407.71 frames. ], batch size: 190, lr: 4.23e-03, grad_scale: 4.0 2023-04-30 09:10:23,537 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([5.0508, 5.3632, 5.0031, 5.0527, 4.8919, 4.7677, 4.6871, 5.4324], device='cuda:1'), covar=tensor([0.1133, 0.0738, 0.1016, 0.0789, 0.0814, 0.0841, 0.1139, 0.0863], device='cuda:1'), in_proj_covar=tensor([0.0605, 0.0744, 0.0611, 0.0547, 0.0467, 0.0478, 0.0619, 0.0571], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:1') 2023-04-30 09:10:38,281 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([1.7046, 2.6059, 2.4405, 4.4002, 3.1351, 3.9248, 1.4952, 2.8735], device='cuda:1'), covar=tensor([0.1318, 0.0827, 0.1275, 0.0157, 0.0270, 0.0448, 0.1647, 0.0827], device='cuda:1'), in_proj_covar=tensor([0.0160, 0.0166, 0.0189, 0.0170, 0.0202, 0.0212, 0.0191, 0.0187], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-04-30 09:10:45,030 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.35 vs. limit=2.0 2023-04-30 09:10:51,961 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.2574, 3.1554, 3.4228, 1.7408, 3.5847, 3.6328, 2.7825, 2.6959], device='cuda:1'), covar=tensor([0.0844, 0.0242, 0.0191, 0.1237, 0.0072, 0.0154, 0.0450, 0.0493], device='cuda:1'), in_proj_covar=tensor([0.0147, 0.0105, 0.0091, 0.0137, 0.0073, 0.0116, 0.0124, 0.0128], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:1') 2023-04-30 09:11:26,090 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=158896.0, num_to_drop=1, layers_to_drop={0} 2023-04-30 09:11:34,213 INFO [train.py:904] (1/8) Epoch 16, batch 6650, loss[loss=0.275, simple_loss=0.3389, pruned_loss=0.1056, over 11530.00 frames. ], tot_loss[loss=0.2122, simple_loss=0.2977, pruned_loss=0.06333, over 3108584.68 frames. ], batch size: 248, lr: 4.23e-03, grad_scale: 4.0 2023-04-30 09:12:04,649 INFO [optim.py:368] (1/8) 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 09:12:50,515 INFO [train.py:904] (1/8) Epoch 16, batch 6700, loss[loss=0.2062, simple_loss=0.2887, pruned_loss=0.0619, over 16704.00 frames. ], tot_loss[loss=0.2125, simple_loss=0.2968, pruned_loss=0.06409, over 3102989.18 frames. ], batch size: 124, lr: 4.23e-03, grad_scale: 4.0 2023-04-30 09:13:46,134 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.3257, 4.1431, 4.0815, 2.7076, 3.6780, 4.0918, 3.7550, 2.2419], device='cuda:1'), covar=tensor([0.0486, 0.0034, 0.0041, 0.0358, 0.0079, 0.0104, 0.0067, 0.0409], device='cuda:1'), in_proj_covar=tensor([0.0129, 0.0074, 0.0075, 0.0128, 0.0087, 0.0097, 0.0086, 0.0122], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-04-30 09:14:07,727 INFO [train.py:904] (1/8) Epoch 16, batch 6750, loss[loss=0.2193, simple_loss=0.2876, pruned_loss=0.07551, over 17006.00 frames. ], tot_loss[loss=0.2127, simple_loss=0.296, pruned_loss=0.0647, over 3085076.01 frames. ], batch size: 53, lr: 4.23e-03, grad_scale: 4.0 2023-04-30 09:14:37,800 INFO [optim.py:368] (1/8) Clipping_scale=2.0, grad-norm quartiles 2.101e+02 2.962e+02 3.496e+02 4.058e+02 1.383e+03, threshold=6.992e+02, percent-clipped=2.0 2023-04-30 09:15:05,246 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 16, batch 6800, loss[loss=0.2483, simple_loss=0.3079, pruned_loss=0.09438, over 11676.00 frames. ], tot_loss[loss=0.2124, simple_loss=0.296, pruned_loss=0.0644, over 3089800.13 frames. ], batch size: 248, lr: 4.22e-03, grad_scale: 8.0 2023-04-30 09:15:50,708 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=159069.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 09:16:20,374 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=159088.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 09:16:40,955 INFO [train.py:904] (1/8) Epoch 16, batch 6850, loss[loss=0.212, simple_loss=0.3159, pruned_loss=0.05405, over 16873.00 frames. ], tot_loss[loss=0.2127, simple_loss=0.2966, pruned_loss=0.06446, over 3082040.52 frames. ], batch size: 90, lr: 4.22e-03, grad_scale: 2.0 2023-04-30 09:17:12,961 INFO [optim.py:368] (1/8) Clipping_scale=2.0, grad-norm quartiles 1.918e+02 2.851e+02 3.389e+02 4.147e+02 9.414e+02, threshold=6.778e+02, percent-clipped=4.0 2023-04-30 09:17:22,455 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=159130.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 09:17:26,686 INFO [zipformer.py:625] (1/8) 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] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=159139.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 09:17:40,271 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.58 vs. limit=2.0 2023-04-30 09:17:54,571 INFO [train.py:904] (1/8) Epoch 16, batch 6900, loss[loss=0.2068, simple_loss=0.2985, pruned_loss=0.05752, over 16895.00 frames. ], tot_loss[loss=0.2141, simple_loss=0.2991, pruned_loss=0.06458, over 3082713.56 frames. ], batch size: 109, lr: 4.22e-03, grad_scale: 2.0 2023-04-30 09:18:07,204 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=159159.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 09:18:56,497 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=159191.0, num_to_drop=1, layers_to_drop={0} 2023-04-30 09:19:13,870 INFO [train.py:904] (1/8) Epoch 16, batch 6950, loss[loss=0.183, simple_loss=0.2736, pruned_loss=0.04623, over 16539.00 frames. ], tot_loss[loss=0.2156, simple_loss=0.3, pruned_loss=0.06557, over 3090254.71 frames. ], batch size: 75, lr: 4.22e-03, grad_scale: 2.0 2023-04-30 09:19:42,091 INFO [zipformer.py:625] (1/8) 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] (1/8) 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 09:20:29,895 INFO [train.py:904] (1/8) Epoch 16, batch 7000, loss[loss=0.1998, simple_loss=0.3062, pruned_loss=0.04673, over 16837.00 frames. ], tot_loss[loss=0.2149, simple_loss=0.3001, pruned_loss=0.06489, over 3095656.89 frames. ], batch size: 83, lr: 4.22e-03, grad_scale: 2.0 2023-04-30 09:21:17,213 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.25 vs. limit=2.0 2023-04-30 09:21:17,427 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.77 vs. limit=2.0 2023-04-30 09:21:27,377 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=159290.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 09:21:45,218 INFO [train.py:904] (1/8) Epoch 16, batch 7050, loss[loss=0.2513, simple_loss=0.3169, pruned_loss=0.09284, over 11265.00 frames. ], tot_loss[loss=0.2153, simple_loss=0.3011, pruned_loss=0.0647, over 3092853.79 frames. ], batch size: 247, lr: 4.22e-03, grad_scale: 2.0 2023-04-30 09:22:18,894 INFO [optim.py:368] (1/8) 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 09:23:00,656 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 16, batch 7100, loss[loss=0.2103, simple_loss=0.291, pruned_loss=0.06479, over 15410.00 frames. ], tot_loss[loss=0.2144, simple_loss=0.2998, pruned_loss=0.06449, over 3093064.64 frames. ], batch size: 190, lr: 4.22e-03, grad_scale: 2.0 2023-04-30 09:24:17,488 INFO [train.py:904] (1/8) Epoch 16, batch 7150, loss[loss=0.2055, simple_loss=0.2894, pruned_loss=0.06076, over 16900.00 frames. ], tot_loss[loss=0.2134, simple_loss=0.2981, pruned_loss=0.06435, over 3098291.94 frames. ], batch size: 116, lr: 4.22e-03, grad_scale: 2.0 2023-04-30 09:24:49,380 INFO [optim.py:368] (1/8) 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 09:24:51,596 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=159425.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 09:25:02,361 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=159433.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 09:25:11,745 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=159439.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 09:25:30,654 INFO [train.py:904] (1/8) Epoch 16, batch 7200, loss[loss=0.184, simple_loss=0.2767, pruned_loss=0.04561, over 16182.00 frames. ], tot_loss[loss=0.2099, simple_loss=0.2953, pruned_loss=0.06221, over 3093561.42 frames. ], batch size: 165, lr: 4.22e-03, grad_scale: 4.0 2023-04-30 09:26:15,845 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=159481.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 09:26:26,898 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=159487.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 09:26:33,343 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=159491.0, num_to_drop=1, layers_to_drop={1} 2023-04-30 09:26:50,665 INFO [train.py:904] (1/8) Epoch 16, batch 7250, loss[loss=0.1801, simple_loss=0.2692, pruned_loss=0.04544, over 16652.00 frames. ], tot_loss[loss=0.2082, simple_loss=0.2935, pruned_loss=0.06148, over 3060763.87 frames. ], batch size: 134, lr: 4.22e-03, grad_scale: 4.0 2023-04-30 09:26:51,220 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.9591, 4.9523, 4.7588, 4.1180, 4.8123, 1.9074, 4.6043, 4.5189], device='cuda:1'), covar=tensor([0.0068, 0.0059, 0.0151, 0.0326, 0.0070, 0.2441, 0.0093, 0.0178], device='cuda:1'), in_proj_covar=tensor([0.0146, 0.0135, 0.0182, 0.0168, 0.0154, 0.0194, 0.0168, 0.0161], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-30 09:27:11,557 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=159515.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 09:27:21,048 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([1.6874, 1.8155, 1.5787, 1.5917, 1.9523, 1.6600, 1.6678, 1.9174], device='cuda:1'), covar=tensor([0.0152, 0.0244, 0.0369, 0.0319, 0.0185, 0.0226, 0.0159, 0.0176], device='cuda:1'), in_proj_covar=tensor([0.0177, 0.0217, 0.0211, 0.0211, 0.0217, 0.0217, 0.0220, 0.0212], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-30 09:27:23,390 INFO [optim.py:368] (1/8) 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 09:27:28,926 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.2453, 2.2920, 2.4249, 4.1207, 2.1854, 2.7023, 2.3821, 2.4832], device='cuda:1'), covar=tensor([0.1258, 0.3523, 0.2551, 0.0459, 0.4023, 0.2411, 0.3528, 0.3313], device='cuda:1'), in_proj_covar=tensor([0.0378, 0.0418, 0.0347, 0.0315, 0.0423, 0.0481, 0.0387, 0.0487], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-30 09:27:40,642 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.1375, 2.4546, 1.9618, 2.2164, 2.7456, 2.4259, 2.8490, 2.9794], device='cuda:1'), covar=tensor([0.0125, 0.0341, 0.0469, 0.0395, 0.0244, 0.0343, 0.0214, 0.0225], device='cuda:1'), in_proj_covar=tensor([0.0177, 0.0216, 0.0210, 0.0211, 0.0217, 0.0217, 0.0219, 0.0212], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-30 09:27:46,672 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=159539.0, num_to_drop=1, layers_to_drop={0} 2023-04-30 09:28:06,273 INFO [train.py:904] (1/8) Epoch 16, batch 7300, loss[loss=0.2064, simple_loss=0.2891, pruned_loss=0.06179, over 16647.00 frames. ], tot_loss[loss=0.2076, simple_loss=0.2928, pruned_loss=0.06116, over 3073650.23 frames. ], batch size: 62, lr: 4.22e-03, grad_scale: 4.0 2023-04-30 09:29:22,389 INFO [train.py:904] (1/8) Epoch 16, batch 7350, loss[loss=0.2607, simple_loss=0.3254, pruned_loss=0.098, over 11000.00 frames. ], tot_loss[loss=0.2096, simple_loss=0.2942, pruned_loss=0.06248, over 3042845.27 frames. ], batch size: 246, lr: 4.22e-03, grad_scale: 4.0 2023-04-30 09:29:33,085 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.5442, 4.5570, 4.4110, 4.1274, 4.1179, 4.5032, 4.2565, 4.2036], device='cuda:1'), covar=tensor([0.0546, 0.0422, 0.0257, 0.0294, 0.0823, 0.0398, 0.0488, 0.0621], device='cuda:1'), in_proj_covar=tensor([0.0261, 0.0364, 0.0305, 0.0294, 0.0318, 0.0342, 0.0211, 0.0364], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-30 09:29:56,735 INFO [optim.py:368] (1/8) 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 09:30:06,080 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.23 vs. limit=2.0 2023-04-30 09:30:07,692 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=159631.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 09:30:27,893 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.7036, 3.6952, 3.0059, 2.3167, 2.7783, 2.4937, 4.0201, 3.4618], device='cuda:1'), covar=tensor([0.2807, 0.0808, 0.1645, 0.2528, 0.2277, 0.1836, 0.0508, 0.1231], device='cuda:1'), in_proj_covar=tensor([0.0317, 0.0264, 0.0297, 0.0298, 0.0290, 0.0241, 0.0282, 0.0319], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:1') 2023-04-30 09:30:31,262 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=159646.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 09:30:39,217 INFO [train.py:904] (1/8) Epoch 16, batch 7400, loss[loss=0.2213, simple_loss=0.3046, pruned_loss=0.069, over 15435.00 frames. ], tot_loss[loss=0.211, simple_loss=0.2953, pruned_loss=0.06332, over 3044699.11 frames. ], batch size: 191, lr: 4.22e-03, grad_scale: 4.0 2023-04-30 09:31:41,293 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=159692.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 09:31:57,479 INFO [train.py:904] (1/8) Epoch 16, batch 7450, loss[loss=0.2232, simple_loss=0.3132, pruned_loss=0.06663, over 16780.00 frames. ], tot_loss[loss=0.213, simple_loss=0.2968, pruned_loss=0.06461, over 3047619.83 frames. ], batch size: 124, lr: 4.22e-03, grad_scale: 4.0 2023-04-30 09:32:33,422 INFO [optim.py:368] (1/8) 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] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=159725.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 09:33:17,754 INFO [train.py:904] (1/8) Epoch 16, batch 7500, loss[loss=0.2061, simple_loss=0.2914, pruned_loss=0.06041, over 16923.00 frames. ], tot_loss[loss=0.2116, simple_loss=0.2962, pruned_loss=0.06348, over 3060593.53 frames. ], batch size: 116, lr: 4.22e-03, grad_scale: 4.0 2023-04-30 09:33:50,299 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=159773.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 09:33:52,430 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.4072, 1.5785, 2.0479, 2.3640, 2.3964, 2.7116, 1.8126, 2.6590], device='cuda:1'), covar=tensor([0.0199, 0.0526, 0.0310, 0.0300, 0.0305, 0.0184, 0.0486, 0.0127], device='cuda:1'), in_proj_covar=tensor([0.0173, 0.0184, 0.0168, 0.0173, 0.0183, 0.0140, 0.0183, 0.0133], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-30 09:33:54,977 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.6831, 4.6552, 4.4524, 3.5181, 4.5268, 1.5542, 4.2752, 4.1948], device='cuda:1'), covar=tensor([0.0100, 0.0080, 0.0205, 0.0502, 0.0110, 0.3001, 0.0135, 0.0265], device='cuda:1'), in_proj_covar=tensor([0.0145, 0.0133, 0.0180, 0.0165, 0.0152, 0.0191, 0.0166, 0.0160], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-30 09:34:37,012 INFO [train.py:904] (1/8) Epoch 16, batch 7550, loss[loss=0.1883, simple_loss=0.2814, pruned_loss=0.04763, over 16839.00 frames. ], tot_loss[loss=0.2098, simple_loss=0.2944, pruned_loss=0.06263, over 3085955.99 frames. ], batch size: 116, lr: 4.21e-03, grad_scale: 4.0 2023-04-30 09:34:38,821 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.6011, 1.6213, 2.1976, 2.5172, 2.5308, 2.8573, 1.8827, 2.7901], device='cuda:1'), covar=tensor([0.0164, 0.0496, 0.0291, 0.0266, 0.0269, 0.0155, 0.0479, 0.0121], device='cuda:1'), in_proj_covar=tensor([0.0173, 0.0184, 0.0168, 0.0172, 0.0183, 0.0140, 0.0183, 0.0133], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-30 09:34:58,292 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=159815.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 09:35:10,140 INFO [optim.py:368] (1/8) 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 09:35:21,327 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.32 vs. limit=2.0 2023-04-30 09:35:53,836 INFO [train.py:904] (1/8) Epoch 16, batch 7600, loss[loss=0.2528, simple_loss=0.3163, pruned_loss=0.09464, over 11480.00 frames. ], tot_loss[loss=0.2099, simple_loss=0.2937, pruned_loss=0.06303, over 3079290.92 frames. ], batch size: 248, lr: 4.21e-03, grad_scale: 8.0 2023-04-30 09:35:59,095 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=159855.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 09:36:10,414 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=159863.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 09:37:09,858 INFO [train.py:904] (1/8) Epoch 16, batch 7650, loss[loss=0.2817, simple_loss=0.3308, pruned_loss=0.1163, over 11366.00 frames. ], tot_loss[loss=0.2115, simple_loss=0.2948, pruned_loss=0.06411, over 3079550.62 frames. ], batch size: 248, lr: 4.21e-03, grad_scale: 8.0 2023-04-30 09:37:21,645 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=159910.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 09:37:30,436 INFO [zipformer.py:625] (1/8) 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] (1/8) Clipping_scale=2.0, grad-norm quartiles 1.780e+02 3.207e+02 4.068e+02 5.310e+02 2.281e+03, threshold=8.135e+02, percent-clipped=15.0 2023-04-30 09:38:13,630 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=159946.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 09:38:22,086 INFO [train.py:904] (1/8) Epoch 16, batch 7700, loss[loss=0.2177, simple_loss=0.2968, pruned_loss=0.06933, over 16751.00 frames. ], tot_loss[loss=0.2122, simple_loss=0.2953, pruned_loss=0.06459, over 3075256.90 frames. ], batch size: 124, lr: 4.21e-03, grad_scale: 4.0 2023-04-30 09:38:41,836 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([1.8809, 3.0093, 2.7384, 5.2740, 4.1130, 4.3898, 1.6367, 3.1517], device='cuda:1'), covar=tensor([0.1279, 0.0722, 0.1206, 0.0139, 0.0409, 0.0407, 0.1582, 0.0850], device='cuda:1'), in_proj_covar=tensor([0.0161, 0.0166, 0.0188, 0.0169, 0.0203, 0.0212, 0.0193, 0.0188], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-04-30 09:38:50,557 INFO [zipformer.py:625] (1/8) 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] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=159987.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 09:39:26,410 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=159994.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 09:39:39,463 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.9673, 5.2512, 5.5180, 5.2682, 5.3149, 5.8574, 5.2782, 5.0916], device='cuda:1'), covar=tensor([0.0875, 0.1831, 0.2151, 0.1685, 0.2131, 0.0833, 0.1453, 0.2146], device='cuda:1'), in_proj_covar=tensor([0.0384, 0.0551, 0.0601, 0.0463, 0.0613, 0.0630, 0.0473, 0.0621], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-04-30 09:39:41,049 INFO [train.py:904] (1/8) Epoch 16, batch 7750, loss[loss=0.1953, simple_loss=0.2848, pruned_loss=0.05286, over 16542.00 frames. ], tot_loss[loss=0.2117, simple_loss=0.2948, pruned_loss=0.06427, over 3075271.20 frames. ], batch size: 68, lr: 4.21e-03, grad_scale: 4.0 2023-04-30 09:40:13,568 INFO [optim.py:368] (1/8) 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 09:40:33,200 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.0264, 4.0708, 4.4023, 4.3863, 4.3995, 4.1023, 4.1226, 4.0546], device='cuda:1'), covar=tensor([0.0346, 0.0641, 0.0427, 0.0456, 0.0471, 0.0491, 0.1004, 0.0552], device='cuda:1'), in_proj_covar=tensor([0.0377, 0.0408, 0.0397, 0.0377, 0.0449, 0.0422, 0.0516, 0.0334], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:1') 2023-04-30 09:40:41,466 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.31 vs. limit=2.0 2023-04-30 09:40:53,467 INFO [train.py:904] (1/8) Epoch 16, batch 7800, loss[loss=0.2249, simple_loss=0.3118, pruned_loss=0.06901, over 16906.00 frames. ], tot_loss[loss=0.2121, simple_loss=0.2956, pruned_loss=0.06428, over 3095057.00 frames. ], batch size: 109, lr: 4.21e-03, grad_scale: 4.0 2023-04-30 09:42:08,841 INFO [train.py:904] (1/8) Epoch 16, batch 7850, loss[loss=0.2059, simple_loss=0.294, pruned_loss=0.05888, over 16765.00 frames. ], tot_loss[loss=0.2126, simple_loss=0.2967, pruned_loss=0.06422, over 3088671.30 frames. ], batch size: 124, lr: 4.21e-03, grad_scale: 4.0 2023-04-30 09:42:12,591 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.30 vs. limit=2.0 2023-04-30 09:42:43,406 INFO [optim.py:368] (1/8) 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 09:42:44,063 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.1275, 3.3582, 3.3903, 2.2307, 3.1256, 3.3466, 3.1877, 1.8352], device='cuda:1'), covar=tensor([0.0463, 0.0056, 0.0052, 0.0382, 0.0092, 0.0111, 0.0082, 0.0452], device='cuda:1'), in_proj_covar=tensor([0.0130, 0.0073, 0.0074, 0.0128, 0.0087, 0.0097, 0.0086, 0.0121], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-04-30 09:43:03,298 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.31 vs. limit=2.0 2023-04-30 09:43:25,067 INFO [train.py:904] (1/8) Epoch 16, batch 7900, loss[loss=0.2081, simple_loss=0.2992, pruned_loss=0.05851, over 15353.00 frames. ], tot_loss[loss=0.211, simple_loss=0.2954, pruned_loss=0.06335, over 3090867.64 frames. ], batch size: 191, lr: 4.21e-03, grad_scale: 4.0 2023-04-30 09:43:54,612 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.8327, 5.1173, 5.3563, 5.1086, 5.1605, 5.7136, 5.1820, 5.0073], device='cuda:1'), covar=tensor([0.0973, 0.1975, 0.2004, 0.1786, 0.2273, 0.0887, 0.1516, 0.2229], device='cuda:1'), in_proj_covar=tensor([0.0387, 0.0556, 0.0605, 0.0466, 0.0619, 0.0636, 0.0477, 0.0628], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-04-30 09:44:43,662 INFO [train.py:904] (1/8) Epoch 16, batch 7950, loss[loss=0.1896, simple_loss=0.2771, pruned_loss=0.05103, over 16873.00 frames. ], tot_loss[loss=0.2111, simple_loss=0.2959, pruned_loss=0.06316, over 3104415.22 frames. ], batch size: 96, lr: 4.21e-03, grad_scale: 4.0 2023-04-30 09:44:56,864 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=160211.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 09:45:16,536 INFO [optim.py:368] (1/8) 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] (1/8) Epoch 16, batch 8000, loss[loss=0.2079, simple_loss=0.2942, pruned_loss=0.06083, over 16689.00 frames. ], tot_loss[loss=0.212, simple_loss=0.2963, pruned_loss=0.06382, over 3105457.32 frames. ], batch size: 57, lr: 4.21e-03, grad_scale: 8.0 2023-04-30 09:46:17,933 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=160266.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 09:46:50,255 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=160287.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 09:47:12,591 INFO [train.py:904] (1/8) Epoch 16, batch 8050, loss[loss=0.1976, simple_loss=0.29, pruned_loss=0.05262, over 16598.00 frames. ], tot_loss[loss=0.2113, simple_loss=0.296, pruned_loss=0.06331, over 3115553.37 frames. ], batch size: 68, lr: 4.21e-03, grad_scale: 8.0 2023-04-30 09:47:14,302 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.0078, 3.7071, 3.4523, 4.0717, 4.1089, 3.9739, 4.1818, 4.1485], device='cuda:1'), covar=tensor([0.1561, 0.1734, 0.3450, 0.1333, 0.1417, 0.2112, 0.1552, 0.1770], device='cuda:1'), in_proj_covar=tensor([0.0581, 0.0715, 0.0842, 0.0718, 0.0543, 0.0570, 0.0584, 0.0677], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-30 09:47:20,318 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=160307.0, num_to_drop=1, layers_to_drop={1} 2023-04-30 09:47:42,422 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.9743, 2.0276, 2.1463, 3.5022, 2.0229, 2.3747, 2.1594, 2.1946], device='cuda:1'), covar=tensor([0.1255, 0.3614, 0.2631, 0.0543, 0.4118, 0.2515, 0.3266, 0.3287], device='cuda:1'), in_proj_covar=tensor([0.0378, 0.0421, 0.0347, 0.0317, 0.0426, 0.0483, 0.0389, 0.0489], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-30 09:47:47,809 INFO [optim.py:368] (1/8) 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] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=160335.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 09:48:29,600 INFO [train.py:904] (1/8) Epoch 16, batch 8100, loss[loss=0.2149, simple_loss=0.2979, pruned_loss=0.06595, over 16626.00 frames. ], tot_loss[loss=0.2111, simple_loss=0.2958, pruned_loss=0.06321, over 3099502.56 frames. ], batch size: 57, lr: 4.21e-03, grad_scale: 8.0 2023-04-30 09:48:54,911 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=160368.0, num_to_drop=1, layers_to_drop={0} 2023-04-30 09:49:00,292 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([1.8372, 2.7158, 2.6420, 1.9722, 2.5471, 2.6693, 2.5830, 1.8656], device='cuda:1'), covar=tensor([0.0398, 0.0071, 0.0069, 0.0320, 0.0106, 0.0106, 0.0101, 0.0376], device='cuda:1'), in_proj_covar=tensor([0.0131, 0.0074, 0.0075, 0.0130, 0.0088, 0.0098, 0.0086, 0.0122], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-04-30 09:49:46,004 INFO [train.py:904] (1/8) Epoch 16, batch 8150, loss[loss=0.18, simple_loss=0.2635, pruned_loss=0.04821, over 16729.00 frames. ], tot_loss[loss=0.2084, simple_loss=0.2929, pruned_loss=0.06191, over 3124893.59 frames. ], batch size: 89, lr: 4.21e-03, grad_scale: 8.0 2023-04-30 09:50:21,735 INFO [optim.py:368] (1/8) 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 09:50:37,271 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.8815, 2.3687, 1.9113, 2.2048, 2.7218, 2.4255, 2.7666, 2.9109], device='cuda:1'), covar=tensor([0.0154, 0.0391, 0.0507, 0.0408, 0.0234, 0.0340, 0.0213, 0.0237], device='cuda:1'), in_proj_covar=tensor([0.0176, 0.0216, 0.0210, 0.0210, 0.0215, 0.0216, 0.0220, 0.0212], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-30 09:50:50,859 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.1913, 3.4225, 3.5663, 3.5408, 3.5500, 3.3621, 3.4172, 3.4379], device='cuda:1'), covar=tensor([0.0400, 0.0662, 0.0438, 0.0475, 0.0504, 0.0542, 0.0791, 0.0511], device='cuda:1'), in_proj_covar=tensor([0.0377, 0.0408, 0.0399, 0.0376, 0.0449, 0.0421, 0.0516, 0.0334], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:1') 2023-04-30 09:51:05,065 INFO [train.py:904] (1/8) Epoch 16, batch 8200, loss[loss=0.1947, simple_loss=0.2867, pruned_loss=0.05137, over 16861.00 frames. ], tot_loss[loss=0.2059, simple_loss=0.2901, pruned_loss=0.06088, over 3135591.98 frames. ], batch size: 116, lr: 4.21e-03, grad_scale: 8.0 2023-04-30 09:51:08,982 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.52 vs. limit=2.0 2023-04-30 09:51:12,513 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.0616, 1.9865, 2.1733, 3.6582, 1.9538, 2.3949, 2.0783, 2.1639], device='cuda:1'), covar=tensor([0.1223, 0.3666, 0.2687, 0.0502, 0.4382, 0.2528, 0.3704, 0.3371], device='cuda:1'), in_proj_covar=tensor([0.0378, 0.0420, 0.0347, 0.0317, 0.0426, 0.0482, 0.0388, 0.0489], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-30 09:52:27,311 INFO [train.py:904] (1/8) Epoch 16, batch 8250, loss[loss=0.1933, simple_loss=0.2862, pruned_loss=0.05022, over 12033.00 frames. ], tot_loss[loss=0.2032, simple_loss=0.2889, pruned_loss=0.0587, over 3108980.38 frames. ], batch size: 247, lr: 4.21e-03, grad_scale: 8.0 2023-04-30 09:52:42,369 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=160511.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 09:53:04,606 INFO [optim.py:368] (1/8) 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 09:53:18,858 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([1.7310, 3.2943, 2.9723, 5.1540, 4.1112, 4.7175, 1.7880, 3.3907], device='cuda:1'), covar=tensor([0.1332, 0.0614, 0.0948, 0.0140, 0.0210, 0.0271, 0.1466, 0.0628], device='cuda:1'), in_proj_covar=tensor([0.0160, 0.0165, 0.0187, 0.0169, 0.0202, 0.0210, 0.0191, 0.0187], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-04-30 09:53:35,724 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.32 vs. limit=2.0 2023-04-30 09:53:49,472 INFO [train.py:904] (1/8) Epoch 16, batch 8300, loss[loss=0.179, simple_loss=0.2613, pruned_loss=0.04833, over 11698.00 frames. ], tot_loss[loss=0.1985, simple_loss=0.286, pruned_loss=0.05554, over 3100177.94 frames. ], batch size: 246, lr: 4.21e-03, grad_scale: 8.0 2023-04-30 09:53:57,027 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.83 vs. limit=2.0 2023-04-30 09:54:00,901 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=160559.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 09:54:13,047 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 16, batch 8350, loss[loss=0.1818, simple_loss=0.277, pruned_loss=0.04334, over 16726.00 frames. ], tot_loss[loss=0.1967, simple_loss=0.2855, pruned_loss=0.05397, over 3087439.07 frames. ], batch size: 124, lr: 4.20e-03, grad_scale: 8.0 2023-04-30 09:55:30,914 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=160614.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 09:55:48,785 INFO [optim.py:368] (1/8) 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 09:56:33,083 INFO [train.py:904] (1/8) Epoch 16, batch 8400, loss[loss=0.188, simple_loss=0.2886, pruned_loss=0.04372, over 16876.00 frames. ], tot_loss[loss=0.1936, simple_loss=0.2829, pruned_loss=0.05214, over 3067697.62 frames. ], batch size: 116, lr: 4.20e-03, grad_scale: 8.0 2023-04-30 09:56:50,650 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=160663.0, num_to_drop=1, layers_to_drop={0} 2023-04-30 09:57:29,540 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=2.89 vs. limit=5.0 2023-04-30 09:57:50,231 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.5799, 3.5676, 3.5032, 2.7067, 3.4568, 1.9528, 3.2837, 2.8887], device='cuda:1'), covar=tensor([0.0139, 0.0117, 0.0188, 0.0213, 0.0110, 0.2466, 0.0133, 0.0225], device='cuda:1'), in_proj_covar=tensor([0.0143, 0.0132, 0.0178, 0.0163, 0.0150, 0.0190, 0.0165, 0.0157], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-30 09:57:54,559 INFO [train.py:904] (1/8) Epoch 16, batch 8450, loss[loss=0.1758, simple_loss=0.2766, pruned_loss=0.03747, over 16721.00 frames. ], tot_loss[loss=0.1919, simple_loss=0.2818, pruned_loss=0.05096, over 3061546.76 frames. ], batch size: 89, lr: 4.20e-03, grad_scale: 8.0 2023-04-30 09:58:31,817 INFO [optim.py:368] (1/8) 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 09:58:35,057 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.4058, 3.3467, 3.4676, 3.5213, 3.5963, 3.3061, 3.5436, 3.6275], device='cuda:1'), covar=tensor([0.1190, 0.0927, 0.0989, 0.0589, 0.0557, 0.2576, 0.0870, 0.0756], device='cuda:1'), in_proj_covar=tensor([0.0562, 0.0697, 0.0817, 0.0704, 0.0530, 0.0554, 0.0565, 0.0663], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-30 09:59:15,545 INFO [train.py:904] (1/8) Epoch 16, batch 8500, loss[loss=0.1922, simple_loss=0.2679, pruned_loss=0.05822, over 12176.00 frames. ], tot_loss[loss=0.1876, simple_loss=0.2779, pruned_loss=0.04866, over 3048471.95 frames. ], batch size: 247, lr: 4.20e-03, grad_scale: 8.0 2023-04-30 09:59:57,085 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.0510, 3.1188, 1.9692, 3.2868, 2.3726, 3.3354, 2.1087, 2.6718], device='cuda:1'), covar=tensor([0.0312, 0.0351, 0.1512, 0.0278, 0.0835, 0.0513, 0.1463, 0.0690], device='cuda:1'), in_proj_covar=tensor([0.0159, 0.0166, 0.0187, 0.0143, 0.0169, 0.0204, 0.0197, 0.0173], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0003, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-04-30 09:59:59,694 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.1230, 3.2904, 3.3546, 2.2591, 3.0219, 3.3408, 3.1986, 2.0214], device='cuda:1'), covar=tensor([0.0480, 0.0059, 0.0044, 0.0361, 0.0099, 0.0080, 0.0073, 0.0452], device='cuda:1'), in_proj_covar=tensor([0.0131, 0.0074, 0.0075, 0.0129, 0.0088, 0.0097, 0.0086, 0.0122], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-04-30 10:00:20,459 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.6000, 3.6823, 2.8886, 2.1452, 2.3460, 2.3875, 3.8671, 3.2739], device='cuda:1'), covar=tensor([0.2773, 0.0603, 0.1606, 0.2924, 0.2999, 0.2017, 0.0428, 0.1290], device='cuda:1'), in_proj_covar=tensor([0.0309, 0.0255, 0.0288, 0.0290, 0.0280, 0.0235, 0.0274, 0.0310], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-30 10:00:42,445 INFO [train.py:904] (1/8) Epoch 16, batch 8550, loss[loss=0.2065, simple_loss=0.3126, pruned_loss=0.05017, over 15420.00 frames. ], tot_loss[loss=0.1855, simple_loss=0.2759, pruned_loss=0.04752, over 3048477.45 frames. ], batch size: 190, lr: 4.20e-03, grad_scale: 8.0 2023-04-30 10:01:04,505 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.6801, 3.0269, 3.4018, 1.8719, 2.7083, 2.0280, 3.2666, 3.2307], device='cuda:1'), covar=tensor([0.0259, 0.0860, 0.0489, 0.2053, 0.0869, 0.1101, 0.0667, 0.1018], device='cuda:1'), in_proj_covar=tensor([0.0146, 0.0152, 0.0159, 0.0145, 0.0138, 0.0124, 0.0138, 0.0162], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:1') 2023-04-30 10:01:27,141 INFO [optim.py:368] (1/8) Clipping_scale=2.0, grad-norm quartiles 1.280e+02 2.208e+02 2.559e+02 3.052e+02 5.214e+02, threshold=5.118e+02, percent-clipped=0.0 2023-04-30 10:01:34,393 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.18 vs. limit=2.0 2023-04-30 10:02:22,610 INFO [train.py:904] (1/8) Epoch 16, batch 8600, loss[loss=0.1955, simple_loss=0.2885, pruned_loss=0.05122, over 16703.00 frames. ], tot_loss[loss=0.1848, simple_loss=0.2765, pruned_loss=0.04655, over 3063001.07 frames. ], batch size: 124, lr: 4.20e-03, grad_scale: 8.0 2023-04-30 10:02:54,902 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.8061, 3.8432, 2.4835, 4.4272, 2.9161, 4.3489, 2.5441, 3.1991], device='cuda:1'), covar=tensor([0.0250, 0.0314, 0.1451, 0.0159, 0.0736, 0.0453, 0.1498, 0.0635], device='cuda:1'), in_proj_covar=tensor([0.0158, 0.0165, 0.0186, 0.0142, 0.0168, 0.0203, 0.0196, 0.0172], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0003, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-04-30 10:03:46,725 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.1823, 2.0780, 2.1488, 3.8165, 2.0105, 2.4375, 2.1968, 2.2076], device='cuda:1'), covar=tensor([0.1145, 0.3818, 0.2816, 0.0486, 0.4310, 0.2553, 0.3568, 0.3496], device='cuda:1'), in_proj_covar=tensor([0.0370, 0.0412, 0.0342, 0.0309, 0.0417, 0.0470, 0.0380, 0.0477], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-30 10:04:02,597 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 16, batch 8650, loss[loss=0.1798, simple_loss=0.264, pruned_loss=0.04782, over 12284.00 frames. ], tot_loss[loss=0.1821, simple_loss=0.2742, pruned_loss=0.04503, over 3052899.56 frames. ], batch size: 248, lr: 4.20e-03, grad_scale: 8.0 2023-04-30 10:04:24,135 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.8196, 3.9196, 2.6163, 4.6451, 2.9911, 4.5535, 2.6550, 3.2467], device='cuda:1'), covar=tensor([0.0274, 0.0335, 0.1459, 0.0164, 0.0824, 0.0380, 0.1497, 0.0687], device='cuda:1'), in_proj_covar=tensor([0.0158, 0.0165, 0.0186, 0.0142, 0.0168, 0.0203, 0.0196, 0.0172], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0003, 0.0004, 0.0004, 0.0003], device='cuda:1') 2023-04-30 10:04:56,266 INFO [optim.py:368] (1/8) 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 10:05:52,475 INFO [train.py:904] (1/8) Epoch 16, batch 8700, loss[loss=0.1647, simple_loss=0.2515, pruned_loss=0.03889, over 12366.00 frames. ], tot_loss[loss=0.1797, simple_loss=0.2716, pruned_loss=0.04391, over 3053571.01 frames. ], batch size: 248, lr: 4.20e-03, grad_scale: 8.0 2023-04-30 10:06:13,093 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=160962.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 10:06:15,080 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=160963.0, num_to_drop=1, layers_to_drop={0} 2023-04-30 10:06:16,923 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.61 vs. limit=2.0 2023-04-30 10:06:18,473 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=160965.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 10:06:18,641 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.2549, 2.1041, 2.1706, 4.0418, 2.0576, 2.5210, 2.2369, 2.2886], device='cuda:1'), covar=tensor([0.1111, 0.3872, 0.2874, 0.0399, 0.4222, 0.2454, 0.3598, 0.3232], device='cuda:1'), in_proj_covar=tensor([0.0368, 0.0409, 0.0340, 0.0307, 0.0414, 0.0467, 0.0378, 0.0474], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-30 10:06:20,567 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.6795, 4.0423, 2.9750, 2.1940, 2.6075, 2.4051, 4.2912, 3.3772], device='cuda:1'), covar=tensor([0.2899, 0.0542, 0.1820, 0.2860, 0.2877, 0.2072, 0.0389, 0.1249], device='cuda:1'), in_proj_covar=tensor([0.0309, 0.0255, 0.0288, 0.0290, 0.0278, 0.0235, 0.0274, 0.0309], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-30 10:07:29,336 INFO [train.py:904] (1/8) Epoch 16, batch 8750, loss[loss=0.1517, simple_loss=0.2433, pruned_loss=0.03007, over 12397.00 frames. ], tot_loss[loss=0.1794, simple_loss=0.2717, pruned_loss=0.04348, over 3072216.46 frames. ], batch size: 248, lr: 4.20e-03, grad_scale: 4.0 2023-04-30 10:07:53,442 INFO [zipformer.py:625] (1/8) 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] (1/8) 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 10:08:28,314 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=161026.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 10:09:20,776 INFO [train.py:904] (1/8) Epoch 16, batch 8800, loss[loss=0.1668, simple_loss=0.2637, pruned_loss=0.03491, over 16903.00 frames. ], tot_loss[loss=0.1779, simple_loss=0.2704, pruned_loss=0.04271, over 3065556.15 frames. ], batch size: 96, lr: 4.20e-03, grad_scale: 8.0 2023-04-30 10:10:20,324 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.9882, 2.9873, 1.7965, 3.2419, 2.1263, 3.3001, 1.9426, 2.4373], device='cuda:1'), covar=tensor([0.0359, 0.0434, 0.1963, 0.0354, 0.1119, 0.0572, 0.1839, 0.1023], device='cuda:1'), in_proj_covar=tensor([0.0158, 0.0165, 0.0187, 0.0142, 0.0168, 0.0202, 0.0197, 0.0172], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0003, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-04-30 10:10:26,490 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.9152, 1.9765, 2.1539, 3.4747, 1.9084, 2.2209, 2.1167, 2.0729], device='cuda:1'), covar=tensor([0.1214, 0.3928, 0.2776, 0.0535, 0.4489, 0.2726, 0.3524, 0.3671], device='cuda:1'), in_proj_covar=tensor([0.0370, 0.0411, 0.0342, 0.0309, 0.0417, 0.0470, 0.0380, 0.0477], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-30 10:11:05,639 INFO [train.py:904] (1/8) Epoch 16, batch 8850, loss[loss=0.1609, simple_loss=0.2496, pruned_loss=0.03604, over 12382.00 frames. ], tot_loss[loss=0.1783, simple_loss=0.2729, pruned_loss=0.04186, over 3067379.58 frames. ], batch size: 248, lr: 4.20e-03, grad_scale: 8.0 2023-04-30 10:11:29,865 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.8088, 2.1954, 1.7659, 1.9234, 2.5436, 2.2005, 2.3505, 2.6596], device='cuda:1'), covar=tensor([0.0132, 0.0403, 0.0520, 0.0478, 0.0266, 0.0367, 0.0203, 0.0246], device='cuda:1'), in_proj_covar=tensor([0.0171, 0.0213, 0.0207, 0.0207, 0.0211, 0.0212, 0.0213, 0.0206], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-30 10:11:55,942 INFO [optim.py:368] (1/8) 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 10:12:29,541 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.6095, 3.6552, 3.4561, 3.1311, 3.2620, 3.5615, 3.3351, 3.3974], device='cuda:1'), covar=tensor([0.0508, 0.0534, 0.0266, 0.0256, 0.0531, 0.0428, 0.1200, 0.0487], device='cuda:1'), in_proj_covar=tensor([0.0254, 0.0356, 0.0297, 0.0288, 0.0308, 0.0336, 0.0207, 0.0356], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-30 10:12:51,027 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.31 vs. limit=2.0 2023-04-30 10:12:53,421 INFO [train.py:904] (1/8) Epoch 16, batch 8900, loss[loss=0.1869, simple_loss=0.2677, pruned_loss=0.05302, over 12607.00 frames. ], tot_loss[loss=0.1779, simple_loss=0.2732, pruned_loss=0.04136, over 3069463.67 frames. ], batch size: 248, lr: 4.20e-03, grad_scale: 8.0 2023-04-30 10:13:42,141 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.8791, 2.3560, 1.8942, 2.1043, 2.6301, 2.3604, 2.5393, 2.8171], device='cuda:1'), covar=tensor([0.0149, 0.0389, 0.0522, 0.0472, 0.0244, 0.0354, 0.0231, 0.0261], device='cuda:1'), in_proj_covar=tensor([0.0172, 0.0213, 0.0207, 0.0207, 0.0211, 0.0212, 0.0213, 0.0206], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-30 10:14:59,662 INFO [train.py:904] (1/8) Epoch 16, batch 8950, loss[loss=0.1717, simple_loss=0.2596, pruned_loss=0.04188, over 12952.00 frames. ], tot_loss[loss=0.1779, simple_loss=0.2725, pruned_loss=0.04166, over 3078998.77 frames. ], batch size: 248, lr: 4.20e-03, grad_scale: 8.0 2023-04-30 10:15:29,202 INFO [zipformer.py:625] (1/8) 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] (1/8) Clipping_scale=2.0, grad-norm quartiles 1.408e+02 2.276e+02 2.813e+02 3.338e+02 5.711e+02, threshold=5.626e+02, percent-clipped=0.0 2023-04-30 10:16:01,813 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=161231.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 10:16:22,924 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.67 vs. limit=2.0 2023-04-30 10:16:46,880 INFO [train.py:904] (1/8) Epoch 16, batch 9000, loss[loss=0.1806, simple_loss=0.2595, pruned_loss=0.05086, over 12252.00 frames. ], tot_loss[loss=0.1743, simple_loss=0.2686, pruned_loss=0.03995, over 3090436.24 frames. ], batch size: 248, lr: 4.20e-03, grad_scale: 8.0 2023-04-30 10:16:46,880 INFO [train.py:929] (1/8) Computing validation loss 2023-04-30 10:16:56,909 INFO [train.py:938] (1/8) Epoch 16, validation: loss=0.1491, simple_loss=0.2531, pruned_loss=0.02259, over 944034.00 frames. 2023-04-30 10:16:56,910 INFO [train.py:939] (1/8) Maximum memory allocated so far is 17857MB 2023-04-30 10:17:08,071 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=161257.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 10:17:48,806 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=161277.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 10:17:56,575 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.4909, 4.5357, 4.3676, 3.9914, 4.0306, 4.4199, 4.2273, 4.1503], device='cuda:1'), covar=tensor([0.0585, 0.0503, 0.0313, 0.0317, 0.0948, 0.0507, 0.0501, 0.0660], device='cuda:1'), in_proj_covar=tensor([0.0256, 0.0359, 0.0300, 0.0290, 0.0309, 0.0338, 0.0208, 0.0358], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-30 10:18:18,475 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.1451, 2.0934, 2.2462, 3.6118, 2.0676, 2.4087, 2.2432, 2.2020], device='cuda:1'), covar=tensor([0.1084, 0.3515, 0.2775, 0.0492, 0.4297, 0.2340, 0.3354, 0.3570], device='cuda:1'), in_proj_covar=tensor([0.0372, 0.0413, 0.0345, 0.0310, 0.0419, 0.0472, 0.0382, 0.0479], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-30 10:18:20,258 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=161292.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 10:18:25,879 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.5123, 4.8733, 4.4189, 4.8017, 4.4938, 4.3607, 4.4384, 4.9354], device='cuda:1'), covar=tensor([0.2041, 0.1592, 0.2415, 0.1126, 0.1508, 0.1820, 0.1886, 0.1588], device='cuda:1'), in_proj_covar=tensor([0.0588, 0.0717, 0.0585, 0.0528, 0.0454, 0.0470, 0.0598, 0.0552], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:1') 2023-04-30 10:18:40,780 INFO [train.py:904] (1/8) Epoch 16, batch 9050, loss[loss=0.1555, simple_loss=0.2473, pruned_loss=0.0318, over 16901.00 frames. ], tot_loss[loss=0.1749, simple_loss=0.2692, pruned_loss=0.0403, over 3081100.78 frames. ], batch size: 96, lr: 4.20e-03, grad_scale: 8.0 2023-04-30 10:18:55,663 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.32 vs. limit=2.0 2023-04-30 10:19:19,399 INFO [zipformer.py:625] (1/8) 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] (1/8) 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 10:20:22,809 INFO [train.py:904] (1/8) Epoch 16, batch 9100, loss[loss=0.1882, simple_loss=0.287, pruned_loss=0.04465, over 16244.00 frames. ], tot_loss[loss=0.1748, simple_loss=0.2686, pruned_loss=0.0405, over 3071195.68 frames. ], batch size: 165, lr: 4.19e-03, grad_scale: 4.0 2023-04-30 10:20:31,289 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=3.31 vs. limit=5.0 2023-04-30 10:21:00,707 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.4645, 3.4078, 3.5071, 3.5845, 3.6350, 3.3050, 3.5910, 3.6700], device='cuda:1'), covar=tensor([0.1215, 0.0828, 0.1082, 0.0590, 0.0551, 0.2212, 0.0765, 0.0650], device='cuda:1'), in_proj_covar=tensor([0.0553, 0.0681, 0.0800, 0.0690, 0.0520, 0.0543, 0.0553, 0.0648], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:1') 2023-04-30 10:22:19,261 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 16, batch 9150, loss[loss=0.1502, simple_loss=0.2522, pruned_loss=0.02411, over 16901.00 frames. ], tot_loss[loss=0.1747, simple_loss=0.2692, pruned_loss=0.04008, over 3070066.64 frames. ], batch size: 96, lr: 4.19e-03, grad_scale: 4.0 2023-04-30 10:22:51,233 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([1.7800, 1.2199, 1.5993, 1.6204, 1.8270, 1.8632, 1.4874, 1.8274], device='cuda:1'), covar=tensor([0.0241, 0.0400, 0.0217, 0.0293, 0.0265, 0.0190, 0.0428, 0.0119], device='cuda:1'), in_proj_covar=tensor([0.0170, 0.0180, 0.0167, 0.0168, 0.0180, 0.0137, 0.0182, 0.0129], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-30 10:23:13,971 INFO [optim.py:368] (1/8) Clipping_scale=2.0, grad-norm quartiles 1.610e+02 2.279e+02 2.699e+02 3.193e+02 4.519e+02, threshold=5.398e+02, percent-clipped=0.0 2023-04-30 10:24:04,517 INFO [train.py:904] (1/8) Epoch 16, batch 9200, loss[loss=0.1542, simple_loss=0.2405, pruned_loss=0.034, over 11921.00 frames. ], tot_loss[loss=0.172, simple_loss=0.2648, pruned_loss=0.03956, over 3058114.81 frames. ], batch size: 247, lr: 4.19e-03, grad_scale: 8.0 2023-04-30 10:24:20,241 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=161460.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 10:24:24,461 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=161462.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 10:25:42,685 INFO [train.py:904] (1/8) Epoch 16, batch 9250, loss[loss=0.1749, simple_loss=0.2667, pruned_loss=0.04154, over 16688.00 frames. ], tot_loss[loss=0.1722, simple_loss=0.2651, pruned_loss=0.03964, over 3063977.30 frames. ], batch size: 134, lr: 4.19e-03, grad_scale: 8.0 2023-04-30 10:25:47,774 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.0533, 4.0116, 3.8820, 3.1135, 3.9640, 1.7182, 3.7492, 3.4199], device='cuda:1'), covar=tensor([0.0099, 0.0093, 0.0195, 0.0288, 0.0099, 0.2676, 0.0141, 0.0269], device='cuda:1'), in_proj_covar=tensor([0.0140, 0.0128, 0.0174, 0.0157, 0.0148, 0.0188, 0.0161, 0.0153], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-30 10:25:56,623 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.82 vs. limit=2.0 2023-04-30 10:26:15,796 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([1.8765, 1.3581, 1.7498, 1.6545, 1.8873, 1.9500, 1.6386, 1.9088], device='cuda:1'), covar=tensor([0.0256, 0.0357, 0.0197, 0.0272, 0.0247, 0.0180, 0.0361, 0.0111], device='cuda:1'), in_proj_covar=tensor([0.0169, 0.0179, 0.0166, 0.0167, 0.0179, 0.0136, 0.0180, 0.0129], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-30 10:26:22,888 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=161521.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 10:26:36,794 INFO [optim.py:368] (1/8) 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 10:27:34,802 INFO [train.py:904] (1/8) Epoch 16, batch 9300, loss[loss=0.1774, simple_loss=0.2662, pruned_loss=0.04431, over 16966.00 frames. ], tot_loss[loss=0.1709, simple_loss=0.2633, pruned_loss=0.03927, over 3046081.40 frames. ], batch size: 109, lr: 4.19e-03, grad_scale: 8.0 2023-04-30 10:27:45,406 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=161557.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 10:28:23,208 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=161572.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 10:28:27,051 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.45 vs. limit=2.0 2023-04-30 10:28:54,061 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=161587.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 10:29:05,622 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.20 vs. limit=2.0 2023-04-30 10:29:21,683 INFO [train.py:904] (1/8) Epoch 16, batch 9350, loss[loss=0.1861, simple_loss=0.2778, pruned_loss=0.04716, over 15429.00 frames. ], tot_loss[loss=0.1708, simple_loss=0.263, pruned_loss=0.03927, over 3019977.27 frames. ], batch size: 191, lr: 4.19e-03, grad_scale: 8.0 2023-04-30 10:29:28,430 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=161605.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 10:30:02,091 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=161621.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 10:30:12,559 INFO [optim.py:368] (1/8) Clipping_scale=2.0, grad-norm quartiles 1.368e+02 2.307e+02 2.790e+02 3.265e+02 7.040e+02, threshold=5.579e+02, percent-clipped=1.0 2023-04-30 10:31:03,994 INFO [train.py:904] (1/8) Epoch 16, batch 9400, loss[loss=0.176, simple_loss=0.2803, pruned_loss=0.03591, over 16690.00 frames. ], tot_loss[loss=0.1702, simple_loss=0.2631, pruned_loss=0.03861, over 3031405.44 frames. ], batch size: 89, lr: 4.19e-03, grad_scale: 8.0 2023-04-30 10:31:22,058 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.6220, 3.0742, 3.3122, 1.8921, 2.8354, 2.1601, 3.1531, 3.1833], device='cuda:1'), covar=tensor([0.0289, 0.0762, 0.0436, 0.1971, 0.0756, 0.0934, 0.0679, 0.0944], device='cuda:1'), in_proj_covar=tensor([0.0145, 0.0147, 0.0158, 0.0145, 0.0136, 0.0123, 0.0137, 0.0159], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:1') 2023-04-30 10:31:38,657 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=161669.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 10:32:44,156 INFO [train.py:904] (1/8) Epoch 16, batch 9450, loss[loss=0.1753, simple_loss=0.2692, pruned_loss=0.04075, over 16689.00 frames. ], tot_loss[loss=0.1706, simple_loss=0.2639, pruned_loss=0.03864, over 3017052.43 frames. ], batch size: 89, lr: 4.19e-03, grad_scale: 8.0 2023-04-30 10:33:33,849 INFO [optim.py:368] (1/8) 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 10:33:43,831 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.22 vs. limit=2.0 2023-04-30 10:33:57,104 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.3859, 4.4196, 4.2704, 3.9189, 3.9731, 4.3788, 4.1039, 4.1066], device='cuda:1'), covar=tensor([0.0565, 0.0449, 0.0308, 0.0326, 0.0918, 0.0485, 0.0635, 0.0662], device='cuda:1'), in_proj_covar=tensor([0.0256, 0.0356, 0.0298, 0.0287, 0.0308, 0.0333, 0.0206, 0.0354], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002], device='cuda:1') 2023-04-30 10:34:23,920 INFO [train.py:904] (1/8) Epoch 16, batch 9500, loss[loss=0.1768, simple_loss=0.2627, pruned_loss=0.04546, over 17011.00 frames. ], tot_loss[loss=0.1703, simple_loss=0.2639, pruned_loss=0.03837, over 3052064.28 frames. ], batch size: 109, lr: 4.19e-03, grad_scale: 8.0 2023-04-30 10:34:37,543 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=161757.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 10:36:08,227 INFO [train.py:904] (1/8) Epoch 16, batch 9550, loss[loss=0.1692, simple_loss=0.268, pruned_loss=0.03518, over 16872.00 frames. ], tot_loss[loss=0.1702, simple_loss=0.2639, pruned_loss=0.0383, over 3080778.35 frames. ], batch size: 116, lr: 4.19e-03, grad_scale: 8.0 2023-04-30 10:36:38,122 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=161815.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 10:36:40,595 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=161816.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 10:37:00,498 INFO [optim.py:368] (1/8) 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 10:37:13,706 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.9017, 2.0713, 2.3241, 3.2590, 2.1339, 2.2471, 2.2689, 2.1467], device='cuda:1'), covar=tensor([0.1083, 0.3545, 0.2493, 0.0635, 0.4291, 0.2653, 0.3289, 0.3874], device='cuda:1'), in_proj_covar=tensor([0.0367, 0.0406, 0.0341, 0.0307, 0.0414, 0.0465, 0.0376, 0.0471], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-30 10:37:41,726 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.2864, 3.4035, 3.6440, 3.6182, 3.6492, 3.4528, 3.5237, 3.5183], device='cuda:1'), covar=tensor([0.0401, 0.0754, 0.0597, 0.0645, 0.0522, 0.0598, 0.0741, 0.0460], device='cuda:1'), in_proj_covar=tensor([0.0356, 0.0382, 0.0378, 0.0352, 0.0421, 0.0395, 0.0482, 0.0313], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003], device='cuda:1') 2023-04-30 10:37:51,434 INFO [train.py:904] (1/8) Epoch 16, batch 9600, loss[loss=0.1711, simple_loss=0.2575, pruned_loss=0.04238, over 12186.00 frames. ], tot_loss[loss=0.172, simple_loss=0.2652, pruned_loss=0.03937, over 3067491.53 frames. ], batch size: 250, lr: 4.19e-03, grad_scale: 8.0 2023-04-30 10:38:29,893 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=161872.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 10:38:38,428 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=161876.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 10:39:02,363 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 16, batch 9650, loss[loss=0.1712, simple_loss=0.2558, pruned_loss=0.04332, over 12408.00 frames. ], tot_loss[loss=0.1738, simple_loss=0.2676, pruned_loss=0.04003, over 3086173.90 frames. ], batch size: 248, lr: 4.19e-03, grad_scale: 8.0 2023-04-30 10:40:21,562 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=161920.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 10:40:36,070 INFO [optim.py:368] (1/8) 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 10:40:49,470 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=161935.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 10:41:23,666 INFO [train.py:904] (1/8) Epoch 16, batch 9700, loss[loss=0.16, simple_loss=0.2567, pruned_loss=0.03161, over 16746.00 frames. ], tot_loss[loss=0.1722, simple_loss=0.266, pruned_loss=0.03918, over 3086812.51 frames. ], batch size: 76, lr: 4.19e-03, grad_scale: 8.0 2023-04-30 10:41:38,175 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.7460, 3.7821, 4.1320, 4.0917, 4.1091, 3.8788, 3.9281, 3.9226], device='cuda:1'), covar=tensor([0.0386, 0.1061, 0.0564, 0.0626, 0.0604, 0.0605, 0.0803, 0.0448], device='cuda:1'), in_proj_covar=tensor([0.0358, 0.0385, 0.0380, 0.0354, 0.0423, 0.0398, 0.0486, 0.0315], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003], device='cuda:1') 2023-04-30 10:42:24,847 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([1.8209, 1.4029, 1.7355, 1.7784, 1.8897, 1.9759, 1.6014, 1.9248], device='cuda:1'), covar=tensor([0.0217, 0.0332, 0.0186, 0.0233, 0.0231, 0.0170, 0.0359, 0.0112], device='cuda:1'), in_proj_covar=tensor([0.0167, 0.0176, 0.0164, 0.0165, 0.0177, 0.0134, 0.0178, 0.0127], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-30 10:43:08,495 INFO [train.py:904] (1/8) Epoch 16, batch 9750, loss[loss=0.1665, simple_loss=0.267, pruned_loss=0.03297, over 15470.00 frames. ], tot_loss[loss=0.1725, simple_loss=0.2652, pruned_loss=0.03993, over 3059657.39 frames. ], batch size: 191, lr: 4.19e-03, grad_scale: 8.0 2023-04-30 10:43:43,717 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.5314, 4.5809, 4.7556, 4.6118, 4.6543, 5.1473, 4.6585, 4.4184], device='cuda:1'), covar=tensor([0.1143, 0.2148, 0.2374, 0.1757, 0.2167, 0.0937, 0.1544, 0.2323], device='cuda:1'), in_proj_covar=tensor([0.0355, 0.0516, 0.0565, 0.0431, 0.0573, 0.0598, 0.0446, 0.0579], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-04-30 10:43:58,494 INFO [optim.py:368] (1/8) Clipping_scale=2.0, grad-norm quartiles 1.537e+02 2.156e+02 2.615e+02 3.105e+02 5.589e+02, threshold=5.230e+02, percent-clipped=0.0 2023-04-30 10:44:46,279 INFO [train.py:904] (1/8) Epoch 16, batch 9800, loss[loss=0.1839, simple_loss=0.2866, pruned_loss=0.04058, over 16272.00 frames. ], tot_loss[loss=0.1714, simple_loss=0.2647, pruned_loss=0.039, over 3039364.15 frames. ], batch size: 165, lr: 4.19e-03, grad_scale: 8.0 2023-04-30 10:44:56,615 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=162057.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 10:45:34,302 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=162077.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 10:46:31,269 INFO [train.py:904] (1/8) Epoch 16, batch 9850, loss[loss=0.167, simple_loss=0.2662, pruned_loss=0.03393, over 16806.00 frames. ], tot_loss[loss=0.1716, simple_loss=0.2662, pruned_loss=0.03855, over 3050993.11 frames. ], batch size: 83, lr: 4.18e-03, grad_scale: 8.0 2023-04-30 10:46:38,614 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=162105.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 10:46:59,762 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=162116.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 10:47:23,356 INFO [optim.py:368] (1/8) 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 10:47:51,516 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=162138.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 10:48:22,257 INFO [train.py:904] (1/8) Epoch 16, batch 9900, loss[loss=0.1741, simple_loss=0.2773, pruned_loss=0.03542, over 16689.00 frames. ], tot_loss[loss=0.172, simple_loss=0.2667, pruned_loss=0.03861, over 3056049.95 frames. ], batch size: 76, lr: 4.18e-03, grad_scale: 8.0 2023-04-30 10:48:52,697 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=162164.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 10:49:08,583 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=162171.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 10:49:36,226 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.61 vs. limit=2.0 2023-04-30 10:50:18,899 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.9246, 3.9519, 4.3069, 4.2598, 4.2721, 4.0350, 4.0317, 4.0093], device='cuda:1'), covar=tensor([0.0322, 0.0627, 0.0439, 0.0454, 0.0429, 0.0390, 0.0802, 0.0431], device='cuda:1'), in_proj_covar=tensor([0.0357, 0.0384, 0.0379, 0.0354, 0.0422, 0.0398, 0.0484, 0.0316], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003], device='cuda:1') 2023-04-30 10:50:22,154 INFO [train.py:904] (1/8) Epoch 16, batch 9950, loss[loss=0.1552, simple_loss=0.2545, pruned_loss=0.0279, over 16721.00 frames. ], tot_loss[loss=0.1728, simple_loss=0.2679, pruned_loss=0.03885, over 3048715.58 frames. ], batch size: 83, lr: 4.18e-03, grad_scale: 8.0 2023-04-30 10:51:26,523 INFO [optim.py:368] (1/8) 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 10:51:36,405 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=162231.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 10:52:05,936 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.1648, 3.0104, 3.1262, 1.7781, 3.3052, 3.3817, 2.7956, 2.6569], device='cuda:1'), covar=tensor([0.0811, 0.0262, 0.0246, 0.1243, 0.0081, 0.0190, 0.0411, 0.0442], device='cuda:1'), in_proj_covar=tensor([0.0144, 0.0103, 0.0088, 0.0135, 0.0071, 0.0113, 0.0122, 0.0125], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:1') 2023-04-30 10:52:23,901 INFO [train.py:904] (1/8) Epoch 16, batch 10000, loss[loss=0.1538, simple_loss=0.2481, pruned_loss=0.02973, over 12730.00 frames. ], tot_loss[loss=0.1713, simple_loss=0.2663, pruned_loss=0.03815, over 3070543.97 frames. ], batch size: 250, lr: 4.18e-03, grad_scale: 8.0 2023-04-30 10:53:44,542 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([1.5860, 2.3687, 2.3939, 3.7352, 2.3867, 3.8432, 1.3543, 2.7666], device='cuda:1'), covar=tensor([0.1445, 0.0812, 0.1168, 0.0125, 0.0140, 0.0312, 0.1723, 0.0754], device='cuda:1'), in_proj_covar=tensor([0.0158, 0.0161, 0.0184, 0.0161, 0.0188, 0.0203, 0.0189, 0.0183], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0003, 0.0004], device='cuda:1') 2023-04-30 10:53:46,606 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=162292.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 10:54:04,078 INFO [train.py:904] (1/8) Epoch 16, batch 10050, loss[loss=0.1582, simple_loss=0.2547, pruned_loss=0.03092, over 12245.00 frames. ], tot_loss[loss=0.1711, simple_loss=0.2661, pruned_loss=0.03804, over 3073913.64 frames. ], batch size: 247, lr: 4.18e-03, grad_scale: 8.0 2023-04-30 10:54:54,676 INFO [optim.py:368] (1/8) Clipping_scale=2.0, grad-norm quartiles 1.562e+02 2.225e+02 2.711e+02 3.158e+02 7.314e+02, threshold=5.421e+02, percent-clipped=6.0 2023-04-30 10:55:38,457 INFO [train.py:904] (1/8) Epoch 16, batch 10100, loss[loss=0.1707, simple_loss=0.2605, pruned_loss=0.04046, over 16958.00 frames. ], tot_loss[loss=0.1721, simple_loss=0.2669, pruned_loss=0.03862, over 3075527.20 frames. ], batch size: 109, lr: 4.18e-03, grad_scale: 8.0 2023-04-30 10:56:11,327 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.9136, 1.8718, 2.4401, 2.8485, 2.6103, 3.0886, 2.1354, 3.0690], device='cuda:1'), covar=tensor([0.0179, 0.0471, 0.0297, 0.0255, 0.0285, 0.0148, 0.0474, 0.0152], device='cuda:1'), in_proj_covar=tensor([0.0170, 0.0180, 0.0166, 0.0168, 0.0181, 0.0136, 0.0181, 0.0129], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-30 10:57:23,031 INFO [train.py:904] (1/8) Epoch 17, batch 0, loss[loss=0.2017, simple_loss=0.2872, pruned_loss=0.05807, over 16784.00 frames. ], tot_loss[loss=0.2017, simple_loss=0.2872, pruned_loss=0.05807, over 16784.00 frames. ], batch size: 57, lr: 4.05e-03, grad_scale: 8.0 2023-04-30 10:57:23,031 INFO [train.py:929] (1/8) Computing validation loss 2023-04-30 10:57:30,749 INFO [train.py:938] (1/8) Epoch 17, validation: loss=0.1481, simple_loss=0.2518, pruned_loss=0.02217, over 944034.00 frames. 2023-04-30 10:57:30,749 INFO [train.py:939] (1/8) Maximum memory allocated so far is 17857MB 2023-04-30 10:58:09,502 INFO [optim.py:368] (1/8) 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 10:58:14,143 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=162433.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 10:58:39,851 INFO [train.py:904] (1/8) Epoch 17, batch 50, loss[loss=0.2039, simple_loss=0.2748, pruned_loss=0.06652, over 16878.00 frames. ], tot_loss[loss=0.1908, simple_loss=0.2741, pruned_loss=0.05378, over 747555.56 frames. ], batch size: 96, lr: 4.05e-03, grad_scale: 2.0 2023-04-30 10:59:06,409 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=162471.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 10:59:40,127 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.7858, 2.7279, 2.2790, 2.7382, 3.1244, 2.9144, 3.5748, 3.3776], device='cuda:1'), covar=tensor([0.0134, 0.0362, 0.0523, 0.0388, 0.0263, 0.0355, 0.0220, 0.0236], device='cuda:1'), in_proj_covar=tensor([0.0177, 0.0220, 0.0213, 0.0212, 0.0218, 0.0219, 0.0219, 0.0210], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-30 10:59:47,951 INFO [train.py:904] (1/8) Epoch 17, batch 100, loss[loss=0.1929, simple_loss=0.2673, pruned_loss=0.05927, over 16843.00 frames. ], tot_loss[loss=0.1861, simple_loss=0.2697, pruned_loss=0.05124, over 1314000.23 frames. ], batch size: 116, lr: 4.05e-03, grad_scale: 2.0 2023-04-30 11:00:00,713 INFO [zipformer.py:625] (1/8) 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] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=162519.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 11:00:26,385 INFO [optim.py:368] (1/8) 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 11:00:55,207 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.9570, 1.9449, 2.5071, 2.8236, 2.8311, 2.8205, 2.0070, 3.1458], device='cuda:1'), covar=tensor([0.0147, 0.0406, 0.0300, 0.0251, 0.0238, 0.0247, 0.0492, 0.0107], device='cuda:1'), in_proj_covar=tensor([0.0173, 0.0182, 0.0170, 0.0172, 0.0183, 0.0139, 0.0184, 0.0131], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-30 11:00:56,552 INFO [train.py:904] (1/8) Epoch 17, batch 150, loss[loss=0.1616, simple_loss=0.2413, pruned_loss=0.04091, over 16707.00 frames. ], tot_loss[loss=0.1848, simple_loss=0.2696, pruned_loss=0.05003, over 1762607.37 frames. ], batch size: 89, lr: 4.05e-03, grad_scale: 2.0 2023-04-30 11:01:23,624 INFO [zipformer.py:625] (1/8) 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] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=162587.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 11:01:47,497 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([5.0398, 4.9597, 4.7983, 4.2433, 4.8615, 1.7700, 4.6608, 4.7427], device='cuda:1'), covar=tensor([0.0085, 0.0089, 0.0189, 0.0394, 0.0103, 0.2736, 0.0131, 0.0206], device='cuda:1'), in_proj_covar=tensor([0.0143, 0.0132, 0.0178, 0.0160, 0.0152, 0.0193, 0.0165, 0.0157], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-30 11:01:53,247 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.16 vs. limit=2.0 2023-04-30 11:02:05,939 INFO [train.py:904] (1/8) Epoch 17, batch 200, loss[loss=0.1543, simple_loss=0.2434, pruned_loss=0.03258, over 17201.00 frames. ], tot_loss[loss=0.1852, simple_loss=0.2693, pruned_loss=0.05057, over 2108890.99 frames. ], batch size: 43, lr: 4.05e-03, grad_scale: 2.0 2023-04-30 11:02:16,251 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.1466, 2.0518, 2.2636, 3.9078, 2.1276, 2.4697, 2.1513, 2.2493], device='cuda:1'), covar=tensor([0.1277, 0.3729, 0.2808, 0.0610, 0.3687, 0.2449, 0.3689, 0.2891], device='cuda:1'), in_proj_covar=tensor([0.0375, 0.0414, 0.0349, 0.0315, 0.0422, 0.0475, 0.0386, 0.0482], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-30 11:02:43,589 INFO [optim.py:368] (1/8) 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 11:02:59,997 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=4.48 vs. limit=5.0 2023-04-30 11:03:12,317 INFO [train.py:904] (1/8) Epoch 17, batch 250, loss[loss=0.1621, simple_loss=0.2526, pruned_loss=0.03579, over 16842.00 frames. ], tot_loss[loss=0.1838, simple_loss=0.267, pruned_loss=0.05028, over 2377806.71 frames. ], batch size: 42, lr: 4.05e-03, grad_scale: 2.0 2023-04-30 11:04:20,337 INFO [train.py:904] (1/8) Epoch 17, batch 300, loss[loss=0.1915, simple_loss=0.2661, pruned_loss=0.05849, over 16455.00 frames. ], tot_loss[loss=0.182, simple_loss=0.2646, pruned_loss=0.04966, over 2586837.00 frames. ], batch size: 146, lr: 4.05e-03, grad_scale: 2.0 2023-04-30 11:04:55,529 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([1.8356, 2.4071, 2.4594, 3.4518, 2.7022, 3.6497, 1.4575, 2.8177], device='cuda:1'), covar=tensor([0.1318, 0.0727, 0.1127, 0.0225, 0.0154, 0.0430, 0.1676, 0.0792], device='cuda:1'), in_proj_covar=tensor([0.0161, 0.0164, 0.0187, 0.0168, 0.0194, 0.0209, 0.0193, 0.0187], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-04-30 11:04:59,735 INFO [optim.py:368] (1/8) 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 11:05:03,709 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=162733.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 11:05:29,585 INFO [train.py:904] (1/8) Epoch 17, batch 350, loss[loss=0.1542, simple_loss=0.2416, pruned_loss=0.03338, over 17233.00 frames. ], tot_loss[loss=0.1784, simple_loss=0.2616, pruned_loss=0.04757, over 2754141.55 frames. ], batch size: 45, lr: 4.05e-03, grad_scale: 2.0 2023-04-30 11:06:07,974 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=162781.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 11:06:27,276 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.9695, 2.0666, 2.5728, 2.8996, 2.7077, 3.4204, 2.2128, 3.2319], device='cuda:1'), covar=tensor([0.0219, 0.0448, 0.0295, 0.0303, 0.0308, 0.0153, 0.0430, 0.0168], device='cuda:1'), in_proj_covar=tensor([0.0175, 0.0183, 0.0170, 0.0174, 0.0185, 0.0141, 0.0185, 0.0133], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-30 11:06:36,764 INFO [train.py:904] (1/8) Epoch 17, batch 400, loss[loss=0.1724, simple_loss=0.2684, pruned_loss=0.03823, over 17124.00 frames. ], tot_loss[loss=0.1757, simple_loss=0.2589, pruned_loss=0.04621, over 2877265.63 frames. ], batch size: 48, lr: 4.05e-03, grad_scale: 4.0 2023-04-30 11:06:37,286 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.7929, 4.2180, 3.1865, 2.3477, 2.6786, 2.5223, 4.5645, 3.6881], device='cuda:1'), covar=tensor([0.2717, 0.0620, 0.1575, 0.2713, 0.2741, 0.1903, 0.0355, 0.1154], device='cuda:1'), in_proj_covar=tensor([0.0311, 0.0257, 0.0290, 0.0292, 0.0279, 0.0237, 0.0277, 0.0314], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:1') 2023-04-30 11:06:56,217 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.7927, 3.7320, 3.8747, 4.0004, 4.0581, 3.6582, 3.9124, 4.0851], device='cuda:1'), covar=tensor([0.1620, 0.1160, 0.1324, 0.0680, 0.0619, 0.1797, 0.1866, 0.0655], device='cuda:1'), in_proj_covar=tensor([0.0589, 0.0727, 0.0858, 0.0729, 0.0553, 0.0580, 0.0595, 0.0688], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-30 11:06:56,339 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([1.7202, 2.2700, 2.3351, 3.3296, 2.6344, 3.6519, 1.4618, 2.7276], device='cuda:1'), covar=tensor([0.1464, 0.0833, 0.1249, 0.0222, 0.0164, 0.0424, 0.1712, 0.0866], device='cuda:1'), in_proj_covar=tensor([0.0160, 0.0164, 0.0186, 0.0167, 0.0194, 0.0208, 0.0192, 0.0186], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-04-30 11:07:10,965 INFO [zipformer.py:625] (1/8) 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] (1/8) 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 11:07:34,477 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.18 vs. limit=2.0 2023-04-30 11:07:35,469 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.3974, 3.4223, 3.4416, 3.5379, 3.5843, 3.2896, 3.5155, 3.6371], device='cuda:1'), covar=tensor([0.1276, 0.0874, 0.1078, 0.0616, 0.0596, 0.2472, 0.1215, 0.0725], device='cuda:1'), in_proj_covar=tensor([0.0589, 0.0728, 0.0859, 0.0729, 0.0553, 0.0579, 0.0595, 0.0688], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-30 11:07:42,788 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.5106, 2.8374, 3.0361, 2.0323, 2.7651, 2.2050, 3.1160, 3.1205], device='cuda:1'), covar=tensor([0.0251, 0.0869, 0.0563, 0.1889, 0.0804, 0.0964, 0.0548, 0.0883], device='cuda:1'), in_proj_covar=tensor([0.0148, 0.0151, 0.0160, 0.0149, 0.0139, 0.0126, 0.0139, 0.0163], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:1') 2023-04-30 11:07:46,583 INFO [train.py:904] (1/8) Epoch 17, batch 450, loss[loss=0.1491, simple_loss=0.2425, pruned_loss=0.02788, over 16854.00 frames. ], tot_loss[loss=0.1745, simple_loss=0.2585, pruned_loss=0.04525, over 2967220.04 frames. ], batch size: 42, lr: 4.05e-03, grad_scale: 4.0 2023-04-30 11:08:06,776 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=162867.0, num_to_drop=1, layers_to_drop={2} 2023-04-30 11:08:34,317 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=162887.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 11:08:35,603 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=162888.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 11:08:55,330 INFO [train.py:904] (1/8) Epoch 17, batch 500, loss[loss=0.1621, simple_loss=0.2534, pruned_loss=0.0354, over 17189.00 frames. ], tot_loss[loss=0.1731, simple_loss=0.2572, pruned_loss=0.04448, over 3050509.33 frames. ], batch size: 46, lr: 4.05e-03, grad_scale: 4.0 2023-04-30 11:09:32,597 INFO [optim.py:368] (1/8) 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 11:09:39,868 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 17, batch 550, loss[loss=0.157, simple_loss=0.2495, pruned_loss=0.03227, over 17250.00 frames. ], tot_loss[loss=0.173, simple_loss=0.2571, pruned_loss=0.04447, over 3103834.73 frames. ], batch size: 43, lr: 4.05e-03, grad_scale: 4.0 2023-04-30 11:11:10,218 INFO [train.py:904] (1/8) Epoch 17, batch 600, loss[loss=0.1644, simple_loss=0.2613, pruned_loss=0.03375, over 17113.00 frames. ], tot_loss[loss=0.1734, simple_loss=0.2571, pruned_loss=0.04486, over 3153594.58 frames. ], batch size: 48, lr: 4.05e-03, grad_scale: 4.0 2023-04-30 11:11:47,293 INFO [optim.py:368] (1/8) 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 11:12:09,277 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([5.0184, 5.5562, 5.7226, 5.4105, 5.5027, 6.1203, 5.6457, 5.3626], device='cuda:1'), covar=tensor([0.0988, 0.2238, 0.1939, 0.1993, 0.2643, 0.0909, 0.1376, 0.2197], device='cuda:1'), in_proj_covar=tensor([0.0388, 0.0564, 0.0619, 0.0474, 0.0632, 0.0652, 0.0489, 0.0634], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-04-30 11:12:16,979 INFO [train.py:904] (1/8) Epoch 17, batch 650, loss[loss=0.1649, simple_loss=0.255, pruned_loss=0.0374, over 16699.00 frames. ], tot_loss[loss=0.1722, simple_loss=0.2556, pruned_loss=0.04437, over 3189313.93 frames. ], batch size: 62, lr: 4.05e-03, grad_scale: 4.0 2023-04-30 11:13:05,156 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.12 vs. limit=2.0 2023-04-30 11:13:06,033 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([5.4983, 5.9589, 5.6719, 5.7670, 5.2880, 5.3598, 5.3750, 6.0440], device='cuda:1'), covar=tensor([0.1430, 0.0922, 0.1104, 0.0808, 0.1004, 0.0711, 0.1179, 0.0920], device='cuda:1'), in_proj_covar=tensor([0.0629, 0.0776, 0.0631, 0.0570, 0.0489, 0.0495, 0.0647, 0.0599], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:1') 2023-04-30 11:13:09,383 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=163090.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 11:13:18,958 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.0284, 2.0884, 2.5655, 2.9234, 2.7372, 3.3251, 2.2807, 3.3528], device='cuda:1'), covar=tensor([0.0198, 0.0431, 0.0312, 0.0284, 0.0310, 0.0202, 0.0408, 0.0137], device='cuda:1'), in_proj_covar=tensor([0.0176, 0.0185, 0.0172, 0.0176, 0.0186, 0.0142, 0.0186, 0.0134], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-30 11:13:25,533 INFO [train.py:904] (1/8) Epoch 17, batch 700, loss[loss=0.1695, simple_loss=0.2437, pruned_loss=0.04769, over 16675.00 frames. ], tot_loss[loss=0.1715, simple_loss=0.2551, pruned_loss=0.04396, over 3213323.02 frames. ], batch size: 89, lr: 4.05e-03, grad_scale: 4.0 2023-04-30 11:13:47,860 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.21 vs. limit=2.0 2023-04-30 11:14:04,486 INFO [optim.py:368] (1/8) 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 11:14:34,403 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 17, batch 750, loss[loss=0.1533, simple_loss=0.2369, pruned_loss=0.03486, over 16968.00 frames. ], tot_loss[loss=0.1707, simple_loss=0.2545, pruned_loss=0.04348, over 3241139.93 frames. ], batch size: 41, lr: 4.05e-03, grad_scale: 4.0 2023-04-30 11:14:57,201 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=163167.0, num_to_drop=1, layers_to_drop={0} 2023-04-30 11:15:01,201 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([5.4506, 5.8633, 5.6158, 5.6656, 5.2817, 5.2912, 5.2623, 5.9690], device='cuda:1'), covar=tensor([0.1376, 0.1050, 0.1071, 0.0850, 0.0802, 0.0679, 0.1178, 0.0937], device='cuda:1'), in_proj_covar=tensor([0.0632, 0.0781, 0.0634, 0.0573, 0.0491, 0.0497, 0.0649, 0.0602], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:1') 2023-04-30 11:15:16,230 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=4.87 vs. limit=5.0 2023-04-30 11:15:18,002 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=163183.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 11:15:44,390 INFO [train.py:904] (1/8) Epoch 17, batch 800, loss[loss=0.1888, simple_loss=0.265, pruned_loss=0.05628, over 12189.00 frames. ], tot_loss[loss=0.1704, simple_loss=0.2547, pruned_loss=0.04311, over 3251479.02 frames. ], batch size: 247, lr: 4.04e-03, grad_scale: 8.0 2023-04-30 11:16:03,280 INFO [zipformer.py:625] (1/8) 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] (1/8) 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 11:16:36,600 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=163240.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 11:16:40,919 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.5453, 2.2407, 2.2281, 4.4814, 2.2613, 2.6948, 2.3709, 2.4253], device='cuda:1'), covar=tensor([0.1146, 0.3657, 0.3040, 0.0433, 0.4139, 0.2602, 0.3637, 0.3689], device='cuda:1'), in_proj_covar=tensor([0.0383, 0.0422, 0.0354, 0.0323, 0.0428, 0.0485, 0.0393, 0.0493], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-30 11:16:53,792 INFO [train.py:904] (1/8) Epoch 17, batch 850, loss[loss=0.179, simple_loss=0.2496, pruned_loss=0.05419, over 16777.00 frames. ], tot_loss[loss=0.1695, simple_loss=0.2541, pruned_loss=0.04241, over 3277622.23 frames. ], batch size: 124, lr: 4.04e-03, grad_scale: 8.0 2023-04-30 11:17:01,565 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([5.1302, 5.7161, 5.7596, 5.5580, 5.6011, 6.1667, 5.6347, 5.3940], device='cuda:1'), covar=tensor([0.0907, 0.2135, 0.2182, 0.2016, 0.2535, 0.0994, 0.1405, 0.2283], device='cuda:1'), in_proj_covar=tensor([0.0391, 0.0562, 0.0621, 0.0473, 0.0631, 0.0655, 0.0490, 0.0634], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-04-30 11:18:01,100 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 17, batch 900, loss[loss=0.1678, simple_loss=0.2446, pruned_loss=0.04554, over 16913.00 frames. ], tot_loss[loss=0.1688, simple_loss=0.2537, pruned_loss=0.04198, over 3291479.77 frames. ], batch size: 96, lr: 4.04e-03, grad_scale: 8.0 2023-04-30 11:18:26,582 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.9465, 4.0369, 4.2889, 4.2826, 4.3039, 4.0322, 4.0869, 3.9895], device='cuda:1'), covar=tensor([0.0399, 0.0617, 0.0395, 0.0393, 0.0485, 0.0426, 0.0714, 0.0559], device='cuda:1'), in_proj_covar=tensor([0.0387, 0.0418, 0.0408, 0.0383, 0.0456, 0.0431, 0.0525, 0.0341], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:1') 2023-04-30 11:18:40,393 INFO [optim.py:368] (1/8) 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 11:18:42,626 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.7982, 2.4415, 1.9706, 2.2343, 2.9099, 2.6512, 2.9588, 3.0001], device='cuda:1'), covar=tensor([0.0213, 0.0424, 0.0505, 0.0439, 0.0215, 0.0306, 0.0210, 0.0247], device='cuda:1'), in_proj_covar=tensor([0.0190, 0.0230, 0.0221, 0.0220, 0.0228, 0.0230, 0.0234, 0.0222], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-30 11:18:52,696 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.13 vs. limit=2.0 2023-04-30 11:19:09,573 INFO [train.py:904] (1/8) Epoch 17, batch 950, loss[loss=0.1541, simple_loss=0.2445, pruned_loss=0.03186, over 17196.00 frames. ], tot_loss[loss=0.1687, simple_loss=0.2534, pruned_loss=0.04198, over 3293340.60 frames. ], batch size: 46, lr: 4.04e-03, grad_scale: 8.0 2023-04-30 11:19:33,354 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.9140, 4.9645, 5.3678, 5.3402, 5.3790, 5.0629, 5.0119, 4.7593], device='cuda:1'), covar=tensor([0.0358, 0.0529, 0.0453, 0.0459, 0.0423, 0.0357, 0.0872, 0.0440], device='cuda:1'), in_proj_covar=tensor([0.0388, 0.0419, 0.0411, 0.0385, 0.0459, 0.0432, 0.0528, 0.0344], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:1') 2023-04-30 11:20:00,130 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.5223, 3.6228, 2.1765, 3.8127, 2.7951, 3.8253, 2.2927, 2.8427], device='cuda:1'), covar=tensor([0.0245, 0.0343, 0.1446, 0.0330, 0.0726, 0.0657, 0.1294, 0.0687], device='cuda:1'), in_proj_covar=tensor([0.0165, 0.0171, 0.0192, 0.0152, 0.0173, 0.0212, 0.0201, 0.0178], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-04-30 11:20:12,986 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.25 vs. limit=2.0 2023-04-30 11:20:17,960 INFO [train.py:904] (1/8) Epoch 17, batch 1000, loss[loss=0.151, simple_loss=0.236, pruned_loss=0.03301, over 15339.00 frames. ], tot_loss[loss=0.1686, simple_loss=0.2527, pruned_loss=0.04231, over 3289750.36 frames. ], batch size: 190, lr: 4.04e-03, grad_scale: 8.0 2023-04-30 11:20:54,826 INFO [optim.py:368] (1/8) 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] (1/8) attn_weights_entropy = tensor([2.9166, 4.3802, 3.2322, 2.3343, 2.7580, 2.5438, 4.7048, 3.6711], device='cuda:1'), covar=tensor([0.2670, 0.0506, 0.1604, 0.2781, 0.2773, 0.1911, 0.0354, 0.1267], device='cuda:1'), in_proj_covar=tensor([0.0317, 0.0262, 0.0295, 0.0298, 0.0286, 0.0241, 0.0281, 0.0320], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:1') 2023-04-30 11:21:18,943 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=163446.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 11:21:26,451 INFO [train.py:904] (1/8) Epoch 17, batch 1050, loss[loss=0.1509, simple_loss=0.2345, pruned_loss=0.03364, over 15473.00 frames. ], tot_loss[loss=0.1687, simple_loss=0.2526, pruned_loss=0.04241, over 3296456.02 frames. ], batch size: 191, lr: 4.04e-03, grad_scale: 8.0 2023-04-30 11:21:32,057 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.3317, 1.6416, 2.0618, 2.1776, 2.3085, 2.3230, 1.7563, 2.2856], device='cuda:1'), covar=tensor([0.0216, 0.0404, 0.0234, 0.0273, 0.0260, 0.0243, 0.0441, 0.0167], device='cuda:1'), in_proj_covar=tensor([0.0179, 0.0187, 0.0174, 0.0178, 0.0188, 0.0145, 0.0188, 0.0137], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-30 11:22:10,650 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 17, batch 1100, loss[loss=0.1365, simple_loss=0.223, pruned_loss=0.02493, over 16989.00 frames. ], tot_loss[loss=0.1687, simple_loss=0.2525, pruned_loss=0.04239, over 3307421.21 frames. ], batch size: 41, lr: 4.04e-03, grad_scale: 8.0 2023-04-30 11:23:14,262 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.2959, 2.1866, 2.3951, 4.0999, 2.1953, 2.5755, 2.2630, 2.3261], device='cuda:1'), covar=tensor([0.1385, 0.3759, 0.2732, 0.0581, 0.3905, 0.2601, 0.3693, 0.3158], device='cuda:1'), in_proj_covar=tensor([0.0386, 0.0425, 0.0357, 0.0325, 0.0430, 0.0489, 0.0396, 0.0497], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-30 11:23:14,822 INFO [optim.py:368] (1/8) 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] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=163531.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 11:23:43,925 INFO [train.py:904] (1/8) Epoch 17, batch 1150, loss[loss=0.159, simple_loss=0.2557, pruned_loss=0.03112, over 17262.00 frames. ], tot_loss[loss=0.1674, simple_loss=0.2514, pruned_loss=0.0417, over 3306468.86 frames. ], batch size: 52, lr: 4.04e-03, grad_scale: 8.0 2023-04-30 11:24:08,196 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=3.59 vs. limit=5.0 2023-04-30 11:24:27,961 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.8444, 4.0523, 2.1657, 4.7145, 3.1990, 4.6253, 2.3705, 3.2426], device='cuda:1'), covar=tensor([0.0306, 0.0353, 0.1885, 0.0203, 0.0738, 0.0389, 0.1760, 0.0672], device='cuda:1'), in_proj_covar=tensor([0.0166, 0.0173, 0.0193, 0.0154, 0.0174, 0.0215, 0.0202, 0.0179], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-04-30 11:24:43,132 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=163596.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 11:24:52,260 INFO [train.py:904] (1/8) Epoch 17, batch 1200, loss[loss=0.1462, simple_loss=0.2353, pruned_loss=0.02852, over 17155.00 frames. ], tot_loss[loss=0.1668, simple_loss=0.2509, pruned_loss=0.04136, over 3298992.15 frames. ], batch size: 46, lr: 4.04e-03, grad_scale: 8.0 2023-04-30 11:25:29,955 INFO [optim.py:368] (1/8) 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] (1/8) Epoch 17, batch 1250, loss[loss=0.1712, simple_loss=0.2651, pruned_loss=0.03868, over 16654.00 frames. ], tot_loss[loss=0.168, simple_loss=0.2512, pruned_loss=0.0424, over 3293479.15 frames. ], batch size: 57, lr: 4.04e-03, grad_scale: 8.0 2023-04-30 11:27:06,150 INFO [train.py:904] (1/8) Epoch 17, batch 1300, loss[loss=0.1779, simple_loss=0.2577, pruned_loss=0.04907, over 16504.00 frames. ], tot_loss[loss=0.1679, simple_loss=0.2511, pruned_loss=0.04236, over 3299292.39 frames. ], batch size: 75, lr: 4.04e-03, grad_scale: 8.0 2023-04-30 11:27:44,996 INFO [optim.py:368] (1/8) 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 11:28:08,484 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 17, batch 1350, loss[loss=0.172, simple_loss=0.2701, pruned_loss=0.03702, over 17042.00 frames. ], tot_loss[loss=0.1674, simple_loss=0.2512, pruned_loss=0.04182, over 3309923.31 frames. ], batch size: 55, lr: 4.04e-03, grad_scale: 8.0 2023-04-30 11:28:48,526 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=4.77 vs. limit=5.0 2023-04-30 11:29:15,149 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 17, batch 1400, loss[loss=0.1558, simple_loss=0.2324, pruned_loss=0.03965, over 16839.00 frames. ], tot_loss[loss=0.1681, simple_loss=0.2515, pruned_loss=0.04232, over 3317646.65 frames. ], batch size: 96, lr: 4.04e-03, grad_scale: 8.0 2023-04-30 11:29:52,560 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.5679, 2.1226, 2.3058, 4.3874, 2.2125, 2.4078, 2.2685, 2.3124], device='cuda:1'), covar=tensor([0.1308, 0.4218, 0.3008, 0.0534, 0.5049, 0.3223, 0.3805, 0.4303], device='cuda:1'), in_proj_covar=tensor([0.0386, 0.0424, 0.0357, 0.0325, 0.0429, 0.0490, 0.0395, 0.0497], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-30 11:30:05,128 INFO [optim.py:368] (1/8) 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] (1/8) Epoch 17, batch 1450, loss[loss=0.1789, simple_loss=0.2526, pruned_loss=0.0526, over 16527.00 frames. ], tot_loss[loss=0.167, simple_loss=0.2498, pruned_loss=0.04212, over 3309879.98 frames. ], batch size: 75, lr: 4.04e-03, grad_scale: 8.0 2023-04-30 11:31:26,372 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.6121, 4.5283, 4.5500, 4.2330, 4.1755, 4.5865, 4.4145, 4.3211], device='cuda:1'), covar=tensor([0.0606, 0.0710, 0.0302, 0.0316, 0.0994, 0.0504, 0.0526, 0.0732], device='cuda:1'), in_proj_covar=tensor([0.0284, 0.0395, 0.0331, 0.0322, 0.0344, 0.0373, 0.0228, 0.0398], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002], device='cuda:1') 2023-04-30 11:31:38,096 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=163896.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 11:31:45,954 INFO [train.py:904] (1/8) Epoch 17, batch 1500, loss[loss=0.1775, simple_loss=0.2492, pruned_loss=0.0529, over 16780.00 frames. ], tot_loss[loss=0.1673, simple_loss=0.2501, pruned_loss=0.04223, over 3310850.70 frames. ], batch size: 124, lr: 4.04e-03, grad_scale: 8.0 2023-04-30 11:32:24,622 INFO [optim.py:368] (1/8) 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 11:32:45,101 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=163944.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 11:32:56,202 INFO [train.py:904] (1/8) Epoch 17, batch 1550, loss[loss=0.1755, simple_loss=0.2677, pruned_loss=0.0416, over 17285.00 frames. ], tot_loss[loss=0.1697, simple_loss=0.2524, pruned_loss=0.04346, over 3305836.48 frames. ], batch size: 52, lr: 4.04e-03, grad_scale: 8.0 2023-04-30 11:33:00,286 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.4548, 2.3153, 2.3637, 4.3915, 2.2913, 2.6874, 2.3604, 2.4650], device='cuda:1'), covar=tensor([0.1158, 0.3560, 0.2807, 0.0466, 0.4024, 0.2492, 0.3315, 0.3618], device='cuda:1'), in_proj_covar=tensor([0.0388, 0.0426, 0.0360, 0.0328, 0.0432, 0.0493, 0.0397, 0.0501], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-30 11:33:14,949 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.16 vs. limit=2.0 2023-04-30 11:34:07,075 INFO [train.py:904] (1/8) Epoch 17, batch 1600, loss[loss=0.1484, simple_loss=0.2395, pruned_loss=0.02862, over 17107.00 frames. ], tot_loss[loss=0.1727, simple_loss=0.2557, pruned_loss=0.04489, over 3305148.22 frames. ], batch size: 47, lr: 4.03e-03, grad_scale: 8.0 2023-04-30 11:34:45,102 INFO [optim.py:368] (1/8) 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 11:35:13,146 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.2284, 4.2249, 4.4827, 2.3384, 4.6584, 4.7928, 3.3613, 3.8021], device='cuda:1'), covar=tensor([0.0621, 0.0185, 0.0173, 0.1022, 0.0071, 0.0109, 0.0372, 0.0302], device='cuda:1'), in_proj_covar=tensor([0.0148, 0.0107, 0.0094, 0.0139, 0.0075, 0.0120, 0.0126, 0.0130], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0004, 0.0003], device='cuda:1') 2023-04-30 11:35:15,602 INFO [train.py:904] (1/8) Epoch 17, batch 1650, loss[loss=0.1728, simple_loss=0.2522, pruned_loss=0.04671, over 16827.00 frames. ], tot_loss[loss=0.1735, simple_loss=0.2568, pruned_loss=0.04511, over 3314372.45 frames. ], batch size: 83, lr: 4.03e-03, grad_scale: 8.0 2023-04-30 11:36:00,128 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.70 vs. limit=2.0 2023-04-30 11:36:08,177 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=164090.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 11:36:24,909 INFO [train.py:904] (1/8) Epoch 17, batch 1700, loss[loss=0.1381, simple_loss=0.2258, pruned_loss=0.02526, over 16853.00 frames. ], tot_loss[loss=0.1732, simple_loss=0.2573, pruned_loss=0.04458, over 3324224.42 frames. ], batch size: 42, lr: 4.03e-03, grad_scale: 8.0 2023-04-30 11:36:48,851 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([5.0761, 5.5671, 5.6969, 5.5163, 5.4171, 6.0888, 5.5942, 5.2622], device='cuda:1'), covar=tensor([0.0918, 0.1669, 0.2053, 0.1794, 0.2798, 0.1019, 0.1461, 0.2492], device='cuda:1'), in_proj_covar=tensor([0.0396, 0.0569, 0.0629, 0.0480, 0.0651, 0.0664, 0.0497, 0.0645], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-04-30 11:37:01,949 INFO [optim.py:368] (1/8) 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 11:37:27,588 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.24 vs. limit=2.0 2023-04-30 11:37:31,592 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 17, batch 1750, loss[loss=0.1656, simple_loss=0.2653, pruned_loss=0.03295, over 17255.00 frames. ], tot_loss[loss=0.1732, simple_loss=0.2579, pruned_loss=0.04424, over 3331087.61 frames. ], batch size: 52, lr: 4.03e-03, grad_scale: 8.0 2023-04-30 11:37:36,744 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([5.6216, 6.0400, 5.6991, 5.8208, 5.4604, 5.3264, 5.3650, 6.0982], device='cuda:1'), covar=tensor([0.1340, 0.0897, 0.1110, 0.0783, 0.0817, 0.0699, 0.1190, 0.0881], device='cuda:1'), in_proj_covar=tensor([0.0645, 0.0791, 0.0645, 0.0587, 0.0502, 0.0506, 0.0658, 0.0613], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:1') 2023-04-30 11:37:36,790 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=164155.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 11:38:05,243 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.35 vs. limit=2.0 2023-04-30 11:38:37,734 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.3438, 4.6639, 4.4373, 4.4815, 4.2270, 4.1876, 4.2557, 4.6861], device='cuda:1'), covar=tensor([0.1186, 0.0905, 0.1000, 0.0806, 0.0746, 0.1421, 0.1126, 0.0895], device='cuda:1'), in_proj_covar=tensor([0.0646, 0.0792, 0.0647, 0.0588, 0.0503, 0.0506, 0.0660, 0.0613], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:1') 2023-04-30 11:38:41,937 INFO [train.py:904] (1/8) Epoch 17, batch 1800, loss[loss=0.219, simple_loss=0.3017, pruned_loss=0.06821, over 11902.00 frames. ], tot_loss[loss=0.1737, simple_loss=0.2589, pruned_loss=0.0442, over 3324673.36 frames. ], batch size: 246, lr: 4.03e-03, grad_scale: 8.0 2023-04-30 11:38:57,660 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.8969, 4.4904, 4.5461, 3.2273, 3.7070, 4.4881, 4.0475, 2.6229], device='cuda:1'), covar=tensor([0.0438, 0.0062, 0.0033, 0.0326, 0.0110, 0.0071, 0.0077, 0.0436], device='cuda:1'), in_proj_covar=tensor([0.0134, 0.0078, 0.0077, 0.0132, 0.0091, 0.0101, 0.0089, 0.0125], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-04-30 11:39:00,488 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.8193, 4.0899, 2.5505, 4.5692, 3.1054, 4.6044, 2.6613, 3.2505], device='cuda:1'), covar=tensor([0.0286, 0.0333, 0.1425, 0.0260, 0.0774, 0.0432, 0.1296, 0.0683], device='cuda:1'), in_proj_covar=tensor([0.0168, 0.0175, 0.0195, 0.0157, 0.0175, 0.0218, 0.0204, 0.0181], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-04-30 11:39:01,602 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=164216.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 11:39:19,767 INFO [optim.py:368] (1/8) 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 11:39:49,957 INFO [train.py:904] (1/8) Epoch 17, batch 1850, loss[loss=0.1718, simple_loss=0.2531, pruned_loss=0.04519, over 16773.00 frames. ], tot_loss[loss=0.1743, simple_loss=0.2599, pruned_loss=0.04431, over 3321606.14 frames. ], batch size: 102, lr: 4.03e-03, grad_scale: 8.0 2023-04-30 11:41:01,368 INFO [train.py:904] (1/8) Epoch 17, batch 1900, loss[loss=0.1448, simple_loss=0.2354, pruned_loss=0.02713, over 16209.00 frames. ], tot_loss[loss=0.1736, simple_loss=0.2594, pruned_loss=0.04391, over 3326058.10 frames. ], batch size: 36, lr: 4.03e-03, grad_scale: 8.0 2023-04-30 11:41:26,303 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.6013, 2.5699, 2.1643, 2.4215, 2.9425, 2.6548, 3.2179, 3.0818], device='cuda:1'), covar=tensor([0.0127, 0.0414, 0.0488, 0.0445, 0.0250, 0.0388, 0.0257, 0.0269], device='cuda:1'), in_proj_covar=tensor([0.0194, 0.0231, 0.0222, 0.0221, 0.0230, 0.0232, 0.0236, 0.0225], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-30 11:41:39,856 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.5402, 2.5582, 2.0560, 2.3018, 2.9456, 2.6421, 3.1888, 3.1615], device='cuda:1'), covar=tensor([0.0132, 0.0409, 0.0535, 0.0486, 0.0246, 0.0367, 0.0273, 0.0251], device='cuda:1'), in_proj_covar=tensor([0.0193, 0.0230, 0.0221, 0.0221, 0.0230, 0.0231, 0.0236, 0.0225], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-30 11:41:41,204 INFO [optim.py:368] (1/8) Clipping_scale=2.0, grad-norm quartiles 1.647e+02 2.257e+02 2.647e+02 3.150e+02 4.785e+02, threshold=5.294e+02, percent-clipped=0.0 2023-04-30 11:42:08,488 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.2699, 2.5376, 2.1062, 2.3831, 2.9472, 2.6477, 3.1471, 3.0489], device='cuda:1'), covar=tensor([0.0181, 0.0368, 0.0483, 0.0420, 0.0229, 0.0354, 0.0221, 0.0245], device='cuda:1'), in_proj_covar=tensor([0.0193, 0.0230, 0.0221, 0.0221, 0.0230, 0.0231, 0.0236, 0.0225], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-30 11:42:12,315 INFO [train.py:904] (1/8) Epoch 17, batch 1950, loss[loss=0.2082, simple_loss=0.2918, pruned_loss=0.06236, over 15503.00 frames. ], tot_loss[loss=0.1736, simple_loss=0.2596, pruned_loss=0.04373, over 3320554.27 frames. ], batch size: 190, lr: 4.03e-03, grad_scale: 8.0 2023-04-30 11:42:24,728 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=164360.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 11:42:47,949 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.26 vs. limit=2.0 2023-04-30 11:42:56,749 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.3702, 3.4170, 1.9081, 3.5840, 2.6450, 3.5601, 2.0216, 2.7435], device='cuda:1'), covar=tensor([0.0269, 0.0393, 0.1646, 0.0337, 0.0727, 0.0774, 0.1517, 0.0705], device='cuda:1'), in_proj_covar=tensor([0.0168, 0.0175, 0.0195, 0.0158, 0.0175, 0.0218, 0.0204, 0.0181], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-04-30 11:43:23,961 INFO [train.py:904] (1/8) Epoch 17, batch 2000, loss[loss=0.1658, simple_loss=0.2631, pruned_loss=0.03423, over 17162.00 frames. ], tot_loss[loss=0.1732, simple_loss=0.2594, pruned_loss=0.04344, over 3322787.32 frames. ], batch size: 49, lr: 4.03e-03, grad_scale: 8.0 2023-04-30 11:43:51,002 INFO [zipformer.py:625] (1/8) 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] (1/8) 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 11:44:25,523 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=164446.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 11:44:32,479 INFO [train.py:904] (1/8) Epoch 17, batch 2050, loss[loss=0.1939, simple_loss=0.2655, pruned_loss=0.06116, over 16719.00 frames. ], tot_loss[loss=0.1748, simple_loss=0.2602, pruned_loss=0.04467, over 3324097.68 frames. ], batch size: 124, lr: 4.03e-03, grad_scale: 16.0 2023-04-30 11:44:47,272 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.7905, 3.9437, 2.5673, 4.6395, 2.9933, 4.5498, 2.6902, 3.3419], device='cuda:1'), covar=tensor([0.0283, 0.0389, 0.1395, 0.0195, 0.0808, 0.0517, 0.1285, 0.0605], device='cuda:1'), in_proj_covar=tensor([0.0168, 0.0175, 0.0195, 0.0157, 0.0176, 0.0218, 0.0204, 0.0181], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-04-30 11:44:52,742 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.6428, 2.3217, 2.3419, 4.6193, 2.3430, 2.7601, 2.4257, 2.5481], device='cuda:1'), covar=tensor([0.1049, 0.3626, 0.2881, 0.0367, 0.3997, 0.2499, 0.3366, 0.3721], device='cuda:1'), in_proj_covar=tensor([0.0388, 0.0428, 0.0358, 0.0326, 0.0431, 0.0493, 0.0397, 0.0500], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-30 11:45:07,654 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=4.65 vs. limit=5.0 2023-04-30 11:45:41,559 INFO [train.py:904] (1/8) Epoch 17, batch 2100, loss[loss=0.1799, simple_loss=0.2614, pruned_loss=0.04915, over 16747.00 frames. ], tot_loss[loss=0.1752, simple_loss=0.2608, pruned_loss=0.04477, over 3313623.07 frames. ], batch size: 124, lr: 4.03e-03, grad_scale: 16.0 2023-04-30 11:45:54,964 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=164511.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 11:46:20,650 INFO [optim.py:368] (1/8) Clipping_scale=2.0, grad-norm quartiles 1.600e+02 2.320e+02 2.783e+02 3.356e+02 6.919e+02, threshold=5.567e+02, percent-clipped=1.0 2023-04-30 11:46:38,719 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.19 vs. limit=2.0 2023-04-30 11:46:50,966 INFO [train.py:904] (1/8) Epoch 17, batch 2150, loss[loss=0.1837, simple_loss=0.2706, pruned_loss=0.04844, over 17139.00 frames. ], tot_loss[loss=0.1764, simple_loss=0.2619, pruned_loss=0.04548, over 3322853.09 frames. ], batch size: 49, lr: 4.03e-03, grad_scale: 16.0 2023-04-30 11:47:13,615 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=164568.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 11:47:27,027 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=164578.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 11:47:31,364 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.31 vs. limit=2.0 2023-04-30 11:47:58,403 INFO [train.py:904] (1/8) Epoch 17, batch 2200, loss[loss=0.2359, simple_loss=0.308, pruned_loss=0.08194, over 11859.00 frames. ], tot_loss[loss=0.1771, simple_loss=0.2627, pruned_loss=0.04569, over 3307289.65 frames. ], batch size: 248, lr: 4.03e-03, grad_scale: 16.0 2023-04-30 11:48:36,607 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=164629.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 11:48:37,254 INFO [optim.py:368] (1/8) 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 11:48:44,222 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.5929, 3.9186, 4.1316, 2.9611, 3.5755, 4.1361, 3.7611, 2.4105], device='cuda:1'), covar=tensor([0.0469, 0.0247, 0.0045, 0.0325, 0.0106, 0.0089, 0.0084, 0.0406], device='cuda:1'), in_proj_covar=tensor([0.0134, 0.0077, 0.0077, 0.0132, 0.0091, 0.0101, 0.0089, 0.0125], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-04-30 11:48:44,469 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=3.13 vs. limit=5.0 2023-04-30 11:48:50,052 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=164639.0, num_to_drop=1, layers_to_drop={0} 2023-04-30 11:48:51,342 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.64 vs. limit=2.0 2023-04-30 11:48:57,547 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.5676, 3.4996, 3.8425, 1.9011, 3.9454, 3.9304, 3.0747, 2.9127], device='cuda:1'), covar=tensor([0.0705, 0.0225, 0.0167, 0.1204, 0.0086, 0.0225, 0.0395, 0.0421], device='cuda:1'), in_proj_covar=tensor([0.0148, 0.0108, 0.0095, 0.0139, 0.0075, 0.0122, 0.0127, 0.0131], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0004, 0.0003], device='cuda:1') 2023-04-30 11:49:02,760 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.15 vs. limit=2.0 2023-04-30 11:49:06,809 INFO [train.py:904] (1/8) Epoch 17, batch 2250, loss[loss=0.1529, simple_loss=0.2407, pruned_loss=0.03254, over 17214.00 frames. ], tot_loss[loss=0.1776, simple_loss=0.263, pruned_loss=0.04608, over 3309437.74 frames. ], batch size: 44, lr: 4.03e-03, grad_scale: 16.0 2023-04-30 11:49:15,021 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 17, batch 2300, loss[loss=0.1763, simple_loss=0.2705, pruned_loss=0.04102, over 16731.00 frames. ], tot_loss[loss=0.178, simple_loss=0.2634, pruned_loss=0.04624, over 3299572.63 frames. ], batch size: 57, lr: 4.03e-03, grad_scale: 16.0 2023-04-30 11:50:34,575 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=164716.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 11:50:38,871 INFO [zipformer.py:625] (1/8) 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] (1/8) 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 11:51:16,720 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=164746.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 11:51:16,848 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=164746.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 11:51:23,973 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.3462, 2.3133, 2.3059, 4.2616, 2.1818, 2.6639, 2.3464, 2.4027], device='cuda:1'), covar=tensor([0.1237, 0.3585, 0.2885, 0.0506, 0.4160, 0.2627, 0.3443, 0.3524], device='cuda:1'), in_proj_covar=tensor([0.0390, 0.0429, 0.0359, 0.0327, 0.0431, 0.0496, 0.0398, 0.0501], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-30 11:51:24,576 INFO [train.py:904] (1/8) Epoch 17, batch 2350, loss[loss=0.1782, simple_loss=0.275, pruned_loss=0.04064, over 16665.00 frames. ], tot_loss[loss=0.1775, simple_loss=0.2632, pruned_loss=0.04589, over 3310267.21 frames. ], batch size: 62, lr: 4.03e-03, grad_scale: 16.0 2023-04-30 11:52:23,521 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=164794.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 11:52:24,206 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.42 vs. limit=2.0 2023-04-30 11:52:34,377 INFO [train.py:904] (1/8) Epoch 17, batch 2400, loss[loss=0.1911, simple_loss=0.2738, pruned_loss=0.05421, over 16837.00 frames. ], tot_loss[loss=0.178, simple_loss=0.2639, pruned_loss=0.046, over 3322400.79 frames. ], batch size: 96, lr: 4.03e-03, grad_scale: 16.0 2023-04-30 11:52:41,594 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=164807.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 11:52:46,683 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=164811.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 11:53:12,671 INFO [optim.py:368] (1/8) 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 11:53:15,538 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([1.7873, 3.0221, 2.8549, 5.0092, 4.1112, 4.4310, 1.6990, 3.2163], device='cuda:1'), covar=tensor([0.1336, 0.0733, 0.1084, 0.0169, 0.0247, 0.0373, 0.1557, 0.0756], device='cuda:1'), in_proj_covar=tensor([0.0158, 0.0164, 0.0185, 0.0172, 0.0198, 0.0210, 0.0190, 0.0186], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-04-30 11:53:19,569 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.7012, 3.7874, 2.2959, 4.4313, 2.9447, 4.4054, 2.4370, 3.1644], device='cuda:1'), covar=tensor([0.0291, 0.0347, 0.1561, 0.0213, 0.0815, 0.0447, 0.1499, 0.0710], device='cuda:1'), in_proj_covar=tensor([0.0167, 0.0174, 0.0193, 0.0157, 0.0173, 0.0217, 0.0202, 0.0180], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-04-30 11:53:41,598 INFO [train.py:904] (1/8) Epoch 17, batch 2450, loss[loss=0.1597, simple_loss=0.2473, pruned_loss=0.03608, over 16247.00 frames. ], tot_loss[loss=0.1777, simple_loss=0.2645, pruned_loss=0.04546, over 3329105.44 frames. ], batch size: 36, lr: 4.02e-03, grad_scale: 8.0 2023-04-30 11:53:47,294 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([5.4281, 5.8598, 5.5887, 5.6598, 5.2436, 5.1612, 5.2124, 6.0122], device='cuda:1'), covar=tensor([0.1246, 0.0876, 0.0956, 0.0798, 0.0833, 0.0739, 0.1055, 0.0827], device='cuda:1'), in_proj_covar=tensor([0.0647, 0.0797, 0.0649, 0.0593, 0.0506, 0.0509, 0.0662, 0.0618], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:1') 2023-04-30 11:53:51,211 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 17, batch 2500, loss[loss=0.1558, simple_loss=0.2404, pruned_loss=0.03563, over 16755.00 frames. ], tot_loss[loss=0.1766, simple_loss=0.2635, pruned_loss=0.04488, over 3326763.34 frames. ], batch size: 89, lr: 4.02e-03, grad_scale: 8.0 2023-04-30 11:55:12,208 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.7706, 2.4424, 2.4975, 4.7351, 2.4503, 2.8322, 2.5261, 2.6051], device='cuda:1'), covar=tensor([0.1046, 0.3488, 0.2781, 0.0373, 0.3867, 0.2419, 0.3363, 0.3584], device='cuda:1'), in_proj_covar=tensor([0.0388, 0.0427, 0.0357, 0.0326, 0.0429, 0.0494, 0.0396, 0.0500], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-30 11:55:17,582 INFO [zipformer.py:625] (1/8) 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] (1/8) 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 11:55:29,625 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.39 vs. limit=2.0 2023-04-30 11:55:30,296 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=164934.0, num_to_drop=1, layers_to_drop={3} 2023-04-30 11:55:55,231 INFO [train.py:904] (1/8) Epoch 17, batch 2550, loss[loss=0.2035, simple_loss=0.2909, pruned_loss=0.05806, over 16164.00 frames. ], tot_loss[loss=0.1762, simple_loss=0.2628, pruned_loss=0.04475, over 3328254.16 frames. ], batch size: 35, lr: 4.02e-03, grad_scale: 8.0 2023-04-30 11:57:02,085 INFO [train.py:904] (1/8) Epoch 17, batch 2600, loss[loss=0.1796, simple_loss=0.2628, pruned_loss=0.04824, over 15594.00 frames. ], tot_loss[loss=0.1758, simple_loss=0.2622, pruned_loss=0.04466, over 3336610.59 frames. ], batch size: 191, lr: 4.02e-03, grad_scale: 8.0 2023-04-30 11:57:17,257 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=165014.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 11:57:20,602 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=165016.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 11:57:31,759 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([1.9126, 3.2350, 2.9918, 5.0816, 4.2297, 4.5665, 1.8109, 3.3945], device='cuda:1'), covar=tensor([0.1235, 0.0657, 0.1014, 0.0166, 0.0224, 0.0329, 0.1473, 0.0666], device='cuda:1'), in_proj_covar=tensor([0.0159, 0.0166, 0.0187, 0.0174, 0.0199, 0.0212, 0.0192, 0.0187], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-04-30 11:57:41,410 INFO [optim.py:368] (1/8) 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] (1/8) attn_weights_entropy = tensor([4.8192, 5.1817, 4.8796, 4.9642, 4.6875, 4.6412, 4.6118, 5.2645], device='cuda:1'), covar=tensor([0.1200, 0.0857, 0.1088, 0.0824, 0.0895, 0.1167, 0.1195, 0.0910], device='cuda:1'), in_proj_covar=tensor([0.0647, 0.0800, 0.0651, 0.0594, 0.0508, 0.0509, 0.0663, 0.0618], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:1') 2023-04-30 11:57:56,052 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.31 vs. limit=2.0 2023-04-30 11:58:08,462 INFO [train.py:904] (1/8) Epoch 17, batch 2650, loss[loss=0.1842, simple_loss=0.2772, pruned_loss=0.04559, over 17172.00 frames. ], tot_loss[loss=0.176, simple_loss=0.2634, pruned_loss=0.04429, over 3341197.03 frames. ], batch size: 46, lr: 4.02e-03, grad_scale: 8.0 2023-04-30 11:58:26,259 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=165064.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 11:58:57,882 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([5.1349, 5.0966, 4.8989, 3.9224, 4.9882, 1.7708, 4.6872, 4.8410], device='cuda:1'), covar=tensor([0.0107, 0.0094, 0.0219, 0.0606, 0.0120, 0.3046, 0.0172, 0.0239], device='cuda:1'), in_proj_covar=tensor([0.0154, 0.0144, 0.0192, 0.0175, 0.0165, 0.0202, 0.0180, 0.0172], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-30 11:59:18,027 INFO [train.py:904] (1/8) Epoch 17, batch 2700, loss[loss=0.1544, simple_loss=0.2451, pruned_loss=0.03183, over 17209.00 frames. ], tot_loss[loss=0.1754, simple_loss=0.263, pruned_loss=0.04393, over 3338473.55 frames. ], batch size: 46, lr: 4.02e-03, grad_scale: 8.0 2023-04-30 11:59:18,299 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=165102.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 11:59:47,945 INFO [zipformer.py:625] (1/8) 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] (1/8) 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 12:00:28,486 INFO [train.py:904] (1/8) Epoch 17, batch 2750, loss[loss=0.1575, simple_loss=0.2474, pruned_loss=0.03374, over 17206.00 frames. ], tot_loss[loss=0.1749, simple_loss=0.2628, pruned_loss=0.04351, over 3332965.97 frames. ], batch size: 44, lr: 4.02e-03, grad_scale: 8.0 2023-04-30 12:01:12,623 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=165184.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 12:01:38,386 INFO [train.py:904] (1/8) Epoch 17, batch 2800, loss[loss=0.1794, simple_loss=0.2818, pruned_loss=0.03849, over 17037.00 frames. ], tot_loss[loss=0.1749, simple_loss=0.2623, pruned_loss=0.04375, over 3328673.20 frames. ], batch size: 50, lr: 4.02e-03, grad_scale: 8.0 2023-04-30 12:02:09,599 INFO [zipformer.py:625] (1/8) 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] (1/8) 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 12:02:24,058 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 17, batch 2850, loss[loss=0.1555, simple_loss=0.2434, pruned_loss=0.03377, over 15763.00 frames. ], tot_loss[loss=0.1738, simple_loss=0.2609, pruned_loss=0.04336, over 3331316.20 frames. ], batch size: 35, lr: 4.02e-03, grad_scale: 8.0 2023-04-30 12:03:18,271 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=165272.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 12:03:31,430 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=165282.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 12:03:58,711 INFO [train.py:904] (1/8) Epoch 17, batch 2900, loss[loss=0.1818, simple_loss=0.2509, pruned_loss=0.05641, over 16845.00 frames. ], tot_loss[loss=0.174, simple_loss=0.2606, pruned_loss=0.0437, over 3335159.02 frames. ], batch size: 116, lr: 4.02e-03, grad_scale: 8.0 2023-04-30 12:04:16,908 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=165314.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 12:04:39,537 INFO [optim.py:368] (1/8) 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 12:04:55,341 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=4.59 vs. limit=5.0 2023-04-30 12:05:09,243 INFO [train.py:904] (1/8) Epoch 17, batch 2950, loss[loss=0.1801, simple_loss=0.282, pruned_loss=0.03908, over 17124.00 frames. ], tot_loss[loss=0.175, simple_loss=0.261, pruned_loss=0.04452, over 3327765.62 frames. ], batch size: 49, lr: 4.02e-03, grad_scale: 8.0 2023-04-30 12:05:23,905 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=165362.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 12:06:02,031 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.3213, 4.2522, 4.5652, 2.4540, 4.9077, 4.8881, 3.4827, 3.8777], device='cuda:1'), covar=tensor([0.0641, 0.0206, 0.0210, 0.1016, 0.0052, 0.0135, 0.0374, 0.0298], device='cuda:1'), in_proj_covar=tensor([0.0147, 0.0106, 0.0094, 0.0137, 0.0075, 0.0121, 0.0126, 0.0129], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0004, 0.0003], device='cuda:1') 2023-04-30 12:06:20,614 INFO [train.py:904] (1/8) Epoch 17, batch 3000, loss[loss=0.1621, simple_loss=0.24, pruned_loss=0.04208, over 16684.00 frames. ], tot_loss[loss=0.1754, simple_loss=0.2611, pruned_loss=0.04483, over 3333813.00 frames. ], batch size: 89, lr: 4.02e-03, grad_scale: 8.0 2023-04-30 12:06:20,614 INFO [train.py:929] (1/8) Computing validation loss 2023-04-30 12:06:29,126 INFO [train.py:938] (1/8) Epoch 17, validation: loss=0.1364, simple_loss=0.2422, pruned_loss=0.0153, over 944034.00 frames. 2023-04-30 12:06:29,127 INFO [train.py:939] (1/8) Maximum memory allocated so far is 17857MB 2023-04-30 12:06:29,417 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=165402.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 12:06:46,649 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.4976, 3.3752, 2.6755, 2.0870, 2.3132, 2.2269, 3.5195, 3.0990], device='cuda:1'), covar=tensor([0.2655, 0.0676, 0.1746, 0.2842, 0.2590, 0.2047, 0.0612, 0.1437], device='cuda:1'), in_proj_covar=tensor([0.0317, 0.0263, 0.0296, 0.0298, 0.0289, 0.0242, 0.0282, 0.0323], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:1') 2023-04-30 12:06:51,033 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=4.90 vs. limit=5.0 2023-04-30 12:07:09,773 INFO [optim.py:368] (1/8) Clipping_scale=2.0, grad-norm quartiles 1.715e+02 2.395e+02 2.828e+02 3.269e+02 8.319e+02, threshold=5.656e+02, percent-clipped=3.0 2023-04-30 12:07:36,651 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 17, batch 3050, loss[loss=0.1672, simple_loss=0.2592, pruned_loss=0.03758, over 17169.00 frames. ], tot_loss[loss=0.1751, simple_loss=0.2607, pruned_loss=0.04479, over 3336300.35 frames. ], batch size: 46, lr: 4.02e-03, grad_scale: 4.0 2023-04-30 12:07:59,187 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([5.0840, 5.4802, 5.5894, 5.2686, 5.3908, 5.9846, 5.4947, 5.2361], device='cuda:1'), covar=tensor([0.1058, 0.2086, 0.2513, 0.2352, 0.2850, 0.1119, 0.1477, 0.2633], device='cuda:1'), in_proj_covar=tensor([0.0402, 0.0578, 0.0637, 0.0489, 0.0657, 0.0671, 0.0503, 0.0655], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-04-30 12:08:12,761 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.48 vs. limit=2.0 2023-04-30 12:08:15,873 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 17, batch 3100, loss[loss=0.1922, simple_loss=0.2781, pruned_loss=0.05313, over 17025.00 frames. ], tot_loss[loss=0.1751, simple_loss=0.2607, pruned_loss=0.04477, over 3335395.36 frames. ], batch size: 55, lr: 4.02e-03, grad_scale: 4.0 2023-04-30 12:09:27,660 INFO [optim.py:368] (1/8) 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 12:09:54,943 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=165551.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 12:09:55,706 INFO [train.py:904] (1/8) Epoch 17, batch 3150, loss[loss=0.1957, simple_loss=0.2645, pruned_loss=0.0635, over 16885.00 frames. ], tot_loss[loss=0.1753, simple_loss=0.2608, pruned_loss=0.04486, over 3329119.90 frames. ], batch size: 109, lr: 4.02e-03, grad_scale: 4.0 2023-04-30 12:10:17,940 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.96 vs. limit=2.0 2023-04-30 12:11:06,274 INFO [train.py:904] (1/8) Epoch 17, batch 3200, loss[loss=0.1761, simple_loss=0.2688, pruned_loss=0.04171, over 16508.00 frames. ], tot_loss[loss=0.1744, simple_loss=0.2598, pruned_loss=0.04444, over 3322939.75 frames. ], batch size: 68, lr: 4.02e-03, grad_scale: 8.0 2023-04-30 12:11:20,426 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=165612.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 12:11:23,498 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.3226, 2.5717, 2.1569, 2.3709, 2.9243, 2.6916, 3.1770, 3.1482], device='cuda:1'), covar=tensor([0.0214, 0.0370, 0.0471, 0.0392, 0.0248, 0.0360, 0.0256, 0.0233], device='cuda:1'), in_proj_covar=tensor([0.0197, 0.0231, 0.0221, 0.0222, 0.0233, 0.0232, 0.0238, 0.0228], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-30 12:11:49,641 INFO [optim.py:368] (1/8) 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 12:12:15,439 INFO [train.py:904] (1/8) Epoch 17, batch 3250, loss[loss=0.1969, simple_loss=0.2797, pruned_loss=0.05709, over 15472.00 frames. ], tot_loss[loss=0.1747, simple_loss=0.26, pruned_loss=0.04471, over 3329791.05 frames. ], batch size: 190, lr: 4.01e-03, grad_scale: 4.0 2023-04-30 12:12:49,776 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.46 vs. limit=2.0 2023-04-30 12:13:23,327 INFO [train.py:904] (1/8) Epoch 17, batch 3300, loss[loss=0.1672, simple_loss=0.2508, pruned_loss=0.04184, over 16502.00 frames. ], tot_loss[loss=0.1752, simple_loss=0.2604, pruned_loss=0.04496, over 3326345.23 frames. ], batch size: 75, lr: 4.01e-03, grad_scale: 4.0 2023-04-30 12:14:06,777 INFO [optim.py:368] (1/8) 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 12:14:30,175 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 17, batch 3350, loss[loss=0.1622, simple_loss=0.2495, pruned_loss=0.03746, over 16816.00 frames. ], tot_loss[loss=0.1753, simple_loss=0.2612, pruned_loss=0.04476, over 3326132.52 frames. ], batch size: 83, lr: 4.01e-03, grad_scale: 4.0 2023-04-30 12:15:08,289 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=2.98 vs. limit=5.0 2023-04-30 12:15:11,004 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=165779.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 12:15:42,358 INFO [train.py:904] (1/8) Epoch 17, batch 3400, loss[loss=0.1713, simple_loss=0.2641, pruned_loss=0.03925, over 16747.00 frames. ], tot_loss[loss=0.1756, simple_loss=0.2611, pruned_loss=0.04501, over 3323473.53 frames. ], batch size: 62, lr: 4.01e-03, grad_scale: 4.0 2023-04-30 12:15:44,609 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=165803.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 12:15:56,215 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=165811.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 12:16:18,378 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=165827.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 12:16:27,735 INFO [optim.py:368] (1/8) 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 12:16:54,380 INFO [train.py:904] (1/8) Epoch 17, batch 3450, loss[loss=0.153, simple_loss=0.2369, pruned_loss=0.0345, over 15832.00 frames. ], tot_loss[loss=0.1739, simple_loss=0.2595, pruned_loss=0.04415, over 3325346.82 frames. ], batch size: 35, lr: 4.01e-03, grad_scale: 4.0 2023-04-30 12:17:11,499 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=165864.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 12:18:05,607 INFO [train.py:904] (1/8) Epoch 17, batch 3500, loss[loss=0.1739, simple_loss=0.2495, pruned_loss=0.04919, over 16697.00 frames. ], tot_loss[loss=0.1728, simple_loss=0.2582, pruned_loss=0.0437, over 3320649.85 frames. ], batch size: 124, lr: 4.01e-03, grad_scale: 4.0 2023-04-30 12:18:13,074 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=165907.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 12:18:45,783 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=3.54 vs. limit=5.0 2023-04-30 12:18:49,909 INFO [optim.py:368] (1/8) 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] (1/8) Epoch 17, batch 3550, loss[loss=0.1902, simple_loss=0.2639, pruned_loss=0.0582, over 16230.00 frames. ], tot_loss[loss=0.1723, simple_loss=0.257, pruned_loss=0.04376, over 3307727.48 frames. ], batch size: 164, lr: 4.01e-03, grad_scale: 4.0 2023-04-30 12:19:59,866 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.4548, 2.3696, 2.4592, 4.2397, 2.2562, 2.7668, 2.4308, 2.5700], device='cuda:1'), covar=tensor([0.1247, 0.3381, 0.2583, 0.0529, 0.3840, 0.2428, 0.3563, 0.3026], device='cuda:1'), in_proj_covar=tensor([0.0391, 0.0429, 0.0357, 0.0327, 0.0430, 0.0498, 0.0399, 0.0503], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-30 12:20:27,933 INFO [train.py:904] (1/8) Epoch 17, batch 3600, loss[loss=0.1687, simple_loss=0.2482, pruned_loss=0.04463, over 15444.00 frames. ], tot_loss[loss=0.172, simple_loss=0.2568, pruned_loss=0.04364, over 3310431.73 frames. ], batch size: 191, lr: 4.01e-03, grad_scale: 8.0 2023-04-30 12:21:12,106 INFO [optim.py:368] (1/8) Clipping_scale=2.0, grad-norm quartiles 1.506e+02 2.087e+02 2.490e+02 2.982e+02 5.501e+02, threshold=4.979e+02, percent-clipped=1.0 2023-04-30 12:21:40,964 INFO [train.py:904] (1/8) Epoch 17, batch 3650, loss[loss=0.1914, simple_loss=0.2646, pruned_loss=0.0591, over 16845.00 frames. ], tot_loss[loss=0.1724, simple_loss=0.2565, pruned_loss=0.04414, over 3309356.66 frames. ], batch size: 83, lr: 4.01e-03, grad_scale: 8.0 2023-04-30 12:21:43,852 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=166054.0, num_to_drop=1, layers_to_drop={1} 2023-04-30 12:22:16,717 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.7463, 1.8064, 2.3457, 2.5812, 2.6469, 2.5516, 1.9178, 2.8252], device='cuda:1'), covar=tensor([0.0127, 0.0433, 0.0258, 0.0245, 0.0237, 0.0297, 0.0430, 0.0160], device='cuda:1'), in_proj_covar=tensor([0.0182, 0.0189, 0.0175, 0.0179, 0.0189, 0.0146, 0.0189, 0.0140], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-30 12:22:55,839 INFO [train.py:904] (1/8) Epoch 17, batch 3700, loss[loss=0.1933, simple_loss=0.2634, pruned_loss=0.06162, over 16824.00 frames. ], tot_loss[loss=0.1727, simple_loss=0.2549, pruned_loss=0.04529, over 3284740.00 frames. ], batch size: 109, lr: 4.01e-03, grad_scale: 8.0 2023-04-30 12:23:02,184 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=166106.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 12:23:15,457 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=166115.0, num_to_drop=1, layers_to_drop={0} 2023-04-30 12:23:29,641 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.8352, 2.5591, 2.1541, 2.2719, 3.0072, 2.7548, 3.0626, 3.0625], device='cuda:1'), covar=tensor([0.0157, 0.0343, 0.0444, 0.0433, 0.0194, 0.0283, 0.0183, 0.0231], device='cuda:1'), in_proj_covar=tensor([0.0197, 0.0231, 0.0221, 0.0223, 0.0233, 0.0232, 0.0237, 0.0228], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-30 12:23:42,388 INFO [optim.py:368] (1/8) 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 12:24:03,608 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=166147.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 12:24:10,328 INFO [train.py:904] (1/8) Epoch 17, batch 3750, loss[loss=0.1782, simple_loss=0.2551, pruned_loss=0.05067, over 16799.00 frames. ], tot_loss[loss=0.1749, simple_loss=0.2558, pruned_loss=0.04704, over 3274392.45 frames. ], batch size: 102, lr: 4.01e-03, grad_scale: 8.0 2023-04-30 12:24:21,421 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=166159.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 12:24:48,212 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=3.80 vs. limit=5.0 2023-04-30 12:25:24,424 INFO [train.py:904] (1/8) Epoch 17, batch 3800, loss[loss=0.1738, simple_loss=0.2515, pruned_loss=0.04808, over 16795.00 frames. ], tot_loss[loss=0.1764, simple_loss=0.2565, pruned_loss=0.04818, over 3266279.92 frames. ], batch size: 83, lr: 4.01e-03, grad_scale: 8.0 2023-04-30 12:25:32,318 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=166207.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 12:25:33,758 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=166208.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 12:25:50,885 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.7616, 3.0275, 2.9572, 2.0733, 2.6635, 2.2039, 3.3585, 3.3940], device='cuda:1'), covar=tensor([0.0300, 0.0872, 0.0704, 0.1829, 0.0928, 0.1032, 0.0573, 0.0832], device='cuda:1'), in_proj_covar=tensor([0.0151, 0.0160, 0.0164, 0.0150, 0.0141, 0.0127, 0.0140, 0.0170], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:1') 2023-04-30 12:26:10,618 INFO [optim.py:368] (1/8) 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 12:26:38,502 INFO [train.py:904] (1/8) Epoch 17, batch 3850, loss[loss=0.1899, simple_loss=0.2584, pruned_loss=0.06072, over 16841.00 frames. ], tot_loss[loss=0.1779, simple_loss=0.2576, pruned_loss=0.04911, over 3260304.23 frames. ], batch size: 96, lr: 4.01e-03, grad_scale: 8.0 2023-04-30 12:26:44,004 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=166255.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 12:27:52,935 INFO [train.py:904] (1/8) Epoch 17, batch 3900, loss[loss=0.1642, simple_loss=0.2381, pruned_loss=0.04514, over 16476.00 frames. ], tot_loss[loss=0.1785, simple_loss=0.2572, pruned_loss=0.04986, over 3266438.32 frames. ], batch size: 68, lr: 4.01e-03, grad_scale: 8.0 2023-04-30 12:28:00,978 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=166307.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 12:28:01,202 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.66 vs. limit=2.0 2023-04-30 12:28:37,861 INFO [optim.py:368] (1/8) 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 12:29:07,066 INFO [train.py:904] (1/8) Epoch 17, batch 3950, loss[loss=0.1906, simple_loss=0.2698, pruned_loss=0.05572, over 16314.00 frames. ], tot_loss[loss=0.1784, simple_loss=0.2564, pruned_loss=0.05014, over 3279152.61 frames. ], batch size: 165, lr: 4.01e-03, grad_scale: 8.0 2023-04-30 12:29:30,587 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=166368.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 12:29:44,858 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.96 vs. limit=2.0 2023-04-30 12:30:18,532 INFO [train.py:904] (1/8) Epoch 17, batch 4000, loss[loss=0.181, simple_loss=0.262, pruned_loss=0.05, over 16229.00 frames. ], tot_loss[loss=0.1791, simple_loss=0.257, pruned_loss=0.05063, over 3281314.87 frames. ], batch size: 165, lr: 4.01e-03, grad_scale: 8.0 2023-04-30 12:30:25,168 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=166406.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 12:30:31,135 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=166410.0, num_to_drop=1, layers_to_drop={2} 2023-04-30 12:31:03,721 INFO [optim.py:368] (1/8) 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] (1/8) Epoch 17, batch 4050, loss[loss=0.175, simple_loss=0.2593, pruned_loss=0.04538, over 17170.00 frames. ], tot_loss[loss=0.178, simple_loss=0.2573, pruned_loss=0.04939, over 3279513.74 frames. ], batch size: 46, lr: 4.01e-03, grad_scale: 8.0 2023-04-30 12:31:34,056 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=166454.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 12:31:41,937 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=166459.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 12:31:46,072 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.6932, 3.8381, 4.2276, 1.7848, 4.5108, 4.5808, 3.1744, 3.2607], device='cuda:1'), covar=tensor([0.0837, 0.0285, 0.0199, 0.1406, 0.0067, 0.0078, 0.0394, 0.0466], device='cuda:1'), in_proj_covar=tensor([0.0149, 0.0108, 0.0095, 0.0139, 0.0077, 0.0123, 0.0128, 0.0132], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0004, 0.0004], device='cuda:1') 2023-04-30 12:32:14,530 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=166481.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 12:32:44,111 INFO [train.py:904] (1/8) Epoch 17, batch 4100, loss[loss=0.2184, simple_loss=0.3053, pruned_loss=0.06572, over 16202.00 frames. ], tot_loss[loss=0.1782, simple_loss=0.2588, pruned_loss=0.04883, over 3274626.78 frames. ], batch size: 165, lr: 4.00e-03, grad_scale: 8.0 2023-04-30 12:32:46,284 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=166503.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 12:32:51,045 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=166507.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 12:32:54,580 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=166509.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 12:33:29,255 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=166531.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 12:33:31,670 INFO [optim.py:368] (1/8) 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 12:33:45,777 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 17, batch 4150, loss[loss=0.2268, simple_loss=0.3042, pruned_loss=0.0747, over 11461.00 frames. ], tot_loss[loss=0.1841, simple_loss=0.2657, pruned_loss=0.05126, over 3228969.11 frames. ], batch size: 246, lr: 4.00e-03, grad_scale: 8.0 2023-04-30 12:34:28,471 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=166570.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 12:35:00,487 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=166592.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 12:35:04,676 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.6975, 3.7034, 3.8681, 3.6189, 3.7583, 4.1883, 3.8542, 3.5668], device='cuda:1'), covar=tensor([0.2109, 0.2206, 0.1949, 0.2403, 0.2693, 0.1550, 0.1527, 0.2531], device='cuda:1'), in_proj_covar=tensor([0.0391, 0.0562, 0.0615, 0.0474, 0.0633, 0.0653, 0.0488, 0.0634], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-04-30 12:35:14,002 INFO [train.py:904] (1/8) Epoch 17, batch 4200, loss[loss=0.1993, simple_loss=0.2894, pruned_loss=0.05462, over 16709.00 frames. ], tot_loss[loss=0.1894, simple_loss=0.2724, pruned_loss=0.05319, over 3183392.46 frames. ], batch size: 89, lr: 4.00e-03, grad_scale: 8.0 2023-04-30 12:36:00,018 INFO [optim.py:368] (1/8) Clipping_scale=2.0, grad-norm quartiles 1.639e+02 2.403e+02 2.795e+02 3.348e+02 5.640e+02, threshold=5.591e+02, percent-clipped=5.0 2023-04-30 12:36:13,704 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.37 vs. limit=2.0 2023-04-30 12:36:27,721 INFO [train.py:904] (1/8) Epoch 17, batch 4250, loss[loss=0.1836, simple_loss=0.2779, pruned_loss=0.0446, over 16364.00 frames. ], tot_loss[loss=0.1912, simple_loss=0.2762, pruned_loss=0.05312, over 3182036.45 frames. ], batch size: 146, lr: 4.00e-03, grad_scale: 8.0 2023-04-30 12:36:42,896 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=166663.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 12:36:49,771 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.8619, 4.9872, 5.2659, 5.2409, 5.2600, 4.9604, 4.8908, 4.5731], device='cuda:1'), covar=tensor([0.0251, 0.0390, 0.0382, 0.0415, 0.0369, 0.0295, 0.0841, 0.0429], device='cuda:1'), in_proj_covar=tensor([0.0378, 0.0415, 0.0404, 0.0381, 0.0450, 0.0426, 0.0524, 0.0339], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:1') 2023-04-30 12:37:33,351 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.1636, 2.3540, 1.8343, 2.1635, 2.7697, 2.4034, 2.8265, 3.0325], device='cuda:1'), covar=tensor([0.0127, 0.0455, 0.0613, 0.0505, 0.0241, 0.0398, 0.0236, 0.0211], device='cuda:1'), in_proj_covar=tensor([0.0191, 0.0225, 0.0216, 0.0218, 0.0227, 0.0226, 0.0230, 0.0222], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-30 12:37:39,011 INFO [train.py:904] (1/8) Epoch 17, batch 4300, loss[loss=0.2058, simple_loss=0.2998, pruned_loss=0.05588, over 16876.00 frames. ], tot_loss[loss=0.191, simple_loss=0.2776, pruned_loss=0.05224, over 3175910.19 frames. ], batch size: 42, lr: 4.00e-03, grad_scale: 8.0 2023-04-30 12:37:51,443 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=166710.0, num_to_drop=1, layers_to_drop={1} 2023-04-30 12:38:24,693 INFO [optim.py:368] (1/8) 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 12:38:28,866 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=166735.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 12:38:52,932 INFO [train.py:904] (1/8) Epoch 17, batch 4350, loss[loss=0.2169, simple_loss=0.3028, pruned_loss=0.06549, over 16877.00 frames. ], tot_loss[loss=0.1939, simple_loss=0.2809, pruned_loss=0.05343, over 3154063.62 frames. ], batch size: 109, lr: 4.00e-03, grad_scale: 8.0 2023-04-30 12:39:01,858 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=166758.0, num_to_drop=1, layers_to_drop={0} 2023-04-30 12:39:10,351 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=166764.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 12:39:57,546 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=166796.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 12:40:05,343 INFO [train.py:904] (1/8) Epoch 17, batch 4400, loss[loss=0.2058, simple_loss=0.297, pruned_loss=0.0573, over 16422.00 frames. ], tot_loss[loss=0.1955, simple_loss=0.2828, pruned_loss=0.0541, over 3167522.34 frames. ], batch size: 75, lr: 4.00e-03, grad_scale: 8.0 2023-04-30 12:40:06,864 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=166803.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 12:40:37,929 INFO [zipformer.py:625] (1/8) 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] (1/8) 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] (1/8) 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] (1/8) 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] (1/8) Epoch 17, batch 4450, loss[loss=0.2237, simple_loss=0.3145, pruned_loss=0.06642, over 16773.00 frames. ], tot_loss[loss=0.1976, simple_loss=0.2854, pruned_loss=0.05489, over 3179435.08 frames. ], batch size: 124, lr: 4.00e-03, grad_scale: 8.0 2023-04-30 12:41:36,423 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=166865.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 12:42:07,487 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=166887.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 12:42:28,861 INFO [train.py:904] (1/8) Epoch 17, batch 4500, loss[loss=0.2137, simple_loss=0.2949, pruned_loss=0.06629, over 15456.00 frames. ], tot_loss[loss=0.1991, simple_loss=0.2865, pruned_loss=0.05579, over 3174791.26 frames. ], batch size: 191, lr: 4.00e-03, grad_scale: 8.0 2023-04-30 12:43:07,234 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.8368, 4.6515, 4.8771, 5.0310, 5.1683, 4.6268, 5.1814, 5.2033], device='cuda:1'), covar=tensor([0.1592, 0.1040, 0.1289, 0.0584, 0.0485, 0.0824, 0.0450, 0.0463], device='cuda:1'), in_proj_covar=tensor([0.0601, 0.0741, 0.0878, 0.0754, 0.0566, 0.0602, 0.0597, 0.0701], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-30 12:43:07,262 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=166929.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 12:43:12,781 INFO [optim.py:368] (1/8) Clipping_scale=2.0, grad-norm quartiles 1.487e+02 1.901e+02 2.204e+02 2.572e+02 5.344e+02, threshold=4.409e+02, percent-clipped=1.0 2023-04-30 12:43:31,330 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.0095, 1.9180, 2.5142, 2.9381, 2.8548, 3.2871, 1.9454, 3.2649], device='cuda:1'), covar=tensor([0.0160, 0.0432, 0.0251, 0.0202, 0.0217, 0.0134, 0.0466, 0.0107], device='cuda:1'), in_proj_covar=tensor([0.0178, 0.0187, 0.0173, 0.0176, 0.0186, 0.0143, 0.0187, 0.0137], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-30 12:43:40,951 INFO [train.py:904] (1/8) Epoch 17, batch 4550, loss[loss=0.1945, simple_loss=0.2861, pruned_loss=0.05145, over 16781.00 frames. ], tot_loss[loss=0.1998, simple_loss=0.2867, pruned_loss=0.05649, over 3196971.37 frames. ], batch size: 124, lr: 4.00e-03, grad_scale: 8.0 2023-04-30 12:43:45,680 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.3946, 4.4586, 4.5910, 4.3886, 4.4795, 5.0105, 4.5040, 4.1988], device='cuda:1'), covar=tensor([0.1323, 0.1965, 0.2062, 0.1971, 0.2585, 0.1022, 0.1472, 0.2477], device='cuda:1'), in_proj_covar=tensor([0.0386, 0.0557, 0.0606, 0.0465, 0.0624, 0.0643, 0.0477, 0.0624], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-04-30 12:43:57,119 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=166963.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 12:44:35,665 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 17, batch 4600, loss[loss=0.22, simple_loss=0.3041, pruned_loss=0.06795, over 17017.00 frames. ], tot_loss[loss=0.2005, simple_loss=0.2874, pruned_loss=0.05674, over 3216645.52 frames. ], batch size: 41, lr: 4.00e-03, grad_scale: 8.0 2023-04-30 12:45:07,098 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=167011.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 12:45:11,650 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.69 vs. limit=2.0 2023-04-30 12:45:38,257 INFO [optim.py:368] (1/8) Clipping_scale=2.0, grad-norm quartiles 1.244e+02 1.923e+02 2.253e+02 2.642e+02 5.543e+02, threshold=4.506e+02, percent-clipped=2.0 2023-04-30 12:46:05,380 INFO [train.py:904] (1/8) Epoch 17, batch 4650, loss[loss=0.187, simple_loss=0.2693, pruned_loss=0.0524, over 17205.00 frames. ], tot_loss[loss=0.2012, simple_loss=0.2875, pruned_loss=0.05741, over 3199275.98 frames. ], batch size: 45, lr: 4.00e-03, grad_scale: 8.0 2023-04-30 12:46:42,795 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.0279, 4.9724, 4.8355, 3.4676, 4.1554, 4.7634, 4.2061, 2.9172], device='cuda:1'), covar=tensor([0.0383, 0.0014, 0.0019, 0.0269, 0.0068, 0.0063, 0.0058, 0.0310], device='cuda:1'), in_proj_covar=tensor([0.0131, 0.0075, 0.0076, 0.0129, 0.0090, 0.0099, 0.0088, 0.0122], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-04-30 12:46:58,419 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.44 vs. limit=2.0 2023-04-30 12:47:00,060 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=167091.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 12:47:13,896 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 17, batch 4700, loss[loss=0.1693, simple_loss=0.2543, pruned_loss=0.0422, over 16438.00 frames. ], tot_loss[loss=0.1986, simple_loss=0.2848, pruned_loss=0.0562, over 3213815.06 frames. ], batch size: 68, lr: 4.00e-03, grad_scale: 8.0 2023-04-30 12:47:18,955 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=167104.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 12:47:42,084 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=167120.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 12:48:00,930 INFO [optim.py:368] (1/8) 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 12:48:06,767 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 17, batch 4750, loss[loss=0.1616, simple_loss=0.2508, pruned_loss=0.03621, over 16772.00 frames. ], tot_loss[loss=0.1943, simple_loss=0.2805, pruned_loss=0.05401, over 3205608.81 frames. ], batch size: 83, lr: 4.00e-03, grad_scale: 8.0 2023-04-30 12:48:41,438 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=167161.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 12:48:44,942 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([1.8112, 2.8794, 2.8026, 4.8956, 3.6559, 4.1889, 1.6960, 3.0729], device='cuda:1'), covar=tensor([0.1302, 0.0744, 0.1157, 0.0132, 0.0291, 0.0374, 0.1579, 0.0806], device='cuda:1'), in_proj_covar=tensor([0.0162, 0.0167, 0.0189, 0.0175, 0.0203, 0.0211, 0.0192, 0.0189], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-04-30 12:48:48,041 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=167165.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 12:48:48,139 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=167165.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 12:49:16,408 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=167185.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 12:49:18,954 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 17, batch 4800, loss[loss=0.1656, simple_loss=0.2462, pruned_loss=0.04256, over 17031.00 frames. ], tot_loss[loss=0.1905, simple_loss=0.2766, pruned_loss=0.0522, over 3201040.70 frames. ], batch size: 53, lr: 4.00e-03, grad_scale: 8.0 2023-04-30 12:49:58,978 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=167213.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 12:50:27,973 INFO [optim.py:368] (1/8) 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 12:50:31,912 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=167235.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 12:50:58,072 INFO [train.py:904] (1/8) Epoch 17, batch 4850, loss[loss=0.1925, simple_loss=0.2805, pruned_loss=0.05222, over 16594.00 frames. ], tot_loss[loss=0.1891, simple_loss=0.2768, pruned_loss=0.05072, over 3214255.60 frames. ], batch size: 62, lr: 4.00e-03, grad_scale: 8.0 2023-04-30 12:51:42,105 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.3396, 3.4335, 1.7660, 3.7466, 2.4365, 3.6959, 1.9763, 2.7074], device='cuda:1'), covar=tensor([0.0236, 0.0312, 0.1824, 0.0135, 0.0890, 0.0550, 0.1753, 0.0755], device='cuda:1'), in_proj_covar=tensor([0.0161, 0.0169, 0.0187, 0.0148, 0.0169, 0.0208, 0.0195, 0.0173], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0003, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-04-30 12:51:46,661 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=167285.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 12:52:12,188 INFO [train.py:904] (1/8) Epoch 17, batch 4900, loss[loss=0.1938, simple_loss=0.284, pruned_loss=0.05178, over 16430.00 frames. ], tot_loss[loss=0.188, simple_loss=0.2765, pruned_loss=0.04976, over 3196318.58 frames. ], batch size: 146, lr: 3.99e-03, grad_scale: 8.0 2023-04-30 12:52:55,966 INFO [optim.py:368] (1/8) 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 12:53:24,978 INFO [train.py:904] (1/8) Epoch 17, batch 4950, loss[loss=0.1821, simple_loss=0.2835, pruned_loss=0.04035, over 16752.00 frames. ], tot_loss[loss=0.1875, simple_loss=0.2762, pruned_loss=0.04945, over 3205598.00 frames. ], batch size: 89, lr: 3.99e-03, grad_scale: 8.0 2023-04-30 12:54:22,548 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=167391.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 12:54:37,763 INFO [train.py:904] (1/8) Epoch 17, batch 5000, loss[loss=0.1802, simple_loss=0.2719, pruned_loss=0.04423, over 16610.00 frames. ], tot_loss[loss=0.1882, simple_loss=0.2777, pruned_loss=0.04941, over 3200710.30 frames. ], batch size: 62, lr: 3.99e-03, grad_scale: 8.0 2023-04-30 12:55:02,756 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=167420.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 12:55:19,961 INFO [zipformer.py:625] (1/8) 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] (1/8) 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 12:55:22,643 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=3.24 vs. limit=5.0 2023-04-30 12:55:29,660 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=167439.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 12:55:46,692 INFO [train.py:904] (1/8) Epoch 17, batch 5050, loss[loss=0.1814, simple_loss=0.2761, pruned_loss=0.04333, over 16309.00 frames. ], tot_loss[loss=0.188, simple_loss=0.278, pruned_loss=0.04899, over 3208982.88 frames. ], batch size: 165, lr: 3.99e-03, grad_scale: 8.0 2023-04-30 12:55:51,604 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=167456.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 12:55:57,166 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=167460.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 12:56:08,480 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=167468.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 12:56:27,443 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.8350, 4.6738, 4.8469, 5.0782, 5.2815, 4.7132, 5.2677, 5.2645], device='cuda:1'), covar=tensor([0.1723, 0.1309, 0.1621, 0.0685, 0.0465, 0.0849, 0.0461, 0.0531], device='cuda:1'), in_proj_covar=tensor([0.0596, 0.0733, 0.0870, 0.0746, 0.0559, 0.0592, 0.0593, 0.0696], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-30 12:56:43,510 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 17, batch 5100, loss[loss=0.1732, simple_loss=0.2662, pruned_loss=0.04011, over 16379.00 frames. ], tot_loss[loss=0.186, simple_loss=0.2758, pruned_loss=0.04809, over 3217894.58 frames. ], batch size: 146, lr: 3.99e-03, grad_scale: 8.0 2023-04-30 12:57:39,789 INFO [optim.py:368] (1/8) 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 12:58:08,924 INFO [train.py:904] (1/8) Epoch 17, batch 5150, loss[loss=0.199, simple_loss=0.2988, pruned_loss=0.04961, over 16286.00 frames. ], tot_loss[loss=0.1853, simple_loss=0.2761, pruned_loss=0.04728, over 3214928.52 frames. ], batch size: 165, lr: 3.99e-03, grad_scale: 8.0 2023-04-30 12:58:09,469 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.8157, 3.2498, 3.3764, 1.9634, 2.9425, 2.2531, 3.4240, 3.4543], device='cuda:1'), covar=tensor([0.0244, 0.0704, 0.0560, 0.1825, 0.0750, 0.0930, 0.0540, 0.0712], device='cuda:1'), in_proj_covar=tensor([0.0151, 0.0159, 0.0165, 0.0150, 0.0142, 0.0127, 0.0142, 0.0168], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:1') 2023-04-30 12:58:56,856 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=167585.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 12:59:21,204 INFO [train.py:904] (1/8) Epoch 17, batch 5200, loss[loss=0.1849, simple_loss=0.2684, pruned_loss=0.05075, over 16275.00 frames. ], tot_loss[loss=0.1845, simple_loss=0.2751, pruned_loss=0.04698, over 3196029.79 frames. ], batch size: 165, lr: 3.99e-03, grad_scale: 8.0 2023-04-30 12:59:38,320 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.9116, 5.3207, 5.5014, 5.2646, 5.3412, 5.8737, 5.2866, 5.0237], device='cuda:1'), covar=tensor([0.0960, 0.1818, 0.2102, 0.1982, 0.2544, 0.1101, 0.1509, 0.2581], device='cuda:1'), in_proj_covar=tensor([0.0384, 0.0550, 0.0596, 0.0460, 0.0618, 0.0640, 0.0471, 0.0619], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-04-30 13:00:07,260 INFO [optim.py:368] (1/8) 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] (1/8) 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] (1/8) Epoch 17, batch 5250, loss[loss=0.1792, simple_loss=0.2555, pruned_loss=0.05145, over 16861.00 frames. ], tot_loss[loss=0.1831, simple_loss=0.2724, pruned_loss=0.04688, over 3203505.76 frames. ], batch size: 42, lr: 3.99e-03, grad_scale: 16.0 2023-04-30 13:00:55,132 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.8847, 5.1808, 4.9466, 4.9724, 4.7206, 4.6554, 4.6505, 5.2617], device='cuda:1'), covar=tensor([0.1172, 0.0755, 0.0875, 0.0711, 0.0751, 0.0926, 0.1088, 0.0777], device='cuda:1'), in_proj_covar=tensor([0.0619, 0.0759, 0.0621, 0.0560, 0.0479, 0.0486, 0.0629, 0.0588], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:1') 2023-04-30 13:01:14,783 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.39 vs. limit=2.0 2023-04-30 13:01:48,349 INFO [train.py:904] (1/8) Epoch 17, batch 5300, loss[loss=0.158, simple_loss=0.2548, pruned_loss=0.03063, over 16876.00 frames. ], tot_loss[loss=0.1809, simple_loss=0.2693, pruned_loss=0.0462, over 3195642.82 frames. ], batch size: 96, lr: 3.99e-03, grad_scale: 8.0 2023-04-30 13:02:12,766 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=167719.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 13:02:35,321 INFO [optim.py:368] (1/8) 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] (1/8) Epoch 17, batch 5350, loss[loss=0.174, simple_loss=0.2645, pruned_loss=0.04178, over 16834.00 frames. ], tot_loss[loss=0.1788, simple_loss=0.2675, pruned_loss=0.04504, over 3201219.02 frames. ], batch size: 116, lr: 3.99e-03, grad_scale: 8.0 2023-04-30 13:03:08,220 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=167756.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 13:03:13,806 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=167760.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 13:03:23,324 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=167767.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 13:03:41,690 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=167780.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 13:03:53,561 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=167788.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 13:04:14,434 INFO [train.py:904] (1/8) Epoch 17, batch 5400, loss[loss=0.2013, simple_loss=0.294, pruned_loss=0.05433, over 15511.00 frames. ], tot_loss[loss=0.1808, simple_loss=0.27, pruned_loss=0.04579, over 3192434.65 frames. ], batch size: 190, lr: 3.99e-03, grad_scale: 8.0 2023-04-30 13:04:17,677 INFO [zipformer.py:625] (1/8) 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] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=167808.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 13:04:54,080 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=167828.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 13:05:02,680 INFO [optim.py:368] (1/8) 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 13:05:11,159 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.2947, 3.6788, 4.0475, 1.6438, 4.1991, 4.2913, 3.0707, 2.9022], device='cuda:1'), covar=tensor([0.1182, 0.0190, 0.0154, 0.1462, 0.0070, 0.0120, 0.0411, 0.0601], device='cuda:1'), in_proj_covar=tensor([0.0147, 0.0105, 0.0092, 0.0137, 0.0075, 0.0118, 0.0125, 0.0130], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:1') 2023-04-30 13:05:31,682 INFO [train.py:904] (1/8) Epoch 17, batch 5450, loss[loss=0.2152, simple_loss=0.304, pruned_loss=0.06315, over 16632.00 frames. ], tot_loss[loss=0.1843, simple_loss=0.2735, pruned_loss=0.04756, over 3203489.58 frames. ], batch size: 62, lr: 3.99e-03, grad_scale: 4.0 2023-04-30 13:06:12,843 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.2363, 3.4284, 3.5620, 3.5629, 3.5710, 3.4013, 3.4078, 3.4697], device='cuda:1'), covar=tensor([0.0425, 0.0657, 0.0609, 0.0485, 0.0553, 0.0625, 0.0932, 0.0509], device='cuda:1'), in_proj_covar=tensor([0.0375, 0.0409, 0.0401, 0.0375, 0.0446, 0.0420, 0.0519, 0.0334], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:1') 2023-04-30 13:06:48,606 INFO [train.py:904] (1/8) Epoch 17, batch 5500, loss[loss=0.2071, simple_loss=0.3009, pruned_loss=0.05668, over 16833.00 frames. ], tot_loss[loss=0.1926, simple_loss=0.2807, pruned_loss=0.05232, over 3164275.27 frames. ], batch size: 102, lr: 3.99e-03, grad_scale: 4.0 2023-04-30 13:07:39,670 INFO [optim.py:368] (1/8) 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 13:08:05,266 INFO [train.py:904] (1/8) Epoch 17, batch 5550, loss[loss=0.25, simple_loss=0.3281, pruned_loss=0.08591, over 15326.00 frames. ], tot_loss[loss=0.2021, simple_loss=0.2881, pruned_loss=0.05805, over 3122648.06 frames. ], batch size: 191, lr: 3.99e-03, grad_scale: 4.0 2023-04-30 13:09:31,455 INFO [train.py:904] (1/8) Epoch 17, batch 5600, loss[loss=0.2888, simple_loss=0.3368, pruned_loss=0.1204, over 10592.00 frames. ], tot_loss[loss=0.2087, simple_loss=0.293, pruned_loss=0.06224, over 3083283.96 frames. ], batch size: 248, lr: 3.99e-03, grad_scale: 8.0 2023-04-30 13:10:27,345 INFO [optim.py:368] (1/8) 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 13:10:27,900 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=168035.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 13:10:41,009 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=4.55 vs. limit=5.0 2023-04-30 13:10:54,415 INFO [train.py:904] (1/8) Epoch 17, batch 5650, loss[loss=0.2188, simple_loss=0.3084, pruned_loss=0.06455, over 16590.00 frames. ], tot_loss[loss=0.2157, simple_loss=0.2984, pruned_loss=0.06653, over 3040741.37 frames. ], batch size: 75, lr: 3.99e-03, grad_scale: 8.0 2023-04-30 13:11:30,931 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=168075.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 13:11:50,818 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=168088.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 13:12:04,671 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 17, batch 5700, loss[loss=0.2335, simple_loss=0.3206, pruned_loss=0.07318, over 15417.00 frames. ], tot_loss[loss=0.2165, simple_loss=0.299, pruned_loss=0.06699, over 3055760.63 frames. ], batch size: 191, lr: 3.99e-03, grad_scale: 8.0 2023-04-30 13:12:23,319 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=3.14 vs. limit=5.0 2023-04-30 13:12:24,996 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([1.6247, 2.5038, 2.3412, 3.6769, 2.7187, 3.8025, 1.4225, 2.7278], device='cuda:1'), covar=tensor([0.1338, 0.0779, 0.1305, 0.0187, 0.0232, 0.0418, 0.1740, 0.0860], device='cuda:1'), in_proj_covar=tensor([0.0161, 0.0168, 0.0189, 0.0175, 0.0202, 0.0210, 0.0193, 0.0189], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-04-30 13:12:45,754 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=168123.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 13:13:04,293 INFO [optim.py:368] (1/8) 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 13:13:06,078 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=168136.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 13:13:31,683 INFO [train.py:904] (1/8) Epoch 17, batch 5750, loss[loss=0.2204, simple_loss=0.3008, pruned_loss=0.07003, over 16601.00 frames. ], tot_loss[loss=0.2202, simple_loss=0.302, pruned_loss=0.06921, over 3007401.40 frames. ], batch size: 57, lr: 3.98e-03, grad_scale: 8.0 2023-04-30 13:14:50,315 INFO [train.py:904] (1/8) Epoch 17, batch 5800, loss[loss=0.2093, simple_loss=0.2957, pruned_loss=0.06142, over 16561.00 frames. ], tot_loss[loss=0.2189, simple_loss=0.3019, pruned_loss=0.06794, over 3019957.42 frames. ], batch size: 68, lr: 3.98e-03, grad_scale: 8.0 2023-04-30 13:15:16,878 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.45 vs. limit=2.0 2023-04-30 13:15:40,502 INFO [zipformer.py:625] (1/8) 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] (1/8) 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 13:16:09,620 INFO [train.py:904] (1/8) Epoch 17, batch 5850, loss[loss=0.2093, simple_loss=0.293, pruned_loss=0.06287, over 16230.00 frames. ], tot_loss[loss=0.2166, simple_loss=0.3002, pruned_loss=0.06645, over 3052693.25 frames. ], batch size: 165, lr: 3.98e-03, grad_scale: 4.0 2023-04-30 13:16:21,726 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.7725, 3.8160, 3.9110, 3.7659, 3.8666, 4.2538, 3.8985, 3.6468], device='cuda:1'), covar=tensor([0.2240, 0.2176, 0.2303, 0.2343, 0.2560, 0.1625, 0.1583, 0.2593], device='cuda:1'), in_proj_covar=tensor([0.0384, 0.0554, 0.0605, 0.0467, 0.0625, 0.0641, 0.0478, 0.0626], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-04-30 13:17:19,314 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=168294.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 13:17:32,945 INFO [train.py:904] (1/8) Epoch 17, batch 5900, loss[loss=0.1836, simple_loss=0.2807, pruned_loss=0.04328, over 16814.00 frames. ], tot_loss[loss=0.2161, simple_loss=0.3, pruned_loss=0.06609, over 3070486.70 frames. ], batch size: 83, lr: 3.98e-03, grad_scale: 4.0 2023-04-30 13:18:02,863 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.75 vs. limit=2.0 2023-04-30 13:18:28,194 INFO [optim.py:368] (1/8) 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 13:18:52,632 INFO [train.py:904] (1/8) Epoch 17, batch 5950, loss[loss=0.2159, simple_loss=0.3118, pruned_loss=0.06, over 16725.00 frames. ], tot_loss[loss=0.2156, simple_loss=0.3011, pruned_loss=0.06511, over 3070478.52 frames. ], batch size: 134, lr: 3.98e-03, grad_scale: 4.0 2023-04-30 13:19:29,602 INFO [zipformer.py:625] (1/8) 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] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=168391.0, num_to_drop=1, layers_to_drop={0} 2023-04-30 13:20:12,027 INFO [train.py:904] (1/8) Epoch 17, batch 6000, loss[loss=0.1944, simple_loss=0.2782, pruned_loss=0.05533, over 17187.00 frames. ], tot_loss[loss=0.213, simple_loss=0.2987, pruned_loss=0.06359, over 3100029.09 frames. ], batch size: 44, lr: 3.98e-03, grad_scale: 8.0 2023-04-30 13:20:12,027 INFO [train.py:929] (1/8) Computing validation loss 2023-04-30 13:20:21,979 INFO [train.py:938] (1/8) Epoch 17, validation: loss=0.1535, simple_loss=0.2667, pruned_loss=0.0202, over 944034.00 frames. 2023-04-30 13:20:21,980 INFO [train.py:939] (1/8) Maximum memory allocated so far is 17857MB 2023-04-30 13:20:53,933 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=168423.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 13:20:53,967 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=168423.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 13:21:15,435 INFO [optim.py:368] (1/8) 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 13:21:27,349 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.18 vs. limit=2.0 2023-04-30 13:21:39,918 INFO [train.py:904] (1/8) Epoch 17, batch 6050, loss[loss=0.2514, simple_loss=0.3352, pruned_loss=0.08386, over 15401.00 frames. ], tot_loss[loss=0.2119, simple_loss=0.2977, pruned_loss=0.06311, over 3097722.84 frames. ], batch size: 191, lr: 3.98e-03, grad_scale: 4.0 2023-04-30 13:22:09,418 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=168471.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 13:22:59,844 INFO [train.py:904] (1/8) Epoch 17, batch 6100, loss[loss=0.2446, simple_loss=0.313, pruned_loss=0.08808, over 11308.00 frames. ], tot_loss[loss=0.211, simple_loss=0.297, pruned_loss=0.06244, over 3098112.32 frames. ], batch size: 247, lr: 3.98e-03, grad_scale: 4.0 2023-04-30 13:23:55,308 INFO [optim.py:368] (1/8) 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] (1/8) Epoch 17, batch 6150, loss[loss=0.194, simple_loss=0.2831, pruned_loss=0.05241, over 16411.00 frames. ], tot_loss[loss=0.2096, simple_loss=0.2951, pruned_loss=0.06205, over 3102363.76 frames. ], batch size: 35, lr: 3.98e-03, grad_scale: 4.0 2023-04-30 13:24:41,101 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.91 vs. limit=2.0 2023-04-30 13:25:09,233 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.80 vs. limit=2.0 2023-04-30 13:25:17,265 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 17, batch 6200, loss[loss=0.2032, simple_loss=0.2898, pruned_loss=0.05827, over 16656.00 frames. ], tot_loss[loss=0.2087, simple_loss=0.2933, pruned_loss=0.06206, over 3099386.75 frames. ], batch size: 62, lr: 3.98e-03, grad_scale: 4.0 2023-04-30 13:26:17,953 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.79 vs. limit=2.0 2023-04-30 13:26:31,026 INFO [optim.py:368] (1/8) 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 13:26:53,709 INFO [train.py:904] (1/8) Epoch 17, batch 6250, loss[loss=0.1928, simple_loss=0.2848, pruned_loss=0.05038, over 16843.00 frames. ], tot_loss[loss=0.2071, simple_loss=0.2922, pruned_loss=0.06096, over 3115571.55 frames. ], batch size: 102, lr: 3.98e-03, grad_scale: 4.0 2023-04-30 13:26:54,927 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.7891, 3.7912, 3.9382, 3.6964, 3.8704, 4.2692, 3.8984, 3.6562], device='cuda:1'), covar=tensor([0.2233, 0.2345, 0.2446, 0.2551, 0.2669, 0.1723, 0.1737, 0.2632], device='cuda:1'), in_proj_covar=tensor([0.0390, 0.0561, 0.0615, 0.0472, 0.0633, 0.0648, 0.0487, 0.0634], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-04-30 13:27:53,907 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 17, batch 6300, loss[loss=0.2031, simple_loss=0.2875, pruned_loss=0.05936, over 16700.00 frames. ], tot_loss[loss=0.2071, simple_loss=0.2925, pruned_loss=0.06085, over 3096621.50 frames. ], batch size: 134, lr: 3.98e-03, grad_scale: 4.0 2023-04-30 13:29:06,430 INFO [optim.py:368] (1/8) 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] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=168739.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 13:29:29,583 INFO [train.py:904] (1/8) Epoch 17, batch 6350, loss[loss=0.192, simple_loss=0.2843, pruned_loss=0.04983, over 16774.00 frames. ], tot_loss[loss=0.2077, simple_loss=0.2931, pruned_loss=0.06113, over 3122409.52 frames. ], batch size: 89, lr: 3.98e-03, grad_scale: 4.0 2023-04-30 13:30:08,835 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.2609, 4.2983, 4.6360, 4.6088, 4.6314, 4.3546, 4.3223, 4.2276], device='cuda:1'), covar=tensor([0.0329, 0.0554, 0.0364, 0.0429, 0.0458, 0.0388, 0.0899, 0.0506], device='cuda:1'), in_proj_covar=tensor([0.0381, 0.0415, 0.0406, 0.0381, 0.0452, 0.0428, 0.0525, 0.0339], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:1') 2023-04-30 13:30:46,693 INFO [train.py:904] (1/8) Epoch 17, batch 6400, loss[loss=0.2189, simple_loss=0.3039, pruned_loss=0.06691, over 16263.00 frames. ], tot_loss[loss=0.2094, simple_loss=0.2935, pruned_loss=0.06266, over 3103575.41 frames. ], batch size: 165, lr: 3.98e-03, grad_scale: 8.0 2023-04-30 13:31:42,104 INFO [optim.py:368] (1/8) 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] (1/8) Epoch 17, batch 6450, loss[loss=0.1905, simple_loss=0.2787, pruned_loss=0.05118, over 16347.00 frames. ], tot_loss[loss=0.2093, simple_loss=0.2936, pruned_loss=0.06256, over 3077130.35 frames. ], batch size: 146, lr: 3.98e-03, grad_scale: 4.0 2023-04-30 13:32:06,433 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=2.57 vs. limit=5.0 2023-04-30 13:33:02,279 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=168889.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 13:33:20,845 INFO [train.py:904] (1/8) Epoch 17, batch 6500, loss[loss=0.2222, simple_loss=0.2924, pruned_loss=0.07601, over 11824.00 frames. ], tot_loss[loss=0.2074, simple_loss=0.2912, pruned_loss=0.0618, over 3074495.24 frames. ], batch size: 246, lr: 3.98e-03, grad_scale: 4.0 2023-04-30 13:33:34,966 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.3240, 1.6076, 1.9597, 2.2004, 2.3599, 2.5532, 1.6239, 2.4421], device='cuda:1'), covar=tensor([0.0189, 0.0427, 0.0268, 0.0279, 0.0261, 0.0173, 0.0451, 0.0128], device='cuda:1'), in_proj_covar=tensor([0.0176, 0.0186, 0.0172, 0.0175, 0.0185, 0.0143, 0.0188, 0.0137], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-30 13:33:43,160 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.4257, 3.4107, 2.6719, 2.1320, 2.2451, 2.2387, 3.5021, 3.1404], device='cuda:1'), covar=tensor([0.2920, 0.0688, 0.1722, 0.2505, 0.2650, 0.2103, 0.0471, 0.1213], device='cuda:1'), in_proj_covar=tensor([0.0317, 0.0259, 0.0295, 0.0298, 0.0288, 0.0241, 0.0282, 0.0318], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:1') 2023-04-30 13:33:46,517 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=4.76 vs. limit=5.0 2023-04-30 13:34:15,420 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=168937.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 13:34:15,502 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([5.2581, 5.5784, 5.2924, 5.3116, 5.0255, 4.9825, 4.8702, 5.6503], device='cuda:1'), covar=tensor([0.1163, 0.0781, 0.0845, 0.0765, 0.0871, 0.0668, 0.1056, 0.0863], device='cuda:1'), in_proj_covar=tensor([0.0631, 0.0772, 0.0632, 0.0574, 0.0486, 0.0497, 0.0642, 0.0597], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:1') 2023-04-30 13:34:15,636 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.7697, 3.8355, 2.4149, 4.2189, 2.7990, 4.1669, 2.3390, 3.0917], device='cuda:1'), covar=tensor([0.0220, 0.0312, 0.1464, 0.0212, 0.0848, 0.0618, 0.1545, 0.0679], device='cuda:1'), in_proj_covar=tensor([0.0162, 0.0170, 0.0191, 0.0150, 0.0171, 0.0210, 0.0198, 0.0174], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-04-30 13:34:16,252 INFO [optim.py:368] (1/8) 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 13:34:22,160 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.6576, 4.8203, 4.9979, 4.7570, 4.9400, 5.4621, 4.9126, 4.6519], device='cuda:1'), covar=tensor([0.1079, 0.2090, 0.2484, 0.2208, 0.2423, 0.0964, 0.1517, 0.2378], device='cuda:1'), in_proj_covar=tensor([0.0387, 0.0558, 0.0609, 0.0468, 0.0629, 0.0643, 0.0482, 0.0628], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-04-30 13:34:40,482 INFO [train.py:904] (1/8) Epoch 17, batch 6550, loss[loss=0.2097, simple_loss=0.3115, pruned_loss=0.05395, over 16383.00 frames. ], tot_loss[loss=0.21, simple_loss=0.2945, pruned_loss=0.06273, over 3082772.65 frames. ], batch size: 146, lr: 3.98e-03, grad_scale: 4.0 2023-04-30 13:35:03,793 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=168967.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 13:35:06,295 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=168969.0, num_to_drop=1, layers_to_drop={1} 2023-04-30 13:35:09,578 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.29 vs. limit=2.0 2023-04-30 13:35:58,162 INFO [train.py:904] (1/8) Epoch 17, batch 6600, loss[loss=0.1876, simple_loss=0.2787, pruned_loss=0.04819, over 16511.00 frames. ], tot_loss[loss=0.2106, simple_loss=0.296, pruned_loss=0.06257, over 3088309.56 frames. ], batch size: 75, lr: 3.97e-03, grad_scale: 4.0 2023-04-30 13:36:17,560 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.1175, 4.1524, 4.5021, 4.4528, 4.4985, 4.2191, 4.2067, 4.1562], device='cuda:1'), covar=tensor([0.0338, 0.0626, 0.0400, 0.0448, 0.0500, 0.0421, 0.0887, 0.0534], device='cuda:1'), in_proj_covar=tensor([0.0382, 0.0416, 0.0407, 0.0383, 0.0453, 0.0429, 0.0527, 0.0341], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:1') 2023-04-30 13:36:37,606 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=169028.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 13:36:39,973 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=169030.0, num_to_drop=1, layers_to_drop={1} 2023-04-30 13:36:50,875 INFO [optim.py:368] (1/8) 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 13:36:52,634 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.0726, 4.3034, 3.9774, 3.7989, 3.3990, 4.2238, 3.9616, 3.8402], device='cuda:1'), covar=tensor([0.0899, 0.0638, 0.0499, 0.0431, 0.1744, 0.0508, 0.0956, 0.0757], device='cuda:1'), in_proj_covar=tensor([0.0273, 0.0384, 0.0322, 0.0310, 0.0332, 0.0362, 0.0220, 0.0384], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-30 13:37:13,177 INFO [train.py:904] (1/8) Epoch 17, batch 6650, loss[loss=0.176, simple_loss=0.2626, pruned_loss=0.04465, over 16708.00 frames. ], tot_loss[loss=0.2118, simple_loss=0.2966, pruned_loss=0.06346, over 3077193.74 frames. ], batch size: 83, lr: 3.97e-03, grad_scale: 4.0 2023-04-30 13:37:59,840 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=169082.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 13:38:30,528 INFO [train.py:904] (1/8) Epoch 17, batch 6700, loss[loss=0.2346, simple_loss=0.3012, pruned_loss=0.08395, over 11268.00 frames. ], tot_loss[loss=0.2105, simple_loss=0.2948, pruned_loss=0.06313, over 3080274.40 frames. ], batch size: 248, lr: 3.97e-03, grad_scale: 4.0 2023-04-30 13:38:53,621 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=169116.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 13:39:26,519 INFO [optim.py:368] (1/8) 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 13:39:34,893 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 17, batch 6750, loss[loss=0.2388, simple_loss=0.3098, pruned_loss=0.08389, over 12220.00 frames. ], tot_loss[loss=0.2096, simple_loss=0.2934, pruned_loss=0.06292, over 3081747.68 frames. ], batch size: 250, lr: 3.97e-03, grad_scale: 4.0 2023-04-30 13:39:58,662 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=169159.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 13:39:58,698 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.4024, 2.1436, 1.6995, 2.0100, 2.4793, 2.1925, 2.3283, 2.6572], device='cuda:1'), covar=tensor([0.0183, 0.0391, 0.0487, 0.0407, 0.0223, 0.0329, 0.0189, 0.0241], device='cuda:1'), in_proj_covar=tensor([0.0185, 0.0224, 0.0217, 0.0217, 0.0225, 0.0223, 0.0226, 0.0219], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-30 13:40:14,511 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.3925, 4.5337, 4.6619, 4.4660, 4.5844, 5.0318, 4.5876, 4.3613], device='cuda:1'), covar=tensor([0.1491, 0.1880, 0.1853, 0.2021, 0.2267, 0.1067, 0.1509, 0.2353], device='cuda:1'), in_proj_covar=tensor([0.0388, 0.0559, 0.0611, 0.0469, 0.0631, 0.0646, 0.0488, 0.0631], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-04-30 13:40:26,173 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=169177.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 13:40:38,897 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([5.0131, 5.5551, 5.7921, 5.4668, 5.5837, 6.1029, 5.5796, 5.3914], device='cuda:1'), covar=tensor([0.0913, 0.1685, 0.1769, 0.1759, 0.2115, 0.0862, 0.1410, 0.2138], device='cuda:1'), in_proj_covar=tensor([0.0388, 0.0559, 0.0611, 0.0468, 0.0632, 0.0646, 0.0487, 0.0631], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-04-30 13:40:59,492 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.4955, 4.4693, 4.3525, 3.6272, 4.4011, 1.6361, 4.1785, 4.0341], device='cuda:1'), covar=tensor([0.0104, 0.0083, 0.0172, 0.0349, 0.0094, 0.2745, 0.0120, 0.0213], device='cuda:1'), in_proj_covar=tensor([0.0149, 0.0139, 0.0186, 0.0170, 0.0159, 0.0196, 0.0172, 0.0164], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-30 13:41:04,380 INFO [train.py:904] (1/8) Epoch 17, batch 6800, loss[loss=0.2139, simple_loss=0.3077, pruned_loss=0.06006, over 16843.00 frames. ], tot_loss[loss=0.2098, simple_loss=0.2937, pruned_loss=0.06299, over 3086638.78 frames. ], batch size: 102, lr: 3.97e-03, grad_scale: 8.0 2023-04-30 13:41:33,103 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=169220.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 13:41:44,497 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.18 vs. limit=2.0 2023-04-30 13:41:58,876 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.0009, 2.7857, 2.7387, 2.0978, 2.6112, 2.1134, 2.6944, 3.0111], device='cuda:1'), covar=tensor([0.0352, 0.0838, 0.0603, 0.1863, 0.0863, 0.1011, 0.0751, 0.0738], device='cuda:1'), in_proj_covar=tensor([0.0150, 0.0159, 0.0165, 0.0151, 0.0143, 0.0127, 0.0142, 0.0168], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:1') 2023-04-30 13:42:02,685 INFO [optim.py:368] (1/8) 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 13:42:08,569 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.5976, 3.7956, 2.8644, 2.2095, 2.5604, 2.4155, 4.0287, 3.4725], device='cuda:1'), covar=tensor([0.2860, 0.0659, 0.1734, 0.2836, 0.2626, 0.1933, 0.0490, 0.1176], device='cuda:1'), in_proj_covar=tensor([0.0318, 0.0261, 0.0297, 0.0299, 0.0290, 0.0243, 0.0284, 0.0321], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:1') 2023-04-30 13:42:23,105 INFO [train.py:904] (1/8) Epoch 17, batch 6850, loss[loss=0.2005, simple_loss=0.2918, pruned_loss=0.05458, over 15517.00 frames. ], tot_loss[loss=0.2102, simple_loss=0.2945, pruned_loss=0.06299, over 3094111.82 frames. ], batch size: 191, lr: 3.97e-03, grad_scale: 8.0 2023-04-30 13:43:37,894 INFO [train.py:904] (1/8) Epoch 17, batch 6900, loss[loss=0.2102, simple_loss=0.2987, pruned_loss=0.06084, over 16217.00 frames. ], tot_loss[loss=0.2099, simple_loss=0.2962, pruned_loss=0.06185, over 3110318.27 frames. ], batch size: 165, lr: 3.97e-03, grad_scale: 8.0 2023-04-30 13:44:02,211 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.27 vs. limit=2.0 2023-04-30 13:44:03,448 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.30 vs. limit=2.0 2023-04-30 13:44:05,830 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.84 vs. limit=2.0 2023-04-30 13:44:10,300 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=169323.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 13:44:13,174 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.6446, 4.5053, 4.6842, 4.8749, 5.0531, 4.5520, 5.0558, 5.0274], device='cuda:1'), covar=tensor([0.1794, 0.1276, 0.1617, 0.0717, 0.0579, 0.0975, 0.0576, 0.0721], device='cuda:1'), in_proj_covar=tensor([0.0586, 0.0724, 0.0861, 0.0740, 0.0558, 0.0587, 0.0594, 0.0690], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-30 13:44:14,314 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=169325.0, num_to_drop=1, layers_to_drop={2} 2023-04-30 13:44:21,002 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.3535, 2.9836, 2.8974, 1.8281, 2.6651, 2.0982, 2.9354, 3.1948], device='cuda:1'), covar=tensor([0.0387, 0.0743, 0.0694, 0.2109, 0.0943, 0.1038, 0.0756, 0.0766], device='cuda:1'), in_proj_covar=tensor([0.0150, 0.0158, 0.0164, 0.0150, 0.0142, 0.0127, 0.0142, 0.0168], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:1') 2023-04-30 13:44:33,264 INFO [optim.py:368] (1/8) 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 13:44:55,629 INFO [train.py:904] (1/8) Epoch 17, batch 6950, loss[loss=0.2091, simple_loss=0.2919, pruned_loss=0.06311, over 16613.00 frames. ], tot_loss[loss=0.2123, simple_loss=0.2978, pruned_loss=0.06335, over 3105282.34 frames. ], batch size: 68, lr: 3.97e-03, grad_scale: 8.0 2023-04-30 13:45:58,963 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.7629, 5.0331, 4.7990, 4.7888, 4.5435, 4.5776, 4.4672, 5.1129], device='cuda:1'), covar=tensor([0.1105, 0.0838, 0.0934, 0.0815, 0.0879, 0.0972, 0.1152, 0.0844], device='cuda:1'), in_proj_covar=tensor([0.0629, 0.0766, 0.0625, 0.0569, 0.0481, 0.0493, 0.0634, 0.0593], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:1') 2023-04-30 13:46:11,848 INFO [train.py:904] (1/8) Epoch 17, batch 7000, loss[loss=0.2077, simple_loss=0.2965, pruned_loss=0.05941, over 15292.00 frames. ], tot_loss[loss=0.2115, simple_loss=0.2977, pruned_loss=0.06265, over 3107090.05 frames. ], batch size: 190, lr: 3.97e-03, grad_scale: 8.0 2023-04-30 13:46:41,112 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=4.80 vs. limit=5.0 2023-04-30 13:47:07,100 INFO [optim.py:368] (1/8) 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 13:47:07,439 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 17, batch 7050, loss[loss=0.2218, simple_loss=0.2963, pruned_loss=0.07361, over 11728.00 frames. ], tot_loss[loss=0.2111, simple_loss=0.298, pruned_loss=0.06205, over 3105538.10 frames. ], batch size: 246, lr: 3.97e-03, grad_scale: 8.0 2023-04-30 13:47:47,083 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.83 vs. limit=2.0 2023-04-30 13:48:00,524 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=169472.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 13:48:47,182 INFO [train.py:904] (1/8) Epoch 17, batch 7100, loss[loss=0.2077, simple_loss=0.2909, pruned_loss=0.06225, over 16441.00 frames. ], tot_loss[loss=0.2098, simple_loss=0.2963, pruned_loss=0.06162, over 3110423.86 frames. ], batch size: 146, lr: 3.97e-03, grad_scale: 8.0 2023-04-30 13:49:08,054 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=169515.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 13:49:22,029 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.30 vs. limit=2.0 2023-04-30 13:49:27,143 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.7293, 3.2575, 3.1342, 1.9163, 2.8101, 2.2653, 3.3051, 3.4975], device='cuda:1'), covar=tensor([0.0326, 0.0761, 0.0736, 0.2180, 0.0922, 0.0966, 0.0742, 0.0864], device='cuda:1'), in_proj_covar=tensor([0.0151, 0.0158, 0.0164, 0.0150, 0.0142, 0.0127, 0.0142, 0.0168], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:1') 2023-04-30 13:49:42,141 INFO [optim.py:368] (1/8) 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] (1/8) Epoch 17, batch 7150, loss[loss=0.2602, simple_loss=0.3099, pruned_loss=0.1053, over 11408.00 frames. ], tot_loss[loss=0.2088, simple_loss=0.2944, pruned_loss=0.06162, over 3110752.96 frames. ], batch size: 246, lr: 3.97e-03, grad_scale: 8.0 2023-04-30 13:50:35,285 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.4370, 3.3036, 2.6682, 2.1032, 2.2973, 2.1987, 3.6184, 3.1126], device='cuda:1'), covar=tensor([0.3093, 0.1026, 0.1937, 0.2730, 0.2706, 0.2216, 0.0551, 0.1368], device='cuda:1'), in_proj_covar=tensor([0.0321, 0.0263, 0.0299, 0.0302, 0.0291, 0.0244, 0.0286, 0.0323], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:1') 2023-04-30 13:50:48,840 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.17 vs. limit=2.0 2023-04-30 13:51:19,558 INFO [train.py:904] (1/8) Epoch 17, batch 7200, loss[loss=0.1726, simple_loss=0.2665, pruned_loss=0.03937, over 16781.00 frames. ], tot_loss[loss=0.2067, simple_loss=0.2927, pruned_loss=0.06035, over 3100664.97 frames. ], batch size: 89, lr: 3.97e-03, grad_scale: 8.0 2023-04-30 13:51:48,102 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=169621.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 13:51:51,462 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=169623.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 13:51:55,188 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=169625.0, num_to_drop=1, layers_to_drop={1} 2023-04-30 13:52:16,449 INFO [optim.py:368] (1/8) 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] (1/8) Epoch 17, batch 7250, loss[loss=0.1981, simple_loss=0.2775, pruned_loss=0.05936, over 16476.00 frames. ], tot_loss[loss=0.2045, simple_loss=0.2905, pruned_loss=0.05925, over 3099781.38 frames. ], batch size: 75, lr: 3.97e-03, grad_scale: 8.0 2023-04-30 13:53:08,593 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=169671.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 13:53:12,519 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=169673.0, num_to_drop=1, layers_to_drop={1} 2023-04-30 13:53:14,878 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.8123, 4.0976, 4.3488, 1.8058, 4.5852, 4.7667, 3.3222, 3.2670], device='cuda:1'), covar=tensor([0.1009, 0.0144, 0.0168, 0.1440, 0.0066, 0.0101, 0.0384, 0.0525], device='cuda:1'), in_proj_covar=tensor([0.0148, 0.0105, 0.0093, 0.0138, 0.0075, 0.0119, 0.0126, 0.0130], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0004, 0.0004], device='cuda:1') 2023-04-30 13:53:25,526 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=169682.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 13:53:36,619 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.6193, 4.2070, 4.1532, 2.8508, 3.7588, 4.1851, 3.7600, 2.4794], device='cuda:1'), covar=tensor([0.0429, 0.0032, 0.0041, 0.0341, 0.0076, 0.0110, 0.0075, 0.0376], device='cuda:1'), in_proj_covar=tensor([0.0133, 0.0076, 0.0077, 0.0131, 0.0090, 0.0102, 0.0089, 0.0124], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-04-30 13:53:57,113 INFO [train.py:904] (1/8) Epoch 17, batch 7300, loss[loss=0.1958, simple_loss=0.2874, pruned_loss=0.05212, over 15346.00 frames. ], tot_loss[loss=0.2046, simple_loss=0.29, pruned_loss=0.05956, over 3090188.24 frames. ], batch size: 190, lr: 3.97e-03, grad_scale: 8.0 2023-04-30 13:54:41,438 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=169729.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 13:54:55,062 INFO [optim.py:368] (1/8) 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 13:54:55,576 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=169738.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 13:55:10,618 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=169747.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 13:55:16,949 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.6167, 4.4966, 4.6587, 4.8220, 4.9664, 4.5433, 4.9481, 4.9903], device='cuda:1'), covar=tensor([0.1709, 0.1183, 0.1529, 0.0643, 0.0593, 0.0877, 0.0571, 0.0569], device='cuda:1'), in_proj_covar=tensor([0.0577, 0.0714, 0.0850, 0.0729, 0.0552, 0.0580, 0.0590, 0.0681], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-30 13:55:17,627 INFO [train.py:904] (1/8) Epoch 17, batch 7350, loss[loss=0.2148, simple_loss=0.3001, pruned_loss=0.06478, over 16571.00 frames. ], tot_loss[loss=0.2062, simple_loss=0.291, pruned_loss=0.06071, over 3074610.04 frames. ], batch size: 75, lr: 3.97e-03, grad_scale: 8.0 2023-04-30 13:55:49,798 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=169772.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 13:56:11,069 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.7512, 3.7495, 2.5951, 2.3842, 2.6084, 2.3037, 3.9477, 3.3234], device='cuda:1'), covar=tensor([0.2820, 0.0901, 0.2167, 0.2389, 0.2585, 0.2154, 0.0603, 0.1298], device='cuda:1'), in_proj_covar=tensor([0.0322, 0.0263, 0.0300, 0.0302, 0.0292, 0.0245, 0.0287, 0.0324], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:1') 2023-04-30 13:56:12,137 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=169786.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 13:56:12,357 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=169786.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 13:56:19,849 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=169790.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 13:56:37,956 INFO [train.py:904] (1/8) Epoch 17, batch 7400, loss[loss=0.202, simple_loss=0.2896, pruned_loss=0.05715, over 16704.00 frames. ], tot_loss[loss=0.2064, simple_loss=0.2918, pruned_loss=0.06054, over 3097811.30 frames. ], batch size: 124, lr: 3.97e-03, grad_scale: 8.0 2023-04-30 13:56:48,611 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=169808.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 13:56:59,820 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=169815.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 13:57:07,983 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=169820.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 13:57:36,683 INFO [optim.py:368] (1/8) 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 13:57:51,854 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=169847.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 13:57:59,465 INFO [train.py:904] (1/8) Epoch 17, batch 7450, loss[loss=0.2441, simple_loss=0.3062, pruned_loss=0.09102, over 11369.00 frames. ], tot_loss[loss=0.2084, simple_loss=0.2929, pruned_loss=0.06198, over 3071546.16 frames. ], batch size: 248, lr: 3.96e-03, grad_scale: 8.0 2023-04-30 13:58:18,999 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=169863.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 13:58:25,464 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=2.71 vs. limit=5.0 2023-04-30 13:59:20,319 INFO [train.py:904] (1/8) Epoch 17, batch 7500, loss[loss=0.204, simple_loss=0.2915, pruned_loss=0.05825, over 15203.00 frames. ], tot_loss[loss=0.2097, simple_loss=0.2943, pruned_loss=0.06253, over 3041587.44 frames. ], batch size: 190, lr: 3.96e-03, grad_scale: 8.0 2023-04-30 14:00:17,626 INFO [optim.py:368] (1/8) 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 14:00:23,191 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.18 vs. limit=2.0 2023-04-30 14:00:39,249 INFO [train.py:904] (1/8) Epoch 17, batch 7550, loss[loss=0.206, simple_loss=0.2983, pruned_loss=0.0569, over 16470.00 frames. ], tot_loss[loss=0.2108, simple_loss=0.2945, pruned_loss=0.06349, over 3029210.81 frames. ], batch size: 68, lr: 3.96e-03, grad_scale: 8.0 2023-04-30 14:01:18,912 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=169977.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 14:02:00,831 INFO [train.py:904] (1/8) Epoch 17, batch 7600, loss[loss=0.266, simple_loss=0.3231, pruned_loss=0.1044, over 11683.00 frames. ], tot_loss[loss=0.2101, simple_loss=0.2936, pruned_loss=0.06328, over 3045014.46 frames. ], batch size: 248, lr: 3.96e-03, grad_scale: 8.0 2023-04-30 14:02:40,033 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.5867, 4.6240, 4.4521, 4.1264, 4.0709, 4.5194, 4.3147, 4.2073], device='cuda:1'), covar=tensor([0.0555, 0.0519, 0.0293, 0.0299, 0.0930, 0.0480, 0.0514, 0.0687], device='cuda:1'), in_proj_covar=tensor([0.0266, 0.0377, 0.0316, 0.0302, 0.0324, 0.0353, 0.0216, 0.0376], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-30 14:02:58,192 INFO [optim.py:368] (1/8) 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 14:03:07,427 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 17, batch 7650, loss[loss=0.2066, simple_loss=0.2924, pruned_loss=0.06038, over 15308.00 frames. ], tot_loss[loss=0.2108, simple_loss=0.294, pruned_loss=0.06377, over 3050383.64 frames. ], batch size: 191, lr: 3.96e-03, grad_scale: 4.0 2023-04-30 14:04:07,074 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=170085.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 14:04:24,199 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.1869, 4.1722, 4.1001, 3.3782, 4.1624, 1.6972, 3.9143, 3.7640], device='cuda:1'), covar=tensor([0.0115, 0.0093, 0.0173, 0.0307, 0.0088, 0.2712, 0.0125, 0.0223], device='cuda:1'), in_proj_covar=tensor([0.0147, 0.0136, 0.0183, 0.0167, 0.0155, 0.0193, 0.0169, 0.0160], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-30 14:04:33,135 INFO [train.py:904] (1/8) Epoch 17, batch 7700, loss[loss=0.248, simple_loss=0.3226, pruned_loss=0.08674, over 11548.00 frames. ], tot_loss[loss=0.2112, simple_loss=0.2944, pruned_loss=0.064, over 3061831.72 frames. ], batch size: 250, lr: 3.96e-03, grad_scale: 4.0 2023-04-30 14:04:34,726 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=170103.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 14:04:39,599 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=170106.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 14:05:29,557 INFO [optim.py:368] (1/8) 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 14:05:34,959 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=170142.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 14:05:50,027 INFO [train.py:904] (1/8) Epoch 17, batch 7750, loss[loss=0.1943, simple_loss=0.2883, pruned_loss=0.05016, over 16799.00 frames. ], tot_loss[loss=0.2111, simple_loss=0.2949, pruned_loss=0.06368, over 3082443.63 frames. ], batch size: 102, lr: 3.96e-03, grad_scale: 4.0 2023-04-30 14:06:26,690 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.43 vs. limit=2.0 2023-04-30 14:06:29,845 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.4366, 2.4206, 2.3266, 4.3442, 2.2155, 2.7836, 2.4049, 2.5078], device='cuda:1'), covar=tensor([0.1188, 0.3440, 0.2754, 0.0410, 0.3991, 0.2445, 0.3530, 0.3145], device='cuda:1'), in_proj_covar=tensor([0.0381, 0.0424, 0.0351, 0.0319, 0.0429, 0.0491, 0.0395, 0.0495], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-30 14:07:07,219 INFO [train.py:904] (1/8) Epoch 17, batch 7800, loss[loss=0.1924, simple_loss=0.2793, pruned_loss=0.05275, over 16268.00 frames. ], tot_loss[loss=0.2117, simple_loss=0.2953, pruned_loss=0.06401, over 3078062.05 frames. ], batch size: 165, lr: 3.96e-03, grad_scale: 4.0 2023-04-30 14:07:22,871 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.8139, 4.7042, 4.8999, 5.0871, 5.2420, 4.7749, 5.2052, 5.2316], device='cuda:1'), covar=tensor([0.2070, 0.1221, 0.1753, 0.0757, 0.0686, 0.0859, 0.0725, 0.0613], device='cuda:1'), in_proj_covar=tensor([0.0584, 0.0721, 0.0857, 0.0733, 0.0557, 0.0584, 0.0595, 0.0689], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-30 14:07:25,030 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.45 vs. limit=2.0 2023-04-30 14:08:03,351 INFO [optim.py:368] (1/8) 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 14:08:23,346 INFO [train.py:904] (1/8) Epoch 17, batch 7850, loss[loss=0.1802, simple_loss=0.271, pruned_loss=0.04469, over 16462.00 frames. ], tot_loss[loss=0.2115, simple_loss=0.2957, pruned_loss=0.06362, over 3078168.12 frames. ], batch size: 68, lr: 3.96e-03, grad_scale: 4.0 2023-04-30 14:09:01,010 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=170277.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 14:09:38,277 INFO [train.py:904] (1/8) Epoch 17, batch 7900, loss[loss=0.181, simple_loss=0.2749, pruned_loss=0.04358, over 17098.00 frames. ], tot_loss[loss=0.2107, simple_loss=0.2949, pruned_loss=0.06328, over 3057453.26 frames. ], batch size: 49, lr: 3.96e-03, grad_scale: 4.0 2023-04-30 14:10:12,457 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=170325.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 14:10:20,345 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.3407, 2.9373, 2.9519, 1.8980, 2.6906, 2.0957, 3.0099, 3.1375], device='cuda:1'), covar=tensor([0.0282, 0.0738, 0.0631, 0.2007, 0.0866, 0.1031, 0.0659, 0.0826], device='cuda:1'), in_proj_covar=tensor([0.0150, 0.0157, 0.0163, 0.0149, 0.0142, 0.0126, 0.0141, 0.0167], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:1') 2023-04-30 14:10:26,381 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.51 vs. limit=2.0 2023-04-30 14:10:30,098 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.1797, 3.9138, 3.9230, 2.4879, 3.4992, 3.9701, 3.6377, 2.1504], device='cuda:1'), covar=tensor([0.0593, 0.0046, 0.0046, 0.0428, 0.0099, 0.0096, 0.0083, 0.0480], device='cuda:1'), in_proj_covar=tensor([0.0135, 0.0077, 0.0078, 0.0132, 0.0091, 0.0103, 0.0090, 0.0125], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-04-30 14:10:36,115 INFO [optim.py:368] (1/8) 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] (1/8) Epoch 17, batch 7950, loss[loss=0.2003, simple_loss=0.2788, pruned_loss=0.06087, over 16174.00 frames. ], tot_loss[loss=0.2111, simple_loss=0.2952, pruned_loss=0.06355, over 3063531.18 frames. ], batch size: 35, lr: 3.96e-03, grad_scale: 4.0 2023-04-30 14:11:49,417 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=170385.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 14:11:53,820 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.5905, 2.6087, 1.8987, 2.6643, 2.1459, 2.7685, 2.1380, 2.4094], device='cuda:1'), covar=tensor([0.0249, 0.0322, 0.1088, 0.0225, 0.0622, 0.0541, 0.1047, 0.0507], device='cuda:1'), in_proj_covar=tensor([0.0162, 0.0170, 0.0192, 0.0150, 0.0172, 0.0211, 0.0200, 0.0175], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-04-30 14:12:13,325 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=170401.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 14:12:14,150 INFO [train.py:904] (1/8) Epoch 17, batch 8000, loss[loss=0.2255, simple_loss=0.3166, pruned_loss=0.06716, over 16761.00 frames. ], tot_loss[loss=0.2122, simple_loss=0.2959, pruned_loss=0.06419, over 3064021.83 frames. ], batch size: 124, lr: 3.96e-03, grad_scale: 8.0 2023-04-30 14:12:15,878 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=170403.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 14:13:00,978 INFO [zipformer.py:625] (1/8) 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] (1/8) 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 14:13:15,187 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=170442.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 14:13:28,429 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 17, batch 8050, loss[loss=0.2047, simple_loss=0.2916, pruned_loss=0.05895, over 16920.00 frames. ], tot_loss[loss=0.2113, simple_loss=0.2955, pruned_loss=0.06358, over 3061462.95 frames. ], batch size: 109, lr: 3.96e-03, grad_scale: 4.0 2023-04-30 14:13:43,690 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=170461.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 14:14:28,130 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=170490.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 14:14:28,401 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.5447, 3.6421, 2.7695, 2.1477, 2.3290, 2.3451, 3.7959, 3.2061], device='cuda:1'), covar=tensor([0.2975, 0.0700, 0.1875, 0.2820, 0.2804, 0.2102, 0.0511, 0.1359], device='cuda:1'), in_proj_covar=tensor([0.0323, 0.0261, 0.0300, 0.0301, 0.0291, 0.0244, 0.0287, 0.0322], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:1') 2023-04-30 14:14:37,488 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.5819, 2.6120, 1.8314, 2.6564, 2.1424, 2.7645, 2.1048, 2.3867], device='cuda:1'), covar=tensor([0.0265, 0.0342, 0.1179, 0.0213, 0.0616, 0.0433, 0.1064, 0.0511], device='cuda:1'), in_proj_covar=tensor([0.0162, 0.0169, 0.0191, 0.0149, 0.0171, 0.0209, 0.0198, 0.0174], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0003, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-04-30 14:14:46,136 INFO [train.py:904] (1/8) Epoch 17, batch 8100, loss[loss=0.2069, simple_loss=0.2887, pruned_loss=0.06251, over 16859.00 frames. ], tot_loss[loss=0.2104, simple_loss=0.2947, pruned_loss=0.06301, over 3065759.16 frames. ], batch size: 116, lr: 3.96e-03, grad_scale: 4.0 2023-04-30 14:14:50,617 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.8304, 4.9423, 5.3112, 5.2728, 5.2909, 4.9200, 4.9557, 4.6470], device='cuda:1'), covar=tensor([0.0293, 0.0483, 0.0344, 0.0387, 0.0424, 0.0361, 0.0847, 0.0476], device='cuda:1'), in_proj_covar=tensor([0.0376, 0.0408, 0.0402, 0.0379, 0.0452, 0.0423, 0.0519, 0.0334], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:1') 2023-04-30 14:14:52,541 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=4.82 vs. limit=5.0 2023-04-30 14:15:16,579 INFO [zipformer.py:625] (1/8) 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] (1/8) 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 14:15:58,982 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.5680, 4.7043, 4.8888, 4.7028, 4.7585, 5.2826, 4.7866, 4.5627], device='cuda:1'), covar=tensor([0.1205, 0.1868, 0.2232, 0.1854, 0.2359, 0.0938, 0.1536, 0.2252], device='cuda:1'), in_proj_covar=tensor([0.0384, 0.0556, 0.0611, 0.0466, 0.0626, 0.0647, 0.0483, 0.0627], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-04-30 14:16:00,979 INFO [train.py:904] (1/8) Epoch 17, batch 8150, loss[loss=0.1957, simple_loss=0.2783, pruned_loss=0.05654, over 17112.00 frames. ], tot_loss[loss=0.2075, simple_loss=0.2918, pruned_loss=0.06165, over 3088179.23 frames. ], batch size: 47, lr: 3.96e-03, grad_scale: 4.0 2023-04-30 14:16:27,569 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=170569.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 14:17:10,877 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.1166, 4.2752, 3.2966, 2.4513, 2.9619, 2.7132, 4.6719, 3.7538], device='cuda:1'), covar=tensor([0.2434, 0.0612, 0.1631, 0.2555, 0.2461, 0.1777, 0.0416, 0.1178], device='cuda:1'), in_proj_covar=tensor([0.0324, 0.0262, 0.0300, 0.0302, 0.0292, 0.0244, 0.0286, 0.0323], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:1') 2023-04-30 14:17:18,497 INFO [train.py:904] (1/8) Epoch 17, batch 8200, loss[loss=0.221, simple_loss=0.3157, pruned_loss=0.06318, over 16618.00 frames. ], tot_loss[loss=0.2055, simple_loss=0.2893, pruned_loss=0.06086, over 3097519.39 frames. ], batch size: 134, lr: 3.96e-03, grad_scale: 4.0 2023-04-30 14:18:05,902 INFO [zipformer.py:625] (1/8) 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] (1/8) 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 14:18:41,070 INFO [train.py:904] (1/8) Epoch 17, batch 8250, loss[loss=0.1686, simple_loss=0.269, pruned_loss=0.03415, over 16839.00 frames. ], tot_loss[loss=0.2026, simple_loss=0.2882, pruned_loss=0.0585, over 3078800.27 frames. ], batch size: 102, lr: 3.96e-03, grad_scale: 4.0 2023-04-30 14:20:04,019 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=170701.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 14:20:05,297 INFO [train.py:904] (1/8) Epoch 17, batch 8300, loss[loss=0.1714, simple_loss=0.2568, pruned_loss=0.04297, over 12330.00 frames. ], tot_loss[loss=0.199, simple_loss=0.2862, pruned_loss=0.05588, over 3084166.22 frames. ], batch size: 248, lr: 3.95e-03, grad_scale: 4.0 2023-04-30 14:21:09,571 INFO [optim.py:368] (1/8) 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 14:21:25,416 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=170749.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 14:21:29,853 INFO [train.py:904] (1/8) Epoch 17, batch 8350, loss[loss=0.1799, simple_loss=0.2761, pruned_loss=0.04188, over 15283.00 frames. ], tot_loss[loss=0.1964, simple_loss=0.285, pruned_loss=0.05391, over 3078715.52 frames. ], batch size: 190, lr: 3.95e-03, grad_scale: 4.0 2023-04-30 14:22:28,225 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=4.35 vs. limit=5.0 2023-04-30 14:22:51,857 INFO [train.py:904] (1/8) Epoch 17, batch 8400, loss[loss=0.1835, simple_loss=0.2775, pruned_loss=0.04478, over 16627.00 frames. ], tot_loss[loss=0.1926, simple_loss=0.2819, pruned_loss=0.05171, over 3064846.03 frames. ], batch size: 62, lr: 3.95e-03, grad_scale: 8.0 2023-04-30 14:23:16,981 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=170817.0, num_to_drop=1, layers_to_drop={2} 2023-04-30 14:23:52,183 INFO [optim.py:368] (1/8) 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] (1/8) Epoch 17, batch 8450, loss[loss=0.1724, simple_loss=0.2693, pruned_loss=0.03775, over 16274.00 frames. ], tot_loss[loss=0.1897, simple_loss=0.28, pruned_loss=0.04976, over 3082560.92 frames. ], batch size: 165, lr: 3.95e-03, grad_scale: 8.0 2023-04-30 14:25:32,473 INFO [train.py:904] (1/8) Epoch 17, batch 8500, loss[loss=0.1532, simple_loss=0.2386, pruned_loss=0.03394, over 12081.00 frames. ], tot_loss[loss=0.186, simple_loss=0.2768, pruned_loss=0.0476, over 3070651.27 frames. ], batch size: 248, lr: 3.95e-03, grad_scale: 8.0 2023-04-30 14:26:10,271 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=170925.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 14:26:14,672 INFO [zipformer.py:625] (1/8) 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] (1/8) 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] (1/8) Epoch 17, batch 8550, loss[loss=0.192, simple_loss=0.2873, pruned_loss=0.0483, over 16889.00 frames. ], tot_loss[loss=0.1835, simple_loss=0.2743, pruned_loss=0.04635, over 3041750.50 frames. ], batch size: 109, lr: 3.95e-03, grad_scale: 8.0 2023-04-30 14:28:08,420 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=170989.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 14:28:33,701 INFO [train.py:904] (1/8) Epoch 17, batch 8600, loss[loss=0.1905, simple_loss=0.2845, pruned_loss=0.04828, over 16262.00 frames. ], tot_loss[loss=0.1823, simple_loss=0.274, pruned_loss=0.04526, over 3026373.76 frames. ], batch size: 165, lr: 3.95e-03, grad_scale: 8.0 2023-04-30 14:29:52,152 INFO [optim.py:368] (1/8) 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] (1/8) Epoch 17, batch 8650, loss[loss=0.1617, simple_loss=0.2598, pruned_loss=0.03183, over 16675.00 frames. ], tot_loss[loss=0.1798, simple_loss=0.272, pruned_loss=0.0438, over 3031414.35 frames. ], batch size: 134, lr: 3.95e-03, grad_scale: 8.0 2023-04-30 14:30:26,299 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.3654, 3.2840, 3.2506, 3.4922, 3.4989, 3.3009, 3.4679, 3.5307], device='cuda:1'), covar=tensor([0.1334, 0.1146, 0.1559, 0.0855, 0.0936, 0.2818, 0.1256, 0.1150], device='cuda:1'), in_proj_covar=tensor([0.0564, 0.0696, 0.0824, 0.0711, 0.0539, 0.0562, 0.0574, 0.0670], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-30 14:31:44,026 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 17, batch 8700, loss[loss=0.1705, simple_loss=0.2633, pruned_loss=0.03886, over 16828.00 frames. ], tot_loss[loss=0.177, simple_loss=0.2691, pruned_loss=0.04249, over 3039763.54 frames. ], batch size: 124, lr: 3.95e-03, grad_scale: 8.0 2023-04-30 14:32:31,769 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=171117.0, num_to_drop=1, layers_to_drop={0} 2023-04-30 14:33:13,274 INFO [optim.py:368] (1/8) 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] (1/8) Epoch 17, batch 8750, loss[loss=0.1969, simple_loss=0.2917, pruned_loss=0.05106, over 16789.00 frames. ], tot_loss[loss=0.1769, simple_loss=0.2694, pruned_loss=0.04217, over 3050580.26 frames. ], batch size: 39, lr: 3.95e-03, grad_scale: 4.0 2023-04-30 14:33:43,316 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=171153.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 14:34:11,474 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.22 vs. limit=2.0 2023-04-30 14:34:13,577 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=171165.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 14:35:21,385 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 17, batch 8800, loss[loss=0.1554, simple_loss=0.2494, pruned_loss=0.03064, over 16633.00 frames. ], tot_loss[loss=0.1745, simple_loss=0.2675, pruned_loss=0.04078, over 3062315.28 frames. ], batch size: 62, lr: 3.95e-03, grad_scale: 8.0 2023-04-30 14:36:22,012 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=171225.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 14:36:54,977 INFO [optim.py:368] (1/8) 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 14:37:18,413 INFO [train.py:904] (1/8) Epoch 17, batch 8850, loss[loss=0.158, simple_loss=0.2574, pruned_loss=0.02927, over 17178.00 frames. ], tot_loss[loss=0.1751, simple_loss=0.2696, pruned_loss=0.04024, over 3050176.30 frames. ], batch size: 46, lr: 3.95e-03, grad_scale: 8.0 2023-04-30 14:37:28,389 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=171257.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 14:37:41,891 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.3928, 1.9214, 1.4115, 1.5412, 2.1419, 1.8120, 1.9729, 2.2865], device='cuda:1'), covar=tensor([0.0203, 0.0419, 0.0659, 0.0549, 0.0302, 0.0430, 0.0210, 0.0320], device='cuda:1'), in_proj_covar=tensor([0.0179, 0.0219, 0.0213, 0.0213, 0.0218, 0.0217, 0.0217, 0.0210], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-30 14:38:03,152 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=171273.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 14:38:17,319 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.1746, 4.1407, 4.5056, 4.4807, 4.5092, 4.2558, 4.2260, 4.1852], device='cuda:1'), covar=tensor([0.0321, 0.0589, 0.0462, 0.0436, 0.0421, 0.0379, 0.0845, 0.0430], device='cuda:1'), in_proj_covar=tensor([0.0366, 0.0398, 0.0392, 0.0372, 0.0438, 0.0411, 0.0504, 0.0325], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:1') 2023-04-30 14:38:27,220 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=171284.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 14:39:01,944 INFO [train.py:904] (1/8) Epoch 17, batch 8900, loss[loss=0.1953, simple_loss=0.2877, pruned_loss=0.05147, over 16916.00 frames. ], tot_loss[loss=0.175, simple_loss=0.2705, pruned_loss=0.03972, over 3070038.77 frames. ], batch size: 116, lr: 3.95e-03, grad_scale: 8.0 2023-04-30 14:39:42,176 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.5565, 4.3688, 4.6331, 4.7345, 4.9195, 4.4298, 4.9190, 4.8974], device='cuda:1'), covar=tensor([0.1732, 0.1115, 0.1474, 0.0763, 0.0520, 0.0902, 0.0470, 0.0598], device='cuda:1'), in_proj_covar=tensor([0.0564, 0.0696, 0.0823, 0.0711, 0.0538, 0.0561, 0.0573, 0.0666], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-30 14:39:46,873 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.2144, 4.2179, 4.6243, 4.5664, 4.5795, 4.3019, 4.2971, 4.2289], device='cuda:1'), covar=tensor([0.0307, 0.0588, 0.0375, 0.0398, 0.0450, 0.0379, 0.0864, 0.0432], device='cuda:1'), in_proj_covar=tensor([0.0364, 0.0395, 0.0389, 0.0369, 0.0435, 0.0409, 0.0502, 0.0324], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:1') 2023-04-30 14:40:38,363 INFO [optim.py:368] (1/8) 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 14:40:50,910 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=171345.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 14:41:04,758 INFO [train.py:904] (1/8) Epoch 17, batch 8950, loss[loss=0.1547, simple_loss=0.2493, pruned_loss=0.03, over 17238.00 frames. ], tot_loss[loss=0.175, simple_loss=0.2702, pruned_loss=0.03991, over 3087352.96 frames. ], batch size: 52, lr: 3.95e-03, grad_scale: 8.0 2023-04-30 14:42:26,268 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.7125, 4.0108, 3.0289, 2.1810, 2.5025, 2.5157, 4.2507, 3.3484], device='cuda:1'), covar=tensor([0.2720, 0.0540, 0.1581, 0.2799, 0.2714, 0.1951, 0.0370, 0.1241], device='cuda:1'), in_proj_covar=tensor([0.0315, 0.0254, 0.0292, 0.0295, 0.0279, 0.0238, 0.0278, 0.0312], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-30 14:42:53,003 INFO [train.py:904] (1/8) Epoch 17, batch 9000, loss[loss=0.1423, simple_loss=0.2367, pruned_loss=0.02392, over 16529.00 frames. ], tot_loss[loss=0.1726, simple_loss=0.2674, pruned_loss=0.03891, over 3083625.94 frames. ], batch size: 68, lr: 3.95e-03, grad_scale: 8.0 2023-04-30 14:42:53,003 INFO [train.py:929] (1/8) Computing validation loss 2023-04-30 14:43:02,948 INFO [train.py:938] (1/8) Epoch 17, validation: loss=0.1478, simple_loss=0.2519, pruned_loss=0.02187, over 944034.00 frames. 2023-04-30 14:43:02,948 INFO [train.py:939] (1/8) Maximum memory allocated so far is 17857MB 2023-04-30 14:43:11,027 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=171406.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 14:43:42,138 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.6402, 2.0468, 1.6204, 1.8192, 2.3678, 2.0146, 2.0998, 2.4424], device='cuda:1'), covar=tensor([0.0154, 0.0422, 0.0572, 0.0497, 0.0297, 0.0404, 0.0210, 0.0313], device='cuda:1'), in_proj_covar=tensor([0.0178, 0.0218, 0.0212, 0.0212, 0.0218, 0.0216, 0.0216, 0.0209], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-30 14:43:57,507 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([5.3592, 5.6806, 5.4713, 5.4789, 5.1883, 5.0719, 5.0976, 5.7330], device='cuda:1'), covar=tensor([0.1054, 0.0816, 0.0756, 0.0780, 0.0733, 0.0732, 0.1118, 0.0778], device='cuda:1'), in_proj_covar=tensor([0.0604, 0.0738, 0.0599, 0.0546, 0.0462, 0.0476, 0.0613, 0.0573], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:1') 2023-04-30 14:44:23,318 INFO [optim.py:368] (1/8) 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 14:44:40,961 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=171448.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 14:44:47,386 INFO [train.py:904] (1/8) Epoch 17, batch 9050, loss[loss=0.149, simple_loss=0.2388, pruned_loss=0.02958, over 16459.00 frames. ], tot_loss[loss=0.1731, simple_loss=0.2681, pruned_loss=0.03908, over 3099281.93 frames. ], batch size: 62, lr: 3.95e-03, grad_scale: 8.0 2023-04-30 14:45:13,333 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.4534, 4.5299, 4.3177, 4.0112, 4.0047, 4.4086, 4.1589, 4.1665], device='cuda:1'), covar=tensor([0.0498, 0.0534, 0.0296, 0.0287, 0.0726, 0.0446, 0.0552, 0.0605], device='cuda:1'), in_proj_covar=tensor([0.0260, 0.0368, 0.0310, 0.0296, 0.0314, 0.0345, 0.0212, 0.0365], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-30 14:46:24,582 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 17, batch 9100, loss[loss=0.1768, simple_loss=0.2791, pruned_loss=0.03722, over 15401.00 frames. ], tot_loss[loss=0.1736, simple_loss=0.2679, pruned_loss=0.03969, over 3089703.05 frames. ], batch size: 191, lr: 3.95e-03, grad_scale: 8.0 2023-04-30 14:46:47,845 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.6703, 3.7499, 2.2913, 4.3017, 2.8056, 4.1718, 2.3379, 2.9424], device='cuda:1'), covar=tensor([0.0266, 0.0327, 0.1550, 0.0155, 0.0767, 0.0476, 0.1511, 0.0756], device='cuda:1'), in_proj_covar=tensor([0.0158, 0.0165, 0.0188, 0.0144, 0.0167, 0.0202, 0.0195, 0.0171], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0003, 0.0004, 0.0004, 0.0003], device='cuda:1') 2023-04-30 14:47:33,876 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([1.9155, 2.8061, 2.6848, 2.0012, 2.5926, 2.8039, 2.7102, 1.9142], device='cuda:1'), covar=tensor([0.0421, 0.0057, 0.0065, 0.0343, 0.0112, 0.0086, 0.0096, 0.0416], device='cuda:1'), in_proj_covar=tensor([0.0132, 0.0075, 0.0076, 0.0130, 0.0089, 0.0100, 0.0087, 0.0123], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-04-30 14:47:58,143 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([1.7759, 1.3856, 1.6861, 1.7182, 1.8398, 1.9009, 1.6157, 1.8235], device='cuda:1'), covar=tensor([0.0253, 0.0367, 0.0202, 0.0270, 0.0271, 0.0177, 0.0408, 0.0124], device='cuda:1'), in_proj_covar=tensor([0.0173, 0.0182, 0.0170, 0.0171, 0.0182, 0.0139, 0.0185, 0.0133], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-30 14:48:02,850 INFO [optim.py:368] (1/8) 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] (1/8) Epoch 17, batch 9150, loss[loss=0.1782, simple_loss=0.272, pruned_loss=0.04221, over 15606.00 frames. ], tot_loss[loss=0.1736, simple_loss=0.2681, pruned_loss=0.03957, over 3073570.29 frames. ], batch size: 192, lr: 3.95e-03, grad_scale: 8.0 2023-04-30 14:48:28,917 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=171552.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 14:48:44,376 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=171559.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 14:48:44,616 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=3.16 vs. limit=5.0 2023-04-30 14:49:18,488 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.87 vs. limit=2.0 2023-04-30 14:49:36,121 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=171584.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 14:49:44,557 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.8600, 5.1438, 4.9194, 4.9444, 4.6760, 4.6212, 4.5310, 5.2099], device='cuda:1'), covar=tensor([0.0941, 0.0841, 0.0740, 0.0733, 0.0713, 0.0930, 0.1060, 0.0811], device='cuda:1'), in_proj_covar=tensor([0.0604, 0.0739, 0.0596, 0.0545, 0.0462, 0.0475, 0.0612, 0.0573], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:1') 2023-04-30 14:50:13,333 INFO [train.py:904] (1/8) Epoch 17, batch 9200, loss[loss=0.1561, simple_loss=0.2511, pruned_loss=0.03055, over 15318.00 frames. ], tot_loss[loss=0.1707, simple_loss=0.2641, pruned_loss=0.03866, over 3070994.42 frames. ], batch size: 191, lr: 3.94e-03, grad_scale: 8.0 2023-04-30 14:50:19,727 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.4108, 4.4921, 4.6324, 4.3815, 4.5131, 5.0330, 4.5246, 4.1566], device='cuda:1'), covar=tensor([0.1280, 0.1819, 0.1742, 0.2253, 0.2589, 0.1019, 0.1583, 0.2593], device='cuda:1'), in_proj_covar=tensor([0.0366, 0.0534, 0.0585, 0.0447, 0.0598, 0.0621, 0.0465, 0.0600], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-04-30 14:51:10,342 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=171632.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 14:51:27,839 INFO [optim.py:368] (1/8) 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 14:51:50,814 INFO [train.py:904] (1/8) Epoch 17, batch 9250, loss[loss=0.1735, simple_loss=0.2648, pruned_loss=0.0411, over 16799.00 frames. ], tot_loss[loss=0.1709, simple_loss=0.264, pruned_loss=0.03888, over 3077680.74 frames. ], batch size: 124, lr: 3.94e-03, grad_scale: 8.0 2023-04-30 14:53:42,093 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=171701.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 14:53:42,880 INFO [train.py:904] (1/8) Epoch 17, batch 9300, loss[loss=0.164, simple_loss=0.2477, pruned_loss=0.04013, over 12182.00 frames. ], tot_loss[loss=0.169, simple_loss=0.2619, pruned_loss=0.03806, over 3060496.16 frames. ], batch size: 248, lr: 3.94e-03, grad_scale: 8.0 2023-04-30 14:55:05,829 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.83 vs. limit=2.0 2023-04-30 14:55:09,819 INFO [optim.py:368] (1/8) 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 14:55:20,706 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=171748.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 14:55:27,769 INFO [train.py:904] (1/8) Epoch 17, batch 9350, loss[loss=0.155, simple_loss=0.2507, pruned_loss=0.02965, over 16469.00 frames. ], tot_loss[loss=0.1685, simple_loss=0.2615, pruned_loss=0.03769, over 3080867.65 frames. ], batch size: 75, lr: 3.94e-03, grad_scale: 4.0 2023-04-30 14:55:42,874 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.8483, 3.6279, 4.0014, 2.0560, 4.0897, 4.1799, 3.1820, 3.1989], device='cuda:1'), covar=tensor([0.0645, 0.0221, 0.0151, 0.1144, 0.0061, 0.0108, 0.0366, 0.0409], device='cuda:1'), in_proj_covar=tensor([0.0139, 0.0100, 0.0087, 0.0131, 0.0070, 0.0111, 0.0118, 0.0123], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:1') 2023-04-30 14:56:24,767 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.29 vs. limit=2.0 2023-04-30 14:56:57,423 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=171796.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 14:57:07,774 INFO [train.py:904] (1/8) Epoch 17, batch 9400, loss[loss=0.1742, simple_loss=0.2716, pruned_loss=0.03842, over 16148.00 frames. ], tot_loss[loss=0.1687, simple_loss=0.2618, pruned_loss=0.03777, over 3074541.76 frames. ], batch size: 165, lr: 3.94e-03, grad_scale: 4.0 2023-04-30 14:58:29,671 INFO [optim.py:368] (1/8) Clipping_scale=2.0, grad-norm quartiles 1.619e+02 2.242e+02 2.701e+02 3.369e+02 7.708e+02, threshold=5.403e+02, percent-clipped=2.0 2023-04-30 14:58:45,813 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.54 vs. limit=2.0 2023-04-30 14:58:48,377 INFO [train.py:904] (1/8) Epoch 17, batch 9450, loss[loss=0.1888, simple_loss=0.2693, pruned_loss=0.05416, over 12305.00 frames. ], tot_loss[loss=0.1697, simple_loss=0.2634, pruned_loss=0.03797, over 3051941.83 frames. ], batch size: 246, lr: 3.94e-03, grad_scale: 4.0 2023-04-30 14:58:48,909 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=171852.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 14:58:52,166 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=171854.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 14:59:05,548 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.4720, 4.4364, 4.2649, 3.7395, 4.3312, 1.5737, 4.1202, 4.0768], device='cuda:1'), covar=tensor([0.0085, 0.0102, 0.0177, 0.0304, 0.0110, 0.2602, 0.0132, 0.0229], device='cuda:1'), in_proj_covar=tensor([0.0143, 0.0133, 0.0176, 0.0159, 0.0152, 0.0190, 0.0164, 0.0155], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-30 15:00:26,650 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=171900.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 15:00:31,019 INFO [train.py:904] (1/8) Epoch 17, batch 9500, loss[loss=0.1575, simple_loss=0.2578, pruned_loss=0.02861, over 16873.00 frames. ], tot_loss[loss=0.1689, simple_loss=0.2628, pruned_loss=0.0375, over 3060619.21 frames. ], batch size: 96, lr: 3.94e-03, grad_scale: 4.0 2023-04-30 15:01:06,827 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.1166, 2.0132, 2.1361, 3.6646, 2.0907, 2.3124, 2.1658, 2.1593], device='cuda:1'), covar=tensor([0.1166, 0.3941, 0.2919, 0.0540, 0.4325, 0.2718, 0.3484, 0.3606], device='cuda:1'), in_proj_covar=tensor([0.0370, 0.0413, 0.0345, 0.0309, 0.0416, 0.0473, 0.0384, 0.0479], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-30 15:01:19,655 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.8803, 4.9960, 4.7832, 4.3628, 4.4263, 4.8557, 4.6898, 4.4970], device='cuda:1'), covar=tensor([0.0586, 0.0375, 0.0292, 0.0323, 0.0920, 0.0445, 0.0371, 0.0629], device='cuda:1'), in_proj_covar=tensor([0.0257, 0.0361, 0.0305, 0.0291, 0.0310, 0.0340, 0.0209, 0.0358], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002], device='cuda:1') 2023-04-30 15:01:41,241 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.0517, 3.0958, 1.8799, 3.3073, 2.2815, 3.2903, 2.0280, 2.6094], device='cuda:1'), covar=tensor([0.0312, 0.0395, 0.1582, 0.0274, 0.0828, 0.0666, 0.1569, 0.0688], device='cuda:1'), in_proj_covar=tensor([0.0159, 0.0165, 0.0187, 0.0145, 0.0168, 0.0204, 0.0196, 0.0171], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0003, 0.0004, 0.0004, 0.0003], device='cuda:1') 2023-04-30 15:01:41,245 INFO [zipformer.py:625] (1/8) 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] (1/8) Clipping_scale=2.0, grad-norm quartiles 1.446e+02 2.125e+02 2.584e+02 3.095e+02 6.778e+02, threshold=5.168e+02, percent-clipped=1.0 2023-04-30 15:02:06,881 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.32 vs. limit=2.0 2023-04-30 15:02:14,948 INFO [train.py:904] (1/8) Epoch 17, batch 9550, loss[loss=0.1816, simple_loss=0.2775, pruned_loss=0.04282, over 16081.00 frames. ], tot_loss[loss=0.169, simple_loss=0.2625, pruned_loss=0.0377, over 3069853.26 frames. ], batch size: 165, lr: 3.94e-03, grad_scale: 4.0 2023-04-30 15:03:26,160 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.76 vs. limit=2.0 2023-04-30 15:03:49,249 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=171998.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 15:03:57,374 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 17, batch 9600, loss[loss=0.1839, simple_loss=0.2813, pruned_loss=0.0432, over 16292.00 frames. ], tot_loss[loss=0.1706, simple_loss=0.2638, pruned_loss=0.03865, over 3066553.17 frames. ], batch size: 165, lr: 3.94e-03, grad_scale: 8.0 2023-04-30 15:05:20,542 INFO [optim.py:368] (1/8) 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 15:05:38,179 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 17, batch 9650, loss[loss=0.1768, simple_loss=0.2723, pruned_loss=0.04063, over 15403.00 frames. ], tot_loss[loss=0.1715, simple_loss=0.2655, pruned_loss=0.03877, over 3071199.15 frames. ], batch size: 192, lr: 3.94e-03, grad_scale: 4.0 2023-04-30 15:06:37,184 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.0221, 3.9113, 4.1286, 4.2301, 4.3597, 3.9857, 4.3528, 4.3927], device='cuda:1'), covar=tensor([0.1810, 0.1191, 0.1418, 0.0692, 0.0568, 0.1212, 0.0558, 0.0704], device='cuda:1'), in_proj_covar=tensor([0.0563, 0.0690, 0.0811, 0.0706, 0.0531, 0.0557, 0.0569, 0.0664], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-30 15:07:32,480 INFO [train.py:904] (1/8) Epoch 17, batch 9700, loss[loss=0.1762, simple_loss=0.263, pruned_loss=0.04467, over 12288.00 frames. ], tot_loss[loss=0.1718, simple_loss=0.2653, pruned_loss=0.03915, over 3061195.69 frames. ], batch size: 248, lr: 3.94e-03, grad_scale: 4.0 2023-04-30 15:08:05,498 INFO [zipformer.py:625] (1/8) 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] (1/8) 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 15:09:14,185 INFO [train.py:904] (1/8) Epoch 17, batch 9750, loss[loss=0.1838, simple_loss=0.2774, pruned_loss=0.04509, over 16818.00 frames. ], tot_loss[loss=0.172, simple_loss=0.2646, pruned_loss=0.03972, over 3059424.54 frames. ], batch size: 124, lr: 3.94e-03, grad_scale: 4.0 2023-04-30 15:09:18,322 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=172154.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 15:09:57,982 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.4414, 3.3976, 3.4987, 3.5447, 3.5981, 3.3148, 3.5763, 3.6448], device='cuda:1'), covar=tensor([0.1303, 0.0872, 0.1052, 0.0659, 0.0616, 0.2410, 0.0821, 0.0716], device='cuda:1'), in_proj_covar=tensor([0.0560, 0.0688, 0.0810, 0.0707, 0.0530, 0.0555, 0.0569, 0.0662], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:1') 2023-04-30 15:10:09,712 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=172179.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 15:10:54,092 INFO [train.py:904] (1/8) Epoch 17, batch 9800, loss[loss=0.1912, simple_loss=0.2899, pruned_loss=0.04625, over 16673.00 frames. ], tot_loss[loss=0.1713, simple_loss=0.2646, pruned_loss=0.03902, over 3047491.45 frames. ], batch size: 134, lr: 3.94e-03, grad_scale: 4.0 2023-04-30 15:10:54,781 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=172202.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 15:12:17,139 INFO [optim.py:368] (1/8) 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 15:12:38,992 INFO [train.py:904] (1/8) Epoch 17, batch 9850, loss[loss=0.168, simple_loss=0.2676, pruned_loss=0.03417, over 16900.00 frames. ], tot_loss[loss=0.1715, simple_loss=0.2659, pruned_loss=0.03852, over 3061204.45 frames. ], batch size: 102, lr: 3.94e-03, grad_scale: 4.0 2023-04-30 15:14:13,934 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=172293.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 15:14:32,003 INFO [train.py:904] (1/8) Epoch 17, batch 9900, loss[loss=0.1709, simple_loss=0.2709, pruned_loss=0.03541, over 16917.00 frames. ], tot_loss[loss=0.1718, simple_loss=0.2665, pruned_loss=0.03853, over 3063184.25 frames. ], batch size: 116, lr: 3.94e-03, grad_scale: 4.0 2023-04-30 15:14:42,677 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.0339, 1.8554, 1.5906, 1.5486, 2.0283, 1.5875, 1.6222, 1.9867], device='cuda:1'), covar=tensor([0.0165, 0.0294, 0.0447, 0.0362, 0.0239, 0.0297, 0.0183, 0.0220], device='cuda:1'), in_proj_covar=tensor([0.0178, 0.0221, 0.0212, 0.0213, 0.0219, 0.0218, 0.0214, 0.0209], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-30 15:16:10,689 INFO [optim.py:368] (1/8) 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] (1/8) Epoch 17, batch 9950, loss[loss=0.1583, simple_loss=0.2619, pruned_loss=0.0273, over 16846.00 frames. ], tot_loss[loss=0.1727, simple_loss=0.2679, pruned_loss=0.03869, over 3054119.27 frames. ], batch size: 102, lr: 3.94e-03, grad_scale: 4.0 2023-04-30 15:16:55,924 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.9159, 3.1408, 3.4230, 1.8843, 2.9638, 2.1455, 3.4300, 3.4172], device='cuda:1'), covar=tensor([0.0210, 0.0743, 0.0591, 0.2054, 0.0715, 0.0986, 0.0536, 0.0788], device='cuda:1'), in_proj_covar=tensor([0.0146, 0.0148, 0.0158, 0.0145, 0.0136, 0.0123, 0.0136, 0.0158], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:1') 2023-04-30 15:18:33,055 INFO [train.py:904] (1/8) Epoch 17, batch 10000, loss[loss=0.1683, simple_loss=0.254, pruned_loss=0.04133, over 17013.00 frames. ], tot_loss[loss=0.1718, simple_loss=0.2669, pruned_loss=0.03835, over 3091800.06 frames. ], batch size: 55, lr: 3.94e-03, grad_scale: 8.0 2023-04-30 15:19:56,011 INFO [optim.py:368] (1/8) 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] (1/8) Epoch 17, batch 10050, loss[loss=0.1707, simple_loss=0.2582, pruned_loss=0.04156, over 12086.00 frames. ], tot_loss[loss=0.172, simple_loss=0.2673, pruned_loss=0.03841, over 3087523.21 frames. ], batch size: 250, lr: 3.93e-03, grad_scale: 8.0 2023-04-30 15:20:58,136 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=172474.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 15:21:17,062 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([1.8272, 1.2970, 1.7179, 1.7332, 1.8066, 1.9133, 1.6274, 1.8619], device='cuda:1'), covar=tensor([0.0213, 0.0381, 0.0203, 0.0263, 0.0291, 0.0191, 0.0385, 0.0140], device='cuda:1'), in_proj_covar=tensor([0.0170, 0.0179, 0.0166, 0.0167, 0.0178, 0.0136, 0.0182, 0.0129], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-30 15:21:47,144 INFO [train.py:904] (1/8) Epoch 17, batch 10100, loss[loss=0.1681, simple_loss=0.2583, pruned_loss=0.03893, over 15273.00 frames. ], tot_loss[loss=0.1729, simple_loss=0.2681, pruned_loss=0.03884, over 3073083.99 frames. ], batch size: 191, lr: 3.93e-03, grad_scale: 8.0 2023-04-30 15:22:57,875 INFO [optim.py:368] (1/8) 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 15:23:33,082 INFO [train.py:904] (1/8) Epoch 18, batch 0, loss[loss=0.1906, simple_loss=0.2814, pruned_loss=0.04995, over 16494.00 frames. ], tot_loss[loss=0.1906, simple_loss=0.2814, pruned_loss=0.04995, over 16494.00 frames. ], batch size: 68, lr: 3.82e-03, grad_scale: 8.0 2023-04-30 15:23:33,082 INFO [train.py:929] (1/8) Computing validation loss 2023-04-30 15:23:40,341 INFO [train.py:938] (1/8) Epoch 18, validation: loss=0.148, simple_loss=0.2515, pruned_loss=0.02223, over 944034.00 frames. 2023-04-30 15:23:40,342 INFO [train.py:939] (1/8) Maximum memory allocated so far is 17857MB 2023-04-30 15:24:36,544 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=172593.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 15:24:44,194 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 18, batch 50, loss[loss=0.1795, simple_loss=0.2784, pruned_loss=0.04028, over 16691.00 frames. ], tot_loss[loss=0.1902, simple_loss=0.2745, pruned_loss=0.05296, over 750063.14 frames. ], batch size: 57, lr: 3.82e-03, grad_scale: 1.0 2023-04-30 15:25:21,077 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=172625.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 15:25:42,961 INFO [zipformer.py:625] (1/8) 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] (1/8) 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 15:25:55,767 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.77 vs. limit=2.0 2023-04-30 15:25:56,033 INFO [train.py:904] (1/8) Epoch 18, batch 100, loss[loss=0.2076, simple_loss=0.2803, pruned_loss=0.06745, over 16529.00 frames. ], tot_loss[loss=0.1865, simple_loss=0.2715, pruned_loss=0.05076, over 1319177.22 frames. ], batch size: 75, lr: 3.82e-03, grad_scale: 1.0 2023-04-30 15:25:59,954 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.31 vs. limit=2.0 2023-04-30 15:26:05,987 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=172659.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 15:26:44,047 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=172686.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 15:26:50,285 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.7335, 5.0430, 5.4998, 5.4121, 5.4337, 5.0402, 4.6709, 4.8247], device='cuda:1'), covar=tensor([0.0669, 0.0634, 0.0518, 0.0722, 0.0780, 0.0582, 0.1594, 0.0612], device='cuda:1'), in_proj_covar=tensor([0.0366, 0.0398, 0.0391, 0.0369, 0.0434, 0.0410, 0.0498, 0.0326], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:1') 2023-04-30 15:27:01,126 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=4.45 vs. limit=5.0 2023-04-30 15:27:02,718 INFO [train.py:904] (1/8) Epoch 18, batch 150, loss[loss=0.1744, simple_loss=0.2705, pruned_loss=0.03909, over 15838.00 frames. ], tot_loss[loss=0.1828, simple_loss=0.2686, pruned_loss=0.04848, over 1767410.47 frames. ], batch size: 35, lr: 3.82e-03, grad_scale: 1.0 2023-04-30 15:27:46,398 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=172734.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 15:28:01,449 INFO [optim.py:368] (1/8) 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] (1/8) Epoch 18, batch 200, loss[loss=0.2091, simple_loss=0.2751, pruned_loss=0.07154, over 16792.00 frames. ], tot_loss[loss=0.1832, simple_loss=0.269, pruned_loss=0.04866, over 2119844.12 frames. ], batch size: 116, lr: 3.82e-03, grad_scale: 1.0 2023-04-30 15:28:20,109 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.8423, 4.2580, 3.0370, 2.2998, 2.7434, 2.3603, 4.5650, 3.6228], device='cuda:1'), covar=tensor([0.2882, 0.0666, 0.1805, 0.2909, 0.2798, 0.2248, 0.0414, 0.1341], device='cuda:1'), in_proj_covar=tensor([0.0316, 0.0257, 0.0296, 0.0297, 0.0280, 0.0242, 0.0281, 0.0317], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:1') 2023-04-30 15:28:29,354 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([5.1738, 5.1007, 5.6552, 5.6205, 5.6654, 5.2658, 5.1738, 5.0613], device='cuda:1'), covar=tensor([0.0318, 0.0582, 0.0377, 0.0483, 0.0524, 0.0385, 0.1046, 0.0441], device='cuda:1'), in_proj_covar=tensor([0.0372, 0.0403, 0.0397, 0.0375, 0.0441, 0.0416, 0.0506, 0.0331], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:1') 2023-04-30 15:28:41,366 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=172774.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 15:29:10,033 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=172795.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 15:29:18,639 INFO [train.py:904] (1/8) Epoch 18, batch 250, loss[loss=0.1588, simple_loss=0.2543, pruned_loss=0.03162, over 17218.00 frames. ], tot_loss[loss=0.181, simple_loss=0.2667, pruned_loss=0.04758, over 2387270.48 frames. ], batch size: 46, lr: 3.82e-03, grad_scale: 1.0 2023-04-30 15:29:47,727 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=172822.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 15:30:18,710 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=172845.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 15:30:19,548 INFO [optim.py:368] (1/8) 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] (1/8) Epoch 18, batch 300, loss[loss=0.1751, simple_loss=0.2527, pruned_loss=0.04876, over 16301.00 frames. ], tot_loss[loss=0.1778, simple_loss=0.2631, pruned_loss=0.04621, over 2595611.25 frames. ], batch size: 165, lr: 3.82e-03, grad_scale: 1.0 2023-04-30 15:31:39,797 INFO [train.py:904] (1/8) Epoch 18, batch 350, loss[loss=0.1978, simple_loss=0.2619, pruned_loss=0.06686, over 16729.00 frames. ], tot_loss[loss=0.1757, simple_loss=0.2609, pruned_loss=0.04523, over 2758648.90 frames. ], batch size: 89, lr: 3.82e-03, grad_scale: 1.0 2023-04-30 15:31:45,415 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=172906.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 15:31:59,081 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.25 vs. limit=2.0 2023-04-30 15:32:07,213 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([5.1504, 5.6872, 5.8689, 5.5835, 5.6702, 6.2235, 5.7104, 5.4081], device='cuda:1'), covar=tensor([0.0918, 0.1977, 0.1998, 0.1903, 0.2507, 0.0930, 0.1647, 0.2385], device='cuda:1'), in_proj_covar=tensor([0.0377, 0.0552, 0.0605, 0.0459, 0.0619, 0.0641, 0.0482, 0.0618], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-04-30 15:32:40,483 INFO [optim.py:368] (1/8) 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 15:32:47,910 INFO [train.py:904] (1/8) Epoch 18, batch 400, loss[loss=0.2019, simple_loss=0.2714, pruned_loss=0.06622, over 16859.00 frames. ], tot_loss[loss=0.1745, simple_loss=0.259, pruned_loss=0.04497, over 2877622.96 frames. ], batch size: 109, lr: 3.82e-03, grad_scale: 2.0 2023-04-30 15:32:51,104 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=172954.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 15:33:22,566 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.58 vs. limit=2.0 2023-04-30 15:33:30,371 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 18, batch 450, loss[loss=0.1626, simple_loss=0.2413, pruned_loss=0.04194, over 16885.00 frames. ], tot_loss[loss=0.1725, simple_loss=0.2566, pruned_loss=0.04414, over 2961914.41 frames. ], batch size: 90, lr: 3.82e-03, grad_scale: 2.0 2023-04-30 15:34:31,824 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=4.11 vs. limit=5.0 2023-04-30 15:35:00,082 INFO [optim.py:368] (1/8) 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 15:35:08,316 INFO [train.py:904] (1/8) Epoch 18, batch 500, loss[loss=0.17, simple_loss=0.2456, pruned_loss=0.04721, over 16737.00 frames. ], tot_loss[loss=0.172, simple_loss=0.2561, pruned_loss=0.04392, over 3042475.46 frames. ], batch size: 89, lr: 3.82e-03, grad_scale: 2.0 2023-04-30 15:35:59,674 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=173090.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 15:36:14,716 INFO [train.py:904] (1/8) Epoch 18, batch 550, loss[loss=0.1738, simple_loss=0.2569, pruned_loss=0.04533, over 16475.00 frames. ], tot_loss[loss=0.171, simple_loss=0.2551, pruned_loss=0.04342, over 3107625.63 frames. ], batch size: 146, lr: 3.82e-03, grad_scale: 2.0 2023-04-30 15:37:14,572 INFO [optim.py:368] (1/8) 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 15:37:20,795 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=173150.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 15:37:22,687 INFO [train.py:904] (1/8) Epoch 18, batch 600, loss[loss=0.1706, simple_loss=0.2459, pruned_loss=0.04765, over 16744.00 frames. ], tot_loss[loss=0.1704, simple_loss=0.2543, pruned_loss=0.04326, over 3160677.53 frames. ], batch size: 124, lr: 3.81e-03, grad_scale: 2.0 2023-04-30 15:37:26,054 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.6810, 4.5911, 4.5483, 4.0653, 4.5926, 1.7291, 4.3323, 4.1891], device='cuda:1'), covar=tensor([0.0136, 0.0125, 0.0197, 0.0319, 0.0125, 0.2867, 0.0176, 0.0219], device='cuda:1'), in_proj_covar=tensor([0.0151, 0.0140, 0.0185, 0.0167, 0.0160, 0.0198, 0.0173, 0.0164], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-30 15:37:33,233 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.6641, 4.7922, 4.9062, 4.8023, 4.8243, 5.3786, 4.8931, 4.6176], device='cuda:1'), covar=tensor([0.1473, 0.2062, 0.2540, 0.2226, 0.2862, 0.1307, 0.1881, 0.2673], device='cuda:1'), in_proj_covar=tensor([0.0385, 0.0563, 0.0618, 0.0467, 0.0634, 0.0653, 0.0492, 0.0630], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-04-30 15:38:30,580 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=173201.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 15:38:31,378 INFO [train.py:904] (1/8) Epoch 18, batch 650, loss[loss=0.1522, simple_loss=0.2373, pruned_loss=0.03357, over 17229.00 frames. ], tot_loss[loss=0.1697, simple_loss=0.2528, pruned_loss=0.04331, over 3197698.16 frames. ], batch size: 45, lr: 3.81e-03, grad_scale: 2.0 2023-04-30 15:38:33,981 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.0161, 3.8991, 4.0621, 4.1858, 4.2681, 3.8073, 4.0735, 4.2509], device='cuda:1'), covar=tensor([0.1618, 0.1181, 0.1283, 0.0673, 0.0657, 0.1883, 0.2250, 0.0939], device='cuda:1'), in_proj_covar=tensor([0.0604, 0.0743, 0.0879, 0.0753, 0.0563, 0.0599, 0.0615, 0.0714], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-30 15:38:44,851 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=173211.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 15:39:32,445 INFO [optim.py:368] (1/8) 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 15:39:40,835 INFO [train.py:904] (1/8) Epoch 18, batch 700, loss[loss=0.1659, simple_loss=0.2461, pruned_loss=0.04285, over 16381.00 frames. ], tot_loss[loss=0.1692, simple_loss=0.2524, pruned_loss=0.04298, over 3222513.60 frames. ], batch size: 146, lr: 3.81e-03, grad_scale: 2.0 2023-04-30 15:39:43,451 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=173254.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 15:40:08,905 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.9421, 4.8354, 4.8006, 4.4332, 4.4230, 4.8486, 4.7411, 4.5386], device='cuda:1'), covar=tensor([0.0736, 0.0817, 0.0313, 0.0296, 0.1035, 0.0576, 0.0419, 0.0738], device='cuda:1'), in_proj_covar=tensor([0.0282, 0.0398, 0.0331, 0.0319, 0.0341, 0.0370, 0.0226, 0.0395], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002], device='cuda:1') 2023-04-30 15:40:22,721 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=173281.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 15:40:46,211 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.7733, 3.8970, 2.4929, 4.4133, 2.9498, 4.4065, 2.3878, 3.0855], device='cuda:1'), covar=tensor([0.0284, 0.0341, 0.1508, 0.0317, 0.0858, 0.0494, 0.1614, 0.0803], device='cuda:1'), in_proj_covar=tensor([0.0166, 0.0173, 0.0195, 0.0155, 0.0174, 0.0213, 0.0203, 0.0177], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-04-30 15:40:50,345 INFO [train.py:904] (1/8) Epoch 18, batch 750, loss[loss=0.2197, simple_loss=0.2821, pruned_loss=0.0786, over 16867.00 frames. ], tot_loss[loss=0.1693, simple_loss=0.2528, pruned_loss=0.04289, over 3247351.60 frames. ], batch size: 109, lr: 3.81e-03, grad_scale: 2.0 2023-04-30 15:40:51,630 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=173302.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 15:41:00,060 INFO [zipformer.py:625] (1/8) 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] (1/8) 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] (1/8) 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] (1/8) Epoch 18, batch 800, loss[loss=0.1895, simple_loss=0.2532, pruned_loss=0.06287, over 16871.00 frames. ], tot_loss[loss=0.1694, simple_loss=0.2532, pruned_loss=0.04278, over 3260843.39 frames. ], batch size: 116, lr: 3.81e-03, grad_scale: 4.0 2023-04-30 15:42:22,672 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=173370.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 15:42:30,198 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.7057, 2.8223, 2.5432, 2.7279, 3.0665, 2.9386, 3.4214, 3.3544], device='cuda:1'), covar=tensor([0.0155, 0.0388, 0.0433, 0.0394, 0.0275, 0.0360, 0.0247, 0.0231], device='cuda:1'), in_proj_covar=tensor([0.0193, 0.0231, 0.0221, 0.0223, 0.0231, 0.0230, 0.0231, 0.0221], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-30 15:42:43,040 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.16 vs. limit=2.0 2023-04-30 15:42:49,276 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=173390.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 15:43:05,169 INFO [train.py:904] (1/8) Epoch 18, batch 850, loss[loss=0.1651, simple_loss=0.2609, pruned_loss=0.03462, over 17040.00 frames. ], tot_loss[loss=0.1683, simple_loss=0.2523, pruned_loss=0.04215, over 3276844.70 frames. ], batch size: 50, lr: 3.81e-03, grad_scale: 4.0 2023-04-30 15:43:55,001 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=173438.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 15:44:03,726 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.8215, 2.4473, 2.5135, 4.5634, 2.4880, 2.8620, 2.5435, 2.6350], device='cuda:1'), covar=tensor([0.1031, 0.3539, 0.2795, 0.0448, 0.4050, 0.2524, 0.3207, 0.3524], device='cuda:1'), in_proj_covar=tensor([0.0386, 0.0426, 0.0355, 0.0322, 0.0427, 0.0490, 0.0396, 0.0498], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-30 15:44:07,464 INFO [optim.py:368] (1/8) 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] (1/8) Epoch 18, batch 900, loss[loss=0.1815, simple_loss=0.2547, pruned_loss=0.05412, over 16292.00 frames. ], tot_loss[loss=0.1683, simple_loss=0.2522, pruned_loss=0.04219, over 3277394.18 frames. ], batch size: 165, lr: 3.81e-03, grad_scale: 4.0 2023-04-30 15:44:40,112 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([1.7848, 2.9689, 2.8724, 5.0143, 4.1203, 4.4584, 1.7282, 3.2008], device='cuda:1'), covar=tensor([0.1389, 0.0714, 0.1088, 0.0212, 0.0265, 0.0421, 0.1591, 0.0783], device='cuda:1'), in_proj_covar=tensor([0.0164, 0.0167, 0.0189, 0.0175, 0.0198, 0.0213, 0.0194, 0.0189], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-04-30 15:45:11,326 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=173491.0, num_to_drop=1, layers_to_drop={0} 2023-04-30 15:45:24,055 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 18, batch 950, loss[loss=0.171, simple_loss=0.2469, pruned_loss=0.04758, over 16809.00 frames. ], tot_loss[loss=0.168, simple_loss=0.2525, pruned_loss=0.04175, over 3297073.74 frames. ], batch size: 102, lr: 3.81e-03, grad_scale: 4.0 2023-04-30 15:45:31,768 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=173506.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 15:45:57,707 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.4274, 2.6166, 2.1564, 2.3548, 2.9464, 2.7027, 3.1331, 3.1227], device='cuda:1'), covar=tensor([0.0198, 0.0438, 0.0576, 0.0514, 0.0297, 0.0356, 0.0339, 0.0279], device='cuda:1'), in_proj_covar=tensor([0.0196, 0.0234, 0.0224, 0.0225, 0.0234, 0.0232, 0.0233, 0.0224], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-30 15:46:26,059 INFO [optim.py:368] (1/8) 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] (1/8) 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] (1/8) Epoch 18, batch 1000, loss[loss=0.1807, simple_loss=0.2544, pruned_loss=0.05352, over 16460.00 frames. ], tot_loss[loss=0.1669, simple_loss=0.2512, pruned_loss=0.04131, over 3297248.74 frames. ], batch size: 146, lr: 3.81e-03, grad_scale: 4.0 2023-04-30 15:46:35,660 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=173552.0, num_to_drop=1, layers_to_drop={0} 2023-04-30 15:47:44,518 INFO [train.py:904] (1/8) Epoch 18, batch 1050, loss[loss=0.1479, simple_loss=0.2377, pruned_loss=0.02909, over 17001.00 frames. ], tot_loss[loss=0.1664, simple_loss=0.2504, pruned_loss=0.04124, over 3301211.75 frames. ], batch size: 41, lr: 3.81e-03, grad_scale: 4.0 2023-04-30 15:48:02,444 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=173614.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 15:48:32,972 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=3.53 vs. limit=5.0 2023-04-30 15:48:46,276 INFO [optim.py:368] (1/8) 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] (1/8) Epoch 18, batch 1100, loss[loss=0.156, simple_loss=0.2513, pruned_loss=0.03031, over 17188.00 frames. ], tot_loss[loss=0.1656, simple_loss=0.2498, pruned_loss=0.04068, over 3298752.38 frames. ], batch size: 46, lr: 3.81e-03, grad_scale: 4.0 2023-04-30 15:49:12,475 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=173665.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 15:49:26,654 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=173675.0, num_to_drop=1, layers_to_drop={3} 2023-04-30 15:49:33,288 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.9522, 1.9376, 2.5026, 2.8987, 2.7060, 3.3616, 2.3599, 3.3183], device='cuda:1'), covar=tensor([0.0209, 0.0482, 0.0319, 0.0289, 0.0296, 0.0174, 0.0426, 0.0143], device='cuda:1'), in_proj_covar=tensor([0.0178, 0.0187, 0.0172, 0.0176, 0.0184, 0.0144, 0.0190, 0.0138], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-30 15:49:34,239 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.9819, 5.5053, 5.6806, 5.3819, 5.4401, 6.0267, 5.5224, 5.2127], device='cuda:1'), covar=tensor([0.0984, 0.1915, 0.2082, 0.2056, 0.2541, 0.0913, 0.1464, 0.2454], device='cuda:1'), in_proj_covar=tensor([0.0393, 0.0577, 0.0630, 0.0477, 0.0646, 0.0661, 0.0498, 0.0645], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-04-30 15:50:01,540 INFO [train.py:904] (1/8) Epoch 18, batch 1150, loss[loss=0.1638, simple_loss=0.255, pruned_loss=0.03635, over 16668.00 frames. ], tot_loss[loss=0.1659, simple_loss=0.2504, pruned_loss=0.04069, over 3305453.05 frames. ], batch size: 62, lr: 3.81e-03, grad_scale: 4.0 2023-04-30 15:51:02,213 INFO [optim.py:368] (1/8) 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] (1/8) Epoch 18, batch 1200, loss[loss=0.1574, simple_loss=0.2406, pruned_loss=0.03704, over 16789.00 frames. ], tot_loss[loss=0.1653, simple_loss=0.25, pruned_loss=0.04031, over 3311266.96 frames. ], batch size: 102, lr: 3.81e-03, grad_scale: 8.0 2023-04-30 15:51:12,954 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.7226, 4.4691, 4.6278, 4.9342, 5.1249, 4.6099, 5.2046, 5.1380], device='cuda:1'), covar=tensor([0.1955, 0.1639, 0.2376, 0.1142, 0.0883, 0.1080, 0.0831, 0.0928], device='cuda:1'), in_proj_covar=tensor([0.0618, 0.0763, 0.0897, 0.0773, 0.0576, 0.0610, 0.0625, 0.0729], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-30 15:52:19,913 INFO [train.py:904] (1/8) Epoch 18, batch 1250, loss[loss=0.1882, simple_loss=0.2614, pruned_loss=0.05749, over 16809.00 frames. ], tot_loss[loss=0.1657, simple_loss=0.2501, pruned_loss=0.04067, over 3308680.73 frames. ], batch size: 124, lr: 3.81e-03, grad_scale: 8.0 2023-04-30 15:52:24,681 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=173806.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 15:52:52,562 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.9323, 2.0324, 2.4827, 2.8670, 2.7142, 3.1966, 2.2575, 3.1183], device='cuda:1'), covar=tensor([0.0207, 0.0448, 0.0318, 0.0275, 0.0322, 0.0201, 0.0438, 0.0174], device='cuda:1'), in_proj_covar=tensor([0.0180, 0.0189, 0.0175, 0.0177, 0.0186, 0.0145, 0.0191, 0.0140], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-30 15:53:21,075 INFO [optim.py:368] (1/8) 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 15:53:22,577 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=173847.0, num_to_drop=1, layers_to_drop={2} 2023-04-30 15:53:28,943 INFO [train.py:904] (1/8) Epoch 18, batch 1300, loss[loss=0.1644, simple_loss=0.2504, pruned_loss=0.03923, over 16830.00 frames. ], tot_loss[loss=0.166, simple_loss=0.2499, pruned_loss=0.04108, over 3306657.61 frames. ], batch size: 42, lr: 3.81e-03, grad_scale: 8.0 2023-04-30 15:53:32,735 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 18, batch 1350, loss[loss=0.1672, simple_loss=0.2642, pruned_loss=0.03511, over 17041.00 frames. ], tot_loss[loss=0.1662, simple_loss=0.2506, pruned_loss=0.04089, over 3314315.43 frames. ], batch size: 50, lr: 3.81e-03, grad_scale: 8.0 2023-04-30 15:54:55,056 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.3598, 4.0798, 4.5344, 2.5315, 4.7718, 4.7309, 3.6444, 3.9861], device='cuda:1'), covar=tensor([0.0557, 0.0213, 0.0202, 0.0935, 0.0050, 0.0148, 0.0317, 0.0273], device='cuda:1'), in_proj_covar=tensor([0.0150, 0.0107, 0.0095, 0.0140, 0.0076, 0.0121, 0.0126, 0.0130], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0004, 0.0004], device='cuda:1') 2023-04-30 15:55:37,173 INFO [optim.py:368] (1/8) 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 15:55:41,041 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.4139, 4.1752, 4.6761, 2.5403, 4.9154, 4.8613, 3.6961, 4.0690], device='cuda:1'), covar=tensor([0.0624, 0.0207, 0.0188, 0.1034, 0.0050, 0.0159, 0.0334, 0.0311], device='cuda:1'), in_proj_covar=tensor([0.0150, 0.0107, 0.0095, 0.0141, 0.0076, 0.0122, 0.0127, 0.0131], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0004, 0.0004], device='cuda:1') 2023-04-30 15:55:44,243 INFO [train.py:904] (1/8) Epoch 18, batch 1400, loss[loss=0.166, simple_loss=0.2437, pruned_loss=0.04413, over 16432.00 frames. ], tot_loss[loss=0.166, simple_loss=0.2506, pruned_loss=0.04074, over 3321096.73 frames. ], batch size: 75, lr: 3.81e-03, grad_scale: 8.0 2023-04-30 15:56:04,268 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=173965.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 15:56:09,763 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=173970.0, num_to_drop=1, layers_to_drop={1} 2023-04-30 15:56:26,377 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([1.6046, 2.6117, 2.4945, 4.7147, 3.8092, 4.3248, 1.7657, 2.9581], device='cuda:1'), covar=tensor([0.1659, 0.0939, 0.1423, 0.0279, 0.0246, 0.0429, 0.1762, 0.0888], device='cuda:1'), in_proj_covar=tensor([0.0163, 0.0167, 0.0187, 0.0175, 0.0197, 0.0212, 0.0192, 0.0188], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-04-30 15:56:37,966 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.85 vs. limit=2.0 2023-04-30 15:56:56,559 INFO [train.py:904] (1/8) Epoch 18, batch 1450, loss[loss=0.1508, simple_loss=0.2285, pruned_loss=0.03653, over 16802.00 frames. ], tot_loss[loss=0.1657, simple_loss=0.2494, pruned_loss=0.04099, over 3314418.25 frames. ], batch size: 102, lr: 3.81e-03, grad_scale: 8.0 2023-04-30 15:57:11,463 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=174013.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 15:57:14,055 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=174015.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 15:57:16,931 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=174017.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 15:57:25,574 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([5.5686, 5.4995, 5.4190, 4.9631, 4.9967, 5.4642, 5.4407, 5.1759], device='cuda:1'), covar=tensor([0.0549, 0.0457, 0.0259, 0.0323, 0.0992, 0.0442, 0.0237, 0.0622], device='cuda:1'), in_proj_covar=tensor([0.0288, 0.0407, 0.0337, 0.0327, 0.0350, 0.0381, 0.0232, 0.0406], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002], device='cuda:1') 2023-04-30 15:57:37,543 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.25 vs. limit=2.0 2023-04-30 15:57:54,752 INFO [optim.py:368] (1/8) Clipping_scale=2.0, grad-norm quartiles 1.387e+02 2.047e+02 2.425e+02 3.116e+02 5.904e+02, threshold=4.850e+02, percent-clipped=1.0 2023-04-30 15:58:03,184 INFO [train.py:904] (1/8) Epoch 18, batch 1500, loss[loss=0.1785, simple_loss=0.2534, pruned_loss=0.05183, over 16375.00 frames. ], tot_loss[loss=0.1661, simple_loss=0.2495, pruned_loss=0.04137, over 3318566.93 frames. ], batch size: 146, lr: 3.80e-03, grad_scale: 8.0 2023-04-30 15:58:18,193 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.29 vs. limit=2.0 2023-04-30 15:58:30,554 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.11 vs. limit=2.0 2023-04-30 15:58:36,138 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=174076.0, num_to_drop=1, layers_to_drop={3} 2023-04-30 15:58:38,876 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=174078.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 15:59:10,941 INFO [train.py:904] (1/8) Epoch 18, batch 1550, loss[loss=0.1851, simple_loss=0.2631, pruned_loss=0.05352, over 16535.00 frames. ], tot_loss[loss=0.1682, simple_loss=0.2512, pruned_loss=0.04264, over 3313283.49 frames. ], batch size: 68, lr: 3.80e-03, grad_scale: 8.0 2023-04-30 16:00:00,522 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=174139.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 16:00:09,836 INFO [optim.py:368] (1/8) 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 16:00:11,980 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=174147.0, num_to_drop=1, layers_to_drop={0} 2023-04-30 16:00:17,418 INFO [train.py:904] (1/8) Epoch 18, batch 1600, loss[loss=0.1589, simple_loss=0.2441, pruned_loss=0.03691, over 15781.00 frames. ], tot_loss[loss=0.1703, simple_loss=0.2534, pruned_loss=0.04356, over 3315942.41 frames. ], batch size: 35, lr: 3.80e-03, grad_scale: 8.0 2023-04-30 16:01:16,143 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=174195.0, num_to_drop=1, layers_to_drop={1} 2023-04-30 16:01:23,788 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=174200.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 16:01:25,620 INFO [train.py:904] (1/8) Epoch 18, batch 1650, loss[loss=0.1458, simple_loss=0.2328, pruned_loss=0.02939, over 17194.00 frames. ], tot_loss[loss=0.1712, simple_loss=0.2544, pruned_loss=0.044, over 3320419.45 frames. ], batch size: 46, lr: 3.80e-03, grad_scale: 8.0 2023-04-30 16:01:40,422 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.5065, 1.6371, 2.1621, 2.3634, 2.3807, 2.4760, 1.8035, 2.6346], device='cuda:1'), covar=tensor([0.0177, 0.0482, 0.0311, 0.0267, 0.0290, 0.0234, 0.0467, 0.0149], device='cuda:1'), in_proj_covar=tensor([0.0182, 0.0190, 0.0176, 0.0180, 0.0188, 0.0147, 0.0193, 0.0141], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-30 16:02:23,837 INFO [optim.py:368] (1/8) 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 16:02:32,761 INFO [train.py:904] (1/8) Epoch 18, batch 1700, loss[loss=0.1901, simple_loss=0.2834, pruned_loss=0.04847, over 15773.00 frames. ], tot_loss[loss=0.1741, simple_loss=0.2575, pruned_loss=0.04528, over 3319419.41 frames. ], batch size: 35, lr: 3.80e-03, grad_scale: 8.0 2023-04-30 16:02:57,301 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=174270.0, num_to_drop=1, layers_to_drop={0} 2023-04-30 16:03:12,400 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.72 vs. limit=2.0 2023-04-30 16:03:15,628 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=174283.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 16:03:24,439 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=174291.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 16:03:24,523 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([1.7373, 2.8557, 2.7506, 4.9601, 3.9809, 4.4618, 1.6228, 3.1423], device='cuda:1'), covar=tensor([0.1358, 0.0775, 0.1202, 0.0228, 0.0270, 0.0402, 0.1551, 0.0793], device='cuda:1'), in_proj_covar=tensor([0.0164, 0.0168, 0.0189, 0.0177, 0.0199, 0.0214, 0.0193, 0.0189], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-04-30 16:03:28,358 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.5356, 2.2505, 1.7122, 2.0590, 2.6084, 2.3914, 2.5994, 2.7033], device='cuda:1'), covar=tensor([0.0210, 0.0449, 0.0629, 0.0522, 0.0258, 0.0354, 0.0231, 0.0305], device='cuda:1'), in_proj_covar=tensor([0.0199, 0.0233, 0.0223, 0.0225, 0.0233, 0.0232, 0.0235, 0.0225], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-30 16:03:40,505 INFO [train.py:904] (1/8) Epoch 18, batch 1750, loss[loss=0.1865, simple_loss=0.2665, pruned_loss=0.05327, over 16720.00 frames. ], tot_loss[loss=0.1743, simple_loss=0.2584, pruned_loss=0.04511, over 3318856.17 frames. ], batch size: 124, lr: 3.80e-03, grad_scale: 8.0 2023-04-30 16:03:56,804 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.1683, 2.2202, 2.7550, 3.2367, 2.9132, 3.6031, 2.6449, 3.5635], device='cuda:1'), covar=tensor([0.0206, 0.0451, 0.0268, 0.0239, 0.0279, 0.0158, 0.0391, 0.0137], device='cuda:1'), in_proj_covar=tensor([0.0181, 0.0189, 0.0175, 0.0179, 0.0187, 0.0147, 0.0192, 0.0140], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-30 16:04:01,306 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=174318.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 16:04:36,724 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=174344.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 16:04:38,536 INFO [optim.py:368] (1/8) 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] (1/8) Epoch 18, batch 1800, loss[loss=0.1704, simple_loss=0.2509, pruned_loss=0.04498, over 16727.00 frames. ], tot_loss[loss=0.1747, simple_loss=0.2594, pruned_loss=0.04496, over 3312031.28 frames. ], batch size: 83, lr: 3.80e-03, grad_scale: 8.0 2023-04-30 16:04:46,536 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=174352.0, num_to_drop=1, layers_to_drop={2} 2023-04-30 16:05:00,216 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.6619, 4.7798, 5.1716, 5.1462, 5.1880, 4.8303, 4.7834, 4.5636], device='cuda:1'), covar=tensor([0.0373, 0.0583, 0.0414, 0.0456, 0.0528, 0.0410, 0.1091, 0.0525], device='cuda:1'), in_proj_covar=tensor([0.0402, 0.0439, 0.0426, 0.0399, 0.0473, 0.0449, 0.0544, 0.0355], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:1') 2023-04-30 16:05:10,957 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=174371.0, num_to_drop=1, layers_to_drop={0} 2023-04-30 16:05:11,531 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.27 vs. limit=2.0 2023-04-30 16:05:13,911 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=174373.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 16:05:33,379 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=174388.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 16:05:51,519 INFO [train.py:904] (1/8) Epoch 18, batch 1850, loss[loss=0.1563, simple_loss=0.245, pruned_loss=0.0338, over 16843.00 frames. ], tot_loss[loss=0.1743, simple_loss=0.2596, pruned_loss=0.04457, over 3316704.62 frames. ], batch size: 42, lr: 3.80e-03, grad_scale: 8.0 2023-04-30 16:06:18,880 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([5.5843, 5.9732, 5.7122, 5.7809, 5.3873, 5.3232, 5.3723, 6.1072], device='cuda:1'), covar=tensor([0.1285, 0.0921, 0.0876, 0.0783, 0.0822, 0.0660, 0.1175, 0.0831], device='cuda:1'), in_proj_covar=tensor([0.0650, 0.0801, 0.0643, 0.0594, 0.0504, 0.0505, 0.0665, 0.0617], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:1') 2023-04-30 16:06:32,177 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.14 vs. limit=2.0 2023-04-30 16:06:37,462 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.47 vs. limit=2.0 2023-04-30 16:06:51,029 INFO [optim.py:368] (1/8) 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 16:06:55,569 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 18, batch 1900, loss[loss=0.1677, simple_loss=0.2459, pruned_loss=0.04472, over 16842.00 frames. ], tot_loss[loss=0.1731, simple_loss=0.2586, pruned_loss=0.04377, over 3315053.04 frames. ], batch size: 116, lr: 3.80e-03, grad_scale: 8.0 2023-04-30 16:07:19,960 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.36 vs. limit=2.0 2023-04-30 16:07:45,485 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.88 vs. limit=2.0 2023-04-30 16:07:57,451 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=174495.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 16:08:06,700 INFO [train.py:904] (1/8) Epoch 18, batch 1950, loss[loss=0.1741, simple_loss=0.2578, pruned_loss=0.04521, over 16409.00 frames. ], tot_loss[loss=0.1726, simple_loss=0.2583, pruned_loss=0.04351, over 3300185.50 frames. ], batch size: 68, lr: 3.80e-03, grad_scale: 8.0 2023-04-30 16:08:28,781 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.1842, 3.9992, 4.4453, 2.2770, 4.6726, 4.6209, 3.4180, 3.6774], device='cuda:1'), covar=tensor([0.0683, 0.0220, 0.0212, 0.1041, 0.0066, 0.0179, 0.0384, 0.0355], device='cuda:1'), in_proj_covar=tensor([0.0149, 0.0107, 0.0094, 0.0140, 0.0076, 0.0122, 0.0126, 0.0131], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0004, 0.0004], device='cuda:1') 2023-04-30 16:08:56,375 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.9210, 4.8598, 4.7227, 4.2445, 4.8007, 1.9862, 4.5253, 4.6204], device='cuda:1'), covar=tensor([0.0116, 0.0100, 0.0207, 0.0374, 0.0110, 0.2464, 0.0149, 0.0206], device='cuda:1'), in_proj_covar=tensor([0.0155, 0.0145, 0.0191, 0.0173, 0.0166, 0.0202, 0.0181, 0.0171], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-30 16:09:05,465 INFO [optim.py:368] (1/8) 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 16:09:14,570 INFO [train.py:904] (1/8) Epoch 18, batch 2000, loss[loss=0.1922, simple_loss=0.2671, pruned_loss=0.05869, over 16750.00 frames. ], tot_loss[loss=0.1725, simple_loss=0.258, pruned_loss=0.04348, over 3308005.39 frames. ], batch size: 134, lr: 3.80e-03, grad_scale: 8.0 2023-04-30 16:09:16,216 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.9375, 4.5274, 2.9961, 2.3573, 2.6327, 2.3068, 4.6403, 3.6202], device='cuda:1'), covar=tensor([0.2696, 0.0581, 0.1930, 0.2847, 0.2999, 0.2261, 0.0378, 0.1306], device='cuda:1'), in_proj_covar=tensor([0.0317, 0.0263, 0.0299, 0.0300, 0.0288, 0.0246, 0.0285, 0.0326], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:1') 2023-04-30 16:09:52,934 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.7615, 4.6235, 4.8315, 4.9965, 5.1469, 4.6039, 5.0682, 5.1140], device='cuda:1'), covar=tensor([0.1629, 0.1111, 0.1387, 0.0677, 0.0546, 0.0925, 0.0683, 0.0596], device='cuda:1'), in_proj_covar=tensor([0.0624, 0.0773, 0.0907, 0.0784, 0.0581, 0.0621, 0.0635, 0.0740], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-30 16:10:23,407 INFO [train.py:904] (1/8) Epoch 18, batch 2050, loss[loss=0.1678, simple_loss=0.2644, pruned_loss=0.03556, over 16711.00 frames. ], tot_loss[loss=0.1732, simple_loss=0.2585, pruned_loss=0.04397, over 3303798.73 frames. ], batch size: 57, lr: 3.80e-03, grad_scale: 16.0 2023-04-30 16:10:35,819 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.32 vs. limit=2.0 2023-04-30 16:11:17,516 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=174639.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 16:11:26,649 INFO [optim.py:368] (1/8) 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 16:11:28,711 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=174647.0, num_to_drop=1, layers_to_drop={0} 2023-04-30 16:11:33,664 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([5.3368, 5.3021, 5.2054, 4.7117, 4.8271, 5.2548, 5.2113, 4.8892], device='cuda:1'), covar=tensor([0.0586, 0.0450, 0.0246, 0.0319, 0.1007, 0.0402, 0.0276, 0.0676], device='cuda:1'), in_proj_covar=tensor([0.0292, 0.0412, 0.0341, 0.0331, 0.0354, 0.0385, 0.0236, 0.0411], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-04-30 16:11:35,057 INFO [train.py:904] (1/8) Epoch 18, batch 2100, loss[loss=0.1529, simple_loss=0.2535, pruned_loss=0.02618, over 17151.00 frames. ], tot_loss[loss=0.174, simple_loss=0.2593, pruned_loss=0.04433, over 3308333.38 frames. ], batch size: 47, lr: 3.80e-03, grad_scale: 16.0 2023-04-30 16:12:02,044 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=174671.0, num_to_drop=1, layers_to_drop={1} 2023-04-30 16:12:05,716 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 18, batch 2150, loss[loss=0.1663, simple_loss=0.2507, pruned_loss=0.04101, over 15986.00 frames. ], tot_loss[loss=0.1745, simple_loss=0.2594, pruned_loss=0.04478, over 3310710.94 frames. ], batch size: 35, lr: 3.80e-03, grad_scale: 8.0 2023-04-30 16:12:50,327 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=174705.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 16:12:52,024 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.53 vs. limit=2.0 2023-04-30 16:13:10,638 INFO [zipformer.py:625] (1/8) 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] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=174721.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 16:13:44,796 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.39 vs. limit=2.0 2023-04-30 16:13:45,406 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=174744.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 16:13:49,506 INFO [optim.py:368] (1/8) 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 16:13:56,940 INFO [train.py:904] (1/8) Epoch 18, batch 2200, loss[loss=0.1984, simple_loss=0.2843, pruned_loss=0.05626, over 15578.00 frames. ], tot_loss[loss=0.1751, simple_loss=0.26, pruned_loss=0.04511, over 3312017.16 frames. ], batch size: 190, lr: 3.80e-03, grad_scale: 8.0 2023-04-30 16:14:16,324 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=174766.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 16:14:56,021 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 18, batch 2250, loss[loss=0.1726, simple_loss=0.2685, pruned_loss=0.0383, over 17066.00 frames. ], tot_loss[loss=0.1757, simple_loss=0.2606, pruned_loss=0.04537, over 3315974.00 frames. ], batch size: 55, lr: 3.80e-03, grad_scale: 8.0 2023-04-30 16:15:10,957 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.8460, 4.7895, 4.6768, 4.2074, 4.7847, 1.8175, 4.4932, 4.4880], device='cuda:1'), covar=tensor([0.0148, 0.0114, 0.0229, 0.0318, 0.0110, 0.2790, 0.0183, 0.0203], device='cuda:1'), in_proj_covar=tensor([0.0155, 0.0146, 0.0192, 0.0173, 0.0167, 0.0202, 0.0181, 0.0172], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-30 16:15:25,038 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.7901, 4.7261, 4.7145, 4.3998, 4.3862, 4.7513, 4.5960, 4.4711], device='cuda:1'), covar=tensor([0.0777, 0.0830, 0.0286, 0.0282, 0.0829, 0.0566, 0.0438, 0.0671], device='cuda:1'), in_proj_covar=tensor([0.0291, 0.0411, 0.0340, 0.0331, 0.0354, 0.0383, 0.0236, 0.0410], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002], device='cuda:1') 2023-04-30 16:16:02,772 INFO [zipformer.py:625] (1/8) 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] (1/8) 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] (1/8) Epoch 18, batch 2300, loss[loss=0.2485, simple_loss=0.3197, pruned_loss=0.08868, over 11843.00 frames. ], tot_loss[loss=0.1747, simple_loss=0.2599, pruned_loss=0.04475, over 3319443.35 frames. ], batch size: 247, lr: 3.80e-03, grad_scale: 8.0 2023-04-30 16:17:22,134 INFO [train.py:904] (1/8) Epoch 18, batch 2350, loss[loss=0.1774, simple_loss=0.2556, pruned_loss=0.04958, over 16810.00 frames. ], tot_loss[loss=0.176, simple_loss=0.261, pruned_loss=0.04556, over 3319619.95 frames. ], batch size: 90, lr: 3.80e-03, grad_scale: 8.0 2023-04-30 16:18:14,617 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=174939.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 16:18:23,297 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.91 vs. limit=2.0 2023-04-30 16:18:25,281 INFO [optim.py:368] (1/8) 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] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=174947.0, num_to_drop=1, layers_to_drop={2} 2023-04-30 16:18:32,031 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.6349, 4.5842, 4.4871, 3.9519, 4.5803, 1.6579, 4.2833, 4.1768], device='cuda:1'), covar=tensor([0.0104, 0.0102, 0.0209, 0.0337, 0.0104, 0.2839, 0.0161, 0.0231], device='cuda:1'), in_proj_covar=tensor([0.0157, 0.0147, 0.0194, 0.0175, 0.0169, 0.0204, 0.0183, 0.0174], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-30 16:18:32,764 INFO [train.py:904] (1/8) Epoch 18, batch 2400, loss[loss=0.1897, simple_loss=0.263, pruned_loss=0.05821, over 16835.00 frames. ], tot_loss[loss=0.1772, simple_loss=0.2623, pruned_loss=0.04604, over 3306332.87 frames. ], batch size: 96, lr: 3.79e-03, grad_scale: 8.0 2023-04-30 16:19:20,827 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=174987.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 16:19:33,334 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 18, batch 2450, loss[loss=0.1779, simple_loss=0.2661, pruned_loss=0.04483, over 17043.00 frames. ], tot_loss[loss=0.1764, simple_loss=0.2625, pruned_loss=0.04517, over 3312463.29 frames. ], batch size: 55, lr: 3.79e-03, grad_scale: 8.0 2023-04-30 16:20:12,723 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.5725, 1.7091, 2.2221, 2.5117, 2.5613, 2.5163, 1.8896, 2.7528], device='cuda:1'), covar=tensor([0.0192, 0.0491, 0.0319, 0.0262, 0.0284, 0.0293, 0.0519, 0.0142], device='cuda:1'), in_proj_covar=tensor([0.0184, 0.0191, 0.0177, 0.0182, 0.0190, 0.0149, 0.0194, 0.0143], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-30 16:20:24,395 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.9468, 4.8590, 4.8239, 4.4762, 4.4673, 4.8764, 4.7087, 4.6084], device='cuda:1'), covar=tensor([0.0683, 0.0758, 0.0333, 0.0362, 0.0988, 0.0572, 0.0440, 0.0725], device='cuda:1'), in_proj_covar=tensor([0.0291, 0.0413, 0.0341, 0.0333, 0.0355, 0.0385, 0.0236, 0.0411], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-04-30 16:20:40,488 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=175044.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 16:20:43,743 INFO [optim.py:368] (1/8) Clipping_scale=2.0, grad-norm quartiles 1.516e+02 2.297e+02 2.808e+02 3.204e+02 6.791e+02, threshold=5.617e+02, percent-clipped=2.0 2023-04-30 16:20:51,500 INFO [train.py:904] (1/8) Epoch 18, batch 2500, loss[loss=0.1774, simple_loss=0.268, pruned_loss=0.04344, over 17078.00 frames. ], tot_loss[loss=0.1761, simple_loss=0.2624, pruned_loss=0.04495, over 3322163.67 frames. ], batch size: 53, lr: 3.79e-03, grad_scale: 8.0 2023-04-30 16:21:04,688 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=175061.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 16:21:46,586 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 18, batch 2550, loss[loss=0.2132, simple_loss=0.2926, pruned_loss=0.06688, over 11996.00 frames. ], tot_loss[loss=0.1772, simple_loss=0.2635, pruned_loss=0.04543, over 3315553.41 frames. ], batch size: 247, lr: 3.79e-03, grad_scale: 8.0 2023-04-30 16:22:30,200 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.8015, 2.4209, 2.5168, 4.7045, 2.4143, 2.8698, 2.4843, 2.7302], device='cuda:1'), covar=tensor([0.1092, 0.3572, 0.2845, 0.0392, 0.3968, 0.2424, 0.3375, 0.3393], device='cuda:1'), in_proj_covar=tensor([0.0393, 0.0432, 0.0361, 0.0327, 0.0432, 0.0500, 0.0401, 0.0505], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-30 16:22:34,735 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.24 vs. limit=2.0 2023-04-30 16:22:51,947 INFO [zipformer.py:625] (1/8) 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] (1/8) 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] (1/8) Epoch 18, batch 2600, loss[loss=0.1684, simple_loss=0.2514, pruned_loss=0.04267, over 16785.00 frames. ], tot_loss[loss=0.1755, simple_loss=0.2625, pruned_loss=0.04425, over 3325326.78 frames. ], batch size: 83, lr: 3.79e-03, grad_scale: 8.0 2023-04-30 16:24:17,450 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 18, batch 2650, loss[loss=0.1953, simple_loss=0.2699, pruned_loss=0.06036, over 16710.00 frames. ], tot_loss[loss=0.1758, simple_loss=0.2632, pruned_loss=0.04418, over 3332526.47 frames. ], batch size: 134, lr: 3.79e-03, grad_scale: 8.0 2023-04-30 16:24:27,536 INFO [zipformer.py:625] (1/8) 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] (1/8) 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] (1/8) Epoch 18, batch 2700, loss[loss=0.1617, simple_loss=0.245, pruned_loss=0.03919, over 16777.00 frames. ], tot_loss[loss=0.1761, simple_loss=0.2639, pruned_loss=0.0441, over 3317584.19 frames. ], batch size: 102, lr: 3.79e-03, grad_scale: 4.0 2023-04-30 16:25:49,504 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=175268.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 16:26:36,865 INFO [train.py:904] (1/8) Epoch 18, batch 2750, loss[loss=0.1441, simple_loss=0.234, pruned_loss=0.02706, over 16939.00 frames. ], tot_loss[loss=0.1752, simple_loss=0.2635, pruned_loss=0.04347, over 3314765.46 frames. ], batch size: 41, lr: 3.79e-03, grad_scale: 4.0 2023-04-30 16:27:40,024 INFO [optim.py:368] (1/8) Clipping_scale=2.0, grad-norm quartiles 1.507e+02 2.063e+02 2.386e+02 2.912e+02 4.096e+02, threshold=4.773e+02, percent-clipped=0.0 2023-04-30 16:27:45,205 INFO [train.py:904] (1/8) Epoch 18, batch 2800, loss[loss=0.1879, simple_loss=0.2612, pruned_loss=0.05732, over 16860.00 frames. ], tot_loss[loss=0.1757, simple_loss=0.2639, pruned_loss=0.04372, over 3310474.61 frames. ], batch size: 116, lr: 3.79e-03, grad_scale: 8.0 2023-04-30 16:27:58,785 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=175361.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 16:28:11,812 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=175370.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 16:28:44,384 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([5.3621, 5.2527, 5.1512, 4.6818, 4.7415, 5.1959, 5.1181, 4.8172], device='cuda:1'), covar=tensor([0.0549, 0.0582, 0.0316, 0.0339, 0.1181, 0.0538, 0.0347, 0.0736], device='cuda:1'), in_proj_covar=tensor([0.0294, 0.0416, 0.0344, 0.0336, 0.0359, 0.0389, 0.0238, 0.0415], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-04-30 16:28:45,631 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=175394.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 16:28:56,405 INFO [train.py:904] (1/8) Epoch 18, batch 2850, loss[loss=0.1473, simple_loss=0.2516, pruned_loss=0.02149, over 17271.00 frames. ], tot_loss[loss=0.1746, simple_loss=0.2625, pruned_loss=0.04338, over 3308099.39 frames. ], batch size: 52, lr: 3.79e-03, grad_scale: 8.0 2023-04-30 16:29:05,423 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=175409.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 16:29:05,649 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.9668, 2.0954, 2.5135, 2.9083, 2.7049, 3.3828, 2.3207, 3.3608], device='cuda:1'), covar=tensor([0.0217, 0.0449, 0.0307, 0.0287, 0.0294, 0.0145, 0.0426, 0.0127], device='cuda:1'), in_proj_covar=tensor([0.0184, 0.0191, 0.0177, 0.0182, 0.0190, 0.0149, 0.0194, 0.0143], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-30 16:29:35,276 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=175431.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 16:29:58,776 INFO [optim.py:368] (1/8) 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] (1/8) Epoch 18, batch 2900, loss[loss=0.1634, simple_loss=0.2403, pruned_loss=0.04324, over 17003.00 frames. ], tot_loss[loss=0.1737, simple_loss=0.2608, pruned_loss=0.04329, over 3315191.46 frames. ], batch size: 41, lr: 3.79e-03, grad_scale: 8.0 2023-04-30 16:30:08,958 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=175455.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 16:30:45,720 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=175481.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 16:30:50,793 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.0338, 3.1296, 2.6694, 2.9459, 3.3657, 3.1185, 3.7072, 3.5575], device='cuda:1'), covar=tensor([0.0104, 0.0293, 0.0432, 0.0340, 0.0224, 0.0329, 0.0216, 0.0227], device='cuda:1'), in_proj_covar=tensor([0.0202, 0.0235, 0.0225, 0.0225, 0.0236, 0.0234, 0.0240, 0.0228], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-30 16:31:04,480 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=175495.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 16:31:13,713 INFO [train.py:904] (1/8) Epoch 18, batch 2950, loss[loss=0.2107, simple_loss=0.2959, pruned_loss=0.06279, over 16772.00 frames. ], tot_loss[loss=0.1742, simple_loss=0.2608, pruned_loss=0.04379, over 3316350.77 frames. ], batch size: 62, lr: 3.79e-03, grad_scale: 8.0 2023-04-30 16:31:50,287 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.0049, 2.1471, 2.5953, 2.9655, 2.8481, 3.4480, 2.3566, 3.4310], device='cuda:1'), covar=tensor([0.0232, 0.0471, 0.0324, 0.0315, 0.0320, 0.0185, 0.0459, 0.0167], device='cuda:1'), in_proj_covar=tensor([0.0185, 0.0192, 0.0179, 0.0185, 0.0192, 0.0150, 0.0195, 0.0145], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-30 16:32:09,651 INFO [zipformer.py:625] (1/8) 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] (1/8) Clipping_scale=2.0, grad-norm quartiles 1.593e+02 2.280e+02 2.587e+02 3.034e+02 5.964e+02, threshold=5.174e+02, percent-clipped=1.0 2023-04-30 16:32:23,317 INFO [train.py:904] (1/8) Epoch 18, batch 3000, loss[loss=0.1909, simple_loss=0.2815, pruned_loss=0.05018, over 16795.00 frames. ], tot_loss[loss=0.1745, simple_loss=0.2607, pruned_loss=0.04415, over 3321757.30 frames. ], batch size: 102, lr: 3.79e-03, grad_scale: 8.0 2023-04-30 16:32:23,317 INFO [train.py:929] (1/8) Computing validation loss 2023-04-30 16:32:32,124 INFO [train.py:938] (1/8) Epoch 18, validation: loss=0.1359, simple_loss=0.2415, pruned_loss=0.01516, over 944034.00 frames. 2023-04-30 16:32:32,125 INFO [train.py:939] (1/8) Maximum memory allocated so far is 17857MB 2023-04-30 16:32:48,742 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=175563.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 16:32:49,037 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.7898, 4.3577, 3.1836, 2.2637, 2.7549, 2.6195, 4.6922, 3.6927], device='cuda:1'), covar=tensor([0.2757, 0.0564, 0.1696, 0.2876, 0.2806, 0.1888, 0.0334, 0.1251], device='cuda:1'), in_proj_covar=tensor([0.0316, 0.0263, 0.0299, 0.0300, 0.0289, 0.0245, 0.0286, 0.0326], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:1') 2023-04-30 16:33:25,223 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.2340, 4.1979, 4.3571, 4.1636, 4.2573, 4.7935, 4.3516, 4.0436], device='cuda:1'), covar=tensor([0.1903, 0.2175, 0.2175, 0.2474, 0.2906, 0.1284, 0.1665, 0.2882], device='cuda:1'), in_proj_covar=tensor([0.0404, 0.0591, 0.0648, 0.0491, 0.0662, 0.0673, 0.0510, 0.0659], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-04-30 16:33:40,721 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=4.90 vs. limit=5.0 2023-04-30 16:33:42,215 INFO [train.py:904] (1/8) Epoch 18, batch 3050, loss[loss=0.1932, simple_loss=0.2634, pruned_loss=0.06149, over 16894.00 frames. ], tot_loss[loss=0.1754, simple_loss=0.2611, pruned_loss=0.04487, over 3324779.87 frames. ], batch size: 109, lr: 3.79e-03, grad_scale: 8.0 2023-04-30 16:34:02,907 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.92 vs. limit=2.0 2023-04-30 16:34:30,957 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([5.0209, 5.4990, 5.5985, 5.3955, 5.3365, 6.0245, 5.5259, 5.2618], device='cuda:1'), covar=tensor([0.0971, 0.2020, 0.2274, 0.1848, 0.2809, 0.0998, 0.1349, 0.2257], device='cuda:1'), in_proj_covar=tensor([0.0401, 0.0587, 0.0643, 0.0488, 0.0658, 0.0669, 0.0506, 0.0656], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-04-30 16:34:46,186 INFO [optim.py:368] (1/8) 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 16:34:51,895 INFO [train.py:904] (1/8) Epoch 18, batch 3100, loss[loss=0.1574, simple_loss=0.2451, pruned_loss=0.03486, over 17210.00 frames. ], tot_loss[loss=0.1743, simple_loss=0.2601, pruned_loss=0.0443, over 3323362.17 frames. ], batch size: 44, lr: 3.79e-03, grad_scale: 8.0 2023-04-30 16:35:06,978 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.30 vs. limit=2.0 2023-04-30 16:36:01,368 INFO [train.py:904] (1/8) Epoch 18, batch 3150, loss[loss=0.1813, simple_loss=0.2553, pruned_loss=0.0537, over 16910.00 frames. ], tot_loss[loss=0.174, simple_loss=0.2599, pruned_loss=0.04407, over 3321932.38 frames. ], batch size: 116, lr: 3.79e-03, grad_scale: 8.0 2023-04-30 16:36:13,016 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=175710.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 16:36:36,340 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=175726.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 16:37:05,953 INFO [optim.py:368] (1/8) Clipping_scale=2.0, grad-norm quartiles 1.459e+02 2.210e+02 2.531e+02 3.035e+02 7.463e+02, threshold=5.061e+02, percent-clipped=4.0 2023-04-30 16:37:06,641 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=2.82 vs. limit=5.0 2023-04-30 16:37:08,589 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 18, batch 3200, loss[loss=0.1793, simple_loss=0.2611, pruned_loss=0.0487, over 16333.00 frames. ], tot_loss[loss=0.1738, simple_loss=0.2589, pruned_loss=0.04436, over 3327721.70 frames. ], batch size: 165, lr: 3.79e-03, grad_scale: 8.0 2023-04-30 16:37:37,212 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=175771.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 16:38:11,031 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=175795.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 16:38:16,132 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.77 vs. limit=2.0 2023-04-30 16:38:20,687 INFO [train.py:904] (1/8) Epoch 18, batch 3250, loss[loss=0.194, simple_loss=0.2746, pruned_loss=0.05672, over 16287.00 frames. ], tot_loss[loss=0.1737, simple_loss=0.259, pruned_loss=0.04422, over 3319650.13 frames. ], batch size: 165, lr: 3.79e-03, grad_scale: 8.0 2023-04-30 16:38:36,726 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([1.7324, 1.7817, 1.5464, 1.5373, 1.9109, 1.6414, 1.7150, 1.9541], device='cuda:1'), covar=tensor([0.0182, 0.0283, 0.0373, 0.0339, 0.0180, 0.0253, 0.0199, 0.0187], device='cuda:1'), in_proj_covar=tensor([0.0201, 0.0235, 0.0225, 0.0224, 0.0235, 0.0233, 0.0240, 0.0228], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-30 16:38:52,188 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.7349, 2.6377, 2.3423, 2.5038, 3.0047, 2.7730, 3.3786, 3.2658], device='cuda:1'), covar=tensor([0.0135, 0.0406, 0.0504, 0.0440, 0.0278, 0.0366, 0.0248, 0.0249], device='cuda:1'), in_proj_covar=tensor([0.0202, 0.0235, 0.0225, 0.0224, 0.0235, 0.0233, 0.0240, 0.0228], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-30 16:39:09,518 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=175837.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 16:39:17,732 INFO [zipformer.py:625] (1/8) 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] (1/8) 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 16:39:29,879 INFO [train.py:904] (1/8) Epoch 18, batch 3300, loss[loss=0.1838, simple_loss=0.2704, pruned_loss=0.04858, over 16490.00 frames. ], tot_loss[loss=0.1751, simple_loss=0.2609, pruned_loss=0.0447, over 3316080.99 frames. ], batch size: 75, lr: 3.79e-03, grad_scale: 8.0 2023-04-30 16:39:45,166 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=175863.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 16:39:49,299 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([5.8661, 6.2192, 5.9104, 5.9694, 5.5840, 5.5712, 5.6420, 6.2853], device='cuda:1'), covar=tensor([0.1167, 0.0769, 0.0934, 0.0782, 0.0834, 0.0580, 0.1050, 0.0841], device='cuda:1'), in_proj_covar=tensor([0.0661, 0.0816, 0.0657, 0.0604, 0.0512, 0.0514, 0.0676, 0.0629], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:1') 2023-04-30 16:40:10,979 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 18, batch 3350, loss[loss=0.171, simple_loss=0.2668, pruned_loss=0.03753, over 16761.00 frames. ], tot_loss[loss=0.1747, simple_loss=0.2605, pruned_loss=0.04445, over 3320870.74 frames. ], batch size: 57, lr: 3.78e-03, grad_scale: 8.0 2023-04-30 16:40:50,862 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=175911.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 16:41:34,509 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=175942.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 16:41:38,089 INFO [zipformer.py:625] (1/8) 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] (1/8) 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] (1/8) Epoch 18, batch 3400, loss[loss=0.1802, simple_loss=0.2582, pruned_loss=0.05111, over 16823.00 frames. ], tot_loss[loss=0.1744, simple_loss=0.2602, pruned_loss=0.04429, over 3307997.64 frames. ], batch size: 102, lr: 3.78e-03, grad_scale: 8.0 2023-04-30 16:42:51,938 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.38 vs. limit=2.0 2023-04-30 16:43:01,858 INFO [train.py:904] (1/8) Epoch 18, batch 3450, loss[loss=0.1897, simple_loss=0.2709, pruned_loss=0.05421, over 15592.00 frames. ], tot_loss[loss=0.1732, simple_loss=0.2593, pruned_loss=0.04354, over 3310074.49 frames. ], batch size: 191, lr: 3.78e-03, grad_scale: 8.0 2023-04-30 16:43:08,055 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=176006.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 16:43:12,473 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.28 vs. limit=2.0 2023-04-30 16:43:35,404 INFO [zipformer.py:625] (1/8) 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] (1/8) 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 16:44:09,488 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=176050.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 16:44:11,309 INFO [train.py:904] (1/8) Epoch 18, batch 3500, loss[loss=0.1559, simple_loss=0.2369, pruned_loss=0.03748, over 16723.00 frames. ], tot_loss[loss=0.1724, simple_loss=0.2581, pruned_loss=0.04333, over 3310777.07 frames. ], batch size: 39, lr: 3.78e-03, grad_scale: 8.0 2023-04-30 16:44:31,121 INFO [zipformer.py:625] (1/8) 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] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=176074.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 16:45:14,939 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=176098.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 16:45:19,384 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.2239, 4.5742, 4.5238, 3.4302, 3.7500, 4.5002, 4.0098, 2.8161], device='cuda:1'), covar=tensor([0.0344, 0.0051, 0.0037, 0.0276, 0.0113, 0.0080, 0.0079, 0.0366], device='cuda:1'), in_proj_covar=tensor([0.0135, 0.0080, 0.0080, 0.0133, 0.0094, 0.0104, 0.0091, 0.0126], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-04-30 16:45:20,521 INFO [train.py:904] (1/8) Epoch 18, batch 3550, loss[loss=0.1623, simple_loss=0.2467, pruned_loss=0.03896, over 17230.00 frames. ], tot_loss[loss=0.171, simple_loss=0.257, pruned_loss=0.0425, over 3319816.78 frames. ], batch size: 43, lr: 3.78e-03, grad_scale: 8.0 2023-04-30 16:46:10,539 INFO [zipformer.py:625] (1/8) 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] (1/8) Clipping_scale=2.0, grad-norm quartiles 1.470e+02 2.026e+02 2.447e+02 2.938e+02 5.668e+02, threshold=4.895e+02, percent-clipped=1.0 2023-04-30 16:46:32,322 INFO [train.py:904] (1/8) Epoch 18, batch 3600, loss[loss=0.1623, simple_loss=0.2613, pruned_loss=0.03164, over 17125.00 frames. ], tot_loss[loss=0.1701, simple_loss=0.2557, pruned_loss=0.04227, over 3317967.25 frames. ], batch size: 49, lr: 3.78e-03, grad_scale: 8.0 2023-04-30 16:47:18,727 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=176185.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 16:47:37,733 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.87 vs. limit=2.0 2023-04-30 16:47:43,326 INFO [train.py:904] (1/8) Epoch 18, batch 3650, loss[loss=0.1668, simple_loss=0.2487, pruned_loss=0.04248, over 15793.00 frames. ], tot_loss[loss=0.1705, simple_loss=0.255, pruned_loss=0.04306, over 3307796.81 frames. ], batch size: 35, lr: 3.78e-03, grad_scale: 8.0 2023-04-30 16:48:30,073 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([1.7592, 2.6677, 2.6076, 4.0369, 3.4231, 4.0437, 1.6056, 2.8144], device='cuda:1'), covar=tensor([0.1328, 0.0706, 0.1088, 0.0169, 0.0148, 0.0365, 0.1523, 0.0820], device='cuda:1'), in_proj_covar=tensor([0.0162, 0.0169, 0.0189, 0.0182, 0.0203, 0.0215, 0.0194, 0.0188], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-04-30 16:48:35,859 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=176237.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 16:48:43,974 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=176242.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 16:48:47,982 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.7849, 3.0324, 3.1332, 2.1304, 2.7094, 2.2055, 3.3666, 3.3476], device='cuda:1'), covar=tensor([0.0253, 0.0903, 0.0632, 0.1825, 0.0867, 0.1041, 0.0577, 0.0861], device='cuda:1'), in_proj_covar=tensor([0.0156, 0.0162, 0.0166, 0.0152, 0.0143, 0.0128, 0.0144, 0.0172], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:1') 2023-04-30 16:48:53,037 INFO [optim.py:368] (1/8) 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] (1/8) Epoch 18, batch 3700, loss[loss=0.1781, simple_loss=0.2514, pruned_loss=0.05233, over 11815.00 frames. ], tot_loss[loss=0.1722, simple_loss=0.2545, pruned_loss=0.04492, over 3277366.96 frames. ], batch size: 247, lr: 3.78e-03, grad_scale: 8.0 2023-04-30 16:49:06,428 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=176257.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 16:50:10,838 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=176301.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 16:50:11,691 INFO [train.py:904] (1/8) Epoch 18, batch 3750, loss[loss=0.1939, simple_loss=0.2651, pruned_loss=0.0614, over 16303.00 frames. ], tot_loss[loss=0.1739, simple_loss=0.2552, pruned_loss=0.04633, over 3262675.41 frames. ], batch size: 68, lr: 3.78e-03, grad_scale: 8.0 2023-04-30 16:50:13,739 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=176303.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 16:50:35,168 INFO [zipformer.py:625] (1/8) 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] (1/8) 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 16:51:23,241 INFO [train.py:904] (1/8) Epoch 18, batch 3800, loss[loss=0.1705, simple_loss=0.2498, pruned_loss=0.0456, over 16716.00 frames. ], tot_loss[loss=0.1759, simple_loss=0.2567, pruned_loss=0.04758, over 3249941.82 frames. ], batch size: 83, lr: 3.78e-03, grad_scale: 8.0 2023-04-30 16:51:44,212 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=176366.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 16:52:35,499 INFO [train.py:904] (1/8) Epoch 18, batch 3850, loss[loss=0.1884, simple_loss=0.2697, pruned_loss=0.05352, over 16129.00 frames. ], tot_loss[loss=0.1759, simple_loss=0.2562, pruned_loss=0.04784, over 3261297.29 frames. ], batch size: 35, lr: 3.78e-03, grad_scale: 8.0 2023-04-30 16:52:53,511 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=176414.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 16:52:54,812 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([5.0254, 4.9463, 5.0633, 5.2746, 5.4922, 4.8632, 5.4061, 5.4812], device='cuda:1'), covar=tensor([0.1804, 0.1109, 0.1671, 0.0767, 0.0460, 0.0798, 0.0572, 0.0509], device='cuda:1'), in_proj_covar=tensor([0.0635, 0.0789, 0.0926, 0.0801, 0.0590, 0.0633, 0.0647, 0.0752], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-30 16:53:42,933 INFO [optim.py:368] (1/8) 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 16:53:49,441 INFO [train.py:904] (1/8) Epoch 18, batch 3900, loss[loss=0.1879, simple_loss=0.2608, pruned_loss=0.05754, over 16810.00 frames. ], tot_loss[loss=0.1754, simple_loss=0.2551, pruned_loss=0.0478, over 3265661.68 frames. ], batch size: 124, lr: 3.78e-03, grad_scale: 8.0 2023-04-30 16:54:10,504 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=176466.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 16:54:13,852 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.30 vs. limit=2.0 2023-04-30 16:55:00,905 INFO [train.py:904] (1/8) Epoch 18, batch 3950, loss[loss=0.1571, simple_loss=0.2488, pruned_loss=0.03263, over 16327.00 frames. ], tot_loss[loss=0.1751, simple_loss=0.2544, pruned_loss=0.04794, over 3278317.27 frames. ], batch size: 35, lr: 3.78e-03, grad_scale: 8.0 2023-04-30 16:55:10,778 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([1.8669, 3.0976, 2.7456, 4.4647, 3.6833, 4.1695, 1.8873, 3.0939], device='cuda:1'), covar=tensor([0.1273, 0.0639, 0.1009, 0.0182, 0.0224, 0.0360, 0.1342, 0.0771], device='cuda:1'), in_proj_covar=tensor([0.0162, 0.0169, 0.0189, 0.0182, 0.0204, 0.0214, 0.0194, 0.0188], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-04-30 16:55:35,236 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=176527.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 16:55:50,447 INFO [zipformer.py:625] (1/8) 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] (1/8) 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 16:56:08,097 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.9199, 2.4881, 1.9275, 2.2736, 2.8394, 2.6191, 2.8969, 3.0181], device='cuda:1'), covar=tensor([0.0212, 0.0373, 0.0595, 0.0484, 0.0227, 0.0333, 0.0275, 0.0251], device='cuda:1'), in_proj_covar=tensor([0.0201, 0.0233, 0.0223, 0.0222, 0.0233, 0.0231, 0.0237, 0.0226], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-30 16:56:12,321 INFO [train.py:904] (1/8) Epoch 18, batch 4000, loss[loss=0.1866, simple_loss=0.2609, pruned_loss=0.05619, over 16730.00 frames. ], tot_loss[loss=0.1759, simple_loss=0.2549, pruned_loss=0.04847, over 3283197.38 frames. ], batch size: 134, lr: 3.78e-03, grad_scale: 8.0 2023-04-30 16:57:00,438 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=176585.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 16:57:18,436 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=176598.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 16:57:22,928 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=176601.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 16:57:24,371 INFO [train.py:904] (1/8) Epoch 18, batch 4050, loss[loss=0.1554, simple_loss=0.2456, pruned_loss=0.0326, over 16826.00 frames. ], tot_loss[loss=0.1753, simple_loss=0.2556, pruned_loss=0.04749, over 3286819.46 frames. ], batch size: 96, lr: 3.78e-03, grad_scale: 8.0 2023-04-30 16:57:39,763 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=176613.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 16:57:57,991 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=176625.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 16:58:09,094 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.6092, 4.7123, 5.0320, 4.9960, 5.0153, 4.6779, 4.6575, 4.4553], device='cuda:1'), covar=tensor([0.0294, 0.0457, 0.0348, 0.0421, 0.0396, 0.0373, 0.0823, 0.0452], device='cuda:1'), in_proj_covar=tensor([0.0399, 0.0436, 0.0425, 0.0397, 0.0469, 0.0445, 0.0541, 0.0354], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:1') 2023-04-30 16:58:30,485 INFO [optim.py:368] (1/8) 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 16:58:33,595 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 18, batch 4100, loss[loss=0.2031, simple_loss=0.2854, pruned_loss=0.06045, over 17213.00 frames. ], tot_loss[loss=0.176, simple_loss=0.2576, pruned_loss=0.04722, over 3281699.33 frames. ], batch size: 46, lr: 3.78e-03, grad_scale: 8.0 2023-04-30 16:58:57,233 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.0658, 4.0496, 3.9752, 3.2746, 3.9425, 1.8229, 3.7445, 3.4263], device='cuda:1'), covar=tensor([0.0098, 0.0081, 0.0166, 0.0257, 0.0072, 0.2678, 0.0113, 0.0231], device='cuda:1'), in_proj_covar=tensor([0.0158, 0.0148, 0.0196, 0.0178, 0.0170, 0.0204, 0.0186, 0.0175], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-30 16:58:57,257 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=176665.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 16:58:59,742 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([1.9417, 3.2444, 2.9952, 5.1953, 4.2647, 4.4502, 2.2244, 3.0348], device='cuda:1'), covar=tensor([0.1285, 0.0686, 0.1144, 0.0123, 0.0338, 0.0361, 0.1326, 0.0930], device='cuda:1'), in_proj_covar=tensor([0.0163, 0.0169, 0.0190, 0.0182, 0.0205, 0.0215, 0.0195, 0.0189], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-04-30 16:59:28,840 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=176686.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 16:59:51,167 INFO [train.py:904] (1/8) Epoch 18, batch 4150, loss[loss=0.2225, simple_loss=0.2995, pruned_loss=0.07274, over 11478.00 frames. ], tot_loss[loss=0.1816, simple_loss=0.2639, pruned_loss=0.04964, over 3238791.41 frames. ], batch size: 247, lr: 3.78e-03, grad_scale: 8.0 2023-04-30 17:00:27,904 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=176726.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 17:00:47,300 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([5.5204, 5.5210, 5.2465, 4.6547, 5.5299, 2.1452, 5.2337, 5.0477], device='cuda:1'), covar=tensor([0.0051, 0.0038, 0.0145, 0.0286, 0.0047, 0.2584, 0.0080, 0.0173], device='cuda:1'), in_proj_covar=tensor([0.0158, 0.0148, 0.0195, 0.0178, 0.0169, 0.0204, 0.0185, 0.0175], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-30 17:01:00,609 INFO [optim.py:368] (1/8) 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 17:01:06,253 INFO [train.py:904] (1/8) Epoch 18, batch 4200, loss[loss=0.2054, simple_loss=0.3015, pruned_loss=0.05469, over 16773.00 frames. ], tot_loss[loss=0.1864, simple_loss=0.2706, pruned_loss=0.05109, over 3212384.96 frames. ], batch size: 89, lr: 3.78e-03, grad_scale: 8.0 2023-04-30 17:02:21,462 INFO [train.py:904] (1/8) Epoch 18, batch 4250, loss[loss=0.1895, simple_loss=0.2841, pruned_loss=0.04745, over 16581.00 frames. ], tot_loss[loss=0.1886, simple_loss=0.2744, pruned_loss=0.0514, over 3201962.36 frames. ], batch size: 62, lr: 3.78e-03, grad_scale: 8.0 2023-04-30 17:02:51,593 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=176822.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 17:03:30,837 INFO [optim.py:368] (1/8) 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] (1/8) Epoch 18, batch 4300, loss[loss=0.2006, simple_loss=0.293, pruned_loss=0.05411, over 11596.00 frames. ], tot_loss[loss=0.1886, simple_loss=0.2758, pruned_loss=0.05068, over 3195175.19 frames. ], batch size: 246, lr: 3.77e-03, grad_scale: 8.0 2023-04-30 17:03:37,825 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=176853.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 17:04:41,124 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.6733, 3.5769, 4.0442, 1.9292, 4.4105, 4.3618, 3.1215, 3.1528], device='cuda:1'), covar=tensor([0.0804, 0.0257, 0.0194, 0.1271, 0.0048, 0.0098, 0.0415, 0.0465], device='cuda:1'), in_proj_covar=tensor([0.0147, 0.0107, 0.0094, 0.0139, 0.0076, 0.0122, 0.0126, 0.0129], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0004, 0.0003], device='cuda:1') 2023-04-30 17:04:44,697 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=176898.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 17:04:49,705 INFO [train.py:904] (1/8) Epoch 18, batch 4350, loss[loss=0.2249, simple_loss=0.313, pruned_loss=0.06845, over 16634.00 frames. ], tot_loss[loss=0.1924, simple_loss=0.28, pruned_loss=0.05242, over 3195219.12 frames. ], batch size: 62, lr: 3.77e-03, grad_scale: 8.0 2023-04-30 17:05:06,467 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=176913.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 17:05:07,724 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=176914.0, num_to_drop=1, layers_to_drop={1} 2023-04-30 17:05:48,296 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.8820, 2.0191, 2.3032, 3.1380, 2.1176, 2.2500, 2.2838, 2.1674], device='cuda:1'), covar=tensor([0.1194, 0.3134, 0.2234, 0.0660, 0.3735, 0.2186, 0.2713, 0.3145], device='cuda:1'), in_proj_covar=tensor([0.0390, 0.0435, 0.0358, 0.0326, 0.0429, 0.0502, 0.0403, 0.0507], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-30 17:05:55,204 INFO [zipformer.py:625] (1/8) 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] (1/8) 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 17:06:03,298 INFO [train.py:904] (1/8) Epoch 18, batch 4400, loss[loss=0.2025, simple_loss=0.2901, pruned_loss=0.05749, over 16472.00 frames. ], tot_loss[loss=0.1941, simple_loss=0.2817, pruned_loss=0.05321, over 3194938.98 frames. ], batch size: 68, lr: 3.77e-03, grad_scale: 8.0 2023-04-30 17:06:06,969 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.0863, 3.1370, 3.1662, 5.1793, 4.2280, 4.3285, 1.9896, 3.4781], device='cuda:1'), covar=tensor([0.1153, 0.0711, 0.0956, 0.0093, 0.0305, 0.0353, 0.1377, 0.0682], device='cuda:1'), in_proj_covar=tensor([0.0163, 0.0169, 0.0189, 0.0181, 0.0204, 0.0213, 0.0194, 0.0189], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-04-30 17:06:17,359 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=176961.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 17:06:45,568 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=176981.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 17:07:03,415 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.27 vs. limit=2.0 2023-04-30 17:07:16,202 INFO [train.py:904] (1/8) Epoch 18, batch 4450, loss[loss=0.2076, simple_loss=0.2992, pruned_loss=0.05797, over 17008.00 frames. ], tot_loss[loss=0.1971, simple_loss=0.2851, pruned_loss=0.05451, over 3219702.87 frames. ], batch size: 41, lr: 3.77e-03, grad_scale: 8.0 2023-04-30 17:07:38,629 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.6058, 4.9284, 4.7018, 4.7045, 4.3633, 4.3402, 4.3791, 5.0103], device='cuda:1'), covar=tensor([0.1212, 0.0830, 0.0951, 0.0777, 0.0860, 0.1142, 0.1137, 0.0781], device='cuda:1'), in_proj_covar=tensor([0.0631, 0.0779, 0.0634, 0.0581, 0.0490, 0.0498, 0.0647, 0.0604], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:1') 2023-04-30 17:07:43,888 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=177021.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 17:08:14,847 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.4708, 4.2852, 4.1850, 2.5815, 3.7301, 4.1231, 3.7111, 2.4974], device='cuda:1'), covar=tensor([0.0472, 0.0027, 0.0037, 0.0397, 0.0082, 0.0088, 0.0077, 0.0368], device='cuda:1'), in_proj_covar=tensor([0.0135, 0.0078, 0.0079, 0.0132, 0.0094, 0.0105, 0.0092, 0.0125], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-04-30 17:08:24,608 INFO [optim.py:368] (1/8) 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] (1/8) Epoch 18, batch 4500, loss[loss=0.1932, simple_loss=0.282, pruned_loss=0.05226, over 16469.00 frames. ], tot_loss[loss=0.1978, simple_loss=0.2854, pruned_loss=0.05514, over 3205889.08 frames. ], batch size: 68, lr: 3.77e-03, grad_scale: 8.0 2023-04-30 17:09:43,613 INFO [train.py:904] (1/8) Epoch 18, batch 4550, loss[loss=0.1887, simple_loss=0.273, pruned_loss=0.05215, over 16382.00 frames. ], tot_loss[loss=0.1981, simple_loss=0.2854, pruned_loss=0.05541, over 3220340.54 frames. ], batch size: 35, lr: 3.77e-03, grad_scale: 8.0 2023-04-30 17:09:53,002 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=177109.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 17:09:55,551 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=177111.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 17:10:11,669 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=177122.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 17:10:31,731 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([5.6780, 5.9959, 5.7298, 5.8134, 5.4473, 5.1786, 5.4154, 6.1420], device='cuda:1'), covar=tensor([0.1218, 0.0750, 0.0982, 0.0783, 0.0836, 0.0666, 0.1052, 0.0820], device='cuda:1'), in_proj_covar=tensor([0.0627, 0.0773, 0.0631, 0.0578, 0.0487, 0.0495, 0.0642, 0.0601], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:1') 2023-04-30 17:10:44,744 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.66 vs. limit=2.0 2023-04-30 17:10:48,535 INFO [optim.py:368] (1/8) 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] (1/8) Epoch 18, batch 4600, loss[loss=0.1766, simple_loss=0.2719, pruned_loss=0.04069, over 16752.00 frames. ], tot_loss[loss=0.1985, simple_loss=0.2862, pruned_loss=0.05537, over 3240477.10 frames. ], batch size: 83, lr: 3.77e-03, grad_scale: 8.0 2023-04-30 17:11:19,982 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=177170.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 17:11:20,172 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=177170.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 17:11:24,047 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=177172.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 17:11:37,403 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.1469, 2.1461, 2.6884, 3.0713, 2.9107, 3.4884, 2.2446, 3.4250], device='cuda:1'), covar=tensor([0.0152, 0.0408, 0.0239, 0.0219, 0.0233, 0.0126, 0.0445, 0.0105], device='cuda:1'), in_proj_covar=tensor([0.0182, 0.0189, 0.0177, 0.0181, 0.0188, 0.0148, 0.0192, 0.0142], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-30 17:12:05,775 INFO [train.py:904] (1/8) Epoch 18, batch 4650, loss[loss=0.1914, simple_loss=0.2727, pruned_loss=0.05508, over 16703.00 frames. ], tot_loss[loss=0.198, simple_loss=0.2851, pruned_loss=0.05539, over 3236134.81 frames. ], batch size: 124, lr: 3.77e-03, grad_scale: 8.0 2023-04-30 17:12:07,418 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.1333, 1.9769, 2.6617, 3.0106, 2.8587, 3.4532, 2.0969, 3.4147], device='cuda:1'), covar=tensor([0.0164, 0.0473, 0.0263, 0.0252, 0.0269, 0.0141, 0.0493, 0.0123], device='cuda:1'), in_proj_covar=tensor([0.0182, 0.0190, 0.0178, 0.0181, 0.0189, 0.0148, 0.0193, 0.0142], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-30 17:12:15,446 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=177209.0, num_to_drop=1, layers_to_drop={3} 2023-04-30 17:13:10,932 INFO [optim.py:368] (1/8) 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] (1/8) Epoch 18, batch 4700, loss[loss=0.2169, simple_loss=0.3057, pruned_loss=0.0641, over 15482.00 frames. ], tot_loss[loss=0.196, simple_loss=0.2831, pruned_loss=0.05448, over 3242327.48 frames. ], batch size: 190, lr: 3.77e-03, grad_scale: 16.0 2023-04-30 17:13:44,384 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.5366, 4.8754, 4.6453, 4.6567, 4.3629, 4.2857, 4.3171, 4.9150], device='cuda:1'), covar=tensor([0.1099, 0.0771, 0.0874, 0.0732, 0.0820, 0.1274, 0.1035, 0.0856], device='cuda:1'), in_proj_covar=tensor([0.0623, 0.0766, 0.0626, 0.0572, 0.0484, 0.0492, 0.0638, 0.0598], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:1') 2023-04-30 17:13:59,446 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 18, batch 4750, loss[loss=0.1809, simple_loss=0.2686, pruned_loss=0.04664, over 16871.00 frames. ], tot_loss[loss=0.1923, simple_loss=0.2791, pruned_loss=0.05272, over 3230189.11 frames. ], batch size: 109, lr: 3.77e-03, grad_scale: 16.0 2023-04-30 17:14:56,101 INFO [zipformer.py:625] (1/8) 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] (1/8) 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] (1/8) 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] (1/8) Epoch 18, batch 4800, loss[loss=0.2012, simple_loss=0.2881, pruned_loss=0.05715, over 15342.00 frames. ], tot_loss[loss=0.1886, simple_loss=0.2759, pruned_loss=0.05068, over 3214716.80 frames. ], batch size: 190, lr: 3.77e-03, grad_scale: 16.0 2023-04-30 17:16:00,903 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.5156, 4.6549, 4.8009, 4.5958, 4.5950, 5.1522, 4.6132, 4.3088], device='cuda:1'), covar=tensor([0.1239, 0.1605, 0.1550, 0.1815, 0.2308, 0.0894, 0.1518, 0.2416], device='cuda:1'), in_proj_covar=tensor([0.0392, 0.0565, 0.0616, 0.0474, 0.0635, 0.0647, 0.0489, 0.0637], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-04-30 17:16:01,082 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.3265, 1.6559, 2.0875, 2.3332, 2.4158, 2.5933, 1.7628, 2.5009], device='cuda:1'), covar=tensor([0.0188, 0.0453, 0.0302, 0.0275, 0.0269, 0.0186, 0.0486, 0.0124], device='cuda:1'), in_proj_covar=tensor([0.0182, 0.0188, 0.0177, 0.0181, 0.0188, 0.0147, 0.0192, 0.0142], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-30 17:16:06,841 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=177369.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 17:16:42,544 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=4.70 vs. limit=5.0 2023-04-30 17:16:57,032 INFO [train.py:904] (1/8) Epoch 18, batch 4850, loss[loss=0.1734, simple_loss=0.2706, pruned_loss=0.03814, over 16737.00 frames. ], tot_loss[loss=0.1884, simple_loss=0.2766, pruned_loss=0.05012, over 3198275.46 frames. ], batch size: 89, lr: 3.77e-03, grad_scale: 16.0 2023-04-30 17:17:03,801 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=4.88 vs. limit=5.0 2023-04-30 17:17:22,955 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([1.6576, 2.3434, 2.3018, 3.5157, 2.4035, 3.7548, 1.5323, 2.7856], device='cuda:1'), covar=tensor([0.1361, 0.0844, 0.1266, 0.0138, 0.0141, 0.0395, 0.1636, 0.0764], device='cuda:1'), in_proj_covar=tensor([0.0163, 0.0170, 0.0191, 0.0180, 0.0204, 0.0214, 0.0195, 0.0189], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-04-30 17:17:28,923 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([1.9867, 3.4350, 3.4409, 2.3236, 3.1395, 3.3819, 3.2427, 1.8866], device='cuda:1'), covar=tensor([0.0553, 0.0044, 0.0047, 0.0366, 0.0102, 0.0101, 0.0086, 0.0502], device='cuda:1'), in_proj_covar=tensor([0.0136, 0.0079, 0.0080, 0.0132, 0.0094, 0.0104, 0.0091, 0.0126], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-04-30 17:17:36,085 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.3236, 3.2968, 1.9437, 3.6627, 2.4258, 3.6604, 2.1116, 2.7072], device='cuda:1'), covar=tensor([0.0279, 0.0383, 0.1655, 0.0123, 0.0937, 0.0476, 0.1618, 0.0718], device='cuda:1'), in_proj_covar=tensor([0.0161, 0.0171, 0.0188, 0.0150, 0.0171, 0.0210, 0.0197, 0.0173], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0003, 0.0004, 0.0004, 0.0003], device='cuda:1') 2023-04-30 17:18:05,879 INFO [optim.py:368] (1/8) 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 17:18:11,207 INFO [train.py:904] (1/8) Epoch 18, batch 4900, loss[loss=0.1753, simple_loss=0.2736, pruned_loss=0.03851, over 16194.00 frames. ], tot_loss[loss=0.1866, simple_loss=0.2757, pruned_loss=0.04879, over 3197111.39 frames. ], batch size: 165, lr: 3.77e-03, grad_scale: 16.0 2023-04-30 17:18:20,329 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.74 vs. limit=2.0 2023-04-30 17:18:30,786 INFO [zipformer.py:625] (1/8) 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] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=177467.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 17:19:25,468 INFO [train.py:904] (1/8) Epoch 18, batch 4950, loss[loss=0.2048, simple_loss=0.3103, pruned_loss=0.04965, over 16787.00 frames. ], tot_loss[loss=0.1861, simple_loss=0.2754, pruned_loss=0.04841, over 3199668.83 frames. ], batch size: 83, lr: 3.77e-03, grad_scale: 16.0 2023-04-30 17:19:34,726 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=177509.0, num_to_drop=1, layers_to_drop={0} 2023-04-30 17:20:14,871 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.0374, 3.8940, 4.0903, 4.2256, 4.3757, 4.0039, 4.2923, 4.3889], device='cuda:1'), covar=tensor([0.1588, 0.1196, 0.1442, 0.0737, 0.0501, 0.1244, 0.0823, 0.0582], device='cuda:1'), in_proj_covar=tensor([0.0600, 0.0743, 0.0872, 0.0761, 0.0558, 0.0598, 0.0611, 0.0706], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-30 17:20:30,521 INFO [optim.py:368] (1/8) 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 17:20:35,814 INFO [train.py:904] (1/8) Epoch 18, batch 5000, loss[loss=0.21, simple_loss=0.3002, pruned_loss=0.05985, over 16811.00 frames. ], tot_loss[loss=0.1875, simple_loss=0.2773, pruned_loss=0.04881, over 3204198.26 frames. ], batch size: 124, lr: 3.77e-03, grad_scale: 16.0 2023-04-30 17:20:37,351 INFO [zipformer.py:625] (1/8) 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] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=177557.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 17:21:47,171 INFO [train.py:904] (1/8) Epoch 18, batch 5050, loss[loss=0.1927, simple_loss=0.2777, pruned_loss=0.05387, over 16536.00 frames. ], tot_loss[loss=0.1868, simple_loss=0.2773, pruned_loss=0.04811, over 3218877.80 frames. ], batch size: 62, lr: 3.77e-03, grad_scale: 16.0 2023-04-30 17:21:49,812 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.8147, 3.6333, 3.9421, 3.6653, 3.8880, 4.2702, 3.9412, 3.5544], device='cuda:1'), covar=tensor([0.2070, 0.2529, 0.2059, 0.2586, 0.2639, 0.1925, 0.1430, 0.2586], device='cuda:1'), in_proj_covar=tensor([0.0391, 0.0562, 0.0615, 0.0474, 0.0633, 0.0645, 0.0485, 0.0635], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-04-30 17:22:04,620 INFO [zipformer.py:625] (1/8) 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] (1/8) 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 17:22:58,434 INFO [train.py:904] (1/8) Epoch 18, batch 5100, loss[loss=0.1808, simple_loss=0.2705, pruned_loss=0.0456, over 15411.00 frames. ], tot_loss[loss=0.1852, simple_loss=0.2755, pruned_loss=0.04748, over 3218920.63 frames. ], batch size: 190, lr: 3.77e-03, grad_scale: 8.0 2023-04-30 17:23:16,196 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=177664.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 17:23:40,942 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.9873, 4.9910, 4.8336, 4.3854, 4.4349, 4.8620, 4.8349, 4.6428], device='cuda:1'), covar=tensor([0.0588, 0.0395, 0.0336, 0.0346, 0.1179, 0.0486, 0.0302, 0.0645], device='cuda:1'), in_proj_covar=tensor([0.0279, 0.0395, 0.0329, 0.0319, 0.0341, 0.0371, 0.0224, 0.0390], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-30 17:24:05,203 INFO [scaling.py:679] (1/8) 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] (1/8) Epoch 18, batch 5150, loss[loss=0.199, simple_loss=0.2952, pruned_loss=0.05141, over 15498.00 frames. ], tot_loss[loss=0.1852, simple_loss=0.276, pruned_loss=0.04722, over 3192821.24 frames. ], batch size: 191, lr: 3.77e-03, grad_scale: 8.0 2023-04-30 17:24:38,695 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.4114, 4.4586, 4.7652, 4.7482, 4.7259, 4.4340, 4.3930, 4.2847], device='cuda:1'), covar=tensor([0.0284, 0.0476, 0.0343, 0.0322, 0.0437, 0.0366, 0.0889, 0.0469], device='cuda:1'), in_proj_covar=tensor([0.0379, 0.0413, 0.0407, 0.0380, 0.0453, 0.0427, 0.0522, 0.0340], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:1') 2023-04-30 17:24:46,069 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=177725.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 17:25:04,047 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.7831, 3.8647, 4.1192, 4.0927, 4.0716, 3.8448, 3.8484, 3.8417], device='cuda:1'), covar=tensor([0.0328, 0.0555, 0.0349, 0.0372, 0.0492, 0.0425, 0.0836, 0.0479], device='cuda:1'), in_proj_covar=tensor([0.0380, 0.0413, 0.0408, 0.0381, 0.0453, 0.0427, 0.0522, 0.0341], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:1') 2023-04-30 17:25:20,503 INFO [optim.py:368] (1/8) 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] (1/8) Epoch 18, batch 5200, loss[loss=0.173, simple_loss=0.2623, pruned_loss=0.04187, over 16890.00 frames. ], tot_loss[loss=0.1835, simple_loss=0.274, pruned_loss=0.04648, over 3196092.03 frames. ], batch size: 96, lr: 3.76e-03, grad_scale: 8.0 2023-04-30 17:25:44,164 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=177765.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 17:25:47,835 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=177767.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 17:26:06,670 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=177780.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 17:26:07,039 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.36 vs. limit=2.0 2023-04-30 17:26:21,208 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.77 vs. limit=2.0 2023-04-30 17:26:31,608 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=177797.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 17:26:39,156 INFO [train.py:904] (1/8) Epoch 18, batch 5250, loss[loss=0.1656, simple_loss=0.2503, pruned_loss=0.04043, over 16635.00 frames. ], tot_loss[loss=0.1824, simple_loss=0.2719, pruned_loss=0.04648, over 3207524.76 frames. ], batch size: 57, lr: 3.76e-03, grad_scale: 8.0 2023-04-30 17:26:55,360 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=177813.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 17:26:57,991 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.1553, 4.2538, 4.0474, 3.7719, 3.7344, 4.1631, 3.8590, 3.9265], device='cuda:1'), covar=tensor([0.0617, 0.0482, 0.0318, 0.0321, 0.0889, 0.0496, 0.0741, 0.0582], device='cuda:1'), in_proj_covar=tensor([0.0280, 0.0395, 0.0329, 0.0320, 0.0340, 0.0370, 0.0224, 0.0389], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-30 17:26:58,849 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=177815.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 17:27:37,103 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.8313, 3.8308, 3.9719, 3.7323, 3.9488, 4.3087, 4.0078, 3.6932], device='cuda:1'), covar=tensor([0.2126, 0.2387, 0.2040, 0.2549, 0.2622, 0.1552, 0.1487, 0.2565], device='cuda:1'), in_proj_covar=tensor([0.0391, 0.0561, 0.0614, 0.0473, 0.0632, 0.0643, 0.0485, 0.0633], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-04-30 17:27:37,226 INFO [zipformer.py:625] (1/8) 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] (1/8) 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 17:27:49,444 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.4302, 3.3943, 2.1119, 3.7820, 2.6150, 3.7681, 2.2869, 2.7960], device='cuda:1'), covar=tensor([0.0262, 0.0343, 0.1656, 0.0141, 0.0847, 0.0481, 0.1497, 0.0718], device='cuda:1'), in_proj_covar=tensor([0.0162, 0.0172, 0.0191, 0.0151, 0.0173, 0.0211, 0.0199, 0.0175], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-04-30 17:27:52,846 INFO [train.py:904] (1/8) Epoch 18, batch 5300, loss[loss=0.1675, simple_loss=0.2584, pruned_loss=0.03834, over 16452.00 frames. ], tot_loss[loss=0.1796, simple_loss=0.2684, pruned_loss=0.04541, over 3201893.54 frames. ], batch size: 146, lr: 3.76e-03, grad_scale: 8.0 2023-04-30 17:28:02,095 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=177858.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 17:29:05,780 INFO [train.py:904] (1/8) Epoch 18, batch 5350, loss[loss=0.1871, simple_loss=0.27, pruned_loss=0.0521, over 17090.00 frames. ], tot_loss[loss=0.1778, simple_loss=0.2662, pruned_loss=0.04473, over 3212766.25 frames. ], batch size: 53, lr: 3.76e-03, grad_scale: 8.0 2023-04-30 17:29:17,491 INFO [zipformer.py:625] (1/8) 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] (1/8) 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 17:30:19,781 INFO [train.py:904] (1/8) Epoch 18, batch 5400, loss[loss=0.2026, simple_loss=0.2939, pruned_loss=0.05558, over 16670.00 frames. ], tot_loss[loss=0.1802, simple_loss=0.2691, pruned_loss=0.0456, over 3210016.50 frames. ], batch size: 134, lr: 3.76e-03, grad_scale: 8.0 2023-04-30 17:30:21,420 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([1.8471, 3.0752, 2.8423, 4.8556, 3.7449, 4.1124, 1.8883, 2.9332], device='cuda:1'), covar=tensor([0.1283, 0.0657, 0.1133, 0.0139, 0.0301, 0.0390, 0.1457, 0.0924], device='cuda:1'), in_proj_covar=tensor([0.0164, 0.0170, 0.0191, 0.0180, 0.0204, 0.0214, 0.0196, 0.0190], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-04-30 17:31:39,766 INFO [train.py:904] (1/8) Epoch 18, batch 5450, loss[loss=0.2161, simple_loss=0.3008, pruned_loss=0.06571, over 12204.00 frames. ], tot_loss[loss=0.183, simple_loss=0.2721, pruned_loss=0.04694, over 3194272.47 frames. ], batch size: 248, lr: 3.76e-03, grad_scale: 8.0 2023-04-30 17:31:52,867 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([1.6045, 2.6204, 2.4238, 3.8842, 2.6308, 3.8609, 1.5355, 2.8339], device='cuda:1'), covar=tensor([0.1426, 0.0780, 0.1233, 0.0168, 0.0233, 0.0418, 0.1664, 0.0823], device='cuda:1'), in_proj_covar=tensor([0.0163, 0.0169, 0.0191, 0.0179, 0.0203, 0.0213, 0.0195, 0.0189], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-04-30 17:31:55,141 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.64 vs. limit=2.0 2023-04-30 17:32:08,122 INFO [zipformer.py:625] (1/8) 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] (1/8) 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 17:32:56,916 INFO [train.py:904] (1/8) Epoch 18, batch 5500, loss[loss=0.2327, simple_loss=0.3141, pruned_loss=0.07568, over 15449.00 frames. ], tot_loss[loss=0.1903, simple_loss=0.2791, pruned_loss=0.0508, over 3177768.50 frames. ], batch size: 191, lr: 3.76e-03, grad_scale: 8.0 2023-04-30 17:33:23,749 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.62 vs. limit=2.0 2023-04-30 17:34:14,385 INFO [train.py:904] (1/8) Epoch 18, batch 5550, loss[loss=0.2813, simple_loss=0.3347, pruned_loss=0.1139, over 11243.00 frames. ], tot_loss[loss=0.1984, simple_loss=0.286, pruned_loss=0.05533, over 3166182.00 frames. ], batch size: 249, lr: 3.76e-03, grad_scale: 8.0 2023-04-30 17:35:05,325 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=3.27 vs. limit=5.0 2023-04-30 17:35:09,027 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=178136.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 17:35:31,152 INFO [optim.py:368] (1/8) 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] (1/8) Epoch 18, batch 5600, loss[loss=0.2481, simple_loss=0.3303, pruned_loss=0.08294, over 15326.00 frames. ], tot_loss[loss=0.2062, simple_loss=0.2917, pruned_loss=0.06037, over 3114779.34 frames. ], batch size: 191, lr: 3.76e-03, grad_scale: 8.0 2023-04-30 17:35:36,601 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=178153.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 17:36:18,312 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.6402, 3.6094, 4.0569, 1.9168, 4.3169, 4.2794, 3.1717, 3.1125], device='cuda:1'), covar=tensor([0.0822, 0.0260, 0.0185, 0.1235, 0.0057, 0.0150, 0.0365, 0.0455], device='cuda:1'), in_proj_covar=tensor([0.0146, 0.0105, 0.0094, 0.0137, 0.0076, 0.0121, 0.0124, 0.0129], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:1') 2023-04-30 17:36:58,587 INFO [train.py:904] (1/8) Epoch 18, batch 5650, loss[loss=0.195, simple_loss=0.2854, pruned_loss=0.05234, over 16636.00 frames. ], tot_loss[loss=0.2134, simple_loss=0.2972, pruned_loss=0.06481, over 3077187.89 frames. ], batch size: 89, lr: 3.76e-03, grad_scale: 4.0 2023-04-30 17:37:09,702 INFO [zipformer.py:625] (1/8) 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] (1/8) 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] (1/8) Epoch 18, batch 5700, loss[loss=0.1923, simple_loss=0.2786, pruned_loss=0.05305, over 17161.00 frames. ], tot_loss[loss=0.2151, simple_loss=0.2982, pruned_loss=0.06597, over 3066865.94 frames. ], batch size: 46, lr: 3.76e-03, grad_scale: 4.0 2023-04-30 17:38:20,140 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([1.9100, 2.7192, 2.6384, 1.9216, 2.5259, 2.6824, 2.5228, 1.9098], device='cuda:1'), covar=tensor([0.0401, 0.0080, 0.0088, 0.0363, 0.0138, 0.0122, 0.0133, 0.0386], device='cuda:1'), in_proj_covar=tensor([0.0134, 0.0078, 0.0079, 0.0130, 0.0093, 0.0103, 0.0090, 0.0124], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-04-30 17:38:25,289 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=178257.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 17:39:39,101 INFO [train.py:904] (1/8) Epoch 18, batch 5750, loss[loss=0.2149, simple_loss=0.3012, pruned_loss=0.06432, over 15357.00 frames. ], tot_loss[loss=0.2182, simple_loss=0.3009, pruned_loss=0.06779, over 3033652.67 frames. ], batch size: 191, lr: 3.76e-03, grad_scale: 4.0 2023-04-30 17:40:08,204 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=178320.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 17:40:56,669 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.65 vs. limit=2.0 2023-04-30 17:40:59,316 INFO [optim.py:368] (1/8) 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] (1/8) Epoch 18, batch 5800, loss[loss=0.187, simple_loss=0.2791, pruned_loss=0.04746, over 16721.00 frames. ], tot_loss[loss=0.2164, simple_loss=0.3002, pruned_loss=0.06634, over 3047295.44 frames. ], batch size: 124, lr: 3.76e-03, grad_scale: 4.0 2023-04-30 17:41:27,554 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=178368.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 17:42:19,759 INFO [train.py:904] (1/8) Epoch 18, batch 5850, loss[loss=0.2185, simple_loss=0.3159, pruned_loss=0.06054, over 16748.00 frames. ], tot_loss[loss=0.2145, simple_loss=0.2984, pruned_loss=0.06525, over 3034637.54 frames. ], batch size: 83, lr: 3.76e-03, grad_scale: 4.0 2023-04-30 17:43:16,098 INFO [zipformer.py:625] (1/8) 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] (1/8) 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] (1/8) Epoch 18, batch 5900, loss[loss=0.2535, simple_loss=0.3148, pruned_loss=0.09608, over 11499.00 frames. ], tot_loss[loss=0.213, simple_loss=0.2973, pruned_loss=0.06435, over 3062913.47 frames. ], batch size: 247, lr: 3.76e-03, grad_scale: 4.0 2023-04-30 17:43:42,239 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=178452.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 17:43:44,118 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=178453.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 17:44:32,739 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=178484.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 17:44:58,649 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 18, batch 5950, loss[loss=0.2001, simple_loss=0.2888, pruned_loss=0.05569, over 17215.00 frames. ], tot_loss[loss=0.2121, simple_loss=0.2984, pruned_loss=0.06293, over 3086327.14 frames. ], batch size: 45, lr: 3.76e-03, grad_scale: 4.0 2023-04-30 17:45:07,286 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.6169, 3.8357, 2.9608, 2.3270, 2.7321, 2.5094, 4.0926, 3.6404], device='cuda:1'), covar=tensor([0.2904, 0.0686, 0.1790, 0.2718, 0.2437, 0.1969, 0.0509, 0.1127], device='cuda:1'), in_proj_covar=tensor([0.0319, 0.0263, 0.0297, 0.0301, 0.0290, 0.0246, 0.0288, 0.0326], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:1') 2023-04-30 17:45:18,582 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=178513.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 17:46:17,847 INFO [optim.py:368] (1/8) 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] (1/8) Epoch 18, batch 6000, loss[loss=0.1862, simple_loss=0.2709, pruned_loss=0.05072, over 16821.00 frames. ], tot_loss[loss=0.2102, simple_loss=0.2966, pruned_loss=0.06192, over 3103161.69 frames. ], batch size: 102, lr: 3.76e-03, grad_scale: 8.0 2023-04-30 17:46:19,086 INFO [train.py:929] (1/8) Computing validation loss 2023-04-30 17:46:29,950 INFO [train.py:938] (1/8) Epoch 18, validation: loss=0.1531, simple_loss=0.2661, pruned_loss=0.02007, over 944034.00 frames. 2023-04-30 17:46:29,951 INFO [train.py:939] (1/8) Maximum memory allocated so far is 17857MB 2023-04-30 17:47:48,055 INFO [train.py:904] (1/8) Epoch 18, batch 6050, loss[loss=0.196, simple_loss=0.2855, pruned_loss=0.05328, over 15393.00 frames. ], tot_loss[loss=0.2083, simple_loss=0.2947, pruned_loss=0.06093, over 3116768.83 frames. ], batch size: 191, lr: 3.76e-03, grad_scale: 8.0 2023-04-30 17:48:14,282 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=178619.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 17:49:06,485 INFO [optim.py:368] (1/8) 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 17:49:06,500 INFO [train.py:904] (1/8) Epoch 18, batch 6100, loss[loss=0.1919, simple_loss=0.2812, pruned_loss=0.05131, over 17165.00 frames. ], tot_loss[loss=0.2076, simple_loss=0.2947, pruned_loss=0.06031, over 3113371.88 frames. ], batch size: 46, lr: 3.76e-03, grad_scale: 4.0 2023-04-30 17:49:15,902 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.65 vs. limit=2.0 2023-04-30 17:49:33,024 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.8642, 4.0083, 2.3635, 4.7988, 3.1545, 4.6751, 2.6110, 3.2380], device='cuda:1'), covar=tensor([0.0285, 0.0378, 0.1758, 0.0166, 0.0766, 0.0466, 0.1468, 0.0720], device='cuda:1'), in_proj_covar=tensor([0.0164, 0.0173, 0.0192, 0.0153, 0.0174, 0.0213, 0.0200, 0.0176], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-04-30 17:49:50,802 INFO [zipformer.py:625] (1/8) 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] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=178694.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 17:50:16,835 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=178697.0, num_to_drop=1, layers_to_drop={1} 2023-04-30 17:50:23,986 INFO [train.py:904] (1/8) Epoch 18, batch 6150, loss[loss=0.1986, simple_loss=0.2814, pruned_loss=0.05794, over 16662.00 frames. ], tot_loss[loss=0.2062, simple_loss=0.2928, pruned_loss=0.05983, over 3114947.83 frames. ], batch size: 62, lr: 3.75e-03, grad_scale: 4.0 2023-04-30 17:51:39,656 INFO [optim.py:368] (1/8) 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] (1/8) Epoch 18, batch 6200, loss[loss=0.1848, simple_loss=0.2767, pruned_loss=0.04641, over 16707.00 frames. ], tot_loss[loss=0.2052, simple_loss=0.291, pruned_loss=0.05971, over 3095235.83 frames. ], batch size: 134, lr: 3.75e-03, grad_scale: 4.0 2023-04-30 17:51:45,189 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=178755.0, num_to_drop=1, layers_to_drop={3} 2023-04-30 17:51:49,727 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=178758.0, num_to_drop=1, layers_to_drop={1} 2023-04-30 17:52:08,995 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=178770.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 17:52:14,253 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.8623, 2.3401, 1.8502, 2.1402, 2.7254, 2.3527, 2.7029, 2.8583], device='cuda:1'), covar=tensor([0.0170, 0.0344, 0.0482, 0.0395, 0.0215, 0.0313, 0.0184, 0.0229], device='cuda:1'), in_proj_covar=tensor([0.0190, 0.0224, 0.0218, 0.0217, 0.0226, 0.0224, 0.0227, 0.0220], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-30 17:52:14,294 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.8935, 3.1507, 3.3606, 1.9154, 2.8553, 2.1439, 3.4099, 3.3946], device='cuda:1'), covar=tensor([0.0233, 0.0842, 0.0555, 0.2056, 0.0887, 0.1016, 0.0647, 0.0980], device='cuda:1'), in_proj_covar=tensor([0.0153, 0.0161, 0.0166, 0.0150, 0.0143, 0.0128, 0.0143, 0.0169], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:1') 2023-04-30 17:52:52,890 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.0303, 4.0036, 3.9642, 3.2203, 3.9770, 1.7318, 3.7661, 3.4812], device='cuda:1'), covar=tensor([0.0140, 0.0114, 0.0180, 0.0311, 0.0106, 0.2844, 0.0155, 0.0260], device='cuda:1'), in_proj_covar=tensor([0.0153, 0.0142, 0.0187, 0.0172, 0.0162, 0.0198, 0.0177, 0.0166], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-30 17:52:56,568 INFO [train.py:904] (1/8) Epoch 18, batch 6250, loss[loss=0.1922, simple_loss=0.2868, pruned_loss=0.04881, over 16634.00 frames. ], tot_loss[loss=0.2052, simple_loss=0.2911, pruned_loss=0.05966, over 3096554.09 frames. ], batch size: 68, lr: 3.75e-03, grad_scale: 2.0 2023-04-30 17:53:08,007 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=178808.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 17:53:41,408 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=178831.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 17:54:15,694 INFO [train.py:904] (1/8) Epoch 18, batch 6300, loss[loss=0.2106, simple_loss=0.2953, pruned_loss=0.06294, over 16846.00 frames. ], tot_loss[loss=0.2051, simple_loss=0.2909, pruned_loss=0.05963, over 3083416.34 frames. ], batch size: 96, lr: 3.75e-03, grad_scale: 2.0 2023-04-30 17:54:17,524 INFO [optim.py:368] (1/8) 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 17:55:22,166 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=178894.0, num_to_drop=1, layers_to_drop={0} 2023-04-30 17:55:34,230 INFO [train.py:904] (1/8) Epoch 18, batch 6350, loss[loss=0.2696, simple_loss=0.3233, pruned_loss=0.1079, over 11647.00 frames. ], tot_loss[loss=0.2065, simple_loss=0.2914, pruned_loss=0.06085, over 3073401.97 frames. ], batch size: 247, lr: 3.75e-03, grad_scale: 2.0 2023-04-30 17:56:52,036 INFO [train.py:904] (1/8) Epoch 18, batch 6400, loss[loss=0.2002, simple_loss=0.2835, pruned_loss=0.05848, over 16632.00 frames. ], tot_loss[loss=0.2084, simple_loss=0.2921, pruned_loss=0.06236, over 3058538.27 frames. ], batch size: 62, lr: 3.75e-03, grad_scale: 4.0 2023-04-30 17:56:53,836 INFO [optim.py:368] (1/8) 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 17:56:57,541 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=178955.0, num_to_drop=1, layers_to_drop={1} 2023-04-30 17:57:27,447 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 18, batch 6450, loss[loss=0.1932, simple_loss=0.2787, pruned_loss=0.05386, over 15372.00 frames. ], tot_loss[loss=0.2071, simple_loss=0.2919, pruned_loss=0.06116, over 3069062.46 frames. ], batch size: 190, lr: 3.75e-03, grad_scale: 4.0 2023-04-30 17:59:24,076 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=179050.0, num_to_drop=1, layers_to_drop={2} 2023-04-30 17:59:26,158 INFO [train.py:904] (1/8) Epoch 18, batch 6500, loss[loss=0.2346, simple_loss=0.2967, pruned_loss=0.08621, over 11354.00 frames. ], tot_loss[loss=0.206, simple_loss=0.2902, pruned_loss=0.06086, over 3054827.27 frames. ], batch size: 247, lr: 3.75e-03, grad_scale: 4.0 2023-04-30 17:59:27,317 INFO [optim.py:368] (1/8) 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] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=179053.0, num_to_drop=1, layers_to_drop={3} 2023-04-30 18:00:44,327 INFO [train.py:904] (1/8) Epoch 18, batch 6550, loss[loss=0.2118, simple_loss=0.3054, pruned_loss=0.05913, over 15245.00 frames. ], tot_loss[loss=0.2083, simple_loss=0.2932, pruned_loss=0.06171, over 3058783.91 frames. ], batch size: 190, lr: 3.75e-03, grad_scale: 4.0 2023-04-30 18:00:54,340 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=179108.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 18:01:20,179 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=179126.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 18:01:58,889 INFO [train.py:904] (1/8) Epoch 18, batch 6600, loss[loss=0.2486, simple_loss=0.3284, pruned_loss=0.08447, over 16887.00 frames. ], tot_loss[loss=0.2095, simple_loss=0.295, pruned_loss=0.06197, over 3062530.61 frames. ], batch size: 116, lr: 3.75e-03, grad_scale: 4.0 2023-04-30 18:02:00,668 INFO [optim.py:368] (1/8) 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 18:02:02,452 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.3323, 2.3283, 2.4488, 4.1484, 2.1737, 2.6961, 2.3976, 2.4980], device='cuda:1'), covar=tensor([0.1192, 0.3314, 0.2581, 0.0468, 0.4085, 0.2423, 0.3254, 0.3147], device='cuda:1'), in_proj_covar=tensor([0.0386, 0.0427, 0.0352, 0.0319, 0.0427, 0.0493, 0.0397, 0.0500], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-30 18:02:05,100 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=179156.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 18:03:06,864 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.9347, 4.0122, 4.3029, 4.2478, 4.2704, 4.0024, 4.0094, 3.9463], device='cuda:1'), covar=tensor([0.0323, 0.0572, 0.0389, 0.0449, 0.0458, 0.0426, 0.0909, 0.0560], device='cuda:1'), in_proj_covar=tensor([0.0384, 0.0419, 0.0409, 0.0386, 0.0458, 0.0430, 0.0527, 0.0345], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:1') 2023-04-30 18:03:17,504 INFO [train.py:904] (1/8) Epoch 18, batch 6650, loss[loss=0.1952, simple_loss=0.2801, pruned_loss=0.05515, over 16801.00 frames. ], tot_loss[loss=0.209, simple_loss=0.2944, pruned_loss=0.06181, over 3085168.36 frames. ], batch size: 124, lr: 3.75e-03, grad_scale: 4.0 2023-04-30 18:04:07,389 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.9431, 4.0151, 4.2989, 4.2384, 4.2707, 3.9974, 4.0047, 3.9307], device='cuda:1'), covar=tensor([0.0360, 0.0598, 0.0413, 0.0458, 0.0472, 0.0429, 0.0991, 0.0608], device='cuda:1'), in_proj_covar=tensor([0.0387, 0.0422, 0.0411, 0.0389, 0.0461, 0.0433, 0.0531, 0.0348], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:1') 2023-04-30 18:04:30,598 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=179250.0, num_to_drop=1, layers_to_drop={3} 2023-04-30 18:04:32,473 INFO [train.py:904] (1/8) Epoch 18, batch 6700, loss[loss=0.2265, simple_loss=0.3085, pruned_loss=0.07225, over 16582.00 frames. ], tot_loss[loss=0.2097, simple_loss=0.2941, pruned_loss=0.06266, over 3067181.92 frames. ], batch size: 68, lr: 3.75e-03, grad_scale: 4.0 2023-04-30 18:04:34,183 INFO [optim.py:368] (1/8) 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 18:05:08,630 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=179275.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 18:05:24,189 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.0689, 2.2616, 2.2901, 2.6552, 1.7296, 3.1408, 1.8181, 2.6501], device='cuda:1'), covar=tensor([0.1156, 0.0723, 0.1147, 0.0186, 0.0131, 0.0439, 0.1476, 0.0769], device='cuda:1'), in_proj_covar=tensor([0.0163, 0.0167, 0.0189, 0.0176, 0.0204, 0.0212, 0.0194, 0.0187], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-04-30 18:05:48,861 INFO [train.py:904] (1/8) Epoch 18, batch 6750, loss[loss=0.1796, simple_loss=0.2658, pruned_loss=0.04676, over 16485.00 frames. ], tot_loss[loss=0.209, simple_loss=0.2929, pruned_loss=0.06261, over 3057067.29 frames. ], batch size: 68, lr: 3.75e-03, grad_scale: 4.0 2023-04-30 18:06:19,477 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=179323.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 18:06:21,435 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.1216, 3.9635, 4.1553, 4.3230, 4.4343, 4.0257, 4.3300, 4.4495], device='cuda:1'), covar=tensor([0.1714, 0.1231, 0.1482, 0.0683, 0.0553, 0.1374, 0.0837, 0.0640], device='cuda:1'), in_proj_covar=tensor([0.0600, 0.0735, 0.0866, 0.0751, 0.0562, 0.0593, 0.0609, 0.0699], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-30 18:06:25,188 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.9933, 3.0302, 1.8204, 3.2218, 2.2655, 3.2962, 2.0264, 2.5076], device='cuda:1'), covar=tensor([0.0360, 0.0393, 0.1619, 0.0211, 0.0847, 0.0537, 0.1600, 0.0777], device='cuda:1'), in_proj_covar=tensor([0.0164, 0.0172, 0.0192, 0.0153, 0.0174, 0.0213, 0.0200, 0.0175], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-04-30 18:06:36,818 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.4840, 3.4701, 3.4566, 2.7462, 3.3615, 2.1328, 3.1068, 2.7248], device='cuda:1'), covar=tensor([0.0144, 0.0124, 0.0176, 0.0224, 0.0102, 0.2265, 0.0126, 0.0228], device='cuda:1'), in_proj_covar=tensor([0.0153, 0.0142, 0.0188, 0.0172, 0.0163, 0.0199, 0.0178, 0.0167], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-30 18:07:00,949 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 18, batch 6800, loss[loss=0.2045, simple_loss=0.29, pruned_loss=0.05953, over 16239.00 frames. ], tot_loss[loss=0.2093, simple_loss=0.2929, pruned_loss=0.0628, over 3068252.60 frames. ], batch size: 165, lr: 3.75e-03, grad_scale: 8.0 2023-04-30 18:07:04,930 INFO [optim.py:368] (1/8) 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 18:07:05,954 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=179353.0, num_to_drop=1, layers_to_drop={2} 2023-04-30 18:07:18,269 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=179361.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 18:08:15,638 INFO [zipformer.py:625] (1/8) 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] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=179401.0, num_to_drop=1, layers_to_drop={1} 2023-04-30 18:08:21,226 INFO [train.py:904] (1/8) Epoch 18, batch 6850, loss[loss=0.2007, simple_loss=0.3017, pruned_loss=0.04988, over 16706.00 frames. ], tot_loss[loss=0.2111, simple_loss=0.2945, pruned_loss=0.06382, over 3050953.39 frames. ], batch size: 62, lr: 3.75e-03, grad_scale: 8.0 2023-04-30 18:08:47,272 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.13 vs. limit=2.0 2023-04-30 18:08:50,198 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=179422.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 18:08:56,260 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=179426.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 18:09:23,517 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=179444.0, num_to_drop=1, layers_to_drop={1} 2023-04-30 18:09:25,648 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.77 vs. limit=2.0 2023-04-30 18:09:34,753 INFO [train.py:904] (1/8) Epoch 18, batch 6900, loss[loss=0.2254, simple_loss=0.3129, pruned_loss=0.06896, over 16237.00 frames. ], tot_loss[loss=0.2112, simple_loss=0.2972, pruned_loss=0.06267, over 3079185.14 frames. ], batch size: 165, lr: 3.75e-03, grad_scale: 4.0 2023-04-30 18:09:38,468 INFO [optim.py:368] (1/8) 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 18:10:10,141 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=179474.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 18:10:53,471 INFO [train.py:904] (1/8) Epoch 18, batch 6950, loss[loss=0.1895, simple_loss=0.2768, pruned_loss=0.05111, over 16909.00 frames. ], tot_loss[loss=0.2123, simple_loss=0.2979, pruned_loss=0.06338, over 3085488.55 frames. ], batch size: 96, lr: 3.75e-03, grad_scale: 4.0 2023-04-30 18:10:58,854 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=179505.0, num_to_drop=1, layers_to_drop={2} 2023-04-30 18:11:06,807 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.4677, 3.3596, 2.6393, 2.0973, 2.2326, 2.1780, 3.3887, 3.0671], device='cuda:1'), covar=tensor([0.2763, 0.0695, 0.1810, 0.2598, 0.2494, 0.2189, 0.0550, 0.1332], device='cuda:1'), in_proj_covar=tensor([0.0320, 0.0264, 0.0299, 0.0305, 0.0291, 0.0248, 0.0286, 0.0327], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:1') 2023-04-30 18:11:21,743 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.1767, 2.0595, 1.6960, 1.7111, 2.2474, 1.8828, 2.0130, 2.3365], device='cuda:1'), covar=tensor([0.0199, 0.0373, 0.0499, 0.0461, 0.0247, 0.0368, 0.0187, 0.0245], device='cuda:1'), in_proj_covar=tensor([0.0191, 0.0225, 0.0220, 0.0220, 0.0227, 0.0225, 0.0227, 0.0222], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-30 18:11:25,614 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.70 vs. limit=2.0 2023-04-30 18:12:07,078 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=179550.0, num_to_drop=1, layers_to_drop={2} 2023-04-30 18:12:09,843 INFO [train.py:904] (1/8) Epoch 18, batch 7000, loss[loss=0.2118, simple_loss=0.3102, pruned_loss=0.05666, over 16359.00 frames. ], tot_loss[loss=0.211, simple_loss=0.2976, pruned_loss=0.06223, over 3105015.93 frames. ], batch size: 165, lr: 3.75e-03, grad_scale: 4.0 2023-04-30 18:12:12,175 INFO [optim.py:368] (1/8) 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 18:12:55,844 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.93 vs. limit=2.0 2023-04-30 18:13:16,940 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=179598.0, num_to_drop=1, layers_to_drop={1} 2023-04-30 18:13:21,678 INFO [train.py:904] (1/8) Epoch 18, batch 7050, loss[loss=0.2366, simple_loss=0.3001, pruned_loss=0.08657, over 11963.00 frames. ], tot_loss[loss=0.2112, simple_loss=0.2985, pruned_loss=0.06199, over 3109842.62 frames. ], batch size: 246, lr: 3.75e-03, grad_scale: 4.0 2023-04-30 18:13:35,900 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.4078, 3.3535, 2.6365, 2.0845, 2.2275, 2.1925, 3.4522, 3.1183], device='cuda:1'), covar=tensor([0.2979, 0.0712, 0.1868, 0.2696, 0.2691, 0.2177, 0.0535, 0.1296], device='cuda:1'), in_proj_covar=tensor([0.0322, 0.0265, 0.0300, 0.0306, 0.0292, 0.0249, 0.0287, 0.0329], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:1') 2023-04-30 18:14:12,930 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 18, batch 7100, loss[loss=0.1914, simple_loss=0.28, pruned_loss=0.05145, over 16629.00 frames. ], tot_loss[loss=0.2105, simple_loss=0.2968, pruned_loss=0.06216, over 3090516.16 frames. ], batch size: 134, lr: 3.75e-03, grad_scale: 4.0 2023-04-30 18:14:40,299 INFO [optim.py:368] (1/8) 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 18:15:33,587 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.3360, 1.5617, 2.0742, 2.2902, 2.4036, 2.6257, 1.8065, 2.5464], device='cuda:1'), covar=tensor([0.0224, 0.0506, 0.0308, 0.0313, 0.0282, 0.0195, 0.0515, 0.0132], device='cuda:1'), in_proj_covar=tensor([0.0181, 0.0189, 0.0175, 0.0178, 0.0187, 0.0145, 0.0191, 0.0141], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-30 18:15:46,903 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=179697.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 18:15:53,679 INFO [train.py:904] (1/8) Epoch 18, batch 7150, loss[loss=0.1996, simple_loss=0.2854, pruned_loss=0.05691, over 17039.00 frames. ], tot_loss[loss=0.2084, simple_loss=0.2941, pruned_loss=0.06135, over 3101367.32 frames. ], batch size: 55, lr: 3.74e-03, grad_scale: 4.0 2023-04-30 18:15:55,535 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=3.09 vs. limit=5.0 2023-04-30 18:16:08,465 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.3499, 2.9475, 2.7003, 2.2919, 2.2824, 2.2650, 2.9164, 2.8770], device='cuda:1'), covar=tensor([0.2305, 0.0728, 0.1419, 0.2144, 0.1999, 0.1917, 0.0467, 0.1101], device='cuda:1'), in_proj_covar=tensor([0.0322, 0.0265, 0.0299, 0.0305, 0.0292, 0.0249, 0.0287, 0.0328], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:1') 2023-04-30 18:16:16,355 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=179717.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 18:16:26,696 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.48 vs. limit=2.0 2023-04-30 18:17:00,437 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.2956, 2.9533, 2.6431, 2.2147, 2.1585, 2.2050, 2.9337, 2.8272], device='cuda:1'), covar=tensor([0.2622, 0.0821, 0.1666, 0.2636, 0.2697, 0.2190, 0.0513, 0.1375], device='cuda:1'), in_proj_covar=tensor([0.0322, 0.0265, 0.0299, 0.0305, 0.0291, 0.0249, 0.0286, 0.0328], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:1') 2023-04-30 18:17:08,205 INFO [train.py:904] (1/8) Epoch 18, batch 7200, loss[loss=0.1792, simple_loss=0.2704, pruned_loss=0.044, over 16757.00 frames. ], tot_loss[loss=0.2056, simple_loss=0.2915, pruned_loss=0.05987, over 3077784.30 frames. ], batch size: 62, lr: 3.74e-03, grad_scale: 8.0 2023-04-30 18:17:10,639 INFO [optim.py:368] (1/8) Clipping_scale=2.0, grad-norm quartiles 1.743e+02 2.765e+02 3.366e+02 4.196e+02 7.871e+02, threshold=6.733e+02, percent-clipped=4.0 2023-04-30 18:17:29,432 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.38 vs. limit=2.0 2023-04-30 18:17:30,266 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([1.7052, 1.7913, 1.6214, 1.5222, 1.9244, 1.5576, 1.6447, 1.8683], device='cuda:1'), covar=tensor([0.0150, 0.0273, 0.0399, 0.0318, 0.0194, 0.0248, 0.0156, 0.0190], device='cuda:1'), in_proj_covar=tensor([0.0188, 0.0222, 0.0217, 0.0217, 0.0225, 0.0222, 0.0225, 0.0219], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-30 18:17:54,631 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=4.82 vs. limit=5.0 2023-04-30 18:18:17,301 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=179796.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 18:18:24,741 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 18, batch 7250, loss[loss=0.1716, simple_loss=0.2555, pruned_loss=0.04383, over 16807.00 frames. ], tot_loss[loss=0.2034, simple_loss=0.2893, pruned_loss=0.05881, over 3076646.04 frames. ], batch size: 83, lr: 3.74e-03, grad_scale: 8.0 2023-04-30 18:18:29,304 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.7643, 3.8896, 4.1342, 4.0878, 4.1028, 3.8619, 3.8995, 3.8486], device='cuda:1'), covar=tensor([0.0339, 0.0580, 0.0407, 0.0458, 0.0449, 0.0433, 0.0792, 0.0497], device='cuda:1'), in_proj_covar=tensor([0.0390, 0.0426, 0.0416, 0.0393, 0.0467, 0.0438, 0.0534, 0.0351], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:1') 2023-04-30 18:19:41,484 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.1960, 2.0858, 2.7359, 3.0682, 2.9276, 3.6523, 2.2817, 3.6329], device='cuda:1'), covar=tensor([0.0166, 0.0443, 0.0258, 0.0251, 0.0265, 0.0125, 0.0470, 0.0104], device='cuda:1'), in_proj_covar=tensor([0.0178, 0.0186, 0.0173, 0.0176, 0.0185, 0.0143, 0.0189, 0.0139], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-30 18:19:42,144 INFO [train.py:904] (1/8) Epoch 18, batch 7300, loss[loss=0.1866, simple_loss=0.2754, pruned_loss=0.04893, over 17187.00 frames. ], tot_loss[loss=0.2024, simple_loss=0.288, pruned_loss=0.05843, over 3080016.31 frames. ], batch size: 44, lr: 3.74e-03, grad_scale: 8.0 2023-04-30 18:19:45,256 INFO [optim.py:368] (1/8) 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 18:19:50,396 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=179857.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 18:20:33,701 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.2117, 4.3539, 4.4870, 4.3151, 4.4086, 4.8278, 4.4073, 4.1441], device='cuda:1'), covar=tensor([0.1667, 0.1810, 0.2033, 0.1953, 0.2224, 0.1053, 0.1622, 0.2541], device='cuda:1'), in_proj_covar=tensor([0.0390, 0.0559, 0.0618, 0.0470, 0.0628, 0.0644, 0.0486, 0.0631], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-04-30 18:20:58,393 INFO [train.py:904] (1/8) Epoch 18, batch 7350, loss[loss=0.2498, simple_loss=0.3128, pruned_loss=0.09343, over 11106.00 frames. ], tot_loss[loss=0.2047, simple_loss=0.2897, pruned_loss=0.05983, over 3050687.80 frames. ], batch size: 246, lr: 3.74e-03, grad_scale: 8.0 2023-04-30 18:21:04,873 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=179906.0, num_to_drop=1, layers_to_drop={0} 2023-04-30 18:21:25,092 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.20 vs. limit=2.0 2023-04-30 18:21:32,181 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=179924.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 18:22:16,601 INFO [train.py:904] (1/8) Epoch 18, batch 7400, loss[loss=0.2294, simple_loss=0.3054, pruned_loss=0.07665, over 11445.00 frames. ], tot_loss[loss=0.2059, simple_loss=0.2908, pruned_loss=0.06051, over 3061542.05 frames. ], batch size: 246, lr: 3.74e-03, grad_scale: 8.0 2023-04-30 18:22:19,967 INFO [optim.py:368] (1/8) 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 18:22:41,155 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=179967.0, num_to_drop=1, layers_to_drop={1} 2023-04-30 18:23:09,663 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=179985.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 18:23:21,186 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 18, batch 7450, loss[loss=0.1788, simple_loss=0.2676, pruned_loss=0.04503, over 17036.00 frames. ], tot_loss[loss=0.2066, simple_loss=0.2915, pruned_loss=0.0608, over 3091510.34 frames. ], batch size: 50, lr: 3.74e-03, grad_scale: 8.0 2023-04-30 18:24:05,954 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=180017.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 18:24:19,165 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=180026.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 18:24:42,562 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.9482, 4.1978, 3.9823, 4.0476, 3.7238, 3.8337, 3.8627, 4.1739], device='cuda:1'), covar=tensor([0.1054, 0.0903, 0.1045, 0.0822, 0.0784, 0.1444, 0.0886, 0.1067], device='cuda:1'), in_proj_covar=tensor([0.0624, 0.0768, 0.0627, 0.0575, 0.0480, 0.0493, 0.0640, 0.0598], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:1') 2023-04-30 18:24:50,243 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.6047, 2.5568, 1.8436, 2.6662, 2.0283, 2.7226, 2.0777, 2.3455], device='cuda:1'), covar=tensor([0.0317, 0.0378, 0.1214, 0.0288, 0.0686, 0.0516, 0.1184, 0.0596], device='cuda:1'), in_proj_covar=tensor([0.0163, 0.0172, 0.0191, 0.0151, 0.0175, 0.0212, 0.0199, 0.0175], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-04-30 18:25:01,488 INFO [train.py:904] (1/8) Epoch 18, batch 7500, loss[loss=0.2377, simple_loss=0.3077, pruned_loss=0.08383, over 11742.00 frames. ], tot_loss[loss=0.2065, simple_loss=0.2919, pruned_loss=0.06049, over 3068243.03 frames. ], batch size: 248, lr: 3.74e-03, grad_scale: 8.0 2023-04-30 18:25:04,520 INFO [optim.py:368] (1/8) 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 18:25:22,814 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=180065.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 18:25:56,688 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=180087.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 18:26:16,418 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 18, batch 7550, loss[loss=0.2132, simple_loss=0.2952, pruned_loss=0.06557, over 15321.00 frames. ], tot_loss[loss=0.2067, simple_loss=0.2913, pruned_loss=0.06105, over 3061626.34 frames. ], batch size: 190, lr: 3.74e-03, grad_scale: 8.0 2023-04-30 18:27:29,785 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 18, batch 7600, loss[loss=0.2306, simple_loss=0.3032, pruned_loss=0.07901, over 11384.00 frames. ], tot_loss[loss=0.2077, simple_loss=0.2916, pruned_loss=0.06188, over 3047662.41 frames. ], batch size: 248, lr: 3.74e-03, grad_scale: 8.0 2023-04-30 18:27:36,875 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=180152.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 18:27:39,407 INFO [optim.py:368] (1/8) 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 18:28:55,652 INFO [train.py:904] (1/8) Epoch 18, batch 7650, loss[loss=0.1924, simple_loss=0.2803, pruned_loss=0.05232, over 16473.00 frames. ], tot_loss[loss=0.2088, simple_loss=0.2928, pruned_loss=0.06247, over 3064292.96 frames. ], batch size: 35, lr: 3.74e-03, grad_scale: 8.0 2023-04-30 18:29:00,754 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=180205.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 18:30:13,288 INFO [train.py:904] (1/8) Epoch 18, batch 7700, loss[loss=0.2022, simple_loss=0.2918, pruned_loss=0.05627, over 16745.00 frames. ], tot_loss[loss=0.2089, simple_loss=0.2929, pruned_loss=0.06243, over 3077106.31 frames. ], batch size: 124, lr: 3.74e-03, grad_scale: 4.0 2023-04-30 18:30:18,206 INFO [optim.py:368] (1/8) 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 18:30:29,174 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=180262.0, num_to_drop=1, layers_to_drop={2} 2023-04-30 18:30:35,129 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=180266.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 18:30:47,578 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.0046, 2.0904, 2.1249, 3.6830, 2.0038, 2.4020, 2.1701, 2.2248], device='cuda:1'), covar=tensor([0.1364, 0.3692, 0.2908, 0.0543, 0.4297, 0.2485, 0.3583, 0.3263], device='cuda:1'), in_proj_covar=tensor([0.0385, 0.0428, 0.0353, 0.0320, 0.0429, 0.0493, 0.0398, 0.0499], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-30 18:30:56,908 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=180280.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 18:31:15,930 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=180292.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 18:31:30,812 INFO [train.py:904] (1/8) Epoch 18, batch 7750, loss[loss=0.1915, simple_loss=0.2914, pruned_loss=0.04581, over 16809.00 frames. ], tot_loss[loss=0.2089, simple_loss=0.2929, pruned_loss=0.06251, over 3053764.59 frames. ], batch size: 102, lr: 3.74e-03, grad_scale: 4.0 2023-04-30 18:31:40,829 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.6919, 4.5036, 4.7372, 4.8907, 5.0619, 4.5936, 5.0561, 5.0411], device='cuda:1'), covar=tensor([0.2043, 0.1269, 0.1672, 0.0785, 0.0572, 0.0926, 0.0621, 0.0617], device='cuda:1'), in_proj_covar=tensor([0.0599, 0.0733, 0.0865, 0.0749, 0.0560, 0.0591, 0.0613, 0.0703], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-30 18:32:23,273 INFO [zipformer.py:625] (1/8) 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] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=180340.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 18:32:46,606 INFO [train.py:904] (1/8) Epoch 18, batch 7800, loss[loss=0.1858, simple_loss=0.283, pruned_loss=0.04433, over 16769.00 frames. ], tot_loss[loss=0.2102, simple_loss=0.2941, pruned_loss=0.06314, over 3051316.06 frames. ], batch size: 83, lr: 3.74e-03, grad_scale: 4.0 2023-04-30 18:32:51,024 INFO [optim.py:368] (1/8) 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 18:33:33,620 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=180382.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 18:33:55,589 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 18, batch 7850, loss[loss=0.2069, simple_loss=0.2934, pruned_loss=0.06015, over 16442.00 frames. ], tot_loss[loss=0.2099, simple_loss=0.2945, pruned_loss=0.06269, over 3066213.05 frames. ], batch size: 146, lr: 3.74e-03, grad_scale: 4.0 2023-04-30 18:34:33,873 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=3.39 vs. limit=5.0 2023-04-30 18:35:16,216 INFO [train.py:904] (1/8) Epoch 18, batch 7900, loss[loss=0.2269, simple_loss=0.3003, pruned_loss=0.07681, over 11322.00 frames. ], tot_loss[loss=0.2081, simple_loss=0.2929, pruned_loss=0.06166, over 3072683.05 frames. ], batch size: 248, lr: 3.74e-03, grad_scale: 4.0 2023-04-30 18:35:17,206 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=180452.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 18:35:20,360 INFO [optim.py:368] (1/8) 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 18:36:31,923 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 18, batch 7950, loss[loss=0.2168, simple_loss=0.2926, pruned_loss=0.0705, over 15296.00 frames. ], tot_loss[loss=0.2091, simple_loss=0.2935, pruned_loss=0.06237, over 3066098.95 frames. ], batch size: 190, lr: 3.74e-03, grad_scale: 4.0 2023-04-30 18:37:52,807 INFO [train.py:904] (1/8) Epoch 18, batch 8000, loss[loss=0.2091, simple_loss=0.2975, pruned_loss=0.06031, over 17016.00 frames. ], tot_loss[loss=0.2121, simple_loss=0.2953, pruned_loss=0.06448, over 3019860.92 frames. ], batch size: 53, lr: 3.74e-03, grad_scale: 8.0 2023-04-30 18:37:57,086 INFO [optim.py:368] (1/8) 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 18:38:07,592 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=180561.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 18:38:08,969 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=180562.0, num_to_drop=1, layers_to_drop={0} 2023-04-30 18:38:21,392 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=180570.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 18:38:35,516 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.4715, 3.5359, 3.2527, 2.9665, 3.1175, 3.4271, 3.3074, 3.2295], device='cuda:1'), covar=tensor([0.0651, 0.0644, 0.0283, 0.0259, 0.0525, 0.0507, 0.1044, 0.0503], device='cuda:1'), in_proj_covar=tensor([0.0275, 0.0392, 0.0321, 0.0312, 0.0332, 0.0365, 0.0221, 0.0385], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-30 18:38:36,574 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 18, batch 8050, loss[loss=0.231, simple_loss=0.3054, pruned_loss=0.07828, over 11668.00 frames. ], tot_loss[loss=0.2131, simple_loss=0.2959, pruned_loss=0.06514, over 2998337.38 frames. ], batch size: 247, lr: 3.74e-03, grad_scale: 4.0 2023-04-30 18:39:23,082 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=180610.0, num_to_drop=1, layers_to_drop={0} 2023-04-30 18:39:37,258 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.43 vs. limit=2.0 2023-04-30 18:39:45,790 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.6397, 1.7568, 2.2985, 2.6298, 2.5532, 3.0961, 1.8514, 3.0591], device='cuda:1'), covar=tensor([0.0223, 0.0507, 0.0318, 0.0330, 0.0325, 0.0157, 0.0507, 0.0148], device='cuda:1'), in_proj_covar=tensor([0.0177, 0.0186, 0.0172, 0.0174, 0.0184, 0.0143, 0.0188, 0.0138], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-30 18:39:50,352 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=180628.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 18:39:54,572 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=180631.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 18:39:56,012 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.4025, 2.9015, 3.0022, 1.9138, 2.6718, 2.1204, 2.8989, 3.1979], device='cuda:1'), covar=tensor([0.0398, 0.0876, 0.0620, 0.2090, 0.0938, 0.0997, 0.0852, 0.0924], device='cuda:1'), in_proj_covar=tensor([0.0151, 0.0158, 0.0163, 0.0149, 0.0141, 0.0127, 0.0140, 0.0166], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:1') 2023-04-30 18:40:26,588 INFO [train.py:904] (1/8) Epoch 18, batch 8100, loss[loss=0.191, simple_loss=0.2817, pruned_loss=0.05015, over 16486.00 frames. ], tot_loss[loss=0.2113, simple_loss=0.2949, pruned_loss=0.06385, over 3024812.86 frames. ], batch size: 146, lr: 3.73e-03, grad_scale: 4.0 2023-04-30 18:40:32,030 INFO [optim.py:368] (1/8) 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 18:40:52,500 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.22 vs. limit=2.0 2023-04-30 18:41:11,940 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=180682.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 18:41:26,890 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=180692.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 18:41:41,227 INFO [train.py:904] (1/8) Epoch 18, batch 8150, loss[loss=0.1851, simple_loss=0.269, pruned_loss=0.05063, over 16714.00 frames. ], tot_loss[loss=0.2078, simple_loss=0.2916, pruned_loss=0.06196, over 3048313.68 frames. ], batch size: 134, lr: 3.73e-03, grad_scale: 4.0 2023-04-30 18:41:57,661 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.7055, 4.7004, 5.1349, 5.0858, 5.1038, 4.8007, 4.7557, 4.5737], device='cuda:1'), covar=tensor([0.0321, 0.0597, 0.0327, 0.0407, 0.0508, 0.0363, 0.0964, 0.0458], device='cuda:1'), in_proj_covar=tensor([0.0382, 0.0418, 0.0407, 0.0385, 0.0454, 0.0429, 0.0524, 0.0342], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:1') 2023-04-30 18:42:21,260 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.27 vs. limit=2.0 2023-04-30 18:42:24,162 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=180730.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 18:42:26,657 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=180732.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 18:42:56,046 INFO [train.py:904] (1/8) Epoch 18, batch 8200, loss[loss=0.2097, simple_loss=0.307, pruned_loss=0.05623, over 16366.00 frames. ], tot_loss[loss=0.2055, simple_loss=0.2889, pruned_loss=0.06099, over 3071147.01 frames. ], batch size: 146, lr: 3.73e-03, grad_scale: 4.0 2023-04-30 18:43:02,071 INFO [optim.py:368] (1/8) 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 18:44:01,797 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=180793.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 18:44:14,887 INFO [train.py:904] (1/8) Epoch 18, batch 8250, loss[loss=0.2018, simple_loss=0.3099, pruned_loss=0.04678, over 16387.00 frames. ], tot_loss[loss=0.2024, simple_loss=0.2881, pruned_loss=0.05836, over 3069399.12 frames. ], batch size: 146, lr: 3.73e-03, grad_scale: 4.0 2023-04-30 18:44:53,366 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=180825.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 18:45:32,726 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.7647, 1.7529, 2.3780, 2.7986, 2.6583, 3.1048, 2.1321, 3.1573], device='cuda:1'), covar=tensor([0.0205, 0.0536, 0.0335, 0.0270, 0.0275, 0.0184, 0.0466, 0.0139], device='cuda:1'), in_proj_covar=tensor([0.0177, 0.0185, 0.0172, 0.0174, 0.0182, 0.0142, 0.0187, 0.0137], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-30 18:45:37,335 INFO [train.py:904] (1/8) Epoch 18, batch 8300, loss[loss=0.1826, simple_loss=0.271, pruned_loss=0.0471, over 16990.00 frames. ], tot_loss[loss=0.1981, simple_loss=0.2854, pruned_loss=0.05542, over 3061488.47 frames. ], batch size: 41, lr: 3.73e-03, grad_scale: 4.0 2023-04-30 18:45:43,842 INFO [optim.py:368] (1/8) 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 18:45:47,943 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=4.62 vs. limit=5.0 2023-04-30 18:45:52,363 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=180861.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 18:46:32,561 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=180886.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 18:46:58,361 INFO [train.py:904] (1/8) Epoch 18, batch 8350, loss[loss=0.1824, simple_loss=0.2797, pruned_loss=0.04253, over 16877.00 frames. ], tot_loss[loss=0.1957, simple_loss=0.2844, pruned_loss=0.05353, over 3042743.12 frames. ], batch size: 116, lr: 3.73e-03, grad_scale: 4.0 2023-04-30 18:47:09,895 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=180909.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 18:47:35,931 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 18, batch 8400, loss[loss=0.1817, simple_loss=0.2795, pruned_loss=0.04198, over 16702.00 frames. ], tot_loss[loss=0.1918, simple_loss=0.2811, pruned_loss=0.05123, over 3036140.19 frames. ], batch size: 124, lr: 3.73e-03, grad_scale: 8.0 2023-04-30 18:48:22,151 INFO [optim.py:368] (1/8) 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 18:48:25,177 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.78 vs. limit=2.0 2023-04-30 18:49:16,522 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=180992.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 18:49:31,293 INFO [train.py:904] (1/8) Epoch 18, batch 8450, loss[loss=0.1619, simple_loss=0.2588, pruned_loss=0.03254, over 16856.00 frames. ], tot_loss[loss=0.1892, simple_loss=0.2795, pruned_loss=0.04942, over 3042589.35 frames. ], batch size: 96, lr: 3.73e-03, grad_scale: 8.0 2023-04-30 18:50:26,693 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.4169, 3.5842, 3.6619, 2.5796, 3.3306, 3.6169, 3.4199, 2.1225], device='cuda:1'), covar=tensor([0.0456, 0.0067, 0.0047, 0.0328, 0.0104, 0.0101, 0.0085, 0.0463], device='cuda:1'), in_proj_covar=tensor([0.0133, 0.0077, 0.0078, 0.0130, 0.0091, 0.0103, 0.0090, 0.0124], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-04-30 18:50:28,521 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=4.87 vs. limit=5.0 2023-04-30 18:50:31,764 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 18, batch 8500, loss[loss=0.1584, simple_loss=0.2541, pruned_loss=0.03136, over 16350.00 frames. ], tot_loss[loss=0.1855, simple_loss=0.2763, pruned_loss=0.04736, over 3059618.19 frames. ], batch size: 146, lr: 3.73e-03, grad_scale: 4.0 2023-04-30 18:50:58,701 INFO [optim.py:368] (1/8) 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 18:51:48,659 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=181088.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 18:52:13,295 INFO [train.py:904] (1/8) Epoch 18, batch 8550, loss[loss=0.1908, simple_loss=0.2846, pruned_loss=0.04851, over 16686.00 frames. ], tot_loss[loss=0.1834, simple_loss=0.274, pruned_loss=0.04641, over 3026806.47 frames. ], batch size: 134, lr: 3.73e-03, grad_scale: 4.0 2023-04-30 18:53:50,704 INFO [train.py:904] (1/8) Epoch 18, batch 8600, loss[loss=0.1899, simple_loss=0.2861, pruned_loss=0.04681, over 16153.00 frames. ], tot_loss[loss=0.1824, simple_loss=0.2743, pruned_loss=0.04529, over 3035519.77 frames. ], batch size: 165, lr: 3.73e-03, grad_scale: 4.0 2023-04-30 18:54:01,221 INFO [optim.py:368] (1/8) 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 18:54:42,930 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.9425, 1.9449, 2.1732, 3.4372, 1.9739, 2.1696, 2.1323, 2.0339], device='cuda:1'), covar=tensor([0.1330, 0.3892, 0.2979, 0.0605, 0.4879, 0.3037, 0.3794, 0.4060], device='cuda:1'), in_proj_covar=tensor([0.0378, 0.0420, 0.0347, 0.0311, 0.0421, 0.0482, 0.0390, 0.0491], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-30 18:54:50,019 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 18, batch 8650, loss[loss=0.1568, simple_loss=0.2622, pruned_loss=0.02564, over 16845.00 frames. ], tot_loss[loss=0.18, simple_loss=0.2722, pruned_loss=0.04387, over 3031358.97 frames. ], batch size: 102, lr: 3.73e-03, grad_scale: 4.0 2023-04-30 18:56:06,851 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.3502, 3.5127, 3.6885, 3.6615, 3.6702, 3.4936, 3.5143, 3.5513], device='cuda:1'), covar=tensor([0.0327, 0.0627, 0.0435, 0.0416, 0.0449, 0.0479, 0.0778, 0.0465], device='cuda:1'), in_proj_covar=tensor([0.0372, 0.0407, 0.0398, 0.0374, 0.0442, 0.0417, 0.0510, 0.0335], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:1') 2023-04-30 18:56:25,647 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=181226.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 18:57:07,000 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.6921, 3.7197, 2.9250, 2.2342, 2.2734, 2.3855, 3.8828, 3.2288], device='cuda:1'), covar=tensor([0.2708, 0.0652, 0.1709, 0.2905, 0.2902, 0.2107, 0.0439, 0.1360], device='cuda:1'), in_proj_covar=tensor([0.0313, 0.0257, 0.0292, 0.0296, 0.0283, 0.0242, 0.0280, 0.0317], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:1') 2023-04-30 18:57:15,663 INFO [train.py:904] (1/8) Epoch 18, batch 8700, loss[loss=0.1729, simple_loss=0.256, pruned_loss=0.04495, over 12348.00 frames. ], tot_loss[loss=0.1775, simple_loss=0.2696, pruned_loss=0.04273, over 3033382.33 frames. ], batch size: 248, lr: 3.73e-03, grad_scale: 4.0 2023-04-30 18:57:21,397 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=3.01 vs. limit=5.0 2023-04-30 18:57:25,073 INFO [optim.py:368] (1/8) 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 18:57:56,330 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=181274.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 18:58:39,136 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=181297.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 18:58:49,004 INFO [train.py:904] (1/8) Epoch 18, batch 8750, loss[loss=0.1789, simple_loss=0.2747, pruned_loss=0.04156, over 12244.00 frames. ], tot_loss[loss=0.1768, simple_loss=0.2693, pruned_loss=0.04208, over 3037886.16 frames. ], batch size: 247, lr: 3.73e-03, grad_scale: 4.0 2023-04-30 18:58:55,471 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.9986, 2.0463, 2.1438, 3.4814, 1.9948, 2.3182, 2.1865, 2.1516], device='cuda:1'), covar=tensor([0.1251, 0.3911, 0.3013, 0.0576, 0.4583, 0.2689, 0.3586, 0.3681], device='cuda:1'), in_proj_covar=tensor([0.0377, 0.0420, 0.0347, 0.0311, 0.0420, 0.0481, 0.0389, 0.0489], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-30 18:59:04,370 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.9077, 5.2462, 5.0108, 5.0189, 4.7806, 4.7402, 4.5933, 5.3002], device='cuda:1'), covar=tensor([0.1083, 0.0754, 0.0901, 0.0745, 0.0737, 0.0868, 0.1155, 0.0815], device='cuda:1'), in_proj_covar=tensor([0.0617, 0.0753, 0.0622, 0.0563, 0.0472, 0.0488, 0.0632, 0.0587], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:1') 2023-04-30 19:00:27,152 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.4559, 4.4585, 4.7891, 4.7660, 4.7782, 4.5239, 4.4735, 4.4216], device='cuda:1'), covar=tensor([0.0237, 0.0548, 0.0385, 0.0387, 0.0391, 0.0356, 0.0878, 0.0391], device='cuda:1'), in_proj_covar=tensor([0.0370, 0.0403, 0.0396, 0.0373, 0.0437, 0.0414, 0.0507, 0.0333], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003], device='cuda:1') 2023-04-30 19:00:41,585 INFO [train.py:904] (1/8) Epoch 18, batch 8800, loss[loss=0.1765, simple_loss=0.2654, pruned_loss=0.0438, over 12785.00 frames. ], tot_loss[loss=0.1745, simple_loss=0.2673, pruned_loss=0.0408, over 3035905.66 frames. ], batch size: 248, lr: 3.73e-03, grad_scale: 8.0 2023-04-30 19:00:51,212 INFO [optim.py:368] (1/8) 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 19:00:54,097 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=181358.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 19:01:57,889 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=181388.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 19:02:27,198 INFO [train.py:904] (1/8) Epoch 18, batch 8850, loss[loss=0.1454, simple_loss=0.2372, pruned_loss=0.02682, over 12397.00 frames. ], tot_loss[loss=0.175, simple_loss=0.2695, pruned_loss=0.04026, over 3018878.34 frames. ], batch size: 248, lr: 3.73e-03, grad_scale: 8.0 2023-04-30 19:03:44,279 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=181436.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 19:04:05,102 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.4641, 2.0743, 1.7760, 1.7591, 2.3526, 2.0438, 2.0416, 2.4084], device='cuda:1'), covar=tensor([0.0167, 0.0329, 0.0431, 0.0421, 0.0211, 0.0314, 0.0177, 0.0222], device='cuda:1'), in_proj_covar=tensor([0.0182, 0.0219, 0.0213, 0.0213, 0.0221, 0.0218, 0.0220, 0.0212], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-30 19:04:15,168 INFO [train.py:904] (1/8) Epoch 18, batch 8900, loss[loss=0.1854, simple_loss=0.2774, pruned_loss=0.04674, over 16400.00 frames. ], tot_loss[loss=0.1746, simple_loss=0.2699, pruned_loss=0.0396, over 3028304.81 frames. ], batch size: 146, lr: 3.73e-03, grad_scale: 8.0 2023-04-30 19:04:16,368 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.8943, 2.0685, 2.3242, 3.2394, 2.0555, 2.2697, 2.2126, 2.1229], device='cuda:1'), covar=tensor([0.1288, 0.3503, 0.2683, 0.0604, 0.4424, 0.2564, 0.3643, 0.3815], device='cuda:1'), in_proj_covar=tensor([0.0379, 0.0422, 0.0349, 0.0312, 0.0422, 0.0483, 0.0392, 0.0490], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-30 19:04:25,752 INFO [optim.py:368] (1/8) 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 19:05:23,181 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 18, batch 8950, loss[loss=0.1623, simple_loss=0.2578, pruned_loss=0.03341, over 16672.00 frames. ], tot_loss[loss=0.1745, simple_loss=0.2696, pruned_loss=0.03968, over 3062829.05 frames. ], batch size: 134, lr: 3.73e-03, grad_scale: 8.0 2023-04-30 19:07:17,051 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=181529.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 19:07:19,774 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=181530.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 19:07:39,989 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.25 vs. limit=2.0 2023-04-30 19:08:08,279 INFO [train.py:904] (1/8) Epoch 18, batch 9000, loss[loss=0.1655, simple_loss=0.2542, pruned_loss=0.03834, over 11994.00 frames. ], tot_loss[loss=0.1726, simple_loss=0.2675, pruned_loss=0.03887, over 3072435.43 frames. ], batch size: 246, lr: 3.73e-03, grad_scale: 8.0 2023-04-30 19:08:08,280 INFO [train.py:929] (1/8) Computing validation loss 2023-04-30 19:08:17,832 INFO [train.py:938] (1/8) Epoch 18, validation: loss=0.1475, simple_loss=0.2516, pruned_loss=0.02169, over 944034.00 frames. 2023-04-30 19:08:17,833 INFO [train.py:939] (1/8) Maximum memory allocated so far is 18145MB 2023-04-30 19:08:27,945 INFO [optim.py:368] (1/8) 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 19:09:39,060 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 18, batch 9050, loss[loss=0.1779, simple_loss=0.265, pruned_loss=0.04541, over 12829.00 frames. ], tot_loss[loss=0.1735, simple_loss=0.2679, pruned_loss=0.03956, over 3061087.39 frames. ], batch size: 246, lr: 3.72e-03, grad_scale: 8.0 2023-04-30 19:10:43,887 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.32 vs. limit=2.0 2023-04-30 19:11:46,775 INFO [train.py:904] (1/8) Epoch 18, batch 9100, loss[loss=0.1628, simple_loss=0.2583, pruned_loss=0.03364, over 16630.00 frames. ], tot_loss[loss=0.1741, simple_loss=0.268, pruned_loss=0.04008, over 3056381.17 frames. ], batch size: 62, lr: 3.72e-03, grad_scale: 8.0 2023-04-30 19:11:50,126 INFO [zipformer.py:625] (1/8) 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] (1/8) Clipping_scale=2.0, grad-norm quartiles 1.522e+02 2.401e+02 2.871e+02 3.524e+02 6.480e+02, threshold=5.743e+02, percent-clipped=6.0 2023-04-30 19:12:51,157 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.0732, 3.9276, 4.1208, 4.2634, 4.3739, 3.9245, 4.3244, 4.3831], device='cuda:1'), covar=tensor([0.1574, 0.1077, 0.1399, 0.0660, 0.0543, 0.1321, 0.0639, 0.0630], device='cuda:1'), in_proj_covar=tensor([0.0572, 0.0703, 0.0824, 0.0723, 0.0538, 0.0570, 0.0586, 0.0676], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-30 19:13:43,587 INFO [train.py:904] (1/8) Epoch 18, batch 9150, loss[loss=0.1578, simple_loss=0.2544, pruned_loss=0.0306, over 16706.00 frames. ], tot_loss[loss=0.1736, simple_loss=0.268, pruned_loss=0.03962, over 3049513.35 frames. ], batch size: 89, lr: 3.72e-03, grad_scale: 8.0 2023-04-30 19:13:54,993 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.37 vs. limit=2.0 2023-04-30 19:14:04,754 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.1057, 2.5608, 2.6577, 1.8981, 2.8091, 2.8555, 2.5111, 2.4980], device='cuda:1'), covar=tensor([0.0604, 0.0213, 0.0200, 0.0936, 0.0086, 0.0191, 0.0418, 0.0386], device='cuda:1'), in_proj_covar=tensor([0.0142, 0.0102, 0.0090, 0.0136, 0.0072, 0.0115, 0.0121, 0.0125], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:1') 2023-04-30 19:15:27,791 INFO [train.py:904] (1/8) Epoch 18, batch 9200, loss[loss=0.1789, simple_loss=0.2682, pruned_loss=0.04482, over 16496.00 frames. ], tot_loss[loss=0.17, simple_loss=0.2635, pruned_loss=0.03824, over 3072921.12 frames. ], batch size: 147, lr: 3.72e-03, grad_scale: 8.0 2023-04-30 19:15:36,945 INFO [optim.py:368] (1/8) 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 19:16:37,782 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.9137, 1.9819, 2.3138, 3.2004, 2.1162, 2.1905, 2.2101, 2.0631], device='cuda:1'), covar=tensor([0.1224, 0.3948, 0.2513, 0.0621, 0.4307, 0.2718, 0.3616, 0.3851], device='cuda:1'), in_proj_covar=tensor([0.0377, 0.0420, 0.0347, 0.0310, 0.0420, 0.0479, 0.0389, 0.0486], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-30 19:16:49,002 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.5416, 3.9092, 3.8287, 2.6715, 3.4940, 3.8145, 3.5658, 2.3182], device='cuda:1'), covar=tensor([0.0420, 0.0033, 0.0039, 0.0318, 0.0084, 0.0079, 0.0064, 0.0392], device='cuda:1'), in_proj_covar=tensor([0.0128, 0.0074, 0.0075, 0.0126, 0.0089, 0.0097, 0.0086, 0.0120], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-04-30 19:16:49,055 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.0506, 2.2759, 1.9368, 2.0296, 2.6498, 2.3724, 2.6550, 2.8213], device='cuda:1'), covar=tensor([0.0145, 0.0447, 0.0550, 0.0503, 0.0310, 0.0417, 0.0216, 0.0286], device='cuda:1'), in_proj_covar=tensor([0.0184, 0.0222, 0.0216, 0.0216, 0.0223, 0.0221, 0.0221, 0.0213], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-30 19:17:05,940 INFO [train.py:904] (1/8) Epoch 18, batch 9250, loss[loss=0.1451, simple_loss=0.2427, pruned_loss=0.02375, over 16924.00 frames. ], tot_loss[loss=0.1701, simple_loss=0.2633, pruned_loss=0.03842, over 3060280.55 frames. ], batch size: 96, lr: 3.72e-03, grad_scale: 8.0 2023-04-30 19:18:57,424 INFO [train.py:904] (1/8) Epoch 18, batch 9300, loss[loss=0.1733, simple_loss=0.261, pruned_loss=0.04284, over 16720.00 frames. ], tot_loss[loss=0.1687, simple_loss=0.2618, pruned_loss=0.03777, over 3059845.64 frames. ], batch size: 134, lr: 3.72e-03, grad_scale: 8.0 2023-04-30 19:19:06,077 INFO [optim.py:368] (1/8) Clipping_scale=2.0, grad-norm quartiles 1.537e+02 2.346e+02 2.677e+02 3.465e+02 6.012e+02, threshold=5.355e+02, percent-clipped=4.0 2023-04-30 19:20:12,871 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 18, batch 9350, loss[loss=0.1897, simple_loss=0.2752, pruned_loss=0.05204, over 15368.00 frames. ], tot_loss[loss=0.1689, simple_loss=0.2621, pruned_loss=0.03786, over 3061339.11 frames. ], batch size: 191, lr: 3.72e-03, grad_scale: 8.0 2023-04-30 19:22:24,027 INFO [train.py:904] (1/8) Epoch 18, batch 9400, loss[loss=0.1794, simple_loss=0.276, pruned_loss=0.04138, over 15291.00 frames. ], tot_loss[loss=0.1685, simple_loss=0.2617, pruned_loss=0.03767, over 3050306.78 frames. ], batch size: 190, lr: 3.72e-03, grad_scale: 8.0 2023-04-30 19:22:28,042 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=181953.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 19:22:33,106 INFO [optim.py:368] (1/8) 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 19:22:47,765 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.58 vs. limit=2.0 2023-04-30 19:23:16,405 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=181978.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 19:23:29,423 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.2786, 3.5601, 3.8875, 2.1612, 3.2126, 2.4815, 3.6878, 3.5929], device='cuda:1'), covar=tensor([0.0235, 0.0771, 0.0435, 0.1872, 0.0698, 0.0875, 0.0600, 0.1070], device='cuda:1'), in_proj_covar=tensor([0.0146, 0.0151, 0.0158, 0.0144, 0.0137, 0.0123, 0.0136, 0.0159], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:1') 2023-04-30 19:23:35,417 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([1.9663, 1.8403, 1.6620, 1.4573, 1.9684, 1.6323, 1.6315, 1.9090], device='cuda:1'), covar=tensor([0.0170, 0.0271, 0.0404, 0.0391, 0.0238, 0.0294, 0.0169, 0.0234], device='cuda:1'), in_proj_covar=tensor([0.0184, 0.0221, 0.0216, 0.0215, 0.0222, 0.0220, 0.0220, 0.0213], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-30 19:24:05,161 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 18, batch 9450, loss[loss=0.1541, simple_loss=0.243, pruned_loss=0.03257, over 12116.00 frames. ], tot_loss[loss=0.1698, simple_loss=0.2633, pruned_loss=0.03816, over 3024680.92 frames. ], batch size: 246, lr: 3.72e-03, grad_scale: 8.0 2023-04-30 19:25:19,113 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=182039.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 19:25:41,960 INFO [train.py:904] (1/8) Epoch 18, batch 9500, loss[loss=0.1727, simple_loss=0.2653, pruned_loss=0.04, over 16662.00 frames. ], tot_loss[loss=0.1691, simple_loss=0.2626, pruned_loss=0.03777, over 3042224.41 frames. ], batch size: 62, lr: 3.72e-03, grad_scale: 8.0 2023-04-30 19:25:55,086 INFO [optim.py:368] (1/8) 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 19:25:59,762 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([1.6287, 2.5895, 2.5031, 4.0642, 2.4196, 3.9979, 1.5003, 2.8386], device='cuda:1'), covar=tensor([0.1528, 0.0839, 0.1207, 0.0215, 0.0144, 0.0367, 0.1817, 0.0794], device='cuda:1'), in_proj_covar=tensor([0.0162, 0.0166, 0.0187, 0.0173, 0.0196, 0.0209, 0.0193, 0.0185], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-04-30 19:26:35,802 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.4012, 3.0514, 2.7230, 2.2893, 2.2267, 2.3114, 2.9828, 2.8776], device='cuda:1'), covar=tensor([0.2369, 0.0732, 0.1510, 0.2472, 0.2463, 0.2013, 0.0417, 0.1243], device='cuda:1'), in_proj_covar=tensor([0.0310, 0.0254, 0.0289, 0.0293, 0.0277, 0.0241, 0.0278, 0.0313], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-30 19:26:44,672 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([1.9998, 2.8344, 2.6859, 2.0482, 2.5487, 2.8231, 2.6294, 1.9699], device='cuda:1'), covar=tensor([0.0356, 0.0060, 0.0059, 0.0272, 0.0119, 0.0097, 0.0093, 0.0382], device='cuda:1'), in_proj_covar=tensor([0.0130, 0.0075, 0.0076, 0.0127, 0.0090, 0.0099, 0.0087, 0.0121], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-04-30 19:27:10,326 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.0400, 2.8022, 2.9718, 2.0878, 2.7506, 2.1384, 2.7383, 2.9371], device='cuda:1'), covar=tensor([0.0269, 0.0843, 0.0492, 0.1720, 0.0704, 0.0897, 0.0624, 0.0797], device='cuda:1'), in_proj_covar=tensor([0.0146, 0.0151, 0.0159, 0.0145, 0.0137, 0.0123, 0.0137, 0.0159], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:1') 2023-04-30 19:27:27,071 INFO [train.py:904] (1/8) Epoch 18, batch 9550, loss[loss=0.1681, simple_loss=0.268, pruned_loss=0.03407, over 16342.00 frames. ], tot_loss[loss=0.1685, simple_loss=0.2619, pruned_loss=0.03761, over 3046838.22 frames. ], batch size: 146, lr: 3.72e-03, grad_scale: 8.0 2023-04-30 19:28:23,514 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.77 vs. limit=2.0 2023-04-30 19:29:06,815 INFO [train.py:904] (1/8) Epoch 18, batch 9600, loss[loss=0.1527, simple_loss=0.2541, pruned_loss=0.02562, over 16547.00 frames. ], tot_loss[loss=0.1696, simple_loss=0.2628, pruned_loss=0.03822, over 3044857.82 frames. ], batch size: 68, lr: 3.72e-03, grad_scale: 8.0 2023-04-30 19:29:15,380 INFO [optim.py:368] (1/8) 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 19:30:14,116 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=182186.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 19:30:52,020 INFO [train.py:904] (1/8) Epoch 18, batch 9650, loss[loss=0.1914, simple_loss=0.2885, pruned_loss=0.04715, over 16833.00 frames. ], tot_loss[loss=0.171, simple_loss=0.2649, pruned_loss=0.03856, over 3032212.22 frames. ], batch size: 124, lr: 3.72e-03, grad_scale: 8.0 2023-04-30 19:32:00,667 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=182234.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 19:32:37,722 INFO [train.py:904] (1/8) Epoch 18, batch 9700, loss[loss=0.1864, simple_loss=0.2763, pruned_loss=0.04819, over 16197.00 frames. ], tot_loss[loss=0.1705, simple_loss=0.2644, pruned_loss=0.03824, over 3058507.92 frames. ], batch size: 165, lr: 3.72e-03, grad_scale: 8.0 2023-04-30 19:32:45,794 INFO [optim.py:368] (1/8) 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 19:33:22,366 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([1.8512, 1.2971, 1.7325, 1.6278, 1.8760, 1.9318, 1.5824, 1.8293], device='cuda:1'), covar=tensor([0.0214, 0.0407, 0.0203, 0.0295, 0.0258, 0.0205, 0.0403, 0.0134], device='cuda:1'), in_proj_covar=tensor([0.0171, 0.0180, 0.0167, 0.0168, 0.0178, 0.0137, 0.0183, 0.0133], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-30 19:33:37,395 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=182281.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 19:34:17,889 INFO [train.py:904] (1/8) Epoch 18, batch 9750, loss[loss=0.1709, simple_loss=0.2643, pruned_loss=0.0387, over 16661.00 frames. ], tot_loss[loss=0.1705, simple_loss=0.2639, pruned_loss=0.03853, over 3060441.24 frames. ], batch size: 134, lr: 3.72e-03, grad_scale: 8.0 2023-04-30 19:35:06,708 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.33 vs. limit=2.0 2023-04-30 19:35:24,325 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=182334.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 19:35:39,072 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 18, batch 9800, loss[loss=0.1819, simple_loss=0.2813, pruned_loss=0.04125, over 16401.00 frames. ], tot_loss[loss=0.17, simple_loss=0.2644, pruned_loss=0.0378, over 3070002.46 frames. ], batch size: 146, lr: 3.72e-03, grad_scale: 8.0 2023-04-30 19:36:05,429 INFO [optim.py:368] (1/8) 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 19:36:24,584 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([1.9950, 3.2693, 3.2135, 2.2209, 2.9401, 3.2354, 3.0430, 2.1157], device='cuda:1'), covar=tensor([0.0485, 0.0042, 0.0049, 0.0326, 0.0115, 0.0072, 0.0078, 0.0386], device='cuda:1'), in_proj_covar=tensor([0.0130, 0.0075, 0.0075, 0.0126, 0.0090, 0.0098, 0.0087, 0.0121], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-04-30 19:37:01,181 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=2.90 vs. limit=5.0 2023-04-30 19:37:32,110 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.33 vs. limit=2.0 2023-04-30 19:37:39,079 INFO [train.py:904] (1/8) Epoch 18, batch 9850, loss[loss=0.159, simple_loss=0.2617, pruned_loss=0.02815, over 16676.00 frames. ], tot_loss[loss=0.17, simple_loss=0.2651, pruned_loss=0.03749, over 3061784.25 frames. ], batch size: 89, lr: 3.72e-03, grad_scale: 8.0 2023-04-30 19:39:29,923 INFO [train.py:904] (1/8) Epoch 18, batch 9900, loss[loss=0.1994, simple_loss=0.2944, pruned_loss=0.05226, over 16948.00 frames. ], tot_loss[loss=0.1699, simple_loss=0.2652, pruned_loss=0.03727, over 3061132.08 frames. ], batch size: 125, lr: 3.72e-03, grad_scale: 8.0 2023-04-30 19:39:40,680 INFO [optim.py:368] (1/8) Clipping_scale=2.0, grad-norm quartiles 1.546e+02 2.082e+02 2.451e+02 2.864e+02 7.377e+02, threshold=4.903e+02, percent-clipped=2.0 2023-04-30 19:40:53,903 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.8291, 3.8393, 2.2791, 4.3684, 2.9600, 4.3018, 2.3755, 3.1503], device='cuda:1'), covar=tensor([0.0204, 0.0284, 0.1526, 0.0193, 0.0803, 0.0369, 0.1607, 0.0651], device='cuda:1'), in_proj_covar=tensor([0.0158, 0.0166, 0.0185, 0.0143, 0.0168, 0.0200, 0.0193, 0.0169], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0003, 0.0004, 0.0004, 0.0003], device='cuda:1') 2023-04-30 19:41:28,674 INFO [train.py:904] (1/8) Epoch 18, batch 9950, loss[loss=0.1759, simple_loss=0.2747, pruned_loss=0.03854, over 16844.00 frames. ], tot_loss[loss=0.1714, simple_loss=0.2675, pruned_loss=0.03764, over 3062115.72 frames. ], batch size: 124, lr: 3.72e-03, grad_scale: 8.0 2023-04-30 19:41:33,581 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.8333, 4.8640, 4.6552, 4.2477, 4.7178, 1.7809, 4.5120, 4.4896], device='cuda:1'), covar=tensor([0.0069, 0.0071, 0.0166, 0.0240, 0.0079, 0.2527, 0.0106, 0.0184], device='cuda:1'), in_proj_covar=tensor([0.0145, 0.0134, 0.0175, 0.0158, 0.0154, 0.0190, 0.0166, 0.0154], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-30 19:41:58,902 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=182514.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 19:42:47,917 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.3259, 3.4389, 3.6682, 3.6369, 3.6641, 3.4501, 3.4967, 3.5594], device='cuda:1'), covar=tensor([0.0462, 0.1097, 0.0549, 0.0529, 0.0512, 0.0641, 0.0754, 0.0449], device='cuda:1'), in_proj_covar=tensor([0.0361, 0.0393, 0.0384, 0.0364, 0.0428, 0.0404, 0.0489, 0.0326], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003], device='cuda:1') 2023-04-30 19:43:31,283 INFO [train.py:904] (1/8) Epoch 18, batch 10000, loss[loss=0.1597, simple_loss=0.2577, pruned_loss=0.03087, over 16576.00 frames. ], tot_loss[loss=0.1705, simple_loss=0.2661, pruned_loss=0.03739, over 3087843.32 frames. ], batch size: 62, lr: 3.72e-03, grad_scale: 8.0 2023-04-30 19:43:42,215 INFO [optim.py:368] (1/8) 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 19:44:16,755 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=182575.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 19:44:21,520 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.4263, 3.3576, 3.4951, 3.5604, 3.5933, 3.3172, 3.5673, 3.6440], device='cuda:1'), covar=tensor([0.1291, 0.0905, 0.1014, 0.0635, 0.0547, 0.1851, 0.0773, 0.0671], device='cuda:1'), in_proj_covar=tensor([0.0568, 0.0696, 0.0814, 0.0719, 0.0537, 0.0561, 0.0585, 0.0673], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:1') 2023-04-30 19:44:27,086 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.0012, 2.2988, 2.3859, 2.9425, 1.8488, 3.3424, 1.6640, 2.7382], device='cuda:1'), covar=tensor([0.1228, 0.0663, 0.1002, 0.0132, 0.0080, 0.0336, 0.1505, 0.0683], device='cuda:1'), in_proj_covar=tensor([0.0162, 0.0166, 0.0188, 0.0172, 0.0194, 0.0207, 0.0192, 0.0185], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-04-30 19:45:02,052 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.2004, 1.5431, 1.9106, 2.1176, 2.2166, 2.3595, 1.6829, 2.3141], device='cuda:1'), covar=tensor([0.0201, 0.0429, 0.0276, 0.0296, 0.0282, 0.0205, 0.0459, 0.0132], device='cuda:1'), in_proj_covar=tensor([0.0171, 0.0180, 0.0169, 0.0168, 0.0179, 0.0137, 0.0183, 0.0133], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-30 19:45:09,945 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.83 vs. limit=2.0 2023-04-30 19:45:14,019 INFO [train.py:904] (1/8) Epoch 18, batch 10050, loss[loss=0.1775, simple_loss=0.2641, pruned_loss=0.04543, over 11805.00 frames. ], tot_loss[loss=0.1706, simple_loss=0.2662, pruned_loss=0.03753, over 3065210.67 frames. ], batch size: 248, lr: 3.71e-03, grad_scale: 8.0 2023-04-30 19:45:16,159 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.9132, 2.1083, 2.3585, 3.2078, 2.1852, 2.3011, 2.3035, 2.1894], device='cuda:1'), covar=tensor([0.1139, 0.3544, 0.2423, 0.0605, 0.4297, 0.2685, 0.3144, 0.3763], device='cuda:1'), in_proj_covar=tensor([0.0376, 0.0417, 0.0347, 0.0308, 0.0419, 0.0475, 0.0386, 0.0484], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-30 19:46:14,995 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=182634.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 19:46:20,796 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=182637.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 19:46:22,244 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.7249, 4.7351, 4.5068, 3.9678, 4.5823, 1.7770, 4.3916, 4.4377], device='cuda:1'), covar=tensor([0.0122, 0.0126, 0.0206, 0.0296, 0.0123, 0.2460, 0.0128, 0.0160], device='cuda:1'), in_proj_covar=tensor([0.0146, 0.0135, 0.0176, 0.0158, 0.0154, 0.0191, 0.0167, 0.0155], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-30 19:46:22,332 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([1.8114, 2.8827, 2.6072, 4.3327, 2.8095, 4.0942, 1.5011, 3.0774], device='cuda:1'), covar=tensor([0.1326, 0.0668, 0.1077, 0.0137, 0.0108, 0.0372, 0.1600, 0.0670], device='cuda:1'), in_proj_covar=tensor([0.0162, 0.0166, 0.0188, 0.0172, 0.0194, 0.0208, 0.0192, 0.0185], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-04-30 19:46:46,717 INFO [train.py:904] (1/8) Epoch 18, batch 10100, loss[loss=0.1685, simple_loss=0.2593, pruned_loss=0.03885, over 16969.00 frames. ], tot_loss[loss=0.1707, simple_loss=0.2663, pruned_loss=0.03752, over 3071014.24 frames. ], batch size: 109, lr: 3.71e-03, grad_scale: 8.0 2023-04-30 19:46:55,748 INFO [optim.py:368] (1/8) 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 19:47:24,196 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.8976, 2.1066, 2.3747, 3.1807, 2.1492, 2.2829, 2.2798, 2.1827], device='cuda:1'), covar=tensor([0.1237, 0.3827, 0.2560, 0.0631, 0.4367, 0.2805, 0.3247, 0.3707], device='cuda:1'), in_proj_covar=tensor([0.0376, 0.0418, 0.0347, 0.0309, 0.0419, 0.0475, 0.0388, 0.0484], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-30 19:47:36,081 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.26 vs. limit=2.0 2023-04-30 19:47:44,811 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=182682.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 19:48:32,141 INFO [train.py:904] (1/8) Epoch 19, batch 0, loss[loss=0.2209, simple_loss=0.2888, pruned_loss=0.07654, over 16363.00 frames. ], tot_loss[loss=0.2209, simple_loss=0.2888, pruned_loss=0.07654, over 16363.00 frames. ], batch size: 146, lr: 3.61e-03, grad_scale: 8.0 2023-04-30 19:48:32,141 INFO [train.py:929] (1/8) Computing validation loss 2023-04-30 19:48:39,774 INFO [train.py:938] (1/8) Epoch 19, validation: loss=0.1468, simple_loss=0.2504, pruned_loss=0.0216, over 944034.00 frames. 2023-04-30 19:48:39,775 INFO [train.py:939] (1/8) Maximum memory allocated so far is 18145MB 2023-04-30 19:49:50,014 INFO [train.py:904] (1/8) Epoch 19, batch 50, loss[loss=0.1858, simple_loss=0.2747, pruned_loss=0.04848, over 17069.00 frames. ], tot_loss[loss=0.186, simple_loss=0.2708, pruned_loss=0.05065, over 750104.69 frames. ], batch size: 53, lr: 3.61e-03, grad_scale: 1.0 2023-04-30 19:49:59,029 INFO [optim.py:368] (1/8) 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 19:50:13,693 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.9814, 3.0747, 3.1568, 2.1151, 2.7857, 2.2233, 3.5259, 3.4579], device='cuda:1'), covar=tensor([0.0223, 0.0900, 0.0622, 0.1888, 0.0877, 0.1010, 0.0518, 0.0853], device='cuda:1'), in_proj_covar=tensor([0.0148, 0.0151, 0.0160, 0.0147, 0.0138, 0.0124, 0.0137, 0.0160], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:1') 2023-04-30 19:50:25,113 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.94 vs. limit=2.0 2023-04-30 19:50:34,130 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([5.1937, 5.1578, 4.9578, 4.5648, 4.9822, 2.0333, 4.7673, 4.9944], device='cuda:1'), covar=tensor([0.0088, 0.0078, 0.0224, 0.0359, 0.0110, 0.2631, 0.0164, 0.0190], device='cuda:1'), in_proj_covar=tensor([0.0149, 0.0137, 0.0179, 0.0161, 0.0157, 0.0194, 0.0170, 0.0157], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-30 19:50:55,407 INFO [train.py:904] (1/8) Epoch 19, batch 100, loss[loss=0.189, simple_loss=0.2663, pruned_loss=0.05591, over 16427.00 frames. ], tot_loss[loss=0.1836, simple_loss=0.2677, pruned_loss=0.04978, over 1318878.73 frames. ], batch size: 68, lr: 3.61e-03, grad_scale: 1.0 2023-04-30 19:51:05,083 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.1833, 4.2055, 4.3347, 4.1517, 4.2465, 4.7867, 4.3583, 4.0395], device='cuda:1'), covar=tensor([0.1899, 0.2319, 0.2533, 0.2582, 0.3186, 0.1396, 0.1809, 0.2906], device='cuda:1'), in_proj_covar=tensor([0.0376, 0.0545, 0.0607, 0.0458, 0.0611, 0.0636, 0.0478, 0.0614], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-04-30 19:51:31,957 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([5.5423, 5.9942, 5.6606, 5.7183, 5.3558, 5.2994, 5.3961, 6.0793], device='cuda:1'), covar=tensor([0.1306, 0.0825, 0.1347, 0.0861, 0.0959, 0.0743, 0.1230, 0.0924], device='cuda:1'), in_proj_covar=tensor([0.0621, 0.0760, 0.0624, 0.0564, 0.0477, 0.0489, 0.0634, 0.0587], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:1') 2023-04-30 19:52:00,685 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([1.6930, 2.8794, 2.7612, 4.9853, 4.0264, 4.3653, 1.6578, 3.0661], device='cuda:1'), covar=tensor([0.1542, 0.0779, 0.1182, 0.0166, 0.0206, 0.0413, 0.1652, 0.0833], device='cuda:1'), in_proj_covar=tensor([0.0165, 0.0169, 0.0190, 0.0175, 0.0197, 0.0211, 0.0194, 0.0188], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-04-30 19:52:03,140 INFO [train.py:904] (1/8) Epoch 19, batch 150, loss[loss=0.1613, simple_loss=0.2388, pruned_loss=0.04194, over 16890.00 frames. ], tot_loss[loss=0.1836, simple_loss=0.2682, pruned_loss=0.04954, over 1746180.08 frames. ], batch size: 96, lr: 3.61e-03, grad_scale: 1.0 2023-04-30 19:52:09,682 INFO [zipformer.py:625] (1/8) 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] (1/8) 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 19:52:18,073 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=182862.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 19:52:28,091 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=182870.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 19:53:13,450 INFO [train.py:904] (1/8) Epoch 19, batch 200, loss[loss=0.1559, simple_loss=0.2387, pruned_loss=0.03653, over 16851.00 frames. ], tot_loss[loss=0.1829, simple_loss=0.2682, pruned_loss=0.04878, over 2091592.26 frames. ], batch size: 39, lr: 3.61e-03, grad_scale: 1.0 2023-04-30 19:53:34,360 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=182917.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 19:53:42,594 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=182923.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 19:53:48,462 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([5.1086, 5.6923, 5.8080, 5.5310, 5.5769, 6.1743, 5.6986, 5.4062], device='cuda:1'), covar=tensor([0.0881, 0.1830, 0.2373, 0.1991, 0.2707, 0.1011, 0.1540, 0.2293], device='cuda:1'), in_proj_covar=tensor([0.0380, 0.0550, 0.0613, 0.0464, 0.0617, 0.0641, 0.0482, 0.0620], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-04-30 19:54:00,511 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=182937.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 19:54:21,122 INFO [train.py:904] (1/8) Epoch 19, batch 250, loss[loss=0.1795, simple_loss=0.2555, pruned_loss=0.05176, over 16766.00 frames. ], tot_loss[loss=0.1804, simple_loss=0.2653, pruned_loss=0.04776, over 2362410.79 frames. ], batch size: 83, lr: 3.61e-03, grad_scale: 1.0 2023-04-30 19:54:32,986 INFO [optim.py:368] (1/8) 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 19:54:35,879 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.86 vs. limit=2.0 2023-04-30 19:54:47,754 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=2.79 vs. limit=5.0 2023-04-30 19:55:08,834 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=182985.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 19:55:30,351 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.55 vs. limit=2.0 2023-04-30 19:55:30,552 INFO [train.py:904] (1/8) Epoch 19, batch 300, loss[loss=0.1634, simple_loss=0.2532, pruned_loss=0.03678, over 17109.00 frames. ], tot_loss[loss=0.178, simple_loss=0.263, pruned_loss=0.04646, over 2576968.05 frames. ], batch size: 47, lr: 3.61e-03, grad_scale: 1.0 2023-04-30 19:56:41,129 INFO [train.py:904] (1/8) Epoch 19, batch 350, loss[loss=0.169, simple_loss=0.2689, pruned_loss=0.03452, over 17056.00 frames. ], tot_loss[loss=0.1757, simple_loss=0.2608, pruned_loss=0.04528, over 2740663.00 frames. ], batch size: 55, lr: 3.61e-03, grad_scale: 1.0 2023-04-30 19:56:52,337 INFO [optim.py:368] (1/8) Clipping_scale=2.0, grad-norm quartiles 1.344e+02 2.193e+02 2.670e+02 3.333e+02 1.052e+03, threshold=5.340e+02, percent-clipped=4.0 2023-04-30 19:57:51,265 INFO [train.py:904] (1/8) Epoch 19, batch 400, loss[loss=0.1638, simple_loss=0.252, pruned_loss=0.03776, over 17255.00 frames. ], tot_loss[loss=0.1741, simple_loss=0.2585, pruned_loss=0.04487, over 2856977.34 frames. ], batch size: 45, lr: 3.61e-03, grad_scale: 2.0 2023-04-30 19:58:15,904 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=4.80 vs. limit=5.0 2023-04-30 19:59:03,157 INFO [train.py:904] (1/8) Epoch 19, batch 450, loss[loss=0.1541, simple_loss=0.2329, pruned_loss=0.03764, over 16724.00 frames. ], tot_loss[loss=0.1724, simple_loss=0.2571, pruned_loss=0.04389, over 2963873.84 frames. ], batch size: 83, lr: 3.61e-03, grad_scale: 2.0 2023-04-30 19:59:14,101 INFO [optim.py:368] (1/8) 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 19:59:28,202 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=183170.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 20:00:09,109 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.31 vs. limit=2.0 2023-04-30 20:00:12,283 INFO [train.py:904] (1/8) Epoch 19, batch 500, loss[loss=0.1547, simple_loss=0.2429, pruned_loss=0.03327, over 16731.00 frames. ], tot_loss[loss=0.1707, simple_loss=0.2556, pruned_loss=0.04292, over 3041388.15 frames. ], batch size: 89, lr: 3.61e-03, grad_scale: 2.0 2023-04-30 20:00:27,802 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=183212.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 20:00:35,919 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=183218.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 20:00:35,938 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=183218.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 20:00:44,696 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.45 vs. limit=2.0 2023-04-30 20:01:18,141 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.11 vs. limit=2.0 2023-04-30 20:01:19,138 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.8102, 2.8566, 2.4835, 2.8251, 3.2217, 3.0269, 3.5806, 3.4551], device='cuda:1'), covar=tensor([0.0148, 0.0402, 0.0501, 0.0383, 0.0264, 0.0345, 0.0269, 0.0255], device='cuda:1'), in_proj_covar=tensor([0.0196, 0.0232, 0.0224, 0.0224, 0.0232, 0.0231, 0.0233, 0.0224], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-30 20:01:23,246 INFO [train.py:904] (1/8) Epoch 19, batch 550, loss[loss=0.1891, simple_loss=0.2684, pruned_loss=0.05489, over 16722.00 frames. ], tot_loss[loss=0.1699, simple_loss=0.255, pruned_loss=0.04234, over 3107917.57 frames. ], batch size: 134, lr: 3.61e-03, grad_scale: 2.0 2023-04-30 20:01:34,915 INFO [optim.py:368] (1/8) Clipping_scale=2.0, grad-norm quartiles 1.382e+02 2.167e+02 2.560e+02 2.885e+02 5.610e+02, threshold=5.120e+02, percent-clipped=1.0 2023-04-30 20:02:32,735 INFO [train.py:904] (1/8) Epoch 19, batch 600, loss[loss=0.1695, simple_loss=0.2661, pruned_loss=0.0364, over 17056.00 frames. ], tot_loss[loss=0.1691, simple_loss=0.2541, pruned_loss=0.04203, over 3158860.35 frames. ], batch size: 53, lr: 3.61e-03, grad_scale: 2.0 2023-04-30 20:03:13,285 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.9570, 3.8855, 4.4390, 2.0731, 4.5452, 4.6758, 3.3548, 3.5342], device='cuda:1'), covar=tensor([0.0745, 0.0286, 0.0196, 0.1179, 0.0074, 0.0140, 0.0404, 0.0404], device='cuda:1'), in_proj_covar=tensor([0.0147, 0.0106, 0.0094, 0.0140, 0.0077, 0.0121, 0.0126, 0.0129], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0004, 0.0003], device='cuda:1') 2023-04-30 20:03:40,889 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([5.1200, 5.1521, 5.6012, 5.5521, 5.5456, 5.2159, 5.1692, 4.9787], device='cuda:1'), covar=tensor([0.0315, 0.0503, 0.0389, 0.0483, 0.0510, 0.0406, 0.0900, 0.0415], device='cuda:1'), in_proj_covar=tensor([0.0392, 0.0429, 0.0418, 0.0394, 0.0464, 0.0441, 0.0530, 0.0351], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:1') 2023-04-30 20:03:42,736 INFO [train.py:904] (1/8) Epoch 19, batch 650, loss[loss=0.1668, simple_loss=0.2613, pruned_loss=0.03619, over 17144.00 frames. ], tot_loss[loss=0.1673, simple_loss=0.2524, pruned_loss=0.04115, over 3194074.64 frames. ], batch size: 47, lr: 3.61e-03, grad_scale: 2.0 2023-04-30 20:03:54,610 INFO [optim.py:368] (1/8) 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 20:04:32,268 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.86 vs. limit=2.0 2023-04-30 20:04:53,074 INFO [train.py:904] (1/8) Epoch 19, batch 700, loss[loss=0.1857, simple_loss=0.2607, pruned_loss=0.05536, over 16689.00 frames. ], tot_loss[loss=0.167, simple_loss=0.2519, pruned_loss=0.04104, over 3221977.15 frames. ], batch size: 89, lr: 3.61e-03, grad_scale: 2.0 2023-04-30 20:05:01,041 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=183408.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 20:05:50,922 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.74 vs. limit=2.0 2023-04-30 20:05:59,948 INFO [train.py:904] (1/8) Epoch 19, batch 750, loss[loss=0.1735, simple_loss=0.2536, pruned_loss=0.04673, over 16682.00 frames. ], tot_loss[loss=0.1673, simple_loss=0.2527, pruned_loss=0.04099, over 3249596.01 frames. ], batch size: 134, lr: 3.61e-03, grad_scale: 2.0 2023-04-30 20:06:11,899 INFO [optim.py:368] (1/8) 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 20:06:25,277 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=183469.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 20:06:39,710 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.7859, 5.1034, 4.8743, 4.8583, 4.6554, 4.6119, 4.5715, 5.1907], device='cuda:1'), covar=tensor([0.1167, 0.0843, 0.0955, 0.0795, 0.0783, 0.1107, 0.1074, 0.0882], device='cuda:1'), in_proj_covar=tensor([0.0654, 0.0801, 0.0658, 0.0593, 0.0502, 0.0511, 0.0665, 0.0619], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:1') 2023-04-30 20:07:10,060 INFO [train.py:904] (1/8) Epoch 19, batch 800, loss[loss=0.1682, simple_loss=0.2456, pruned_loss=0.04539, over 16795.00 frames. ], tot_loss[loss=0.1672, simple_loss=0.2523, pruned_loss=0.04104, over 3265272.55 frames. ], batch size: 83, lr: 3.61e-03, grad_scale: 4.0 2023-04-30 20:07:24,234 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=183512.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 20:07:32,558 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=183518.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 20:08:19,010 INFO [train.py:904] (1/8) Epoch 19, batch 850, loss[loss=0.1737, simple_loss=0.2584, pruned_loss=0.0445, over 16752.00 frames. ], tot_loss[loss=0.167, simple_loss=0.2519, pruned_loss=0.04111, over 3275936.03 frames. ], batch size: 57, lr: 3.60e-03, grad_scale: 4.0 2023-04-30 20:08:20,586 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.4727, 3.6431, 3.6917, 1.8238, 2.9055, 2.2590, 3.8989, 3.9325], device='cuda:1'), covar=tensor([0.0292, 0.0986, 0.0645, 0.2581, 0.1152, 0.1285, 0.0699, 0.1053], device='cuda:1'), in_proj_covar=tensor([0.0152, 0.0158, 0.0164, 0.0150, 0.0142, 0.0127, 0.0142, 0.0167], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:1') 2023-04-30 20:08:29,786 INFO [optim.py:368] (1/8) 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 20:08:30,102 INFO [zipformer.py:625] (1/8) 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] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=183566.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 20:09:11,413 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.6729, 2.3850, 2.3960, 4.5255, 2.3465, 2.8315, 2.5004, 2.5503], device='cuda:1'), covar=tensor([0.1140, 0.3659, 0.2903, 0.0433, 0.4081, 0.2580, 0.3595, 0.3628], device='cuda:1'), in_proj_covar=tensor([0.0391, 0.0434, 0.0360, 0.0323, 0.0433, 0.0498, 0.0405, 0.0507], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-30 20:09:28,459 INFO [train.py:904] (1/8) Epoch 19, batch 900, loss[loss=0.1662, simple_loss=0.2576, pruned_loss=0.03738, over 17011.00 frames. ], tot_loss[loss=0.1662, simple_loss=0.2512, pruned_loss=0.04065, over 3285903.81 frames. ], batch size: 55, lr: 3.60e-03, grad_scale: 4.0 2023-04-30 20:09:44,199 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 19, batch 950, loss[loss=0.1688, simple_loss=0.2381, pruned_loss=0.04975, over 16734.00 frames. ], tot_loss[loss=0.1664, simple_loss=0.2514, pruned_loss=0.04072, over 3296284.00 frames. ], batch size: 124, lr: 3.60e-03, grad_scale: 4.0 2023-04-30 20:10:45,885 INFO [optim.py:368] (1/8) 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 20:11:08,235 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=183675.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 20:11:13,276 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=2.90 vs. limit=5.0 2023-04-30 20:11:40,101 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.31 vs. limit=2.0 2023-04-30 20:11:42,836 INFO [train.py:904] (1/8) Epoch 19, batch 1000, loss[loss=0.1699, simple_loss=0.2457, pruned_loss=0.04701, over 16855.00 frames. ], tot_loss[loss=0.1659, simple_loss=0.2505, pruned_loss=0.04065, over 3296560.24 frames. ], batch size: 96, lr: 3.60e-03, grad_scale: 4.0 2023-04-30 20:12:50,598 INFO [train.py:904] (1/8) Epoch 19, batch 1050, loss[loss=0.1658, simple_loss=0.2413, pruned_loss=0.04512, over 16759.00 frames. ], tot_loss[loss=0.1668, simple_loss=0.2512, pruned_loss=0.04122, over 3311212.00 frames. ], batch size: 124, lr: 3.60e-03, grad_scale: 4.0 2023-04-30 20:13:01,769 INFO [optim.py:368] (1/8) 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 20:13:06,588 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=183764.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 20:14:00,539 INFO [train.py:904] (1/8) Epoch 19, batch 1100, loss[loss=0.1471, simple_loss=0.2331, pruned_loss=0.03057, over 16943.00 frames. ], tot_loss[loss=0.1664, simple_loss=0.2509, pruned_loss=0.04091, over 3318363.60 frames. ], batch size: 41, lr: 3.60e-03, grad_scale: 4.0 2023-04-30 20:14:55,861 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 19, batch 1150, loss[loss=0.1546, simple_loss=0.2375, pruned_loss=0.0359, over 16836.00 frames. ], tot_loss[loss=0.1658, simple_loss=0.2504, pruned_loss=0.04061, over 3322273.98 frames. ], batch size: 102, lr: 3.60e-03, grad_scale: 4.0 2023-04-30 20:15:20,304 INFO [optim.py:368] (1/8) 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 20:16:04,248 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.6986, 4.5839, 4.6148, 4.3188, 4.3017, 4.6446, 4.4482, 4.3886], device='cuda:1'), covar=tensor([0.0651, 0.0782, 0.0326, 0.0293, 0.0899, 0.0471, 0.0536, 0.0705], device='cuda:1'), in_proj_covar=tensor([0.0291, 0.0416, 0.0342, 0.0331, 0.0352, 0.0387, 0.0235, 0.0406], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002], device='cuda:1') 2023-04-30 20:16:18,544 INFO [train.py:904] (1/8) Epoch 19, batch 1200, loss[loss=0.168, simple_loss=0.2411, pruned_loss=0.0475, over 16507.00 frames. ], tot_loss[loss=0.1658, simple_loss=0.2504, pruned_loss=0.04059, over 3323764.76 frames. ], batch size: 146, lr: 3.60e-03, grad_scale: 8.0 2023-04-30 20:16:19,006 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.0612, 2.5561, 2.1214, 2.2946, 2.8967, 2.6562, 2.9888, 3.0073], device='cuda:1'), covar=tensor([0.0214, 0.0378, 0.0512, 0.0467, 0.0264, 0.0366, 0.0259, 0.0274], device='cuda:1'), in_proj_covar=tensor([0.0202, 0.0235, 0.0226, 0.0227, 0.0236, 0.0234, 0.0238, 0.0229], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-30 20:16:20,022 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=183903.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 20:17:25,272 INFO [train.py:904] (1/8) Epoch 19, batch 1250, loss[loss=0.1852, simple_loss=0.259, pruned_loss=0.05564, over 12386.00 frames. ], tot_loss[loss=0.1655, simple_loss=0.2499, pruned_loss=0.04052, over 3318224.23 frames. ], batch size: 247, lr: 3.60e-03, grad_scale: 8.0 2023-04-30 20:17:29,244 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.8880, 4.6556, 4.9013, 5.1054, 5.2898, 4.5879, 5.2656, 5.2983], device='cuda:1'), covar=tensor([0.1937, 0.1344, 0.1924, 0.0837, 0.0568, 0.1058, 0.0658, 0.0665], device='cuda:1'), in_proj_covar=tensor([0.0628, 0.0773, 0.0904, 0.0792, 0.0586, 0.0619, 0.0640, 0.0737], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-30 20:17:35,896 INFO [optim.py:368] (1/8) 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 20:17:50,867 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=183970.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 20:17:54,545 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.9592, 2.5256, 1.9594, 2.4200, 2.9302, 2.7387, 2.9699, 3.0410], device='cuda:1'), covar=tensor([0.0222, 0.0424, 0.0595, 0.0438, 0.0263, 0.0331, 0.0289, 0.0270], device='cuda:1'), in_proj_covar=tensor([0.0202, 0.0235, 0.0226, 0.0227, 0.0236, 0.0233, 0.0237, 0.0228], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-30 20:18:02,070 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.91 vs. limit=2.0 2023-04-30 20:18:17,051 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([5.4914, 5.8683, 5.6571, 5.6890, 5.2380, 5.2606, 5.2105, 6.0039], device='cuda:1'), covar=tensor([0.1401, 0.0877, 0.1080, 0.0889, 0.1071, 0.0758, 0.1280, 0.0887], device='cuda:1'), in_proj_covar=tensor([0.0664, 0.0814, 0.0670, 0.0605, 0.0510, 0.0520, 0.0677, 0.0632], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:1') 2023-04-30 20:18:37,488 INFO [train.py:904] (1/8) Epoch 19, batch 1300, loss[loss=0.148, simple_loss=0.2329, pruned_loss=0.03158, over 16967.00 frames. ], tot_loss[loss=0.1651, simple_loss=0.2493, pruned_loss=0.04045, over 3324999.86 frames. ], batch size: 41, lr: 3.60e-03, grad_scale: 8.0 2023-04-30 20:18:47,718 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.32 vs. limit=2.0 2023-04-30 20:19:44,233 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=184050.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 20:19:46,329 INFO [train.py:904] (1/8) Epoch 19, batch 1350, loss[loss=0.1847, simple_loss=0.2731, pruned_loss=0.04812, over 16757.00 frames. ], tot_loss[loss=0.1656, simple_loss=0.25, pruned_loss=0.04064, over 3321646.29 frames. ], batch size: 62, lr: 3.60e-03, grad_scale: 8.0 2023-04-30 20:19:55,437 INFO [optim.py:368] (1/8) 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 20:20:03,932 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=184064.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 20:20:21,040 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=3.28 vs. limit=5.0 2023-04-30 20:20:26,479 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.45 vs. limit=2.0 2023-04-30 20:20:55,153 INFO [train.py:904] (1/8) Epoch 19, batch 1400, loss[loss=0.1638, simple_loss=0.2422, pruned_loss=0.04268, over 16792.00 frames. ], tot_loss[loss=0.1659, simple_loss=0.2505, pruned_loss=0.04065, over 3316149.85 frames. ], batch size: 83, lr: 3.60e-03, grad_scale: 8.0 2023-04-30 20:21:09,594 INFO [zipformer.py:625] (1/8) 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] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=184112.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 20:21:11,839 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.8702, 3.7337, 4.1628, 2.2067, 4.3463, 4.4413, 3.2170, 3.3555], device='cuda:1'), covar=tensor([0.0742, 0.0271, 0.0226, 0.1132, 0.0078, 0.0165, 0.0431, 0.0433], device='cuda:1'), in_proj_covar=tensor([0.0147, 0.0107, 0.0095, 0.0139, 0.0077, 0.0123, 0.0126, 0.0129], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0004, 0.0003], device='cuda:1') 2023-04-30 20:22:05,079 INFO [train.py:904] (1/8) Epoch 19, batch 1450, loss[loss=0.1682, simple_loss=0.2537, pruned_loss=0.04137, over 16709.00 frames. ], tot_loss[loss=0.1651, simple_loss=0.2493, pruned_loss=0.04043, over 3323287.98 frames. ], batch size: 62, lr: 3.60e-03, grad_scale: 8.0 2023-04-30 20:22:15,457 INFO [optim.py:368] (1/8) 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 20:22:33,135 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.9643, 4.4645, 3.2580, 2.4465, 2.7988, 2.7361, 4.9318, 3.6918], device='cuda:1'), covar=tensor([0.2759, 0.0581, 0.1727, 0.2866, 0.3035, 0.1998, 0.0340, 0.1383], device='cuda:1'), in_proj_covar=tensor([0.0324, 0.0268, 0.0302, 0.0306, 0.0294, 0.0253, 0.0290, 0.0332], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:1') 2023-04-30 20:22:40,230 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.2190, 3.5304, 3.6561, 2.1576, 3.0615, 2.4681, 3.7206, 3.7574], device='cuda:1'), covar=tensor([0.0281, 0.0826, 0.0543, 0.1926, 0.0815, 0.0949, 0.0567, 0.0938], device='cuda:1'), in_proj_covar=tensor([0.0151, 0.0159, 0.0164, 0.0150, 0.0142, 0.0127, 0.0142, 0.0168], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:1') 2023-04-30 20:22:42,480 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([5.2147, 5.1858, 4.9599, 4.0999, 5.0607, 1.9404, 4.7648, 4.8357], device='cuda:1'), covar=tensor([0.0093, 0.0080, 0.0197, 0.0481, 0.0112, 0.2948, 0.0168, 0.0249], device='cuda:1'), in_proj_covar=tensor([0.0159, 0.0148, 0.0192, 0.0176, 0.0170, 0.0205, 0.0184, 0.0171], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-30 20:23:08,682 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=184198.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 20:23:13,297 INFO [train.py:904] (1/8) Epoch 19, batch 1500, loss[loss=0.1841, simple_loss=0.2571, pruned_loss=0.05555, over 16351.00 frames. ], tot_loss[loss=0.1649, simple_loss=0.2491, pruned_loss=0.04032, over 3324535.28 frames. ], batch size: 145, lr: 3.60e-03, grad_scale: 4.0 2023-04-30 20:23:42,405 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.5090, 4.2319, 4.1643, 4.6216, 4.8305, 4.3370, 4.6742, 4.7734], device='cuda:1'), covar=tensor([0.1597, 0.1333, 0.2496, 0.1261, 0.0865, 0.1526, 0.1881, 0.1155], device='cuda:1'), in_proj_covar=tensor([0.0634, 0.0780, 0.0913, 0.0796, 0.0590, 0.0624, 0.0643, 0.0744], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-30 20:24:22,394 INFO [train.py:904] (1/8) Epoch 19, batch 1550, loss[loss=0.128, simple_loss=0.2171, pruned_loss=0.01942, over 16994.00 frames. ], tot_loss[loss=0.1667, simple_loss=0.2508, pruned_loss=0.04133, over 3331580.32 frames. ], batch size: 41, lr: 3.60e-03, grad_scale: 4.0 2023-04-30 20:24:34,823 INFO [optim.py:368] (1/8) 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 20:24:45,596 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.18 vs. limit=2.0 2023-04-30 20:24:47,411 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=184270.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 20:25:31,172 INFO [train.py:904] (1/8) Epoch 19, batch 1600, loss[loss=0.1899, simple_loss=0.2797, pruned_loss=0.0501, over 17082.00 frames. ], tot_loss[loss=0.1692, simple_loss=0.2531, pruned_loss=0.04268, over 3324744.58 frames. ], batch size: 53, lr: 3.60e-03, grad_scale: 8.0 2023-04-30 20:25:39,405 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.7377, 3.7269, 2.8573, 2.2358, 2.4619, 2.3139, 3.8628, 3.2481], device='cuda:1'), covar=tensor([0.2546, 0.0621, 0.1627, 0.2936, 0.2775, 0.2079, 0.0488, 0.1420], device='cuda:1'), in_proj_covar=tensor([0.0323, 0.0266, 0.0301, 0.0305, 0.0293, 0.0252, 0.0289, 0.0330], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:1') 2023-04-30 20:25:53,346 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=184318.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 20:26:01,433 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.7704, 4.3224, 4.4418, 3.1436, 3.7502, 4.3523, 3.9007, 2.3336], device='cuda:1'), covar=tensor([0.0485, 0.0067, 0.0039, 0.0344, 0.0126, 0.0088, 0.0081, 0.0499], device='cuda:1'), in_proj_covar=tensor([0.0137, 0.0082, 0.0080, 0.0133, 0.0096, 0.0105, 0.0092, 0.0127], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-04-30 20:26:15,616 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.24 vs. limit=2.0 2023-04-30 20:26:39,948 INFO [train.py:904] (1/8) Epoch 19, batch 1650, loss[loss=0.1704, simple_loss=0.2532, pruned_loss=0.04381, over 16792.00 frames. ], tot_loss[loss=0.1709, simple_loss=0.2547, pruned_loss=0.04354, over 3322778.25 frames. ], batch size: 102, lr: 3.60e-03, grad_scale: 8.0 2023-04-30 20:26:45,844 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.4752, 3.4822, 2.1144, 3.7099, 2.6902, 3.6327, 2.2134, 2.8448], device='cuda:1'), covar=tensor([0.0282, 0.0472, 0.1576, 0.0346, 0.0795, 0.0929, 0.1445, 0.0750], device='cuda:1'), in_proj_covar=tensor([0.0167, 0.0177, 0.0195, 0.0160, 0.0175, 0.0215, 0.0202, 0.0179], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-04-30 20:26:50,284 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=184359.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 20:26:52,312 INFO [optim.py:368] (1/8) 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 20:27:45,762 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.69 vs. limit=2.0 2023-04-30 20:27:49,666 INFO [train.py:904] (1/8) Epoch 19, batch 1700, loss[loss=0.1798, simple_loss=0.2765, pruned_loss=0.04149, over 17076.00 frames. ], tot_loss[loss=0.1716, simple_loss=0.2556, pruned_loss=0.04382, over 3321917.94 frames. ], batch size: 50, lr: 3.60e-03, grad_scale: 8.0 2023-04-30 20:27:55,023 INFO [zipformer.py:625] (1/8) 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] (1/8) attn_weights_entropy = tensor([4.9098, 4.9810, 5.4100, 5.4032, 5.4020, 5.0374, 5.0343, 4.8266], device='cuda:1'), covar=tensor([0.0362, 0.0537, 0.0412, 0.0500, 0.0546, 0.0450, 0.0877, 0.0442], device='cuda:1'), in_proj_covar=tensor([0.0402, 0.0440, 0.0427, 0.0402, 0.0476, 0.0450, 0.0542, 0.0358], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:1') 2023-04-30 20:28:14,060 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=184420.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 20:28:32,327 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=3.52 vs. limit=5.0 2023-04-30 20:28:58,451 INFO [train.py:904] (1/8) Epoch 19, batch 1750, loss[loss=0.1736, simple_loss=0.2664, pruned_loss=0.04044, over 17256.00 frames. ], tot_loss[loss=0.1714, simple_loss=0.2562, pruned_loss=0.04325, over 3331017.36 frames. ], batch size: 45, lr: 3.60e-03, grad_scale: 8.0 2023-04-30 20:29:10,989 INFO [optim.py:368] (1/8) 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 20:29:45,761 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=184486.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 20:30:00,577 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.9538, 5.0524, 5.4607, 5.4809, 5.4414, 5.1021, 5.0188, 4.8717], device='cuda:1'), covar=tensor([0.0367, 0.0495, 0.0424, 0.0398, 0.0479, 0.0452, 0.0963, 0.0418], device='cuda:1'), in_proj_covar=tensor([0.0404, 0.0441, 0.0429, 0.0404, 0.0479, 0.0452, 0.0546, 0.0360], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:1') 2023-04-30 20:30:02,694 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 19, batch 1800, loss[loss=0.1713, simple_loss=0.2563, pruned_loss=0.04317, over 16895.00 frames. ], tot_loss[loss=0.1721, simple_loss=0.2576, pruned_loss=0.04327, over 3322118.64 frames. ], batch size: 96, lr: 3.60e-03, grad_scale: 8.0 2023-04-30 20:30:47,113 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=3.10 vs. limit=5.0 2023-04-30 20:31:07,262 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=184546.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 20:31:09,249 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=184547.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 20:31:16,692 INFO [train.py:904] (1/8) Epoch 19, batch 1850, loss[loss=0.1641, simple_loss=0.2612, pruned_loss=0.03354, over 17109.00 frames. ], tot_loss[loss=0.1731, simple_loss=0.259, pruned_loss=0.0436, over 3318391.89 frames. ], batch size: 49, lr: 3.59e-03, grad_scale: 8.0 2023-04-30 20:31:20,776 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.47 vs. limit=2.0 2023-04-30 20:31:29,852 INFO [optim.py:368] (1/8) 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 20:32:26,188 INFO [train.py:904] (1/8) Epoch 19, batch 1900, loss[loss=0.2178, simple_loss=0.3038, pruned_loss=0.06593, over 11957.00 frames. ], tot_loss[loss=0.1722, simple_loss=0.2583, pruned_loss=0.04305, over 3309541.73 frames. ], batch size: 246, lr: 3.59e-03, grad_scale: 8.0 2023-04-30 20:32:53,053 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.1096, 3.2077, 3.2946, 2.3201, 3.1115, 3.3857, 3.1699, 1.9540], device='cuda:1'), covar=tensor([0.0468, 0.0124, 0.0059, 0.0364, 0.0126, 0.0105, 0.0094, 0.0459], device='cuda:1'), in_proj_covar=tensor([0.0137, 0.0082, 0.0081, 0.0133, 0.0096, 0.0106, 0.0092, 0.0128], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-04-30 20:33:24,810 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([5.1579, 5.6646, 5.8284, 5.4866, 5.6811, 6.2074, 5.7526, 5.4057], device='cuda:1'), covar=tensor([0.0911, 0.2293, 0.2486, 0.2227, 0.2882, 0.1038, 0.1441, 0.2458], device='cuda:1'), in_proj_covar=tensor([0.0406, 0.0582, 0.0646, 0.0490, 0.0655, 0.0684, 0.0504, 0.0656], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-04-30 20:33:35,921 INFO [train.py:904] (1/8) Epoch 19, batch 1950, loss[loss=0.1477, simple_loss=0.2316, pruned_loss=0.03193, over 15972.00 frames. ], tot_loss[loss=0.1717, simple_loss=0.2584, pruned_loss=0.04245, over 3316889.17 frames. ], batch size: 35, lr: 3.59e-03, grad_scale: 8.0 2023-04-30 20:33:48,870 INFO [optim.py:368] (1/8) 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 20:34:14,600 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.9052, 2.0337, 2.5649, 2.8581, 2.6714, 3.3719, 2.2446, 3.2850], device='cuda:1'), covar=tensor([0.0235, 0.0479, 0.0317, 0.0328, 0.0321, 0.0188, 0.0464, 0.0194], device='cuda:1'), in_proj_covar=tensor([0.0185, 0.0193, 0.0179, 0.0182, 0.0192, 0.0149, 0.0195, 0.0144], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-30 20:34:47,046 INFO [train.py:904] (1/8) Epoch 19, batch 2000, loss[loss=0.2118, simple_loss=0.2757, pruned_loss=0.07393, over 16894.00 frames. ], tot_loss[loss=0.1719, simple_loss=0.2582, pruned_loss=0.0428, over 3308819.77 frames. ], batch size: 109, lr: 3.59e-03, grad_scale: 8.0 2023-04-30 20:34:53,108 INFO [zipformer.py:625] (1/8) 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] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=184715.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 20:35:43,629 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.4912, 3.5058, 3.7243, 2.6406, 3.4202, 3.7794, 3.5249, 2.1546], device='cuda:1'), covar=tensor([0.0469, 0.0155, 0.0058, 0.0367, 0.0108, 0.0093, 0.0085, 0.0463], device='cuda:1'), in_proj_covar=tensor([0.0137, 0.0082, 0.0081, 0.0133, 0.0096, 0.0105, 0.0092, 0.0127], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-04-30 20:35:56,927 INFO [train.py:904] (1/8) Epoch 19, batch 2050, loss[loss=0.1853, simple_loss=0.2749, pruned_loss=0.04784, over 17033.00 frames. ], tot_loss[loss=0.172, simple_loss=0.2578, pruned_loss=0.04312, over 3311897.61 frames. ], batch size: 53, lr: 3.59e-03, grad_scale: 8.0 2023-04-30 20:35:57,461 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.0029, 4.4158, 3.0578, 2.4639, 3.0306, 2.6997, 4.9098, 3.8078], device='cuda:1'), covar=tensor([0.2651, 0.0660, 0.1889, 0.2698, 0.2586, 0.1952, 0.0334, 0.1262], device='cuda:1'), in_proj_covar=tensor([0.0322, 0.0267, 0.0301, 0.0304, 0.0293, 0.0251, 0.0289, 0.0331], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:1') 2023-04-30 20:35:59,926 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=184754.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 20:36:10,012 INFO [optim.py:368] (1/8) 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 20:36:35,802 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.4830, 3.6389, 3.8223, 2.6910, 3.5157, 3.8725, 3.6238, 2.1737], device='cuda:1'), covar=tensor([0.0481, 0.0172, 0.0059, 0.0356, 0.0108, 0.0100, 0.0092, 0.0461], device='cuda:1'), in_proj_covar=tensor([0.0137, 0.0082, 0.0081, 0.0133, 0.0096, 0.0105, 0.0092, 0.0127], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-04-30 20:37:10,295 INFO [train.py:904] (1/8) Epoch 19, batch 2100, loss[loss=0.1845, simple_loss=0.2574, pruned_loss=0.05576, over 16857.00 frames. ], tot_loss[loss=0.1714, simple_loss=0.2572, pruned_loss=0.04284, over 3320697.85 frames. ], batch size: 96, lr: 3.59e-03, grad_scale: 8.0 2023-04-30 20:37:13,246 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.80 vs. limit=2.0 2023-04-30 20:37:18,779 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.9161, 4.9816, 5.4207, 5.4155, 5.4053, 5.0610, 5.0370, 4.7953], device='cuda:1'), covar=tensor([0.0327, 0.0401, 0.0346, 0.0370, 0.0438, 0.0377, 0.0801, 0.0434], device='cuda:1'), in_proj_covar=tensor([0.0401, 0.0437, 0.0426, 0.0402, 0.0473, 0.0451, 0.0542, 0.0357], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:1') 2023-04-30 20:38:05,005 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 19, batch 2150, loss[loss=0.1456, simple_loss=0.2373, pruned_loss=0.027, over 17213.00 frames. ], tot_loss[loss=0.1723, simple_loss=0.2579, pruned_loss=0.04332, over 3323428.43 frames. ], batch size: 46, lr: 3.59e-03, grad_scale: 8.0 2023-04-30 20:38:30,739 INFO [optim.py:368] (1/8) 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 20:38:38,228 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=184866.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 20:38:52,676 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([5.4611, 5.4236, 5.3073, 4.8061, 4.9013, 5.3465, 5.2689, 4.9131], device='cuda:1'), covar=tensor([0.0602, 0.0508, 0.0307, 0.0328, 0.1154, 0.0514, 0.0263, 0.0845], device='cuda:1'), in_proj_covar=tensor([0.0299, 0.0426, 0.0350, 0.0340, 0.0361, 0.0397, 0.0238, 0.0416], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-04-30 20:39:17,946 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.9051, 4.9361, 5.3513, 5.3349, 5.3304, 4.9818, 4.9433, 4.7336], device='cuda:1'), covar=tensor([0.0342, 0.0458, 0.0363, 0.0392, 0.0485, 0.0392, 0.0910, 0.0464], device='cuda:1'), in_proj_covar=tensor([0.0403, 0.0441, 0.0428, 0.0404, 0.0477, 0.0454, 0.0545, 0.0359], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:1') 2023-04-30 20:39:27,103 INFO [train.py:904] (1/8) Epoch 19, batch 2200, loss[loss=0.1542, simple_loss=0.2413, pruned_loss=0.03352, over 16948.00 frames. ], tot_loss[loss=0.173, simple_loss=0.2586, pruned_loss=0.04366, over 3328684.13 frames. ], batch size: 41, lr: 3.59e-03, grad_scale: 8.0 2023-04-30 20:40:01,217 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=184927.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 20:40:17,848 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=184939.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 20:40:31,227 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([5.5335, 5.9131, 5.6457, 5.7167, 5.2893, 5.2733, 5.3014, 6.0488], device='cuda:1'), covar=tensor([0.1342, 0.1002, 0.1115, 0.0903, 0.0961, 0.0689, 0.1226, 0.0960], device='cuda:1'), in_proj_covar=tensor([0.0659, 0.0812, 0.0666, 0.0604, 0.0509, 0.0517, 0.0675, 0.0628], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:1') 2023-04-30 20:40:32,423 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.8543, 4.3522, 2.8317, 2.2288, 2.8498, 2.4076, 4.6816, 3.5644], device='cuda:1'), covar=tensor([0.3024, 0.0629, 0.2174, 0.3185, 0.3038, 0.2239, 0.0412, 0.1563], device='cuda:1'), in_proj_covar=tensor([0.0322, 0.0267, 0.0300, 0.0304, 0.0293, 0.0251, 0.0289, 0.0331], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:1') 2023-04-30 20:40:35,941 INFO [train.py:904] (1/8) Epoch 19, batch 2250, loss[loss=0.1574, simple_loss=0.2481, pruned_loss=0.03332, over 17199.00 frames. ], tot_loss[loss=0.1737, simple_loss=0.2598, pruned_loss=0.0438, over 3329796.11 frames. ], batch size: 46, lr: 3.59e-03, grad_scale: 8.0 2023-04-30 20:40:48,428 INFO [optim.py:368] (1/8) 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 20:41:09,821 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=4.93 vs. limit=5.0 2023-04-30 20:41:44,974 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 19, batch 2300, loss[loss=0.1677, simple_loss=0.2609, pruned_loss=0.03725, over 17043.00 frames. ], tot_loss[loss=0.173, simple_loss=0.2595, pruned_loss=0.04323, over 3329402.81 frames. ], batch size: 55, lr: 3.59e-03, grad_scale: 8.0 2023-04-30 20:42:05,814 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=185015.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 20:42:56,743 INFO [train.py:904] (1/8) Epoch 19, batch 2350, loss[loss=0.2008, simple_loss=0.271, pruned_loss=0.0653, over 16777.00 frames. ], tot_loss[loss=0.1742, simple_loss=0.2604, pruned_loss=0.04399, over 3325409.61 frames. ], batch size: 89, lr: 3.59e-03, grad_scale: 8.0 2023-04-30 20:43:08,918 INFO [zipformer.py:625] (1/8) 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] (1/8) 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 20:43:12,099 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=185063.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 20:43:40,299 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.9329, 5.0363, 5.4036, 5.4363, 5.4108, 5.0877, 4.9891, 4.8777], device='cuda:1'), covar=tensor([0.0324, 0.0451, 0.0407, 0.0370, 0.0486, 0.0375, 0.0937, 0.0403], device='cuda:1'), in_proj_covar=tensor([0.0400, 0.0439, 0.0427, 0.0400, 0.0473, 0.0451, 0.0542, 0.0356], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:1') 2023-04-30 20:44:06,535 INFO [train.py:904] (1/8) Epoch 19, batch 2400, loss[loss=0.1592, simple_loss=0.2544, pruned_loss=0.03202, over 17175.00 frames. ], tot_loss[loss=0.1743, simple_loss=0.2605, pruned_loss=0.04407, over 3325199.58 frames. ], batch size: 46, lr: 3.59e-03, grad_scale: 8.0 2023-04-30 20:44:33,380 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=185121.0, num_to_drop=1, layers_to_drop={3} 2023-04-30 20:44:43,077 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=185129.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 20:45:02,172 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=185142.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 20:45:15,142 INFO [train.py:904] (1/8) Epoch 19, batch 2450, loss[loss=0.1807, simple_loss=0.2587, pruned_loss=0.05141, over 16843.00 frames. ], tot_loss[loss=0.1748, simple_loss=0.2619, pruned_loss=0.04385, over 3330224.40 frames. ], batch size: 116, lr: 3.59e-03, grad_scale: 8.0 2023-04-30 20:45:27,049 INFO [optim.py:368] (1/8) 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 20:45:44,366 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.0488, 4.1044, 2.8493, 4.8402, 3.3387, 4.7595, 2.9855, 3.4788], device='cuda:1'), covar=tensor([0.0253, 0.0344, 0.1309, 0.0214, 0.0696, 0.0377, 0.1268, 0.0618], device='cuda:1'), in_proj_covar=tensor([0.0170, 0.0179, 0.0196, 0.0163, 0.0177, 0.0219, 0.0203, 0.0181], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-04-30 20:46:08,235 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=185190.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 20:46:08,415 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 19, batch 2500, loss[loss=0.2039, simple_loss=0.2885, pruned_loss=0.05961, over 12308.00 frames. ], tot_loss[loss=0.1743, simple_loss=0.2615, pruned_loss=0.04355, over 3330784.30 frames. ], batch size: 246, lr: 3.59e-03, grad_scale: 8.0 2023-04-30 20:46:51,191 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=185222.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 20:47:32,255 INFO [train.py:904] (1/8) Epoch 19, batch 2550, loss[loss=0.1761, simple_loss=0.257, pruned_loss=0.04757, over 16752.00 frames. ], tot_loss[loss=0.174, simple_loss=0.2613, pruned_loss=0.04333, over 3331427.91 frames. ], batch size: 124, lr: 3.59e-03, grad_scale: 8.0 2023-04-30 20:47:44,039 INFO [optim.py:368] (1/8) 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] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=185295.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 20:48:40,169 INFO [train.py:904] (1/8) Epoch 19, batch 2600, loss[loss=0.1735, simple_loss=0.2519, pruned_loss=0.04757, over 16851.00 frames. ], tot_loss[loss=0.1745, simple_loss=0.262, pruned_loss=0.04354, over 3321886.70 frames. ], batch size: 96, lr: 3.59e-03, grad_scale: 8.0 2023-04-30 20:48:52,015 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.7426, 3.8335, 2.2547, 4.2486, 2.8389, 4.2153, 2.4907, 3.0757], device='cuda:1'), covar=tensor([0.0276, 0.0346, 0.1620, 0.0314, 0.0865, 0.0617, 0.1445, 0.0730], device='cuda:1'), in_proj_covar=tensor([0.0171, 0.0178, 0.0196, 0.0163, 0.0177, 0.0219, 0.0203, 0.0181], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-04-30 20:48:59,130 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.3290, 3.5190, 3.6819, 3.6317, 3.6648, 3.4685, 3.5247, 3.5502], device='cuda:1'), covar=tensor([0.0390, 0.0657, 0.0442, 0.0487, 0.0563, 0.0533, 0.0771, 0.0467], device='cuda:1'), in_proj_covar=tensor([0.0403, 0.0444, 0.0431, 0.0403, 0.0476, 0.0453, 0.0548, 0.0358], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:1') 2023-04-30 20:49:50,084 INFO [train.py:904] (1/8) Epoch 19, batch 2650, loss[loss=0.196, simple_loss=0.2829, pruned_loss=0.05458, over 16560.00 frames. ], tot_loss[loss=0.1746, simple_loss=0.2624, pruned_loss=0.04336, over 3321374.31 frames. ], batch size: 68, lr: 3.59e-03, grad_scale: 8.0 2023-04-30 20:50:03,486 INFO [optim.py:368] (1/8) Clipping_scale=2.0, grad-norm quartiles 1.500e+02 2.113e+02 2.462e+02 3.036e+02 8.000e+02, threshold=4.924e+02, percent-clipped=5.0 2023-04-30 20:50:59,019 INFO [train.py:904] (1/8) Epoch 19, batch 2700, loss[loss=0.1583, simple_loss=0.2613, pruned_loss=0.0277, over 17112.00 frames. ], tot_loss[loss=0.1739, simple_loss=0.2624, pruned_loss=0.04274, over 3333449.91 frames. ], batch size: 49, lr: 3.59e-03, grad_scale: 8.0 2023-04-30 20:51:19,123 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=185416.0, num_to_drop=1, layers_to_drop={3} 2023-04-30 20:51:44,475 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.4992, 3.8692, 4.1782, 1.9561, 4.4339, 4.7070, 3.2926, 3.4589], device='cuda:1'), covar=tensor([0.1146, 0.0246, 0.0294, 0.1352, 0.0124, 0.0152, 0.0436, 0.0493], device='cuda:1'), in_proj_covar=tensor([0.0149, 0.0109, 0.0097, 0.0141, 0.0079, 0.0126, 0.0128, 0.0131], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0004, 0.0004], device='cuda:1') 2023-04-30 20:52:08,642 INFO [train.py:904] (1/8) Epoch 19, batch 2750, loss[loss=0.1743, simple_loss=0.2643, pruned_loss=0.04216, over 12474.00 frames. ], tot_loss[loss=0.1736, simple_loss=0.2619, pruned_loss=0.04263, over 3328644.97 frames. ], batch size: 246, lr: 3.59e-03, grad_scale: 8.0 2023-04-30 20:52:20,528 INFO [optim.py:368] (1/8) 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 20:52:30,800 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=185468.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 20:52:54,480 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=185485.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 20:53:18,345 INFO [train.py:904] (1/8) Epoch 19, batch 2800, loss[loss=0.1416, simple_loss=0.2338, pruned_loss=0.02469, over 17144.00 frames. ], tot_loss[loss=0.1731, simple_loss=0.2614, pruned_loss=0.04241, over 3336634.18 frames. ], batch size: 46, lr: 3.59e-03, grad_scale: 8.0 2023-04-30 20:53:40,052 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.90 vs. limit=2.0 2023-04-30 20:53:46,614 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=185522.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 20:53:56,306 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=185529.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 20:54:28,292 INFO [train.py:904] (1/8) Epoch 19, batch 2850, loss[loss=0.155, simple_loss=0.2334, pruned_loss=0.03825, over 16678.00 frames. ], tot_loss[loss=0.1725, simple_loss=0.261, pruned_loss=0.04202, over 3342174.38 frames. ], batch size: 89, lr: 3.59e-03, grad_scale: 8.0 2023-04-30 20:54:33,648 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.8599, 4.1397, 2.7031, 4.6775, 3.2478, 4.6306, 2.7567, 3.3000], device='cuda:1'), covar=tensor([0.0315, 0.0347, 0.1397, 0.0278, 0.0746, 0.0500, 0.1338, 0.0687], device='cuda:1'), in_proj_covar=tensor([0.0170, 0.0179, 0.0195, 0.0163, 0.0177, 0.0220, 0.0203, 0.0180], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-04-30 20:54:41,475 INFO [optim.py:368] (1/8) 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 20:54:53,416 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=185570.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 20:55:28,815 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=185595.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 20:55:39,056 INFO [train.py:904] (1/8) Epoch 19, batch 2900, loss[loss=0.194, simple_loss=0.282, pruned_loss=0.05303, over 16796.00 frames. ], tot_loss[loss=0.1727, simple_loss=0.2602, pruned_loss=0.0426, over 3345787.56 frames. ], batch size: 57, lr: 3.58e-03, grad_scale: 8.0 2023-04-30 20:56:36,416 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=185643.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 20:56:47,761 INFO [train.py:904] (1/8) Epoch 19, batch 2950, loss[loss=0.2271, simple_loss=0.2885, pruned_loss=0.08283, over 11434.00 frames. ], tot_loss[loss=0.1732, simple_loss=0.2602, pruned_loss=0.04307, over 3334983.38 frames. ], batch size: 246, lr: 3.58e-03, grad_scale: 8.0 2023-04-30 20:57:00,933 INFO [optim.py:368] (1/8) 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 20:57:11,855 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.28 vs. limit=2.0 2023-04-30 20:57:16,455 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.34 vs. limit=2.0 2023-04-30 20:57:26,731 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.2176, 3.3487, 3.5309, 2.2579, 2.9277, 2.3897, 3.6654, 3.7259], device='cuda:1'), covar=tensor([0.0219, 0.0829, 0.0551, 0.1843, 0.0828, 0.0989, 0.0488, 0.0780], device='cuda:1'), in_proj_covar=tensor([0.0152, 0.0159, 0.0163, 0.0149, 0.0141, 0.0126, 0.0141, 0.0168], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:1') 2023-04-30 20:57:38,164 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.4272, 4.2482, 4.4721, 4.6534, 4.7494, 4.3251, 4.6190, 4.7395], device='cuda:1'), covar=tensor([0.1783, 0.1241, 0.1546, 0.0706, 0.0639, 0.0966, 0.1730, 0.0705], device='cuda:1'), in_proj_covar=tensor([0.0650, 0.0804, 0.0945, 0.0819, 0.0611, 0.0644, 0.0660, 0.0769], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-30 20:57:58,854 INFO [train.py:904] (1/8) Epoch 19, batch 3000, loss[loss=0.2201, simple_loss=0.2943, pruned_loss=0.07292, over 12503.00 frames. ], tot_loss[loss=0.1735, simple_loss=0.2602, pruned_loss=0.04343, over 3325118.66 frames. ], batch size: 247, lr: 3.58e-03, grad_scale: 8.0 2023-04-30 20:57:58,854 INFO [train.py:929] (1/8) Computing validation loss 2023-04-30 20:58:03,828 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.8315, 4.9267, 5.2444, 5.2108, 5.1727, 4.8695, 4.8258, 4.7420], device='cuda:1'), covar=tensor([0.0307, 0.0380, 0.0310, 0.0335, 0.0441, 0.0338, 0.0906, 0.0364], device='cuda:1'), in_proj_covar=tensor([0.0406, 0.0446, 0.0433, 0.0407, 0.0479, 0.0457, 0.0553, 0.0361], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:1') 2023-04-30 20:58:07,638 INFO [train.py:938] (1/8) Epoch 19, validation: loss=0.1362, simple_loss=0.2416, pruned_loss=0.01538, over 944034.00 frames. 2023-04-30 20:58:07,639 INFO [train.py:939] (1/8) Maximum memory allocated so far is 18145MB 2023-04-30 20:58:27,619 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=185716.0, num_to_drop=1, layers_to_drop={0} 2023-04-30 20:58:44,913 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.32 vs. limit=2.0 2023-04-30 20:59:16,277 INFO [train.py:904] (1/8) Epoch 19, batch 3050, loss[loss=0.1366, simple_loss=0.2305, pruned_loss=0.02134, over 17242.00 frames. ], tot_loss[loss=0.1729, simple_loss=0.2592, pruned_loss=0.04324, over 3326985.54 frames. ], batch size: 45, lr: 3.58e-03, grad_scale: 8.0 2023-04-30 20:59:28,050 INFO [zipformer.py:625] (1/8) 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] (1/8) 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 20:59:35,052 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=185764.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 21:00:04,097 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=185785.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 21:00:23,213 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([5.1230, 5.4530, 5.1968, 5.2299, 4.9404, 4.9274, 4.9101, 5.5818], device='cuda:1'), covar=tensor([0.1260, 0.0971, 0.1104, 0.0974, 0.0858, 0.0925, 0.1183, 0.0986], device='cuda:1'), in_proj_covar=tensor([0.0675, 0.0830, 0.0686, 0.0615, 0.0521, 0.0527, 0.0692, 0.0641], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:1') 2023-04-30 21:00:23,311 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=185799.0, num_to_drop=1, layers_to_drop={0} 2023-04-30 21:00:26,856 INFO [train.py:904] (1/8) Epoch 19, batch 3100, loss[loss=0.1682, simple_loss=0.2676, pruned_loss=0.03442, over 17249.00 frames. ], tot_loss[loss=0.1729, simple_loss=0.2587, pruned_loss=0.04354, over 3332638.75 frames. ], batch size: 52, lr: 3.58e-03, grad_scale: 8.0 2023-04-30 21:00:47,970 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.7098, 4.7822, 4.9139, 4.8015, 4.7855, 5.4318, 4.9533, 4.5992], device='cuda:1'), covar=tensor([0.1351, 0.2114, 0.2477, 0.2150, 0.2855, 0.1030, 0.1619, 0.2625], device='cuda:1'), in_proj_covar=tensor([0.0408, 0.0591, 0.0654, 0.0495, 0.0662, 0.0691, 0.0509, 0.0661], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-04-30 21:00:51,290 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=185820.0, num_to_drop=1, layers_to_drop={3} 2023-04-30 21:00:56,520 INFO [zipformer.py:625] (1/8) 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] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=185833.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 21:01:34,487 INFO [train.py:904] (1/8) Epoch 19, batch 3150, loss[loss=0.187, simple_loss=0.2752, pruned_loss=0.0494, over 16428.00 frames. ], tot_loss[loss=0.1724, simple_loss=0.2579, pruned_loss=0.04349, over 3333981.10 frames. ], batch size: 68, lr: 3.58e-03, grad_scale: 8.0 2023-04-30 21:01:46,053 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=185860.0, num_to_drop=1, layers_to_drop={0} 2023-04-30 21:01:47,103 INFO [optim.py:368] (1/8) 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 21:01:48,304 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.1964, 2.4441, 2.9649, 3.1249, 3.1252, 3.6680, 2.7146, 3.6255], device='cuda:1'), covar=tensor([0.0210, 0.0398, 0.0255, 0.0277, 0.0258, 0.0160, 0.0392, 0.0158], device='cuda:1'), in_proj_covar=tensor([0.0185, 0.0192, 0.0179, 0.0182, 0.0192, 0.0151, 0.0195, 0.0146], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-30 21:02:14,890 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.9811, 4.7367, 4.9616, 5.1613, 5.3378, 4.6456, 5.2885, 5.3216], device='cuda:1'), covar=tensor([0.1616, 0.1159, 0.1529, 0.0746, 0.0529, 0.0934, 0.0549, 0.0584], device='cuda:1'), in_proj_covar=tensor([0.0652, 0.0806, 0.0947, 0.0822, 0.0613, 0.0645, 0.0662, 0.0771], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-30 21:02:40,197 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([5.0039, 4.9788, 4.7809, 4.0885, 4.8575, 1.8438, 4.6011, 4.5687], device='cuda:1'), covar=tensor([0.0105, 0.0091, 0.0220, 0.0465, 0.0128, 0.2870, 0.0158, 0.0264], device='cuda:1'), in_proj_covar=tensor([0.0162, 0.0151, 0.0196, 0.0179, 0.0173, 0.0206, 0.0188, 0.0175], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-30 21:02:43,073 INFO [train.py:904] (1/8) Epoch 19, batch 3200, loss[loss=0.1516, simple_loss=0.2361, pruned_loss=0.03357, over 15963.00 frames. ], tot_loss[loss=0.1711, simple_loss=0.2566, pruned_loss=0.04281, over 3328131.48 frames. ], batch size: 35, lr: 3.58e-03, grad_scale: 8.0 2023-04-30 21:03:02,888 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.9051, 2.0856, 2.6311, 2.8374, 2.7866, 3.4484, 2.3344, 3.4103], device='cuda:1'), covar=tensor([0.0246, 0.0457, 0.0298, 0.0366, 0.0316, 0.0179, 0.0466, 0.0153], device='cuda:1'), in_proj_covar=tensor([0.0186, 0.0193, 0.0180, 0.0182, 0.0193, 0.0152, 0.0196, 0.0146], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-30 21:03:45,248 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.9242, 4.9813, 5.4071, 5.4050, 5.4127, 5.0586, 5.0192, 4.8591], device='cuda:1'), covar=tensor([0.0369, 0.0482, 0.0449, 0.0445, 0.0476, 0.0392, 0.0925, 0.0422], device='cuda:1'), in_proj_covar=tensor([0.0409, 0.0449, 0.0435, 0.0408, 0.0484, 0.0460, 0.0556, 0.0365], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:1') 2023-04-30 21:03:53,535 INFO [train.py:904] (1/8) Epoch 19, batch 3250, loss[loss=0.204, simple_loss=0.2816, pruned_loss=0.06316, over 12214.00 frames. ], tot_loss[loss=0.1717, simple_loss=0.2574, pruned_loss=0.04299, over 3322686.72 frames. ], batch size: 246, lr: 3.58e-03, grad_scale: 8.0 2023-04-30 21:04:06,305 INFO [optim.py:368] (1/8) 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 21:05:07,475 INFO [train.py:904] (1/8) Epoch 19, batch 3300, loss[loss=0.1957, simple_loss=0.2846, pruned_loss=0.05339, over 17021.00 frames. ], tot_loss[loss=0.1727, simple_loss=0.2588, pruned_loss=0.0433, over 3330579.79 frames. ], batch size: 55, lr: 3.58e-03, grad_scale: 8.0 2023-04-30 21:05:25,493 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.9154, 2.0181, 2.5629, 2.8644, 2.7233, 3.4225, 2.2007, 3.3709], device='cuda:1'), covar=tensor([0.0242, 0.0483, 0.0319, 0.0324, 0.0329, 0.0197, 0.0514, 0.0186], device='cuda:1'), in_proj_covar=tensor([0.0186, 0.0193, 0.0180, 0.0183, 0.0193, 0.0152, 0.0197, 0.0147], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-30 21:06:15,806 INFO [train.py:904] (1/8) Epoch 19, batch 3350, loss[loss=0.1496, simple_loss=0.2386, pruned_loss=0.03024, over 16818.00 frames. ], tot_loss[loss=0.1724, simple_loss=0.2587, pruned_loss=0.04307, over 3334136.55 frames. ], batch size: 42, lr: 3.58e-03, grad_scale: 8.0 2023-04-30 21:06:27,998 INFO [optim.py:368] (1/8) Clipping_scale=2.0, grad-norm quartiles 1.470e+02 2.170e+02 2.521e+02 2.996e+02 5.638e+02, threshold=5.041e+02, percent-clipped=4.0 2023-04-30 21:06:59,980 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([1.8969, 3.0421, 2.6685, 5.1164, 4.2122, 4.3798, 1.8647, 3.1494], device='cuda:1'), covar=tensor([0.1213, 0.0691, 0.1182, 0.0200, 0.0206, 0.0479, 0.1386, 0.0771], device='cuda:1'), in_proj_covar=tensor([0.0165, 0.0173, 0.0192, 0.0185, 0.0205, 0.0217, 0.0198, 0.0189], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-04-30 21:07:23,708 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=186100.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 21:07:25,691 INFO [train.py:904] (1/8) Epoch 19, batch 3400, loss[loss=0.2002, simple_loss=0.27, pruned_loss=0.06526, over 16902.00 frames. ], tot_loss[loss=0.1722, simple_loss=0.2586, pruned_loss=0.0429, over 3330540.73 frames. ], batch size: 109, lr: 3.58e-03, grad_scale: 8.0 2023-04-30 21:07:33,828 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.7332, 2.7033, 2.4471, 2.4981, 2.9801, 2.8444, 3.3424, 3.2341], device='cuda:1'), covar=tensor([0.0140, 0.0389, 0.0443, 0.0431, 0.0281, 0.0316, 0.0264, 0.0261], device='cuda:1'), in_proj_covar=tensor([0.0206, 0.0236, 0.0225, 0.0227, 0.0237, 0.0235, 0.0243, 0.0230], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-30 21:07:44,660 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=186115.0, num_to_drop=1, layers_to_drop={3} 2023-04-30 21:07:57,585 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=186124.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 21:08:02,777 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([5.1924, 5.6135, 5.7595, 5.5008, 5.4778, 6.1147, 5.6664, 5.3355], device='cuda:1'), covar=tensor([0.0847, 0.2124, 0.2585, 0.1992, 0.2993, 0.1009, 0.1475, 0.2621], device='cuda:1'), in_proj_covar=tensor([0.0409, 0.0593, 0.0659, 0.0497, 0.0667, 0.0693, 0.0511, 0.0666], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-04-30 21:08:34,723 INFO [train.py:904] (1/8) Epoch 19, batch 3450, loss[loss=0.1745, simple_loss=0.2504, pruned_loss=0.04928, over 16430.00 frames. ], tot_loss[loss=0.1716, simple_loss=0.2574, pruned_loss=0.04294, over 3337119.46 frames. ], batch size: 146, lr: 3.58e-03, grad_scale: 8.0 2023-04-30 21:08:38,534 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=186155.0, num_to_drop=1, layers_to_drop={1} 2023-04-30 21:08:44,058 INFO [zipformer.py:625] (1/8) 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] (1/8) Clipping_scale=2.0, grad-norm quartiles 1.616e+02 2.301e+02 2.793e+02 3.380e+02 7.495e+02, threshold=5.586e+02, percent-clipped=5.0 2023-04-30 21:08:48,329 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=186161.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 21:09:02,955 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=186172.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 21:09:45,134 INFO [train.py:904] (1/8) Epoch 19, batch 3500, loss[loss=0.1528, simple_loss=0.2408, pruned_loss=0.03241, over 15857.00 frames. ], tot_loss[loss=0.1707, simple_loss=0.2562, pruned_loss=0.04262, over 3334524.03 frames. ], batch size: 35, lr: 3.58e-03, grad_scale: 16.0 2023-04-30 21:10:08,949 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=186219.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 21:10:21,901 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.6468, 1.8398, 2.2334, 2.5089, 2.6108, 2.5125, 1.8923, 2.7580], device='cuda:1'), covar=tensor([0.0159, 0.0457, 0.0287, 0.0234, 0.0257, 0.0288, 0.0480, 0.0151], device='cuda:1'), in_proj_covar=tensor([0.0186, 0.0192, 0.0179, 0.0182, 0.0192, 0.0152, 0.0195, 0.0146], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-30 21:10:55,186 INFO [train.py:904] (1/8) Epoch 19, batch 3550, loss[loss=0.1983, simple_loss=0.2798, pruned_loss=0.0584, over 12444.00 frames. ], tot_loss[loss=0.1698, simple_loss=0.2555, pruned_loss=0.04207, over 3324170.94 frames. ], batch size: 247, lr: 3.58e-03, grad_scale: 16.0 2023-04-30 21:11:01,642 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.7689, 2.7149, 2.2838, 2.5517, 3.0195, 2.9106, 3.4583, 3.4184], device='cuda:1'), covar=tensor([0.0162, 0.0410, 0.0571, 0.0417, 0.0294, 0.0361, 0.0265, 0.0260], device='cuda:1'), in_proj_covar=tensor([0.0207, 0.0237, 0.0226, 0.0227, 0.0238, 0.0235, 0.0243, 0.0231], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-30 21:11:05,142 INFO [zipformer.py:625] (1/8) 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] (1/8) Clipping_scale=2.0, grad-norm quartiles 1.235e+02 2.073e+02 2.370e+02 2.856e+02 5.323e+02, threshold=4.740e+02, percent-clipped=0.0 2023-04-30 21:11:29,138 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.8097, 1.9064, 2.3227, 2.6533, 2.6853, 2.6168, 1.9906, 2.8466], device='cuda:1'), covar=tensor([0.0137, 0.0437, 0.0302, 0.0244, 0.0266, 0.0321, 0.0449, 0.0175], device='cuda:1'), in_proj_covar=tensor([0.0186, 0.0191, 0.0179, 0.0182, 0.0192, 0.0152, 0.0195, 0.0146], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-30 21:11:29,207 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.7828, 3.8039, 2.8993, 2.2817, 2.5280, 2.3732, 3.8844, 3.3714], device='cuda:1'), covar=tensor([0.2380, 0.0527, 0.1616, 0.2729, 0.2437, 0.1954, 0.0505, 0.1319], device='cuda:1'), in_proj_covar=tensor([0.0322, 0.0268, 0.0301, 0.0306, 0.0297, 0.0253, 0.0290, 0.0333], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:1') 2023-04-30 21:11:41,975 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.2385, 3.2764, 3.5168, 2.4346, 3.2951, 3.6387, 3.3314, 2.0532], device='cuda:1'), covar=tensor([0.0493, 0.0125, 0.0057, 0.0378, 0.0104, 0.0090, 0.0089, 0.0462], device='cuda:1'), in_proj_covar=tensor([0.0135, 0.0081, 0.0081, 0.0133, 0.0095, 0.0106, 0.0093, 0.0127], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-04-30 21:11:54,369 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.32 vs. limit=2.0 2023-04-30 21:12:03,397 INFO [train.py:904] (1/8) Epoch 19, batch 3600, loss[loss=0.2138, simple_loss=0.281, pruned_loss=0.07328, over 11599.00 frames. ], tot_loss[loss=0.1692, simple_loss=0.2543, pruned_loss=0.04208, over 3325388.18 frames. ], batch size: 248, lr: 3.58e-03, grad_scale: 16.0 2023-04-30 21:12:28,388 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=186320.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 21:13:13,926 INFO [train.py:904] (1/8) Epoch 19, batch 3650, loss[loss=0.1688, simple_loss=0.2445, pruned_loss=0.04661, over 16859.00 frames. ], tot_loss[loss=0.1691, simple_loss=0.2532, pruned_loss=0.04251, over 3321275.56 frames. ], batch size: 116, lr: 3.58e-03, grad_scale: 16.0 2023-04-30 21:13:27,661 INFO [optim.py:368] (1/8) 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 21:14:16,291 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([1.9637, 3.1138, 3.1315, 2.0839, 3.0188, 3.2208, 3.0154, 1.9386], device='cuda:1'), covar=tensor([0.0517, 0.0082, 0.0073, 0.0423, 0.0108, 0.0104, 0.0102, 0.0449], device='cuda:1'), in_proj_covar=tensor([0.0135, 0.0081, 0.0081, 0.0132, 0.0095, 0.0106, 0.0092, 0.0126], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-04-30 21:14:28,525 INFO [train.py:904] (1/8) Epoch 19, batch 3700, loss[loss=0.1729, simple_loss=0.2434, pruned_loss=0.05117, over 16477.00 frames. ], tot_loss[loss=0.1698, simple_loss=0.2521, pruned_loss=0.04373, over 3298912.28 frames. ], batch size: 68, lr: 3.58e-03, grad_scale: 16.0 2023-04-30 21:14:48,795 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=186415.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 21:14:57,074 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=186421.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 21:15:41,467 INFO [train.py:904] (1/8) Epoch 19, batch 3750, loss[loss=0.1873, simple_loss=0.2718, pruned_loss=0.05144, over 15599.00 frames. ], tot_loss[loss=0.1709, simple_loss=0.2522, pruned_loss=0.04476, over 3294908.92 frames. ], batch size: 190, lr: 3.58e-03, grad_scale: 16.0 2023-04-30 21:15:45,991 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=186455.0, num_to_drop=1, layers_to_drop={1} 2023-04-30 21:15:47,759 INFO [zipformer.py:625] (1/8) 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] (1/8) 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] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=186463.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 21:16:00,542 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.7162, 3.7252, 2.8647, 2.3293, 2.5018, 2.4426, 3.7507, 3.3878], device='cuda:1'), covar=tensor([0.2465, 0.0596, 0.1684, 0.2705, 0.2365, 0.1894, 0.0545, 0.1317], device='cuda:1'), in_proj_covar=tensor([0.0322, 0.0267, 0.0301, 0.0306, 0.0298, 0.0253, 0.0290, 0.0333], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:1') 2023-04-30 21:16:15,868 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=186475.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 21:16:18,145 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.3024, 4.3712, 4.6691, 4.6588, 4.7165, 4.3797, 4.4068, 4.2475], device='cuda:1'), covar=tensor([0.0402, 0.0571, 0.0405, 0.0394, 0.0501, 0.0447, 0.0874, 0.0741], device='cuda:1'), in_proj_covar=tensor([0.0407, 0.0448, 0.0433, 0.0406, 0.0478, 0.0457, 0.0552, 0.0365], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:1') 2023-04-30 21:16:23,025 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([5.1305, 5.6660, 5.8445, 5.5557, 5.5291, 6.1646, 5.7331, 5.3973], device='cuda:1'), covar=tensor([0.0879, 0.1600, 0.1631, 0.1666, 0.2234, 0.0867, 0.1390, 0.2274], device='cuda:1'), in_proj_covar=tensor([0.0404, 0.0587, 0.0648, 0.0491, 0.0655, 0.0685, 0.0508, 0.0657], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-04-30 21:16:26,558 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=186482.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 21:16:27,619 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.7152, 4.6364, 4.6159, 4.3241, 4.3455, 4.6566, 4.5051, 4.4543], device='cuda:1'), covar=tensor([0.0709, 0.0852, 0.0348, 0.0306, 0.0891, 0.0584, 0.0443, 0.0676], device='cuda:1'), in_proj_covar=tensor([0.0301, 0.0429, 0.0352, 0.0343, 0.0362, 0.0398, 0.0240, 0.0417], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-04-30 21:16:53,852 INFO [train.py:904] (1/8) Epoch 19, batch 3800, loss[loss=0.1744, simple_loss=0.2654, pruned_loss=0.04173, over 16974.00 frames. ], tot_loss[loss=0.1731, simple_loss=0.2541, pruned_loss=0.04603, over 3278712.67 frames. ], batch size: 41, lr: 3.58e-03, grad_scale: 16.0 2023-04-30 21:16:55,938 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=186503.0, num_to_drop=1, layers_to_drop={1} 2023-04-30 21:17:11,597 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=186514.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 21:17:15,476 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=186517.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 21:17:43,557 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=186536.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 21:17:47,915 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=186539.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 21:18:05,817 INFO [train.py:904] (1/8) Epoch 19, batch 3850, loss[loss=0.1919, simple_loss=0.2619, pruned_loss=0.06098, over 16888.00 frames. ], tot_loss[loss=0.1737, simple_loss=0.2541, pruned_loss=0.04663, over 3278130.73 frames. ], batch size: 109, lr: 3.58e-03, grad_scale: 4.0 2023-04-30 21:18:22,162 INFO [optim.py:368] (1/8) 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 21:18:28,817 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=186567.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 21:18:44,986 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=186578.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 21:18:49,704 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.9496, 4.9243, 4.8189, 4.1530, 4.8994, 1.9046, 4.6239, 4.4541], device='cuda:1'), covar=tensor([0.0117, 0.0095, 0.0171, 0.0338, 0.0085, 0.2700, 0.0134, 0.0215], device='cuda:1'), in_proj_covar=tensor([0.0161, 0.0150, 0.0196, 0.0179, 0.0173, 0.0205, 0.0187, 0.0175], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-30 21:18:56,153 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=186586.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 21:19:11,507 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.84 vs. limit=2.0 2023-04-30 21:19:16,637 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 19, batch 3900, loss[loss=0.1956, simple_loss=0.2645, pruned_loss=0.06338, over 16760.00 frames. ], tot_loss[loss=0.1739, simple_loss=0.2537, pruned_loss=0.04708, over 3276143.13 frames. ], batch size: 124, lr: 3.58e-03, grad_scale: 4.0 2023-04-30 21:19:36,729 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=186615.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 21:19:55,077 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=186628.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 21:20:01,609 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.26 vs. limit=2.0 2023-04-30 21:20:15,231 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.89 vs. limit=2.0 2023-04-30 21:20:20,736 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=186647.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 21:20:28,329 INFO [train.py:904] (1/8) Epoch 19, batch 3950, loss[loss=0.1765, simple_loss=0.2583, pruned_loss=0.04733, over 16845.00 frames. ], tot_loss[loss=0.1746, simple_loss=0.2535, pruned_loss=0.04789, over 3284721.80 frames. ], batch size: 116, lr: 3.57e-03, grad_scale: 4.0 2023-04-30 21:20:42,972 INFO [optim.py:368] (1/8) 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 21:21:39,185 INFO [train.py:904] (1/8) Epoch 19, batch 4000, loss[loss=0.1732, simple_loss=0.2451, pruned_loss=0.05064, over 16767.00 frames. ], tot_loss[loss=0.1756, simple_loss=0.254, pruned_loss=0.04863, over 3274562.46 frames. ], batch size: 124, lr: 3.57e-03, grad_scale: 8.0 2023-04-30 21:21:48,963 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=186709.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 21:22:24,935 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.0861, 2.1838, 2.2083, 3.7520, 2.1571, 2.5261, 2.2791, 2.3589], device='cuda:1'), covar=tensor([0.1399, 0.3589, 0.2840, 0.0572, 0.3855, 0.2466, 0.3475, 0.3089], device='cuda:1'), in_proj_covar=tensor([0.0398, 0.0441, 0.0362, 0.0328, 0.0434, 0.0509, 0.0410, 0.0516], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-30 21:22:49,923 INFO [train.py:904] (1/8) Epoch 19, batch 4050, loss[loss=0.1593, simple_loss=0.2442, pruned_loss=0.03723, over 16408.00 frames. ], tot_loss[loss=0.1745, simple_loss=0.2541, pruned_loss=0.0475, over 3279339.11 frames. ], batch size: 146, lr: 3.57e-03, grad_scale: 8.0 2023-04-30 21:22:57,381 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=186756.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 21:23:06,011 INFO [optim.py:368] (1/8) 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 21:23:16,214 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=186770.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 21:23:26,473 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=186777.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 21:24:03,612 INFO [train.py:904] (1/8) Epoch 19, batch 4100, loss[loss=0.1882, simple_loss=0.281, pruned_loss=0.04768, over 15311.00 frames. ], tot_loss[loss=0.1748, simple_loss=0.2559, pruned_loss=0.04688, over 3267913.31 frames. ], batch size: 190, lr: 3.57e-03, grad_scale: 8.0 2023-04-30 21:24:07,124 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=186804.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 21:24:21,727 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=186814.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 21:24:46,677 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=186831.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 21:24:59,895 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 19, batch 4150, loss[loss=0.2209, simple_loss=0.2975, pruned_loss=0.0721, over 11367.00 frames. ], tot_loss[loss=0.1803, simple_loss=0.2627, pruned_loss=0.04897, over 3248920.79 frames. ], batch size: 246, lr: 3.57e-03, grad_scale: 8.0 2023-04-30 21:25:28,318 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.8171, 3.8017, 3.9539, 3.7218, 3.8693, 4.2566, 3.9327, 3.6385], device='cuda:1'), covar=tensor([0.1925, 0.2239, 0.2062, 0.2276, 0.2513, 0.1658, 0.1489, 0.2377], device='cuda:1'), in_proj_covar=tensor([0.0403, 0.0582, 0.0641, 0.0486, 0.0648, 0.0679, 0.0502, 0.0648], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-04-30 21:25:33,650 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=186862.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 21:25:34,924 INFO [optim.py:368] (1/8) 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 21:25:35,535 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.0575, 3.7692, 3.6664, 2.3460, 3.4272, 3.7105, 3.3640, 2.1427], device='cuda:1'), covar=tensor([0.0599, 0.0043, 0.0054, 0.0432, 0.0096, 0.0115, 0.0092, 0.0477], device='cuda:1'), in_proj_covar=tensor([0.0135, 0.0080, 0.0081, 0.0132, 0.0095, 0.0105, 0.0092, 0.0126], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-04-30 21:25:49,713 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=186873.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 21:26:10,357 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([5.1980, 5.4799, 5.2376, 5.2850, 4.9668, 4.8268, 4.9185, 5.5871], device='cuda:1'), covar=tensor([0.1148, 0.0808, 0.0952, 0.0830, 0.0850, 0.0848, 0.1074, 0.0866], device='cuda:1'), in_proj_covar=tensor([0.0659, 0.0810, 0.0666, 0.0605, 0.0508, 0.0516, 0.0676, 0.0622], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:1') 2023-04-30 21:26:22,974 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=186895.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 21:26:31,963 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=186901.0, num_to_drop=1, layers_to_drop={1} 2023-04-30 21:26:32,668 INFO [train.py:904] (1/8) Epoch 19, batch 4200, loss[loss=0.23, simple_loss=0.316, pruned_loss=0.07196, over 16472.00 frames. ], tot_loss[loss=0.1861, simple_loss=0.2702, pruned_loss=0.05103, over 3218309.85 frames. ], batch size: 35, lr: 3.57e-03, grad_scale: 8.0 2023-04-30 21:26:47,049 INFO [zipformer.py:625] (1/8) 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] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=186915.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 21:27:05,438 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.0008, 2.2936, 2.3138, 2.7800, 2.0812, 3.0993, 1.7913, 2.6116], device='cuda:1'), covar=tensor([0.1186, 0.0650, 0.1113, 0.0172, 0.0163, 0.0360, 0.1487, 0.0774], device='cuda:1'), in_proj_covar=tensor([0.0164, 0.0172, 0.0191, 0.0184, 0.0205, 0.0215, 0.0196, 0.0189], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-04-30 21:27:06,383 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=186923.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 21:27:34,738 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=186942.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 21:27:49,659 INFO [train.py:904] (1/8) Epoch 19, batch 4250, loss[loss=0.187, simple_loss=0.2808, pruned_loss=0.04662, over 16689.00 frames. ], tot_loss[loss=0.1874, simple_loss=0.2734, pruned_loss=0.05072, over 3205539.01 frames. ], batch size: 83, lr: 3.57e-03, grad_scale: 8.0 2023-04-30 21:28:05,621 INFO [optim.py:368] (1/8) 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 21:28:06,529 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=186963.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 21:28:19,160 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=186972.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 21:29:02,884 INFO [train.py:904] (1/8) Epoch 19, batch 4300, loss[loss=0.1921, simple_loss=0.2847, pruned_loss=0.0497, over 16584.00 frames. ], tot_loss[loss=0.1869, simple_loss=0.2743, pruned_loss=0.04979, over 3197126.37 frames. ], batch size: 62, lr: 3.57e-03, grad_scale: 8.0 2023-04-30 21:29:37,033 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=187025.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 21:29:43,312 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.4856, 2.4408, 2.4487, 4.3587, 2.1832, 2.8481, 2.5118, 2.5774], device='cuda:1'), covar=tensor([0.1154, 0.3045, 0.2569, 0.0409, 0.4012, 0.2179, 0.2989, 0.3271], device='cuda:1'), in_proj_covar=tensor([0.0395, 0.0439, 0.0359, 0.0324, 0.0432, 0.0506, 0.0407, 0.0513], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-30 21:29:56,608 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=187038.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 21:30:16,621 INFO [train.py:904] (1/8) Epoch 19, batch 4350, loss[loss=0.1875, simple_loss=0.2793, pruned_loss=0.0478, over 17206.00 frames. ], tot_loss[loss=0.1896, simple_loss=0.2774, pruned_loss=0.05089, over 3186598.37 frames. ], batch size: 45, lr: 3.57e-03, grad_scale: 8.0 2023-04-30 21:30:32,929 INFO [optim.py:368] (1/8) 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] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=187065.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 21:30:53,891 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=187077.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 21:31:07,477 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=187086.0, num_to_drop=1, layers_to_drop={0} 2023-04-30 21:31:26,278 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=187099.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 21:31:31,215 INFO [train.py:904] (1/8) Epoch 19, batch 4400, loss[loss=0.1922, simple_loss=0.2881, pruned_loss=0.0482, over 16835.00 frames. ], tot_loss[loss=0.1914, simple_loss=0.2792, pruned_loss=0.05177, over 3200669.04 frames. ], batch size: 116, lr: 3.57e-03, grad_scale: 8.0 2023-04-30 21:31:35,972 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.3072, 2.3825, 2.9307, 3.2985, 3.1317, 3.8304, 2.4183, 3.7393], device='cuda:1'), covar=tensor([0.0179, 0.0408, 0.0260, 0.0214, 0.0222, 0.0113, 0.0456, 0.0122], device='cuda:1'), in_proj_covar=tensor([0.0181, 0.0188, 0.0177, 0.0179, 0.0190, 0.0148, 0.0192, 0.0143], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-30 21:31:39,073 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([5.4860, 5.4608, 5.3328, 4.9444, 4.9492, 5.3582, 5.2047, 4.9877], device='cuda:1'), covar=tensor([0.0464, 0.0217, 0.0220, 0.0244, 0.0887, 0.0280, 0.0272, 0.0591], device='cuda:1'), in_proj_covar=tensor([0.0287, 0.0408, 0.0338, 0.0327, 0.0347, 0.0379, 0.0230, 0.0400], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002], device='cuda:1') 2023-04-30 21:31:46,406 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.2342, 3.9253, 3.7578, 2.4461, 3.4509, 3.8156, 3.5040, 2.0448], device='cuda:1'), covar=tensor([0.0495, 0.0030, 0.0050, 0.0386, 0.0085, 0.0079, 0.0080, 0.0442], device='cuda:1'), in_proj_covar=tensor([0.0136, 0.0080, 0.0081, 0.0133, 0.0096, 0.0106, 0.0093, 0.0127], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-04-30 21:31:52,421 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.6856, 4.5528, 4.7236, 4.8503, 4.9948, 4.5177, 5.0142, 5.0220], device='cuda:1'), covar=tensor([0.1436, 0.0986, 0.1248, 0.0581, 0.0410, 0.0850, 0.0430, 0.0482], device='cuda:1'), in_proj_covar=tensor([0.0622, 0.0769, 0.0903, 0.0787, 0.0587, 0.0615, 0.0632, 0.0737], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-30 21:32:04,910 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=187125.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 21:32:13,618 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=187131.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 21:32:43,362 INFO [train.py:904] (1/8) Epoch 19, batch 4450, loss[loss=0.203, simple_loss=0.2755, pruned_loss=0.06525, over 11875.00 frames. ], tot_loss[loss=0.1935, simple_loss=0.2818, pruned_loss=0.05262, over 3195855.61 frames. ], batch size: 248, lr: 3.57e-03, grad_scale: 8.0 2023-04-30 21:32:47,781 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([5.2190, 5.1988, 5.5487, 5.5116, 5.6357, 5.1807, 5.1394, 4.8167], device='cuda:1'), covar=tensor([0.0237, 0.0376, 0.0279, 0.0351, 0.0372, 0.0340, 0.1010, 0.0425], device='cuda:1'), in_proj_covar=tensor([0.0393, 0.0432, 0.0419, 0.0392, 0.0461, 0.0443, 0.0534, 0.0349], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:1') 2023-04-30 21:33:00,621 INFO [optim.py:368] (1/8) 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 21:33:15,026 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=187173.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 21:33:24,158 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=187179.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 21:33:47,374 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=187195.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 21:33:49,219 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=187196.0, num_to_drop=1, layers_to_drop={2} 2023-04-30 21:33:57,507 INFO [train.py:904] (1/8) Epoch 19, batch 4500, loss[loss=0.2021, simple_loss=0.2895, pruned_loss=0.05741, over 15363.00 frames. ], tot_loss[loss=0.1947, simple_loss=0.2825, pruned_loss=0.05346, over 3198235.06 frames. ], batch size: 191, lr: 3.57e-03, grad_scale: 8.0 2023-04-30 21:34:24,640 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=187221.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 21:34:27,154 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=187223.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 21:34:39,785 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.4949, 4.5592, 4.3376, 2.8370, 3.8559, 4.4393, 3.8878, 2.2339], device='cuda:1'), covar=tensor([0.0502, 0.0020, 0.0035, 0.0381, 0.0077, 0.0066, 0.0077, 0.0477], device='cuda:1'), in_proj_covar=tensor([0.0135, 0.0080, 0.0081, 0.0133, 0.0095, 0.0106, 0.0093, 0.0127], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-04-30 21:34:54,323 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=187242.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 21:34:55,416 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=187243.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 21:35:08,491 INFO [train.py:904] (1/8) Epoch 19, batch 4550, loss[loss=0.2021, simple_loss=0.2909, pruned_loss=0.05665, over 16724.00 frames. ], tot_loss[loss=0.1961, simple_loss=0.2836, pruned_loss=0.05428, over 3208007.74 frames. ], batch size: 76, lr: 3.57e-03, grad_scale: 8.0 2023-04-30 21:35:24,436 INFO [optim.py:368] (1/8) 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 21:35:30,227 INFO [zipformer.py:625] (1/8) 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] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=187271.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 21:35:43,227 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.8767, 2.1489, 2.4017, 3.1532, 2.1896, 2.3406, 2.3419, 2.2215], device='cuda:1'), covar=tensor([0.1224, 0.3037, 0.2241, 0.0643, 0.3818, 0.2287, 0.2769, 0.3255], device='cuda:1'), in_proj_covar=tensor([0.0395, 0.0438, 0.0358, 0.0323, 0.0432, 0.0505, 0.0407, 0.0511], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-30 21:36:01,884 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=5.06 vs. limit=5.0 2023-04-30 21:36:04,448 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 19, batch 4600, loss[loss=0.192, simple_loss=0.2784, pruned_loss=0.05279, over 16987.00 frames. ], tot_loss[loss=0.197, simple_loss=0.2848, pruned_loss=0.05454, over 3217963.94 frames. ], batch size: 41, lr: 3.57e-03, grad_scale: 8.0 2023-04-30 21:37:33,043 INFO [train.py:904] (1/8) Epoch 19, batch 4650, loss[loss=0.1985, simple_loss=0.2862, pruned_loss=0.05542, over 16650.00 frames. ], tot_loss[loss=0.1969, simple_loss=0.2844, pruned_loss=0.05472, over 3208320.91 frames. ], batch size: 134, lr: 3.57e-03, grad_scale: 8.0 2023-04-30 21:37:49,310 INFO [optim.py:368] (1/8) 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 21:37:52,245 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=187365.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 21:38:01,555 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=187372.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 21:38:14,869 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=187381.0, num_to_drop=1, layers_to_drop={3} 2023-04-30 21:38:33,314 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=187394.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 21:38:43,237 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=3.12 vs. limit=5.0 2023-04-30 21:38:44,846 INFO [train.py:904] (1/8) Epoch 19, batch 4700, loss[loss=0.1704, simple_loss=0.2556, pruned_loss=0.04263, over 16699.00 frames. ], tot_loss[loss=0.1949, simple_loss=0.282, pruned_loss=0.05391, over 3216263.42 frames. ], batch size: 134, lr: 3.57e-03, grad_scale: 8.0 2023-04-30 21:38:53,644 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([5.3132, 5.3118, 5.1417, 4.4325, 5.2422, 1.8464, 4.9753, 4.9108], device='cuda:1'), covar=tensor([0.0089, 0.0089, 0.0136, 0.0419, 0.0094, 0.2729, 0.0115, 0.0223], device='cuda:1'), in_proj_covar=tensor([0.0157, 0.0146, 0.0191, 0.0175, 0.0168, 0.0201, 0.0183, 0.0170], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-30 21:38:58,763 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.15 vs. limit=2.0 2023-04-30 21:39:01,715 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=187413.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 21:39:29,950 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=187433.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 21:39:34,152 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.1105, 3.4612, 3.5274, 3.4917, 3.5075, 3.3904, 3.1344, 3.4410], device='cuda:1'), covar=tensor([0.0645, 0.0716, 0.0622, 0.0697, 0.0719, 0.0673, 0.1308, 0.0621], device='cuda:1'), in_proj_covar=tensor([0.0390, 0.0428, 0.0417, 0.0390, 0.0460, 0.0440, 0.0530, 0.0347], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:1') 2023-04-30 21:39:55,445 INFO [train.py:904] (1/8) Epoch 19, batch 4750, loss[loss=0.1633, simple_loss=0.251, pruned_loss=0.03778, over 16650.00 frames. ], tot_loss[loss=0.1902, simple_loss=0.2774, pruned_loss=0.0515, over 3228165.31 frames. ], batch size: 89, lr: 3.57e-03, grad_scale: 8.0 2023-04-30 21:40:11,092 INFO [optim.py:368] (1/8) 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 21:40:57,698 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 19, batch 4800, loss[loss=0.1719, simple_loss=0.2592, pruned_loss=0.04232, over 17254.00 frames. ], tot_loss[loss=0.1869, simple_loss=0.2744, pruned_loss=0.04971, over 3220638.41 frames. ], batch size: 52, lr: 3.57e-03, grad_scale: 8.0 2023-04-30 21:41:20,895 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=187512.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 21:41:30,339 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=187519.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 21:41:30,638 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=3.00 vs. limit=5.0 2023-04-30 21:41:37,722 INFO [zipformer.py:625] (1/8) 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] (1/8) attn_weights_entropy = tensor([4.1542, 4.1364, 4.1050, 3.2497, 4.0783, 1.5975, 3.8667, 3.7418], device='cuda:1'), covar=tensor([0.0123, 0.0130, 0.0156, 0.0440, 0.0114, 0.2887, 0.0161, 0.0267], device='cuda:1'), in_proj_covar=tensor([0.0156, 0.0145, 0.0190, 0.0174, 0.0167, 0.0200, 0.0181, 0.0169], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-30 21:42:08,454 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=187544.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 21:42:19,532 INFO [train.py:904] (1/8) Epoch 19, batch 4850, loss[loss=0.2081, simple_loss=0.3083, pruned_loss=0.054, over 16239.00 frames. ], tot_loss[loss=0.1868, simple_loss=0.2752, pruned_loss=0.04924, over 3204866.52 frames. ], batch size: 165, lr: 3.57e-03, grad_scale: 8.0 2023-04-30 21:42:36,204 INFO [optim.py:368] (1/8) 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 21:42:41,814 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=187567.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 21:42:50,658 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.7379, 4.8804, 5.0138, 4.8209, 4.8422, 5.4126, 4.9063, 4.5946], device='cuda:1'), covar=tensor([0.1051, 0.1786, 0.1953, 0.1841, 0.2378, 0.0755, 0.1375, 0.2184], device='cuda:1'), in_proj_covar=tensor([0.0395, 0.0566, 0.0623, 0.0474, 0.0632, 0.0658, 0.0488, 0.0635], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-04-30 21:42:50,846 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=187573.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 21:43:01,967 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=187580.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 21:43:07,583 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=187584.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 21:43:32,170 INFO [train.py:904] (1/8) Epoch 19, batch 4900, loss[loss=0.1752, simple_loss=0.2624, pruned_loss=0.04398, over 12026.00 frames. ], tot_loss[loss=0.1856, simple_loss=0.2749, pruned_loss=0.04815, over 3196416.69 frames. ], batch size: 248, lr: 3.57e-03, grad_scale: 8.0 2023-04-30 21:43:51,071 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=187615.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 21:43:56,887 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.32 vs. limit=2.0 2023-04-30 21:44:09,410 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.6982, 3.5526, 4.1061, 1.9427, 4.3121, 4.2645, 3.0242, 2.9547], device='cuda:1'), covar=tensor([0.0736, 0.0255, 0.0147, 0.1206, 0.0045, 0.0095, 0.0421, 0.0471], device='cuda:1'), in_proj_covar=tensor([0.0147, 0.0107, 0.0097, 0.0137, 0.0077, 0.0122, 0.0126, 0.0130], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0004, 0.0003], device='cuda:1') 2023-04-30 21:44:19,392 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=4.14 vs. limit=5.0 2023-04-30 21:44:37,198 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.18 vs. limit=2.0 2023-04-30 21:44:43,105 INFO [train.py:904] (1/8) Epoch 19, batch 4950, loss[loss=0.1701, simple_loss=0.261, pruned_loss=0.03957, over 16897.00 frames. ], tot_loss[loss=0.1856, simple_loss=0.2747, pruned_loss=0.04829, over 3188280.72 frames. ], batch size: 42, lr: 3.57e-03, grad_scale: 8.0 2023-04-30 21:44:58,341 INFO [optim.py:368] (1/8) 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 21:45:24,089 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=187681.0, num_to_drop=1, layers_to_drop={2} 2023-04-30 21:45:43,950 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=187694.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 21:45:47,617 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.7781, 3.8133, 2.3850, 4.5801, 2.9046, 4.4532, 2.4615, 2.9772], device='cuda:1'), covar=tensor([0.0255, 0.0319, 0.1515, 0.0132, 0.0848, 0.0432, 0.1425, 0.0755], device='cuda:1'), in_proj_covar=tensor([0.0165, 0.0173, 0.0190, 0.0154, 0.0172, 0.0211, 0.0196, 0.0176], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-04-30 21:45:55,107 INFO [train.py:904] (1/8) Epoch 19, batch 5000, loss[loss=0.1769, simple_loss=0.2769, pruned_loss=0.03845, over 16773.00 frames. ], tot_loss[loss=0.186, simple_loss=0.2758, pruned_loss=0.04812, over 3201536.06 frames. ], batch size: 102, lr: 3.56e-03, grad_scale: 8.0 2023-04-30 21:46:32,202 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=187728.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 21:46:33,445 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=187729.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 21:46:52,595 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 19, batch 5050, loss[loss=0.1875, simple_loss=0.2765, pruned_loss=0.04926, over 16440.00 frames. ], tot_loss[loss=0.1863, simple_loss=0.2764, pruned_loss=0.04813, over 3197024.64 frames. ], batch size: 146, lr: 3.56e-03, grad_scale: 8.0 2023-04-30 21:47:19,589 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([5.6982, 5.9992, 5.6705, 5.8610, 5.4693, 5.2382, 5.4469, 6.1095], device='cuda:1'), covar=tensor([0.1114, 0.0808, 0.1094, 0.0707, 0.0769, 0.0696, 0.1074, 0.0824], device='cuda:1'), in_proj_covar=tensor([0.0645, 0.0796, 0.0655, 0.0590, 0.0501, 0.0508, 0.0660, 0.0614], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:1') 2023-04-30 21:47:21,707 INFO [optim.py:368] (1/8) 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 21:48:17,853 INFO [train.py:904] (1/8) Epoch 19, batch 5100, loss[loss=0.2076, simple_loss=0.2958, pruned_loss=0.05968, over 12674.00 frames. ], tot_loss[loss=0.1848, simple_loss=0.2746, pruned_loss=0.04748, over 3202532.46 frames. ], batch size: 248, lr: 3.56e-03, grad_scale: 8.0 2023-04-30 21:49:30,691 INFO [train.py:904] (1/8) Epoch 19, batch 5150, loss[loss=0.1618, simple_loss=0.2505, pruned_loss=0.03661, over 17201.00 frames. ], tot_loss[loss=0.184, simple_loss=0.275, pruned_loss=0.04645, over 3217780.17 frames. ], batch size: 45, lr: 3.56e-03, grad_scale: 8.0 2023-04-30 21:49:47,461 INFO [optim.py:368] (1/8) 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 21:49:54,399 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=187868.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 21:50:04,808 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=187875.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 21:50:11,340 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=187879.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 21:50:41,937 INFO [train.py:904] (1/8) Epoch 19, batch 5200, loss[loss=0.1831, simple_loss=0.2678, pruned_loss=0.0492, over 16547.00 frames. ], tot_loss[loss=0.1829, simple_loss=0.274, pruned_loss=0.04588, over 3220409.99 frames. ], batch size: 68, lr: 3.56e-03, grad_scale: 8.0 2023-04-30 21:51:53,326 INFO [train.py:904] (1/8) Epoch 19, batch 5250, loss[loss=0.1866, simple_loss=0.2792, pruned_loss=0.047, over 16351.00 frames. ], tot_loss[loss=0.1806, simple_loss=0.2708, pruned_loss=0.04524, over 3234123.27 frames. ], batch size: 146, lr: 3.56e-03, grad_scale: 8.0 2023-04-30 21:52:08,308 INFO [optim.py:368] (1/8) Clipping_scale=2.0, grad-norm quartiles 1.295e+02 1.924e+02 2.321e+02 2.643e+02 4.137e+02, threshold=4.643e+02, percent-clipped=0.0 2023-04-30 21:52:40,475 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([1.9512, 2.3803, 2.3118, 2.8199, 2.0314, 3.2553, 1.7365, 2.6908], device='cuda:1'), covar=tensor([0.1129, 0.0619, 0.1031, 0.0166, 0.0125, 0.0366, 0.1408, 0.0690], device='cuda:1'), in_proj_covar=tensor([0.0164, 0.0173, 0.0192, 0.0183, 0.0204, 0.0215, 0.0197, 0.0190], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-04-30 21:52:52,889 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.0585, 4.1360, 2.5170, 4.9568, 3.4173, 4.8635, 2.9010, 3.4002], device='cuda:1'), covar=tensor([0.0265, 0.0339, 0.1709, 0.0145, 0.0724, 0.0375, 0.1357, 0.0664], device='cuda:1'), in_proj_covar=tensor([0.0165, 0.0174, 0.0192, 0.0154, 0.0173, 0.0212, 0.0198, 0.0177], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-04-30 21:52:57,555 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.32 vs. limit=2.0 2023-04-30 21:53:08,598 INFO [train.py:904] (1/8) Epoch 19, batch 5300, loss[loss=0.1595, simple_loss=0.2504, pruned_loss=0.03423, over 16889.00 frames. ], tot_loss[loss=0.1785, simple_loss=0.2677, pruned_loss=0.04462, over 3215995.37 frames. ], batch size: 96, lr: 3.56e-03, grad_scale: 8.0 2023-04-30 21:53:46,805 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=188028.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 21:54:02,453 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.28 vs. limit=2.0 2023-04-30 21:54:21,264 INFO [train.py:904] (1/8) Epoch 19, batch 5350, loss[loss=0.1857, simple_loss=0.2725, pruned_loss=0.04947, over 17179.00 frames. ], tot_loss[loss=0.1772, simple_loss=0.2663, pruned_loss=0.04401, over 3208464.83 frames. ], batch size: 46, lr: 3.56e-03, grad_scale: 8.0 2023-04-30 21:54:30,295 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.48 vs. limit=2.0 2023-04-30 21:54:38,012 INFO [optim.py:368] (1/8) 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] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=188076.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 21:55:35,753 INFO [train.py:904] (1/8) Epoch 19, batch 5400, loss[loss=0.1772, simple_loss=0.2725, pruned_loss=0.04094, over 16700.00 frames. ], tot_loss[loss=0.179, simple_loss=0.2687, pruned_loss=0.04465, over 3210430.58 frames. ], batch size: 89, lr: 3.56e-03, grad_scale: 8.0 2023-04-30 21:56:54,045 INFO [train.py:904] (1/8) Epoch 19, batch 5450, loss[loss=0.2124, simple_loss=0.2962, pruned_loss=0.06429, over 16882.00 frames. ], tot_loss[loss=0.1823, simple_loss=0.2721, pruned_loss=0.04628, over 3192521.45 frames. ], batch size: 116, lr: 3.56e-03, grad_scale: 8.0 2023-04-30 21:57:11,918 INFO [optim.py:368] (1/8) 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 21:57:19,760 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=188168.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 21:57:23,916 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.83 vs. limit=2.0 2023-04-30 21:57:32,275 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=188175.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 21:57:37,815 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=188179.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 21:58:00,399 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.8543, 4.8776, 4.7298, 4.3911, 4.3595, 4.7702, 4.6663, 4.5351], device='cuda:1'), covar=tensor([0.0528, 0.0535, 0.0276, 0.0283, 0.0926, 0.0479, 0.0349, 0.0580], device='cuda:1'), in_proj_covar=tensor([0.0284, 0.0408, 0.0335, 0.0326, 0.0346, 0.0382, 0.0230, 0.0399], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002], device='cuda:1') 2023-04-30 21:58:14,499 INFO [train.py:904] (1/8) Epoch 19, batch 5500, loss[loss=0.2921, simple_loss=0.3511, pruned_loss=0.1166, over 12183.00 frames. ], tot_loss[loss=0.1902, simple_loss=0.2793, pruned_loss=0.05061, over 3162195.04 frames. ], batch size: 248, lr: 3.56e-03, grad_scale: 8.0 2023-04-30 21:58:37,952 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=188216.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 21:58:48,994 INFO [zipformer.py:625] (1/8) 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] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=188227.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 21:59:35,425 INFO [train.py:904] (1/8) Epoch 19, batch 5550, loss[loss=0.2702, simple_loss=0.3316, pruned_loss=0.1044, over 11101.00 frames. ], tot_loss[loss=0.1993, simple_loss=0.2867, pruned_loss=0.05593, over 3141346.63 frames. ], batch size: 248, lr: 3.56e-03, grad_scale: 8.0 2023-04-30 21:59:53,552 INFO [optim.py:368] (1/8) 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 22:00:26,045 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=188283.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 22:00:55,106 INFO [train.py:904] (1/8) Epoch 19, batch 5600, loss[loss=0.3005, simple_loss=0.3424, pruned_loss=0.1293, over 11037.00 frames. ], tot_loss[loss=0.2065, simple_loss=0.2919, pruned_loss=0.06057, over 3094170.30 frames. ], batch size: 246, lr: 3.56e-03, grad_scale: 8.0 2023-04-30 22:02:07,220 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 19, batch 5650, loss[loss=0.1857, simple_loss=0.2718, pruned_loss=0.0498, over 16470.00 frames. ], tot_loss[loss=0.2112, simple_loss=0.2956, pruned_loss=0.06337, over 3102500.05 frames. ], batch size: 62, lr: 3.56e-03, grad_scale: 8.0 2023-04-30 22:02:36,991 INFO [optim.py:368] (1/8) 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 22:03:30,525 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=3.76 vs. limit=5.0 2023-04-30 22:03:36,659 INFO [train.py:904] (1/8) Epoch 19, batch 5700, loss[loss=0.2779, simple_loss=0.3396, pruned_loss=0.1081, over 11516.00 frames. ], tot_loss[loss=0.214, simple_loss=0.2974, pruned_loss=0.06533, over 3070737.90 frames. ], batch size: 248, lr: 3.56e-03, grad_scale: 8.0 2023-04-30 22:04:16,428 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.81 vs. limit=2.0 2023-04-30 22:04:32,467 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.32 vs. limit=2.0 2023-04-30 22:04:55,839 INFO [train.py:904] (1/8) Epoch 19, batch 5750, loss[loss=0.2241, simple_loss=0.311, pruned_loss=0.06855, over 16245.00 frames. ], tot_loss[loss=0.2168, simple_loss=0.3, pruned_loss=0.0668, over 3050813.97 frames. ], batch size: 165, lr: 3.56e-03, grad_scale: 8.0 2023-04-30 22:05:12,679 INFO [optim.py:368] (1/8) 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 22:05:45,253 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=188481.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 22:06:19,318 INFO [train.py:904] (1/8) Epoch 19, batch 5800, loss[loss=0.2054, simple_loss=0.3011, pruned_loss=0.05481, over 16731.00 frames. ], tot_loss[loss=0.2155, simple_loss=0.2997, pruned_loss=0.06561, over 3051865.95 frames. ], batch size: 134, lr: 3.56e-03, grad_scale: 8.0 2023-04-30 22:07:24,014 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 19, batch 5850, loss[loss=0.2368, simple_loss=0.3152, pruned_loss=0.07924, over 11580.00 frames. ], tot_loss[loss=0.212, simple_loss=0.2971, pruned_loss=0.06347, over 3066666.05 frames. ], batch size: 249, lr: 3.56e-03, grad_scale: 8.0 2023-04-30 22:07:57,310 INFO [optim.py:368] (1/8) 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 22:08:59,600 INFO [train.py:904] (1/8) Epoch 19, batch 5900, loss[loss=0.2125, simple_loss=0.3042, pruned_loss=0.06043, over 16577.00 frames. ], tot_loss[loss=0.2109, simple_loss=0.2961, pruned_loss=0.06282, over 3081288.55 frames. ], batch size: 68, lr: 3.56e-03, grad_scale: 8.0 2023-04-30 22:09:45,330 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.89 vs. limit=2.0 2023-04-30 22:09:57,620 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.75 vs. limit=2.0 2023-04-30 22:10:01,849 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 19, batch 5950, loss[loss=0.2087, simple_loss=0.2999, pruned_loss=0.05873, over 16857.00 frames. ], tot_loss[loss=0.2105, simple_loss=0.297, pruned_loss=0.06202, over 3062391.74 frames. ], batch size: 83, lr: 3.56e-03, grad_scale: 8.0 2023-04-30 22:10:40,535 INFO [optim.py:368] (1/8) 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] (1/8) Epoch 19, batch 6000, loss[loss=0.2026, simple_loss=0.2846, pruned_loss=0.06031, over 16227.00 frames. ], tot_loss[loss=0.21, simple_loss=0.2965, pruned_loss=0.06174, over 3056389.64 frames. ], batch size: 165, lr: 3.56e-03, grad_scale: 8.0 2023-04-30 22:11:41,786 INFO [train.py:929] (1/8) Computing validation loss 2023-04-30 22:11:49,014 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.7907, 2.0175, 2.4253, 2.7851, 2.5817, 3.2203, 2.3832, 3.2100], device='cuda:1'), covar=tensor([0.0203, 0.0458, 0.0360, 0.0299, 0.0319, 0.0171, 0.0409, 0.0147], device='cuda:1'), in_proj_covar=tensor([0.0179, 0.0187, 0.0175, 0.0177, 0.0188, 0.0147, 0.0190, 0.0142], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-30 22:11:52,558 INFO [train.py:938] (1/8) Epoch 19, validation: loss=0.1523, simple_loss=0.2653, pruned_loss=0.01968, over 944034.00 frames. 2023-04-30 22:11:52,558 INFO [train.py:939] (1/8) Maximum memory allocated so far is 18145MB 2023-04-30 22:12:00,528 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.7191, 4.6002, 4.6834, 4.9003, 5.0304, 4.5616, 5.0377, 5.0398], device='cuda:1'), covar=tensor([0.1881, 0.1270, 0.1747, 0.0820, 0.0715, 0.1025, 0.0690, 0.0816], device='cuda:1'), in_proj_covar=tensor([0.0610, 0.0749, 0.0879, 0.0771, 0.0575, 0.0603, 0.0618, 0.0722], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-30 22:13:13,860 INFO [train.py:904] (1/8) Epoch 19, batch 6050, loss[loss=0.1922, simple_loss=0.2864, pruned_loss=0.04901, over 16203.00 frames. ], tot_loss[loss=0.2076, simple_loss=0.2943, pruned_loss=0.06042, over 3083002.51 frames. ], batch size: 165, lr: 3.55e-03, grad_scale: 8.0 2023-04-30 22:13:33,103 INFO [optim.py:368] (1/8) 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 22:14:01,236 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.4589, 2.9225, 2.7312, 2.3117, 2.3167, 2.3360, 2.9815, 2.9135], device='cuda:1'), covar=tensor([0.2291, 0.0704, 0.1564, 0.2406, 0.2074, 0.1877, 0.0532, 0.1348], device='cuda:1'), in_proj_covar=tensor([0.0322, 0.0265, 0.0298, 0.0304, 0.0291, 0.0249, 0.0288, 0.0327], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:1') 2023-04-30 22:14:32,939 INFO [train.py:904] (1/8) Epoch 19, batch 6100, loss[loss=0.1813, simple_loss=0.2779, pruned_loss=0.04233, over 16518.00 frames. ], tot_loss[loss=0.2061, simple_loss=0.2936, pruned_loss=0.05929, over 3087013.30 frames. ], batch size: 75, lr: 3.55e-03, grad_scale: 8.0 2023-04-30 22:15:33,781 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=188837.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 22:15:49,751 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.1462, 4.4338, 3.3901, 2.6608, 3.1737, 2.9141, 4.8701, 4.1130], device='cuda:1'), covar=tensor([0.2474, 0.0513, 0.1568, 0.2281, 0.2316, 0.1689, 0.0361, 0.0957], device='cuda:1'), in_proj_covar=tensor([0.0323, 0.0266, 0.0299, 0.0305, 0.0293, 0.0250, 0.0289, 0.0328], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:1') 2023-04-30 22:15:57,515 INFO [train.py:904] (1/8) Epoch 19, batch 6150, loss[loss=0.2044, simple_loss=0.2905, pruned_loss=0.05912, over 16248.00 frames. ], tot_loss[loss=0.2049, simple_loss=0.2919, pruned_loss=0.05894, over 3091084.26 frames. ], batch size: 165, lr: 3.55e-03, grad_scale: 8.0 2023-04-30 22:16:16,884 INFO [optim.py:368] (1/8) 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 22:16:19,739 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=4.79 vs. limit=5.0 2023-04-30 22:16:50,078 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.27 vs. limit=2.0 2023-04-30 22:17:16,331 INFO [train.py:904] (1/8) Epoch 19, batch 6200, loss[loss=0.2301, simple_loss=0.2969, pruned_loss=0.08166, over 11462.00 frames. ], tot_loss[loss=0.2032, simple_loss=0.2901, pruned_loss=0.05819, over 3099420.86 frames. ], batch size: 248, lr: 3.55e-03, grad_scale: 8.0 2023-04-30 22:17:54,301 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.8205, 3.1810, 3.3060, 2.0270, 2.8337, 2.2021, 3.3764, 3.3988], device='cuda:1'), covar=tensor([0.0257, 0.0781, 0.0586, 0.1940, 0.0849, 0.0964, 0.0601, 0.0884], device='cuda:1'), in_proj_covar=tensor([0.0155, 0.0161, 0.0167, 0.0152, 0.0145, 0.0129, 0.0144, 0.0172], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:1') 2023-04-30 22:17:58,824 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=188929.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 22:18:01,812 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=188931.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 22:18:14,973 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 19, batch 6250, loss[loss=0.1997, simple_loss=0.285, pruned_loss=0.05723, over 16739.00 frames. ], tot_loss[loss=0.2026, simple_loss=0.2894, pruned_loss=0.05787, over 3115509.09 frames. ], batch size: 124, lr: 3.55e-03, grad_scale: 8.0 2023-04-30 22:18:52,888 INFO [optim.py:368] (1/8) 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 22:19:18,185 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.5073, 4.5389, 4.8911, 4.8642, 4.8743, 4.5881, 4.5425, 4.4354], device='cuda:1'), covar=tensor([0.0375, 0.0654, 0.0433, 0.0432, 0.0489, 0.0461, 0.1014, 0.0565], device='cuda:1'), in_proj_covar=tensor([0.0395, 0.0434, 0.0421, 0.0395, 0.0468, 0.0443, 0.0537, 0.0354], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:1') 2023-04-30 22:19:28,070 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=188987.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 22:19:32,245 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=188990.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 22:19:36,027 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=188992.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 22:19:51,019 INFO [train.py:904] (1/8) Epoch 19, batch 6300, loss[loss=0.2034, simple_loss=0.2951, pruned_loss=0.05584, over 16885.00 frames. ], tot_loss[loss=0.2027, simple_loss=0.2897, pruned_loss=0.05786, over 3106302.74 frames. ], batch size: 109, lr: 3.55e-03, grad_scale: 8.0 2023-04-30 22:20:35,852 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 19, batch 6350, loss[loss=0.1924, simple_loss=0.2833, pruned_loss=0.05073, over 16683.00 frames. ], tot_loss[loss=0.2053, simple_loss=0.2911, pruned_loss=0.05974, over 3065656.15 frames. ], batch size: 76, lr: 3.55e-03, grad_scale: 8.0 2023-04-30 22:21:27,081 INFO [optim.py:368] (1/8) 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 22:21:28,343 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=189064.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 22:21:51,789 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=189079.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 22:22:03,998 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.9011, 2.7872, 2.8975, 2.1852, 2.7460, 2.2336, 2.8489, 2.9801], device='cuda:1'), covar=tensor([0.0228, 0.0796, 0.0506, 0.1692, 0.0696, 0.0992, 0.0473, 0.0590], device='cuda:1'), in_proj_covar=tensor([0.0154, 0.0161, 0.0167, 0.0152, 0.0145, 0.0129, 0.0143, 0.0172], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-04-30 22:22:10,389 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=189091.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 22:22:10,500 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.6988, 3.9351, 3.0174, 2.2811, 2.7106, 2.5312, 4.3031, 3.5960], device='cuda:1'), covar=tensor([0.2739, 0.0605, 0.1670, 0.2629, 0.2588, 0.1937, 0.0391, 0.1125], device='cuda:1'), in_proj_covar=tensor([0.0324, 0.0265, 0.0299, 0.0305, 0.0293, 0.0250, 0.0289, 0.0328], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:1') 2023-04-30 22:22:26,882 INFO [train.py:904] (1/8) Epoch 19, batch 6400, loss[loss=0.2391, simple_loss=0.3138, pruned_loss=0.08221, over 11389.00 frames. ], tot_loss[loss=0.207, simple_loss=0.2916, pruned_loss=0.06122, over 3057443.52 frames. ], batch size: 247, lr: 3.55e-03, grad_scale: 8.0 2023-04-30 22:23:02,667 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=189125.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 22:23:21,438 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=189137.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 22:23:25,620 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=189140.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 22:23:43,275 INFO [train.py:904] (1/8) Epoch 19, batch 6450, loss[loss=0.1842, simple_loss=0.2788, pruned_loss=0.04479, over 16850.00 frames. ], tot_loss[loss=0.2075, simple_loss=0.2924, pruned_loss=0.06132, over 3041597.52 frames. ], batch size: 96, lr: 3.55e-03, grad_scale: 8.0 2023-04-30 22:24:00,999 INFO [optim.py:368] (1/8) 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 22:24:27,833 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.4219, 4.4820, 4.8078, 4.7801, 4.8135, 4.4817, 4.4949, 4.3603], device='cuda:1'), covar=tensor([0.0346, 0.0600, 0.0399, 0.0436, 0.0449, 0.0415, 0.0982, 0.0501], device='cuda:1'), in_proj_covar=tensor([0.0393, 0.0430, 0.0419, 0.0394, 0.0466, 0.0440, 0.0531, 0.0351], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:1') 2023-04-30 22:24:34,446 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=189185.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 22:24:59,912 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.5948, 3.5333, 2.4685, 2.2300, 2.2853, 2.1223, 3.6557, 3.1162], device='cuda:1'), covar=tensor([0.2917, 0.0794, 0.2140, 0.2899, 0.2962, 0.2455, 0.0577, 0.1361], device='cuda:1'), in_proj_covar=tensor([0.0320, 0.0262, 0.0296, 0.0302, 0.0290, 0.0248, 0.0286, 0.0325], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:1') 2023-04-30 22:25:00,511 INFO [train.py:904] (1/8) Epoch 19, batch 6500, loss[loss=0.1936, simple_loss=0.2837, pruned_loss=0.05173, over 16674.00 frames. ], tot_loss[loss=0.2048, simple_loss=0.2897, pruned_loss=0.05997, over 3052623.76 frames. ], batch size: 134, lr: 3.55e-03, grad_scale: 8.0 2023-04-30 22:26:01,054 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=4.07 vs. limit=5.0 2023-04-30 22:26:14,615 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=189248.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 22:26:19,689 INFO [train.py:904] (1/8) Epoch 19, batch 6550, loss[loss=0.2939, simple_loss=0.3512, pruned_loss=0.1183, over 11391.00 frames. ], tot_loss[loss=0.207, simple_loss=0.2924, pruned_loss=0.06073, over 3082992.92 frames. ], batch size: 246, lr: 3.55e-03, grad_scale: 8.0 2023-04-30 22:26:37,017 INFO [optim.py:368] (1/8) 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 22:27:03,375 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.41 vs. limit=2.0 2023-04-30 22:27:09,601 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=189285.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 22:27:12,558 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 19, batch 6600, loss[loss=0.2151, simple_loss=0.3, pruned_loss=0.06509, over 16911.00 frames. ], tot_loss[loss=0.2083, simple_loss=0.2944, pruned_loss=0.06107, over 3074888.27 frames. ], batch size: 116, lr: 3.55e-03, grad_scale: 8.0 2023-04-30 22:27:45,602 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=189309.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 22:28:44,280 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 19, batch 6650, loss[loss=0.1844, simple_loss=0.2756, pruned_loss=0.04658, over 16621.00 frames. ], tot_loss[loss=0.2082, simple_loss=0.294, pruned_loss=0.06119, over 3088448.89 frames. ], batch size: 76, lr: 3.55e-03, grad_scale: 8.0 2023-04-30 22:29:08,220 INFO [optim.py:368] (1/8) 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 22:29:33,228 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.15 vs. limit=2.0 2023-04-30 22:29:42,425 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=189386.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 22:30:06,163 INFO [train.py:904] (1/8) Epoch 19, batch 6700, loss[loss=0.2058, simple_loss=0.292, pruned_loss=0.0598, over 16795.00 frames. ], tot_loss[loss=0.2082, simple_loss=0.293, pruned_loss=0.06174, over 3077958.01 frames. ], batch size: 124, lr: 3.55e-03, grad_scale: 8.0 2023-04-30 22:30:13,018 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.4875, 2.2241, 1.7503, 1.9889, 2.5126, 2.1592, 2.2702, 2.6464], device='cuda:1'), covar=tensor([0.0176, 0.0404, 0.0520, 0.0444, 0.0230, 0.0372, 0.0200, 0.0231], device='cuda:1'), in_proj_covar=tensor([0.0193, 0.0224, 0.0217, 0.0218, 0.0225, 0.0224, 0.0225, 0.0220], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-30 22:30:15,374 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.6274, 2.5709, 1.8535, 2.6965, 2.0805, 2.7648, 2.1001, 2.3273], device='cuda:1'), covar=tensor([0.0271, 0.0324, 0.1278, 0.0210, 0.0655, 0.0458, 0.1111, 0.0562], device='cuda:1'), in_proj_covar=tensor([0.0167, 0.0176, 0.0194, 0.0156, 0.0175, 0.0214, 0.0201, 0.0178], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-04-30 22:30:15,391 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=189408.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 22:30:33,802 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=189420.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 22:30:56,602 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=189435.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 22:30:56,809 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([1.7782, 2.4536, 2.3233, 3.2039, 2.2898, 3.5715, 1.5360, 2.6883], device='cuda:1'), covar=tensor([0.1370, 0.0775, 0.1282, 0.0172, 0.0170, 0.0401, 0.1715, 0.0828], device='cuda:1'), in_proj_covar=tensor([0.0166, 0.0172, 0.0194, 0.0184, 0.0206, 0.0215, 0.0197, 0.0191], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-04-30 22:31:21,940 INFO [train.py:904] (1/8) Epoch 19, batch 6750, loss[loss=0.2013, simple_loss=0.2981, pruned_loss=0.05221, over 16823.00 frames. ], tot_loss[loss=0.2092, simple_loss=0.2935, pruned_loss=0.06246, over 3070974.76 frames. ], batch size: 102, lr: 3.55e-03, grad_scale: 4.0 2023-04-30 22:31:42,235 INFO [optim.py:368] (1/8) 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 22:31:45,122 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.11 vs. limit=2.0 2023-04-30 22:32:10,438 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([5.0068, 5.0802, 5.4214, 5.4042, 5.4313, 5.0335, 5.0360, 4.7204], device='cuda:1'), covar=tensor([0.0359, 0.0582, 0.0351, 0.0369, 0.0584, 0.0409, 0.1016, 0.0481], device='cuda:1'), in_proj_covar=tensor([0.0394, 0.0432, 0.0421, 0.0393, 0.0467, 0.0442, 0.0533, 0.0351], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:1') 2023-04-30 22:32:38,312 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=189501.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 22:32:39,015 INFO [train.py:904] (1/8) Epoch 19, batch 6800, loss[loss=0.2362, simple_loss=0.3126, pruned_loss=0.07991, over 15429.00 frames. ], tot_loss[loss=0.2085, simple_loss=0.2928, pruned_loss=0.06214, over 3084365.79 frames. ], batch size: 190, lr: 3.55e-03, grad_scale: 8.0 2023-04-30 22:33:58,199 INFO [train.py:904] (1/8) Epoch 19, batch 6850, loss[loss=0.2063, simple_loss=0.3112, pruned_loss=0.05071, over 16767.00 frames. ], tot_loss[loss=0.2085, simple_loss=0.2934, pruned_loss=0.06181, over 3106928.95 frames. ], batch size: 83, lr: 3.55e-03, grad_scale: 8.0 2023-04-30 22:34:13,598 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=189562.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 22:34:17,023 INFO [optim.py:368] (1/8) 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 22:34:25,234 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.2670, 3.4983, 3.6166, 3.5950, 3.6052, 3.3940, 3.4463, 3.5161], device='cuda:1'), covar=tensor([0.0434, 0.0645, 0.0436, 0.0418, 0.0536, 0.0539, 0.0853, 0.0502], device='cuda:1'), in_proj_covar=tensor([0.0393, 0.0431, 0.0420, 0.0393, 0.0467, 0.0441, 0.0532, 0.0351], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:1') 2023-04-30 22:34:35,832 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.77 vs. limit=2.0 2023-04-30 22:34:47,095 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=189585.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 22:34:50,436 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=189587.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 22:34:50,757 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.84 vs. limit=2.0 2023-04-30 22:35:13,013 INFO [train.py:904] (1/8) Epoch 19, batch 6900, loss[loss=0.2332, simple_loss=0.3136, pruned_loss=0.07639, over 15350.00 frames. ], tot_loss[loss=0.2084, simple_loss=0.2951, pruned_loss=0.06083, over 3116180.73 frames. ], batch size: 190, lr: 3.55e-03, grad_scale: 8.0 2023-04-30 22:35:16,721 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=189604.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 22:35:41,292 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.7659, 3.6484, 3.8139, 3.5975, 3.8095, 4.2017, 3.8844, 3.6363], device='cuda:1'), covar=tensor([0.2066, 0.2493, 0.2808, 0.2724, 0.2647, 0.1806, 0.1725, 0.2738], device='cuda:1'), in_proj_covar=tensor([0.0397, 0.0572, 0.0632, 0.0479, 0.0636, 0.0662, 0.0494, 0.0641], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-04-30 22:36:00,958 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=189633.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 22:36:04,228 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=189635.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 22:36:12,149 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=189640.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 22:36:25,966 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.15 vs. limit=2.0 2023-04-30 22:36:29,233 INFO [train.py:904] (1/8) Epoch 19, batch 6950, loss[loss=0.1933, simple_loss=0.2782, pruned_loss=0.05418, over 16543.00 frames. ], tot_loss[loss=0.2125, simple_loss=0.2981, pruned_loss=0.0635, over 3092341.29 frames. ], batch size: 68, lr: 3.55e-03, grad_scale: 8.0 2023-04-30 22:36:30,961 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.0721, 2.0173, 2.5745, 2.9487, 2.8031, 3.4230, 2.1558, 3.3450], device='cuda:1'), covar=tensor([0.0195, 0.0493, 0.0355, 0.0316, 0.0315, 0.0145, 0.0532, 0.0155], device='cuda:1'), in_proj_covar=tensor([0.0179, 0.0188, 0.0175, 0.0178, 0.0189, 0.0148, 0.0191, 0.0142], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-30 22:36:36,160 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=2.98 vs. limit=5.0 2023-04-30 22:36:48,881 INFO [optim.py:368] (1/8) 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 22:37:21,274 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=189686.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 22:37:43,839 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=189701.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 22:37:44,554 INFO [train.py:904] (1/8) Epoch 19, batch 7000, loss[loss=0.1937, simple_loss=0.2893, pruned_loss=0.04905, over 16685.00 frames. ], tot_loss[loss=0.2107, simple_loss=0.2971, pruned_loss=0.06217, over 3089608.44 frames. ], batch size: 57, lr: 3.55e-03, grad_scale: 8.0 2023-04-30 22:37:47,740 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=189703.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 22:37:49,126 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=189704.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 22:38:04,485 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.7857, 5.2297, 5.3912, 5.1518, 5.2400, 5.8096, 5.1815, 4.9701], device='cuda:1'), covar=tensor([0.1117, 0.1784, 0.2409, 0.1932, 0.2295, 0.0902, 0.1693, 0.2479], device='cuda:1'), in_proj_covar=tensor([0.0397, 0.0571, 0.0631, 0.0477, 0.0634, 0.0659, 0.0493, 0.0641], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-04-30 22:38:13,452 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=189720.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 22:38:25,561 INFO [zipformer.py:625] (1/8) 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] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=189734.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 22:38:36,179 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=189735.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 22:39:01,644 INFO [train.py:904] (1/8) Epoch 19, batch 7050, loss[loss=0.2146, simple_loss=0.3024, pruned_loss=0.06337, over 16967.00 frames. ], tot_loss[loss=0.2124, simple_loss=0.299, pruned_loss=0.06291, over 3084677.50 frames. ], batch size: 109, lr: 3.55e-03, grad_scale: 8.0 2023-04-30 22:39:22,079 INFO [optim.py:368] (1/8) 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 22:39:22,602 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=189765.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 22:39:27,394 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=189768.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 22:39:50,319 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=189783.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 22:40:00,703 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=189789.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 22:40:18,043 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.1036, 3.3801, 3.5546, 3.5018, 3.5056, 3.3414, 3.2034, 3.4577], device='cuda:1'), covar=tensor([0.0823, 0.1786, 0.0972, 0.1202, 0.1237, 0.1712, 0.1533, 0.1216], device='cuda:1'), in_proj_covar=tensor([0.0394, 0.0433, 0.0420, 0.0395, 0.0469, 0.0443, 0.0534, 0.0353], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:1') 2023-04-30 22:40:19,979 INFO [train.py:904] (1/8) Epoch 19, batch 7100, loss[loss=0.208, simple_loss=0.3005, pruned_loss=0.05777, over 16788.00 frames. ], tot_loss[loss=0.211, simple_loss=0.297, pruned_loss=0.06249, over 3069590.21 frames. ], batch size: 83, lr: 3.54e-03, grad_scale: 8.0 2023-04-30 22:40:40,641 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([5.0188, 5.6792, 5.8410, 5.5092, 5.6542, 6.1887, 5.7012, 5.5128], device='cuda:1'), covar=tensor([0.0902, 0.1768, 0.1944, 0.1859, 0.2117, 0.0900, 0.1414, 0.2199], device='cuda:1'), in_proj_covar=tensor([0.0399, 0.0573, 0.0633, 0.0479, 0.0637, 0.0662, 0.0495, 0.0643], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-04-30 22:40:48,984 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=4.86 vs. limit=5.0 2023-04-30 22:41:38,324 INFO [train.py:904] (1/8) Epoch 19, batch 7150, loss[loss=0.2604, simple_loss=0.3129, pruned_loss=0.104, over 11614.00 frames. ], tot_loss[loss=0.2103, simple_loss=0.2951, pruned_loss=0.06279, over 3064145.22 frames. ], batch size: 246, lr: 3.54e-03, grad_scale: 4.0 2023-04-30 22:41:45,534 INFO [zipformer.py:625] (1/8) 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] (1/8) Clipping_scale=2.0, grad-norm quartiles 1.902e+02 2.755e+02 3.301e+02 3.879e+02 9.632e+02, threshold=6.601e+02, percent-clipped=3.0 2023-04-30 22:42:01,313 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([1.7630, 1.3277, 1.6850, 1.6391, 1.7366, 1.9107, 1.4940, 1.7876], device='cuda:1'), covar=tensor([0.0262, 0.0416, 0.0216, 0.0296, 0.0281, 0.0175, 0.0441, 0.0137], device='cuda:1'), in_proj_covar=tensor([0.0179, 0.0189, 0.0174, 0.0178, 0.0190, 0.0148, 0.0190, 0.0141], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-30 22:42:27,201 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.3802, 2.8321, 3.1271, 1.9614, 2.7112, 2.0794, 3.0735, 3.0830], device='cuda:1'), covar=tensor([0.0286, 0.0849, 0.0548, 0.1938, 0.0849, 0.1030, 0.0645, 0.0897], device='cuda:1'), in_proj_covar=tensor([0.0154, 0.0161, 0.0167, 0.0152, 0.0145, 0.0129, 0.0144, 0.0172], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-04-30 22:42:32,162 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.1534, 2.1478, 2.1216, 3.8693, 2.1219, 2.4788, 2.2275, 2.3281], device='cuda:1'), covar=tensor([0.1283, 0.3755, 0.3022, 0.0509, 0.4039, 0.2452, 0.3626, 0.3345], device='cuda:1'), in_proj_covar=tensor([0.0390, 0.0430, 0.0354, 0.0317, 0.0429, 0.0495, 0.0403, 0.0503], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-30 22:42:51,645 INFO [train.py:904] (1/8) Epoch 19, batch 7200, loss[loss=0.1694, simple_loss=0.2645, pruned_loss=0.03711, over 16477.00 frames. ], tot_loss[loss=0.207, simple_loss=0.2927, pruned_loss=0.06068, over 3065499.97 frames. ], batch size: 68, lr: 3.54e-03, grad_scale: 8.0 2023-04-30 22:42:55,708 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=189904.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 22:43:48,170 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.73 vs. limit=2.0 2023-04-30 22:44:12,493 INFO [train.py:904] (1/8) Epoch 19, batch 7250, loss[loss=0.1825, simple_loss=0.2699, pruned_loss=0.04749, over 16738.00 frames. ], tot_loss[loss=0.2049, simple_loss=0.2904, pruned_loss=0.05965, over 3053639.59 frames. ], batch size: 89, lr: 3.54e-03, grad_scale: 8.0 2023-04-30 22:44:12,830 INFO [zipformer.py:625] (1/8) 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] (1/8) 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 22:45:19,378 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=189996.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 22:45:32,081 INFO [train.py:904] (1/8) Epoch 19, batch 7300, loss[loss=0.2432, simple_loss=0.3065, pruned_loss=0.08993, over 10939.00 frames. ], tot_loss[loss=0.2041, simple_loss=0.2896, pruned_loss=0.05929, over 3062566.68 frames. ], batch size: 248, lr: 3.54e-03, grad_scale: 8.0 2023-04-30 22:45:33,926 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=190003.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 22:46:48,532 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=190051.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 22:46:49,311 INFO [train.py:904] (1/8) Epoch 19, batch 7350, loss[loss=0.1961, simple_loss=0.2824, pruned_loss=0.05488, over 16787.00 frames. ], tot_loss[loss=0.2059, simple_loss=0.2911, pruned_loss=0.0604, over 3053928.77 frames. ], batch size: 76, lr: 3.54e-03, grad_scale: 8.0 2023-04-30 22:47:01,679 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=190060.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 22:47:03,229 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.6889, 1.8444, 2.3409, 2.6410, 2.5517, 3.0585, 1.8442, 2.9612], device='cuda:1'), covar=tensor([0.0213, 0.0468, 0.0297, 0.0287, 0.0288, 0.0149, 0.0512, 0.0140], device='cuda:1'), in_proj_covar=tensor([0.0177, 0.0187, 0.0172, 0.0175, 0.0188, 0.0146, 0.0188, 0.0140], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-30 22:47:10,690 INFO [optim.py:368] (1/8) 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 22:47:25,689 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.5129, 3.5030, 3.4558, 2.7226, 3.3900, 2.0777, 3.1117, 2.6464], device='cuda:1'), covar=tensor([0.0154, 0.0117, 0.0182, 0.0222, 0.0104, 0.2312, 0.0133, 0.0236], device='cuda:1'), in_proj_covar=tensor([0.0157, 0.0143, 0.0189, 0.0173, 0.0164, 0.0199, 0.0178, 0.0166], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-30 22:47:40,424 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=190084.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 22:48:08,858 INFO [train.py:904] (1/8) Epoch 19, batch 7400, loss[loss=0.213, simple_loss=0.3075, pruned_loss=0.05929, over 16942.00 frames. ], tot_loss[loss=0.2071, simple_loss=0.2917, pruned_loss=0.06124, over 3039899.84 frames. ], batch size: 41, lr: 3.54e-03, grad_scale: 8.0 2023-04-30 22:48:45,336 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.2983, 2.3444, 2.9880, 3.2326, 3.0427, 3.8385, 2.4008, 3.7489], device='cuda:1'), covar=tensor([0.0174, 0.0395, 0.0238, 0.0225, 0.0245, 0.0110, 0.0439, 0.0097], device='cuda:1'), in_proj_covar=tensor([0.0177, 0.0188, 0.0173, 0.0176, 0.0188, 0.0146, 0.0189, 0.0140], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-30 22:49:09,870 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=190139.0, num_to_drop=1, layers_to_drop={1} 2023-04-30 22:49:30,218 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 19, batch 7450, loss[loss=0.2126, simple_loss=0.3056, pruned_loss=0.05975, over 15423.00 frames. ], tot_loss[loss=0.2085, simple_loss=0.2927, pruned_loss=0.06215, over 3037257.86 frames. ], batch size: 191, lr: 3.54e-03, grad_scale: 8.0 2023-04-30 22:49:31,648 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.6057, 3.6603, 2.3142, 4.0590, 2.7074, 4.0535, 2.3040, 2.8534], device='cuda:1'), covar=tensor([0.0255, 0.0393, 0.1582, 0.0229, 0.0814, 0.0596, 0.1617, 0.0772], device='cuda:1'), in_proj_covar=tensor([0.0168, 0.0176, 0.0194, 0.0156, 0.0175, 0.0214, 0.0201, 0.0179], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-04-30 22:49:41,033 INFO [zipformer.py:625] (1/8) 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] (1/8) 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 22:50:30,672 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.7185, 2.6938, 2.0914, 2.6404, 3.1887, 2.8872, 3.3768, 3.4253], device='cuda:1'), covar=tensor([0.0093, 0.0401, 0.0553, 0.0378, 0.0236, 0.0331, 0.0206, 0.0197], device='cuda:1'), in_proj_covar=tensor([0.0193, 0.0226, 0.0218, 0.0218, 0.0226, 0.0225, 0.0227, 0.0220], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-30 22:50:51,782 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=190200.0, num_to_drop=1, layers_to_drop={1} 2023-04-30 22:50:54,227 INFO [train.py:904] (1/8) Epoch 19, batch 7500, loss[loss=0.2062, simple_loss=0.2873, pruned_loss=0.06259, over 17032.00 frames. ], tot_loss[loss=0.2081, simple_loss=0.2929, pruned_loss=0.06158, over 3046404.66 frames. ], batch size: 53, lr: 3.54e-03, grad_scale: 8.0 2023-04-30 22:50:59,305 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=190205.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 22:51:10,263 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=190212.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 22:52:06,514 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.17 vs. limit=2.0 2023-04-30 22:52:13,611 INFO [train.py:904] (1/8) Epoch 19, batch 7550, loss[loss=0.1832, simple_loss=0.2701, pruned_loss=0.04821, over 16820.00 frames. ], tot_loss[loss=0.2074, simple_loss=0.2918, pruned_loss=0.06149, over 3054942.18 frames. ], batch size: 83, lr: 3.54e-03, grad_scale: 8.0 2023-04-30 22:52:34,940 INFO [optim.py:368] (1/8) 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 22:53:22,543 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=190296.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 22:53:31,291 INFO [train.py:904] (1/8) Epoch 19, batch 7600, loss[loss=0.2252, simple_loss=0.3081, pruned_loss=0.07115, over 16940.00 frames. ], tot_loss[loss=0.2071, simple_loss=0.2912, pruned_loss=0.06146, over 3063893.15 frames. ], batch size: 109, lr: 3.54e-03, grad_scale: 8.0 2023-04-30 22:53:48,416 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.24 vs. limit=2.0 2023-04-30 22:53:52,937 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=190316.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 22:54:36,981 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 19, batch 7650, loss[loss=0.2236, simple_loss=0.3058, pruned_loss=0.07066, over 16746.00 frames. ], tot_loss[loss=0.2078, simple_loss=0.2916, pruned_loss=0.06199, over 3053316.46 frames. ], batch size: 124, lr: 3.54e-03, grad_scale: 8.0 2023-04-30 22:55:02,235 INFO [zipformer.py:625] (1/8) 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] (1/8) 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 22:55:27,522 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=190377.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 22:55:38,662 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=190384.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 22:56:02,429 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=190399.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 22:56:06,244 INFO [train.py:904] (1/8) Epoch 19, batch 7700, loss[loss=0.2098, simple_loss=0.293, pruned_loss=0.0633, over 16624.00 frames. ], tot_loss[loss=0.2078, simple_loss=0.2912, pruned_loss=0.06218, over 3052547.01 frames. ], batch size: 57, lr: 3.54e-03, grad_scale: 8.0 2023-04-30 22:56:15,128 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=190408.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 22:56:26,092 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.25 vs. limit=2.0 2023-04-30 22:56:50,414 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=190432.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 22:57:01,413 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([5.1078, 5.4442, 5.2447, 5.2208, 4.9589, 4.9254, 4.8685, 5.5839], device='cuda:1'), covar=tensor([0.1320, 0.0863, 0.1007, 0.0881, 0.0828, 0.0798, 0.1221, 0.0913], device='cuda:1'), in_proj_covar=tensor([0.0637, 0.0772, 0.0642, 0.0579, 0.0485, 0.0498, 0.0647, 0.0595], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:1') 2023-04-30 22:57:18,830 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.4881, 4.5053, 4.3426, 3.5562, 4.4199, 1.7111, 4.1705, 3.9779], device='cuda:1'), covar=tensor([0.0116, 0.0093, 0.0194, 0.0346, 0.0098, 0.2698, 0.0148, 0.0242], device='cuda:1'), in_proj_covar=tensor([0.0157, 0.0143, 0.0189, 0.0172, 0.0164, 0.0198, 0.0178, 0.0165], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-30 22:57:21,409 INFO [train.py:904] (1/8) Epoch 19, batch 7750, loss[loss=0.1923, simple_loss=0.2846, pruned_loss=0.05004, over 17007.00 frames. ], tot_loss[loss=0.2079, simple_loss=0.2914, pruned_loss=0.06223, over 3066216.40 frames. ], batch size: 53, lr: 3.54e-03, grad_scale: 8.0 2023-04-30 22:57:35,054 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=190460.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 22:57:43,807 INFO [optim.py:368] (1/8) 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 22:58:30,600 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=190495.0, num_to_drop=1, layers_to_drop={1} 2023-04-30 22:58:39,267 INFO [train.py:904] (1/8) Epoch 19, batch 7800, loss[loss=0.1814, simple_loss=0.283, pruned_loss=0.03989, over 16842.00 frames. ], tot_loss[loss=0.2083, simple_loss=0.292, pruned_loss=0.06232, over 3066369.02 frames. ], batch size: 96, lr: 3.54e-03, grad_scale: 8.0 2023-04-30 22:58:47,776 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 19, batch 7850, loss[loss=0.2026, simple_loss=0.2943, pruned_loss=0.05547, over 16635.00 frames. ], tot_loss[loss=0.2073, simple_loss=0.2925, pruned_loss=0.06112, over 3092226.93 frames. ], batch size: 62, lr: 3.54e-03, grad_scale: 8.0 2023-04-30 23:00:17,836 INFO [optim.py:368] (1/8) 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 23:00:56,973 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=190591.0, num_to_drop=1, layers_to_drop={0} 2023-04-30 23:01:12,372 INFO [train.py:904] (1/8) Epoch 19, batch 7900, loss[loss=0.221, simple_loss=0.325, pruned_loss=0.05852, over 16872.00 frames. ], tot_loss[loss=0.2066, simple_loss=0.2918, pruned_loss=0.06069, over 3087396.91 frames. ], batch size: 116, lr: 3.54e-03, grad_scale: 8.0 2023-04-30 23:01:23,311 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=190609.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 23:02:32,481 INFO [train.py:904] (1/8) Epoch 19, batch 7950, loss[loss=0.2672, simple_loss=0.3315, pruned_loss=0.1015, over 11448.00 frames. ], tot_loss[loss=0.2074, simple_loss=0.2923, pruned_loss=0.0613, over 3077037.87 frames. ], batch size: 248, lr: 3.54e-03, grad_scale: 8.0 2023-04-30 23:02:32,967 INFO [zipformer.py:625] (1/8) 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] (1/8) 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 23:03:00,801 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=190670.0, num_to_drop=1, layers_to_drop={3} 2023-04-30 23:03:04,351 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=190672.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 23:03:08,597 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 19, batch 8000, loss[loss=0.2882, simple_loss=0.3396, pruned_loss=0.1184, over 11379.00 frames. ], tot_loss[loss=0.2086, simple_loss=0.2929, pruned_loss=0.06217, over 3057472.35 frames. ], batch size: 246, lr: 3.54e-03, grad_scale: 8.0 2023-04-30 23:04:27,776 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.82 vs. limit=2.0 2023-04-30 23:04:44,514 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=190736.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 23:05:07,643 INFO [train.py:904] (1/8) Epoch 19, batch 8050, loss[loss=0.235, simple_loss=0.2998, pruned_loss=0.08505, over 11721.00 frames. ], tot_loss[loss=0.208, simple_loss=0.2926, pruned_loss=0.06166, over 3061465.98 frames. ], batch size: 247, lr: 3.54e-03, grad_scale: 8.0 2023-04-30 23:05:12,147 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=190755.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 23:05:28,124 INFO [optim.py:368] (1/8) 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 23:06:12,388 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=190795.0, num_to_drop=1, layers_to_drop={0} 2023-04-30 23:06:22,312 INFO [train.py:904] (1/8) Epoch 19, batch 8100, loss[loss=0.2296, simple_loss=0.3025, pruned_loss=0.07835, over 11869.00 frames. ], tot_loss[loss=0.2074, simple_loss=0.2922, pruned_loss=0.0613, over 3041710.16 frames. ], batch size: 248, lr: 3.54e-03, grad_scale: 8.0 2023-04-30 23:06:23,055 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.67 vs. limit=2.0 2023-04-30 23:06:30,742 INFO [zipformer.py:625] (1/8) 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] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=190843.0, num_to_drop=1, layers_to_drop={0} 2023-04-30 23:07:39,351 INFO [train.py:904] (1/8) Epoch 19, batch 8150, loss[loss=0.1738, simple_loss=0.2596, pruned_loss=0.04403, over 16855.00 frames. ], tot_loss[loss=0.2047, simple_loss=0.2895, pruned_loss=0.05989, over 3049227.70 frames. ], batch size: 42, lr: 3.54e-03, grad_scale: 8.0 2023-04-30 23:07:44,593 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=190855.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 23:08:01,317 INFO [optim.py:368] (1/8) 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 23:08:39,905 INFO [scaling.py:679] (1/8) 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] (1/8) Epoch 19, batch 8200, loss[loss=0.1958, simple_loss=0.2869, pruned_loss=0.05231, over 16239.00 frames. ], tot_loss[loss=0.2024, simple_loss=0.2867, pruned_loss=0.05902, over 3059534.50 frames. ], batch size: 165, lr: 3.53e-03, grad_scale: 8.0 2023-04-30 23:10:11,507 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=190947.0, num_to_drop=1, layers_to_drop={1} 2023-04-30 23:10:19,455 INFO [train.py:904] (1/8) Epoch 19, batch 8250, loss[loss=0.2091, simple_loss=0.2992, pruned_loss=0.05952, over 16806.00 frames. ], tot_loss[loss=0.1996, simple_loss=0.2856, pruned_loss=0.05681, over 3040481.35 frames. ], batch size: 124, lr: 3.53e-03, grad_scale: 8.0 2023-04-30 23:10:30,776 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.9265, 3.7879, 4.0342, 4.1346, 4.2474, 3.7967, 4.1975, 4.2870], device='cuda:1'), covar=tensor([0.1861, 0.1319, 0.1450, 0.0730, 0.0581, 0.1650, 0.0702, 0.0684], device='cuda:1'), in_proj_covar=tensor([0.0599, 0.0744, 0.0869, 0.0760, 0.0575, 0.0597, 0.0619, 0.0715], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-30 23:10:39,905 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=190965.0, num_to_drop=1, layers_to_drop={2} 2023-04-30 23:10:40,147 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.0399, 2.0743, 2.1543, 3.5593, 2.0689, 2.3586, 2.1998, 2.2057], device='cuda:1'), covar=tensor([0.1243, 0.3862, 0.3049, 0.0580, 0.4463, 0.2595, 0.3603, 0.3462], device='cuda:1'), in_proj_covar=tensor([0.0385, 0.0429, 0.0352, 0.0316, 0.0427, 0.0492, 0.0400, 0.0500], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-30 23:10:41,154 INFO [optim.py:368] (1/8) 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 23:10:48,366 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.26 vs. limit=2.0 2023-04-30 23:10:51,837 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=190972.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 23:11:29,753 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.3108, 3.3276, 3.4558, 2.4198, 3.1920, 3.3941, 3.2888, 2.0758], device='cuda:1'), covar=tensor([0.0483, 0.0073, 0.0055, 0.0357, 0.0102, 0.0110, 0.0091, 0.0460], device='cuda:1'), in_proj_covar=tensor([0.0134, 0.0079, 0.0080, 0.0133, 0.0093, 0.0107, 0.0091, 0.0126], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-04-30 23:11:39,003 INFO [train.py:904] (1/8) Epoch 19, batch 8300, loss[loss=0.1812, simple_loss=0.2779, pruned_loss=0.04223, over 16885.00 frames. ], tot_loss[loss=0.1955, simple_loss=0.2832, pruned_loss=0.05388, over 3043132.78 frames. ], batch size: 116, lr: 3.53e-03, grad_scale: 8.0 2023-04-30 23:12:10,266 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=191020.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 23:12:27,702 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=191031.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 23:12:40,244 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=3.26 vs. limit=5.0 2023-04-30 23:13:00,823 INFO [train.py:904] (1/8) Epoch 19, batch 8350, loss[loss=0.1807, simple_loss=0.2773, pruned_loss=0.04207, over 15305.00 frames. ], tot_loss[loss=0.1935, simple_loss=0.2828, pruned_loss=0.05207, over 3041505.96 frames. ], batch size: 191, lr: 3.53e-03, grad_scale: 8.0 2023-04-30 23:13:07,508 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=191055.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 23:13:24,343 INFO [optim.py:368] (1/8) 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 23:13:28,640 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=191068.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 23:13:51,940 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.3708, 4.3306, 4.7367, 4.7162, 4.7010, 4.4586, 4.3924, 4.4136], device='cuda:1'), covar=tensor([0.0375, 0.0809, 0.0449, 0.0463, 0.0520, 0.0439, 0.1173, 0.0551], device='cuda:1'), in_proj_covar=tensor([0.0393, 0.0431, 0.0418, 0.0394, 0.0465, 0.0439, 0.0535, 0.0355], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:1') 2023-04-30 23:14:22,694 INFO [train.py:904] (1/8) Epoch 19, batch 8400, loss[loss=0.1772, simple_loss=0.2715, pruned_loss=0.04141, over 15350.00 frames. ], tot_loss[loss=0.1905, simple_loss=0.2808, pruned_loss=0.05013, over 3043763.18 frames. ], batch size: 191, lr: 3.53e-03, grad_scale: 8.0 2023-04-30 23:14:24,464 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=191103.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 23:15:06,907 INFO [zipformer.py:625] (1/8) 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] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=191132.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 23:15:42,687 INFO [train.py:904] (1/8) Epoch 19, batch 8450, loss[loss=0.1844, simple_loss=0.2637, pruned_loss=0.05258, over 12130.00 frames. ], tot_loss[loss=0.1877, simple_loss=0.2787, pruned_loss=0.04832, over 3043769.66 frames. ], batch size: 246, lr: 3.53e-03, grad_scale: 8.0 2023-04-30 23:16:06,334 INFO [optim.py:368] (1/8) 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 23:16:37,007 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=191185.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 23:16:50,101 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=191193.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 23:17:04,705 INFO [train.py:904] (1/8) Epoch 19, batch 8500, loss[loss=0.1499, simple_loss=0.2479, pruned_loss=0.02589, over 16818.00 frames. ], tot_loss[loss=0.1838, simple_loss=0.2754, pruned_loss=0.04603, over 3063999.81 frames. ], batch size: 102, lr: 3.53e-03, grad_scale: 8.0 2023-04-30 23:18:22,325 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=191246.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 23:18:24,091 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 19, batch 8550, loss[loss=0.1654, simple_loss=0.2529, pruned_loss=0.03892, over 12011.00 frames. ], tot_loss[loss=0.1818, simple_loss=0.273, pruned_loss=0.04527, over 3032691.18 frames. ], batch size: 246, lr: 3.53e-03, grad_scale: 8.0 2023-04-30 23:18:57,145 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=191265.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 23:19:00,614 INFO [optim.py:368] (1/8) 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 23:20:00,163 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=191295.0, num_to_drop=1, layers_to_drop={0} 2023-04-30 23:20:05,704 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([1.5969, 2.3728, 2.4252, 3.9912, 2.3426, 3.9297, 1.3867, 2.9449], device='cuda:1'), covar=tensor([0.1315, 0.0850, 0.1080, 0.0167, 0.0101, 0.0329, 0.1622, 0.0660], device='cuda:1'), in_proj_covar=tensor([0.0163, 0.0169, 0.0189, 0.0179, 0.0201, 0.0209, 0.0193, 0.0186], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-04-30 23:20:13,448 INFO [train.py:904] (1/8) Epoch 19, batch 8600, loss[loss=0.1711, simple_loss=0.2713, pruned_loss=0.03547, over 15302.00 frames. ], tot_loss[loss=0.1806, simple_loss=0.2729, pruned_loss=0.04416, over 3027732.70 frames. ], batch size: 190, lr: 3.53e-03, grad_scale: 8.0 2023-04-30 23:20:37,121 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=191313.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 23:21:03,012 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.4121, 3.0422, 2.6995, 2.1917, 2.1908, 2.2036, 3.0463, 2.8837], device='cuda:1'), covar=tensor([0.2417, 0.0739, 0.1597, 0.2946, 0.2630, 0.2248, 0.0434, 0.1451], device='cuda:1'), in_proj_covar=tensor([0.0318, 0.0259, 0.0294, 0.0299, 0.0286, 0.0246, 0.0282, 0.0321], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:1') 2023-04-30 23:21:11,770 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=191331.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 23:21:52,927 INFO [train.py:904] (1/8) Epoch 19, batch 8650, loss[loss=0.1624, simple_loss=0.2748, pruned_loss=0.025, over 16877.00 frames. ], tot_loss[loss=0.1777, simple_loss=0.2704, pruned_loss=0.04244, over 3003347.50 frames. ], batch size: 102, lr: 3.53e-03, grad_scale: 8.0 2023-04-30 23:22:25,673 INFO [zipformer.py:625] (1/8) 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] (1/8) 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 23:22:55,703 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=191379.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 23:23:39,020 INFO [train.py:904] (1/8) Epoch 19, batch 8700, loss[loss=0.1673, simple_loss=0.2639, pruned_loss=0.03533, over 16787.00 frames. ], tot_loss[loss=0.1749, simple_loss=0.2677, pruned_loss=0.04108, over 3005278.40 frames. ], batch size: 83, lr: 3.53e-03, grad_scale: 2.0 2023-04-30 23:24:20,674 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=191424.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 23:24:23,639 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=191426.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 23:25:14,445 INFO [train.py:904] (1/8) Epoch 19, batch 8750, loss[loss=0.1708, simple_loss=0.2782, pruned_loss=0.03168, over 16893.00 frames. ], tot_loss[loss=0.1739, simple_loss=0.2669, pruned_loss=0.04042, over 3011696.38 frames. ], batch size: 102, lr: 3.53e-03, grad_scale: 2.0 2023-04-30 23:25:37,807 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([1.7033, 3.0926, 2.7666, 4.9986, 3.5204, 4.4151, 1.7568, 3.2050], device='cuda:1'), covar=tensor([0.1423, 0.0755, 0.1159, 0.0164, 0.0250, 0.0377, 0.1613, 0.0731], device='cuda:1'), in_proj_covar=tensor([0.0163, 0.0168, 0.0189, 0.0177, 0.0200, 0.0207, 0.0193, 0.0186], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-04-30 23:25:58,340 INFO [optim.py:368] (1/8) 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] (1/8) 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] (1/8) Epoch 19, batch 8800, loss[loss=0.1852, simple_loss=0.2732, pruned_loss=0.04857, over 12598.00 frames. ], tot_loss[loss=0.1726, simple_loss=0.2658, pruned_loss=0.03976, over 3016344.29 frames. ], batch size: 247, lr: 3.53e-03, grad_scale: 4.0 2023-04-30 23:27:34,733 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.2245, 3.3421, 3.6032, 1.7036, 3.7388, 3.8423, 2.8867, 2.9292], device='cuda:1'), covar=tensor([0.0863, 0.0255, 0.0199, 0.1229, 0.0086, 0.0140, 0.0436, 0.0420], device='cuda:1'), in_proj_covar=tensor([0.0141, 0.0102, 0.0092, 0.0134, 0.0075, 0.0116, 0.0121, 0.0125], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:1') 2023-04-30 23:27:52,115 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.23 vs. limit=2.0 2023-04-30 23:28:04,467 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.40 vs. limit=2.0 2023-04-30 23:28:29,012 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 19, batch 8850, loss[loss=0.171, simple_loss=0.2721, pruned_loss=0.03497, over 16653.00 frames. ], tot_loss[loss=0.1735, simple_loss=0.2686, pruned_loss=0.03922, over 3024211.85 frames. ], batch size: 134, lr: 3.53e-03, grad_scale: 4.0 2023-04-30 23:29:00,223 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.1853, 3.4800, 3.4931, 2.3774, 3.1527, 3.5374, 3.3666, 2.0673], device='cuda:1'), covar=tensor([0.0514, 0.0046, 0.0049, 0.0387, 0.0109, 0.0082, 0.0087, 0.0453], device='cuda:1'), in_proj_covar=tensor([0.0134, 0.0078, 0.0079, 0.0133, 0.0094, 0.0105, 0.0091, 0.0126], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-04-30 23:29:28,627 INFO [optim.py:368] (1/8) 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 23:29:49,282 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.7342, 4.7712, 4.5735, 4.1767, 4.2236, 4.6466, 4.5341, 4.3240], device='cuda:1'), covar=tensor([0.0634, 0.0703, 0.0326, 0.0350, 0.0973, 0.0613, 0.0406, 0.0787], device='cuda:1'), in_proj_covar=tensor([0.0269, 0.0389, 0.0316, 0.0308, 0.0324, 0.0358, 0.0219, 0.0376], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-30 23:30:28,181 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.5405, 3.6599, 2.7529, 2.1022, 2.3252, 2.3579, 3.8542, 3.3543], device='cuda:1'), covar=tensor([0.2957, 0.0639, 0.1764, 0.2803, 0.2661, 0.2034, 0.0456, 0.1154], device='cuda:1'), in_proj_covar=tensor([0.0316, 0.0256, 0.0291, 0.0295, 0.0282, 0.0244, 0.0279, 0.0317], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:1') 2023-04-30 23:30:39,776 INFO [train.py:904] (1/8) Epoch 19, batch 8900, loss[loss=0.175, simple_loss=0.2634, pruned_loss=0.04324, over 12565.00 frames. ], tot_loss[loss=0.1733, simple_loss=0.2691, pruned_loss=0.03878, over 3027389.80 frames. ], batch size: 248, lr: 3.53e-03, grad_scale: 4.0 2023-04-30 23:32:45,267 INFO [train.py:904] (1/8) Epoch 19, batch 8950, loss[loss=0.1618, simple_loss=0.2654, pruned_loss=0.0291, over 16578.00 frames. ], tot_loss[loss=0.1735, simple_loss=0.269, pruned_loss=0.03897, over 3050064.11 frames. ], batch size: 75, lr: 3.53e-03, grad_scale: 4.0 2023-04-30 23:33:16,002 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.7984, 3.8564, 4.1366, 4.1062, 4.1101, 3.9098, 3.9088, 3.9395], device='cuda:1'), covar=tensor([0.0320, 0.0605, 0.0363, 0.0409, 0.0413, 0.0431, 0.0779, 0.0438], device='cuda:1'), in_proj_covar=tensor([0.0380, 0.0414, 0.0405, 0.0381, 0.0447, 0.0425, 0.0514, 0.0343], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:1') 2023-04-30 23:33:21,157 INFO [optim.py:368] (1/8) 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 23:34:34,395 INFO [train.py:904] (1/8) Epoch 19, batch 9000, loss[loss=0.1647, simple_loss=0.2567, pruned_loss=0.03636, over 16283.00 frames. ], tot_loss[loss=0.1705, simple_loss=0.2656, pruned_loss=0.03768, over 3058492.93 frames. ], batch size: 166, lr: 3.53e-03, grad_scale: 4.0 2023-04-30 23:34:34,395 INFO [train.py:929] (1/8) Computing validation loss 2023-04-30 23:34:44,207 INFO [train.py:938] (1/8) Epoch 19, validation: loss=0.1462, simple_loss=0.2506, pruned_loss=0.02087, over 944034.00 frames. 2023-04-30 23:34:44,207 INFO [train.py:939] (1/8) Maximum memory allocated so far is 18145MB 2023-04-30 23:35:25,455 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=191721.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 23:35:31,098 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=191724.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 23:36:27,797 INFO [train.py:904] (1/8) Epoch 19, batch 9050, loss[loss=0.1895, simple_loss=0.2863, pruned_loss=0.0464, over 16996.00 frames. ], tot_loss[loss=0.1711, simple_loss=0.2664, pruned_loss=0.03795, over 3077341.37 frames. ], batch size: 55, lr: 3.53e-03, grad_scale: 4.0 2023-04-30 23:37:04,255 INFO [optim.py:368] (1/8) 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 23:37:09,315 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=191772.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 23:37:37,778 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.2332, 4.2730, 4.0931, 3.7848, 3.8250, 4.1954, 3.9142, 3.9242], device='cuda:1'), covar=tensor([0.0522, 0.0515, 0.0326, 0.0311, 0.0741, 0.0500, 0.0742, 0.0676], device='cuda:1'), in_proj_covar=tensor([0.0270, 0.0390, 0.0316, 0.0308, 0.0323, 0.0359, 0.0219, 0.0377], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-30 23:37:39,985 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=191788.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 23:37:40,029 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.7307, 4.1513, 4.1383, 3.1331, 3.5876, 4.1370, 3.7580, 2.4364], device='cuda:1'), covar=tensor([0.0444, 0.0036, 0.0040, 0.0287, 0.0100, 0.0077, 0.0072, 0.0460], device='cuda:1'), in_proj_covar=tensor([0.0132, 0.0077, 0.0078, 0.0131, 0.0093, 0.0104, 0.0089, 0.0125], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-04-30 23:38:10,475 INFO [train.py:904] (1/8) Epoch 19, batch 9100, loss[loss=0.1614, simple_loss=0.2523, pruned_loss=0.03529, over 12381.00 frames. ], tot_loss[loss=0.1713, simple_loss=0.2659, pruned_loss=0.03839, over 3094583.17 frames. ], batch size: 246, lr: 3.53e-03, grad_scale: 4.0 2023-04-30 23:38:23,501 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.1993, 2.1373, 2.0576, 3.8801, 2.0198, 2.4484, 2.2336, 2.2774], device='cuda:1'), covar=tensor([0.1174, 0.3596, 0.3155, 0.0501, 0.4458, 0.2686, 0.3500, 0.3619], device='cuda:1'), in_proj_covar=tensor([0.0380, 0.0422, 0.0349, 0.0310, 0.0421, 0.0484, 0.0395, 0.0492], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-30 23:38:40,782 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=4.27 vs. limit=5.0 2023-04-30 23:39:32,422 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=191836.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 23:39:43,533 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=191841.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 23:39:43,655 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.4192, 1.7296, 2.0846, 2.3731, 2.3883, 2.6856, 1.8632, 2.6057], device='cuda:1'), covar=tensor([0.0236, 0.0493, 0.0301, 0.0309, 0.0339, 0.0175, 0.0484, 0.0149], device='cuda:1'), in_proj_covar=tensor([0.0173, 0.0184, 0.0169, 0.0172, 0.0184, 0.0142, 0.0187, 0.0137], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-30 23:40:08,177 INFO [train.py:904] (1/8) Epoch 19, batch 9150, loss[loss=0.1665, simple_loss=0.2562, pruned_loss=0.03841, over 11983.00 frames. ], tot_loss[loss=0.1719, simple_loss=0.2669, pruned_loss=0.03844, over 3091116.78 frames. ], batch size: 246, lr: 3.53e-03, grad_scale: 4.0 2023-04-30 23:40:46,340 INFO [optim.py:368] (1/8) 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 23:41:28,239 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.9333, 2.0986, 2.4012, 3.2003, 2.1621, 2.2796, 2.2969, 2.2066], device='cuda:1'), covar=tensor([0.1271, 0.3482, 0.2642, 0.0692, 0.4461, 0.2723, 0.3263, 0.3724], device='cuda:1'), in_proj_covar=tensor([0.0379, 0.0421, 0.0348, 0.0310, 0.0420, 0.0483, 0.0393, 0.0491], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-30 23:41:30,261 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=191889.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 23:41:39,648 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.9871, 4.2816, 4.0889, 4.1388, 3.8470, 3.8499, 3.8952, 4.2708], device='cuda:1'), covar=tensor([0.1179, 0.0990, 0.1013, 0.0762, 0.0818, 0.1903, 0.1007, 0.0977], device='cuda:1'), in_proj_covar=tensor([0.0625, 0.0760, 0.0624, 0.0565, 0.0479, 0.0488, 0.0635, 0.0585], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:1') 2023-04-30 23:41:52,732 INFO [train.py:904] (1/8) Epoch 19, batch 9200, loss[loss=0.1589, simple_loss=0.247, pruned_loss=0.03542, over 16691.00 frames. ], tot_loss[loss=0.1687, simple_loss=0.2626, pruned_loss=0.03737, over 3089618.47 frames. ], batch size: 57, lr: 3.53e-03, grad_scale: 8.0 2023-04-30 23:42:11,329 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.55 vs. limit=2.0 2023-04-30 23:43:29,321 INFO [train.py:904] (1/8) Epoch 19, batch 9250, loss[loss=0.1577, simple_loss=0.2502, pruned_loss=0.03258, over 15317.00 frames. ], tot_loss[loss=0.1684, simple_loss=0.2622, pruned_loss=0.03728, over 3076406.77 frames. ], batch size: 191, lr: 3.52e-03, grad_scale: 8.0 2023-04-30 23:43:36,842 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.5830, 4.8684, 4.6689, 4.6911, 4.3806, 4.4042, 4.3222, 4.9222], device='cuda:1'), covar=tensor([0.1122, 0.0963, 0.0949, 0.0829, 0.0851, 0.1157, 0.1115, 0.0987], device='cuda:1'), in_proj_covar=tensor([0.0625, 0.0760, 0.0624, 0.0566, 0.0479, 0.0489, 0.0638, 0.0585], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:1') 2023-04-30 23:44:05,909 INFO [optim.py:368] (1/8) 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 23:45:23,367 INFO [train.py:904] (1/8) Epoch 19, batch 9300, loss[loss=0.1489, simple_loss=0.2432, pruned_loss=0.02732, over 16404.00 frames. ], tot_loss[loss=0.1667, simple_loss=0.2605, pruned_loss=0.03648, over 3076533.20 frames. ], batch size: 146, lr: 3.52e-03, grad_scale: 4.0 2023-04-30 23:46:09,690 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=192021.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 23:47:09,782 INFO [train.py:904] (1/8) Epoch 19, batch 9350, loss[loss=0.1572, simple_loss=0.2596, pruned_loss=0.02742, over 16844.00 frames. ], tot_loss[loss=0.1663, simple_loss=0.2601, pruned_loss=0.0363, over 3075917.72 frames. ], batch size: 102, lr: 3.52e-03, grad_scale: 4.0 2023-04-30 23:47:46,318 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=192069.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 23:47:47,129 INFO [optim.py:368] (1/8) 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 23:48:25,118 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.5196, 3.4832, 3.5059, 2.8644, 3.4118, 1.9418, 3.0837, 2.8300], device='cuda:1'), covar=tensor([0.0130, 0.0116, 0.0161, 0.0202, 0.0102, 0.2393, 0.0129, 0.0220], device='cuda:1'), in_proj_covar=tensor([0.0151, 0.0139, 0.0181, 0.0163, 0.0159, 0.0194, 0.0171, 0.0159], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-30 23:48:48,601 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.91 vs. limit=2.0 2023-04-30 23:48:49,308 INFO [train.py:904] (1/8) Epoch 19, batch 9400, loss[loss=0.147, simple_loss=0.237, pruned_loss=0.02848, over 12238.00 frames. ], tot_loss[loss=0.166, simple_loss=0.2599, pruned_loss=0.03606, over 3076310.09 frames. ], batch size: 246, lr: 3.52e-03, grad_scale: 4.0 2023-04-30 23:48:50,356 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=192102.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 23:49:10,047 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.3076, 2.4464, 2.0087, 2.0691, 2.7792, 2.4637, 2.7584, 2.9955], device='cuda:1'), covar=tensor([0.0123, 0.0428, 0.0555, 0.0553, 0.0283, 0.0405, 0.0234, 0.0241], device='cuda:1'), in_proj_covar=tensor([0.0188, 0.0223, 0.0215, 0.0216, 0.0225, 0.0222, 0.0221, 0.0214], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-30 23:50:12,465 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.87 vs. limit=2.0 2023-04-30 23:50:18,051 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([1.6494, 2.7298, 2.6627, 4.3879, 2.8335, 4.1522, 1.6668, 2.9636], device='cuda:1'), covar=tensor([0.1554, 0.0788, 0.1180, 0.0192, 0.0145, 0.0344, 0.1746, 0.0774], device='cuda:1'), in_proj_covar=tensor([0.0163, 0.0167, 0.0188, 0.0175, 0.0195, 0.0206, 0.0193, 0.0185], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0003], device='cuda:1') 2023-04-30 23:50:29,958 INFO [train.py:904] (1/8) Epoch 19, batch 9450, loss[loss=0.1607, simple_loss=0.2559, pruned_loss=0.03276, over 12403.00 frames. ], tot_loss[loss=0.1677, simple_loss=0.2621, pruned_loss=0.03659, over 3070363.54 frames. ], batch size: 247, lr: 3.52e-03, grad_scale: 4.0 2023-04-30 23:50:35,970 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=192155.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 23:50:51,020 INFO [zipformer.py:625] (1/8) 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] (1/8) 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 23:51:42,730 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.33 vs. limit=2.0 2023-04-30 23:52:10,463 INFO [train.py:904] (1/8) Epoch 19, batch 9500, loss[loss=0.1623, simple_loss=0.2588, pruned_loss=0.03287, over 16762.00 frames. ], tot_loss[loss=0.1676, simple_loss=0.2617, pruned_loss=0.0367, over 3064282.01 frames. ], batch size: 124, lr: 3.52e-03, grad_scale: 4.0 2023-04-30 23:52:39,636 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=192216.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 23:52:45,115 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=192219.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 23:52:49,390 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=192221.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 23:53:55,147 INFO [train.py:904] (1/8) Epoch 19, batch 9550, loss[loss=0.189, simple_loss=0.279, pruned_loss=0.04945, over 16870.00 frames. ], tot_loss[loss=0.1671, simple_loss=0.2608, pruned_loss=0.03672, over 3068982.24 frames. ], batch size: 116, lr: 3.52e-03, grad_scale: 4.0 2023-04-30 23:54:34,513 INFO [optim.py:368] (1/8) Clipping_scale=2.0, grad-norm quartiles 1.443e+02 2.006e+02 2.351e+02 2.778e+02 5.702e+02, threshold=4.701e+02, percent-clipped=1.0 2023-04-30 23:54:55,502 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=192280.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 23:55:00,682 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=192282.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 23:55:38,438 INFO [train.py:904] (1/8) Epoch 19, batch 9600, loss[loss=0.2001, simple_loss=0.2977, pruned_loss=0.05121, over 15428.00 frames. ], tot_loss[loss=0.169, simple_loss=0.2626, pruned_loss=0.03766, over 3065916.77 frames. ], batch size: 191, lr: 3.52e-03, grad_scale: 8.0 2023-04-30 23:57:21,284 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.2741, 2.1546, 2.0965, 3.9758, 2.0896, 2.5771, 2.2847, 2.3376], device='cuda:1'), covar=tensor([0.1176, 0.3817, 0.3169, 0.0482, 0.4501, 0.2535, 0.3611, 0.3739], device='cuda:1'), in_proj_covar=tensor([0.0381, 0.0423, 0.0351, 0.0310, 0.0423, 0.0485, 0.0396, 0.0493], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-30 23:57:27,640 INFO [train.py:904] (1/8) Epoch 19, batch 9650, loss[loss=0.1736, simple_loss=0.2724, pruned_loss=0.03741, over 15396.00 frames. ], tot_loss[loss=0.1703, simple_loss=0.2646, pruned_loss=0.03805, over 3070709.79 frames. ], batch size: 190, lr: 3.52e-03, grad_scale: 8.0 2023-04-30 23:58:09,764 INFO [optim.py:368] (1/8) 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 23:58:53,969 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=192392.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 23:59:09,436 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=192399.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 23:59:15,172 INFO [train.py:904] (1/8) Epoch 19, batch 9700, loss[loss=0.161, simple_loss=0.2484, pruned_loss=0.03686, over 12314.00 frames. ], tot_loss[loss=0.1693, simple_loss=0.2632, pruned_loss=0.03763, over 3061915.38 frames. ], batch size: 250, lr: 3.52e-03, grad_scale: 8.0 2023-04-30 23:59:40,213 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.9056, 4.0046, 2.6323, 4.6837, 3.1391, 4.5564, 2.7131, 3.2982], device='cuda:1'), covar=tensor([0.0271, 0.0331, 0.1469, 0.0165, 0.0763, 0.0405, 0.1425, 0.0628], device='cuda:1'), in_proj_covar=tensor([0.0161, 0.0168, 0.0187, 0.0148, 0.0169, 0.0203, 0.0195, 0.0171], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0003, 0.0004, 0.0004, 0.0003], device='cuda:1') 2023-05-01 00:00:19,330 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=192433.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 00:00:57,137 INFO [train.py:904] (1/8) Epoch 19, batch 9750, loss[loss=0.17, simple_loss=0.265, pruned_loss=0.03749, over 16892.00 frames. ], tot_loss[loss=0.1689, simple_loss=0.2624, pruned_loss=0.03769, over 3062755.76 frames. ], batch size: 116, lr: 3.52e-03, grad_scale: 8.0 2023-05-01 00:00:58,972 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=192453.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 00:01:03,444 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.33 vs. limit=2.0 2023-05-01 00:01:09,806 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=192458.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 00:01:14,121 INFO [zipformer.py:625] (1/8) 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] (1/8) 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 00:02:23,612 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=192494.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 00:02:37,068 INFO [train.py:904] (1/8) Epoch 19, batch 9800, loss[loss=0.1836, simple_loss=0.2824, pruned_loss=0.04243, over 16643.00 frames. ], tot_loss[loss=0.1681, simple_loss=0.2626, pruned_loss=0.03674, over 3075593.39 frames. ], batch size: 134, lr: 3.52e-03, grad_scale: 8.0 2023-05-01 00:02:55,445 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=192511.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 00:04:23,527 INFO [train.py:904] (1/8) Epoch 19, batch 9850, loss[loss=0.1504, simple_loss=0.2427, pruned_loss=0.02905, over 12136.00 frames. ], tot_loss[loss=0.1685, simple_loss=0.2639, pruned_loss=0.03651, over 3075964.92 frames. ], batch size: 248, lr: 3.52e-03, grad_scale: 8.0 2023-05-01 00:04:33,057 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.7847, 3.7798, 2.4250, 4.5301, 2.9064, 4.3783, 2.5647, 3.2671], device='cuda:1'), covar=tensor([0.0274, 0.0368, 0.1478, 0.0160, 0.0865, 0.0385, 0.1406, 0.0631], device='cuda:1'), in_proj_covar=tensor([0.0159, 0.0166, 0.0185, 0.0147, 0.0168, 0.0201, 0.0193, 0.0170], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0003, 0.0004, 0.0004, 0.0003], device='cuda:1') 2023-05-01 00:04:57,498 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.4456, 2.8327, 3.1034, 1.9260, 2.7296, 2.0888, 2.9799, 3.0489], device='cuda:1'), covar=tensor([0.0264, 0.0782, 0.0520, 0.1975, 0.0772, 0.1006, 0.0657, 0.0854], device='cuda:1'), in_proj_covar=tensor([0.0148, 0.0151, 0.0160, 0.0147, 0.0139, 0.0124, 0.0138, 0.0162], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:1') 2023-05-01 00:05:00,568 INFO [optim.py:368] (1/8) 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 00:05:11,432 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=192575.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 00:05:15,468 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=192577.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 00:06:14,628 INFO [train.py:904] (1/8) Epoch 19, batch 9900, loss[loss=0.1639, simple_loss=0.2696, pruned_loss=0.02906, over 16889.00 frames. ], tot_loss[loss=0.1688, simple_loss=0.2646, pruned_loss=0.03651, over 3077010.10 frames. ], batch size: 102, lr: 3.52e-03, grad_scale: 8.0 2023-05-01 00:06:39,149 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.3638, 3.3674, 3.5776, 1.8494, 3.6982, 3.8355, 2.8899, 2.9713], device='cuda:1'), covar=tensor([0.0820, 0.0284, 0.0220, 0.1291, 0.0094, 0.0152, 0.0473, 0.0452], device='cuda:1'), in_proj_covar=tensor([0.0139, 0.0101, 0.0088, 0.0131, 0.0073, 0.0113, 0.0119, 0.0123], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:1') 2023-05-01 00:06:46,557 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 19, batch 9950, loss[loss=0.1746, simple_loss=0.2656, pruned_loss=0.04176, over 12425.00 frames. ], tot_loss[loss=0.1702, simple_loss=0.2665, pruned_loss=0.03696, over 3075227.56 frames. ], batch size: 248, lr: 3.52e-03, grad_scale: 8.0 2023-05-01 00:08:54,784 INFO [optim.py:368] (1/8) 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 00:09:04,404 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=192673.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 00:09:14,969 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=192677.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 00:10:14,419 INFO [train.py:904] (1/8) Epoch 19, batch 10000, loss[loss=0.172, simple_loss=0.2734, pruned_loss=0.03534, over 16955.00 frames. ], tot_loss[loss=0.1697, simple_loss=0.2655, pruned_loss=0.03691, over 3066393.66 frames. ], batch size: 109, lr: 3.52e-03, grad_scale: 8.0 2023-05-01 00:10:27,875 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.24 vs. limit=2.0 2023-05-01 00:11:22,447 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=192734.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 00:11:25,272 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=192735.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 00:11:50,830 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=192748.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 00:11:56,957 INFO [train.py:904] (1/8) Epoch 19, batch 10050, loss[loss=0.1865, simple_loss=0.2839, pruned_loss=0.04453, over 16667.00 frames. ], tot_loss[loss=0.1689, simple_loss=0.265, pruned_loss=0.03646, over 3082880.07 frames. ], batch size: 134, lr: 3.52e-03, grad_scale: 8.0 2023-05-01 00:12:02,951 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=192755.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 00:12:09,549 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=192758.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 00:12:31,268 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.3909, 3.3378, 1.9445, 3.7810, 2.4143, 3.6805, 2.2039, 2.7796], device='cuda:1'), covar=tensor([0.0272, 0.0412, 0.1678, 0.0216, 0.0925, 0.0645, 0.1505, 0.0766], device='cuda:1'), in_proj_covar=tensor([0.0158, 0.0166, 0.0184, 0.0146, 0.0167, 0.0200, 0.0193, 0.0169], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0003, 0.0004, 0.0004, 0.0003], device='cuda:1') 2023-05-01 00:12:32,889 INFO [optim.py:368] (1/8) Clipping_scale=2.0, grad-norm quartiles 1.427e+02 2.015e+02 2.519e+02 2.909e+02 8.183e+02, threshold=5.037e+02, percent-clipped=1.0 2023-05-01 00:12:40,003 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([5.0407, 5.3812, 5.1825, 5.1528, 4.8533, 4.8580, 4.7459, 5.4745], device='cuda:1'), covar=tensor([0.1258, 0.0864, 0.0949, 0.0751, 0.0760, 0.0872, 0.1242, 0.0782], device='cuda:1'), in_proj_covar=tensor([0.0614, 0.0755, 0.0614, 0.0561, 0.0474, 0.0482, 0.0629, 0.0579], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:1') 2023-05-01 00:12:44,276 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.74 vs. limit=2.0 2023-05-01 00:13:08,085 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=192789.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 00:13:21,267 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=192796.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 00:13:30,579 INFO [train.py:904] (1/8) Epoch 19, batch 10100, loss[loss=0.152, simple_loss=0.2372, pruned_loss=0.03337, over 12585.00 frames. ], tot_loss[loss=0.1691, simple_loss=0.2648, pruned_loss=0.03674, over 3062613.26 frames. ], batch size: 248, lr: 3.52e-03, grad_scale: 8.0 2023-05-01 00:13:38,853 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=192806.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 00:13:48,333 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=192811.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 00:14:16,236 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.5846, 3.5985, 2.7720, 2.1939, 2.2354, 2.3015, 3.8043, 3.1604], device='cuda:1'), covar=tensor([0.3005, 0.0643, 0.1726, 0.2847, 0.3012, 0.2197, 0.0436, 0.1374], device='cuda:1'), in_proj_covar=tensor([0.0313, 0.0254, 0.0288, 0.0292, 0.0274, 0.0242, 0.0275, 0.0313], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-01 00:14:28,794 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=192831.0, num_to_drop=1, layers_to_drop={1} 2023-05-01 00:14:44,312 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.33 vs. limit=2.0 2023-05-01 00:15:13,689 INFO [train.py:904] (1/8) Epoch 20, batch 0, loss[loss=0.2375, simple_loss=0.3147, pruned_loss=0.08012, over 15491.00 frames. ], tot_loss[loss=0.2375, simple_loss=0.3147, pruned_loss=0.08012, over 15491.00 frames. ], batch size: 191, lr: 3.43e-03, grad_scale: 8.0 2023-05-01 00:15:13,689 INFO [train.py:929] (1/8) Computing validation loss 2023-05-01 00:15:21,162 INFO [train.py:938] (1/8) Epoch 20, validation: loss=0.146, simple_loss=0.2496, pruned_loss=0.02121, over 944034.00 frames. 2023-05-01 00:15:21,162 INFO [train.py:939] (1/8) Maximum memory allocated so far is 18145MB 2023-05-01 00:15:31,953 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=192859.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 00:15:45,141 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.4643, 2.3963, 2.2455, 4.3747, 2.2349, 2.7729, 2.3935, 2.4687], device='cuda:1'), covar=tensor([0.1197, 0.3619, 0.3045, 0.0439, 0.4122, 0.2440, 0.3478, 0.3425], device='cuda:1'), in_proj_covar=tensor([0.0383, 0.0423, 0.0352, 0.0310, 0.0423, 0.0484, 0.0395, 0.0492], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-01 00:15:49,198 INFO [optim.py:368] (1/8) 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 00:15:53,719 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=192875.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 00:15:54,057 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.59 vs. limit=2.0 2023-05-01 00:15:56,034 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=192877.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 00:16:03,249 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.31 vs. limit=2.0 2023-05-01 00:16:17,611 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 20, batch 50, loss[loss=0.1698, simple_loss=0.2462, pruned_loss=0.04667, over 16738.00 frames. ], tot_loss[loss=0.1877, simple_loss=0.2723, pruned_loss=0.05159, over 751200.40 frames. ], batch size: 89, lr: 3.43e-03, grad_scale: 2.0 2023-05-01 00:16:48,424 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([1.9794, 3.0315, 2.8704, 5.1701, 4.3536, 4.6481, 1.7490, 3.3570], device='cuda:1'), covar=tensor([0.1287, 0.0757, 0.1167, 0.0196, 0.0229, 0.0372, 0.1602, 0.0757], device='cuda:1'), in_proj_covar=tensor([0.0165, 0.0169, 0.0190, 0.0177, 0.0195, 0.0208, 0.0195, 0.0187], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-05-01 00:16:59,571 INFO [zipformer.py:625] (1/8) 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] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=192925.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 00:17:17,317 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.8474, 2.0531, 2.4486, 2.8457, 2.7634, 3.1882, 2.2515, 3.2687], device='cuda:1'), covar=tensor([0.0244, 0.0491, 0.0365, 0.0313, 0.0334, 0.0221, 0.0481, 0.0154], device='cuda:1'), in_proj_covar=tensor([0.0176, 0.0187, 0.0172, 0.0173, 0.0186, 0.0143, 0.0188, 0.0138], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-01 00:17:17,515 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.18 vs. limit=2.0 2023-05-01 00:17:25,182 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.9891, 4.1532, 4.0047, 3.8280, 3.4293, 4.2255, 4.0130, 3.8598], device='cuda:1'), covar=tensor([0.1112, 0.0979, 0.0637, 0.0514, 0.1697, 0.0649, 0.0915, 0.0847], device='cuda:1'), in_proj_covar=tensor([0.0268, 0.0383, 0.0313, 0.0304, 0.0320, 0.0355, 0.0216, 0.0373], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002], device='cuda:1') 2023-05-01 00:17:38,938 INFO [train.py:904] (1/8) Epoch 20, batch 100, loss[loss=0.1895, simple_loss=0.2594, pruned_loss=0.05985, over 16704.00 frames. ], tot_loss[loss=0.1832, simple_loss=0.2694, pruned_loss=0.04853, over 1323843.40 frames. ], batch size: 83, lr: 3.43e-03, grad_scale: 2.0 2023-05-01 00:18:07,333 INFO [optim.py:368] (1/8) 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] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=192972.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 00:18:15,967 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([5.2226, 5.1416, 5.6336, 5.6184, 5.6997, 5.2798, 5.2411, 5.0360], device='cuda:1'), covar=tensor([0.0345, 0.0447, 0.0354, 0.0483, 0.0472, 0.0410, 0.1098, 0.0458], device='cuda:1'), in_proj_covar=tensor([0.0382, 0.0416, 0.0407, 0.0381, 0.0452, 0.0427, 0.0517, 0.0346], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:1') 2023-05-01 00:18:48,432 INFO [train.py:904] (1/8) Epoch 20, batch 150, loss[loss=0.1879, simple_loss=0.269, pruned_loss=0.05343, over 16229.00 frames. ], tot_loss[loss=0.1818, simple_loss=0.2674, pruned_loss=0.04807, over 1759368.11 frames. ], batch size: 165, lr: 3.42e-03, grad_scale: 2.0 2023-05-01 00:19:27,067 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=193029.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 00:19:43,950 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=193041.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 00:19:52,905 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=193048.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 00:19:54,447 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.17 vs. limit=2.0 2023-05-01 00:19:58,179 INFO [train.py:904] (1/8) Epoch 20, batch 200, loss[loss=0.2117, simple_loss=0.2792, pruned_loss=0.07211, over 16921.00 frames. ], tot_loss[loss=0.1833, simple_loss=0.269, pruned_loss=0.04881, over 2099094.72 frames. ], batch size: 109, lr: 3.42e-03, grad_scale: 2.0 2023-05-01 00:20:03,596 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=193055.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 00:20:27,492 INFO [optim.py:368] (1/8) 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 00:20:39,527 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([5.0985, 5.7523, 5.8528, 5.5747, 5.6958, 6.2433, 5.6997, 5.4215], device='cuda:1'), covar=tensor([0.0984, 0.1877, 0.2532, 0.1933, 0.2699, 0.0974, 0.1635, 0.2354], device='cuda:1'), in_proj_covar=tensor([0.0390, 0.0562, 0.0622, 0.0469, 0.0627, 0.0651, 0.0489, 0.0624], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-05-01 00:20:42,410 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=4.43 vs. limit=5.0 2023-05-01 00:20:50,739 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=193089.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 00:20:53,816 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=193091.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 00:20:57,974 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.3659, 4.6601, 4.4857, 4.4744, 4.2273, 4.1661, 4.2394, 4.7088], device='cuda:1'), covar=tensor([0.1196, 0.0998, 0.1081, 0.0847, 0.0833, 0.1487, 0.1064, 0.1085], device='cuda:1'), in_proj_covar=tensor([0.0638, 0.0787, 0.0638, 0.0583, 0.0492, 0.0501, 0.0655, 0.0601], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:1') 2023-05-01 00:21:00,957 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 20, batch 250, loss[loss=0.1747, simple_loss=0.2527, pruned_loss=0.0484, over 16374.00 frames. ], tot_loss[loss=0.181, simple_loss=0.2662, pruned_loss=0.04796, over 2364460.57 frames. ], batch size: 146, lr: 3.42e-03, grad_scale: 2.0 2023-05-01 00:21:09,307 INFO [zipformer.py:625] (1/8) 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] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=193103.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 00:21:57,541 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=193137.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 00:22:07,371 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.4514, 2.2878, 2.2093, 4.3059, 2.3454, 2.6333, 2.3124, 2.4405], device='cuda:1'), covar=tensor([0.1233, 0.3900, 0.3170, 0.0495, 0.3985, 0.2744, 0.3652, 0.3816], device='cuda:1'), in_proj_covar=tensor([0.0388, 0.0431, 0.0358, 0.0318, 0.0429, 0.0494, 0.0402, 0.0502], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-01 00:22:17,119 INFO [train.py:904] (1/8) Epoch 20, batch 300, loss[loss=0.18, simple_loss=0.2661, pruned_loss=0.04689, over 16538.00 frames. ], tot_loss[loss=0.1778, simple_loss=0.2627, pruned_loss=0.04642, over 2584358.51 frames. ], batch size: 68, lr: 3.42e-03, grad_scale: 1.0 2023-05-01 00:22:46,394 INFO [optim.py:368] (1/8) 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 00:23:05,586 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=193187.0, num_to_drop=1, layers_to_drop={1} 2023-05-01 00:23:10,363 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.4838, 3.4718, 2.1512, 3.6981, 2.7461, 3.6505, 2.2033, 2.7415], device='cuda:1'), covar=tensor([0.0264, 0.0403, 0.1526, 0.0351, 0.0777, 0.0757, 0.1385, 0.0718], device='cuda:1'), in_proj_covar=tensor([0.0167, 0.0176, 0.0193, 0.0157, 0.0176, 0.0212, 0.0202, 0.0177], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-05-01 00:23:28,161 INFO [train.py:904] (1/8) Epoch 20, batch 350, loss[loss=0.1649, simple_loss=0.238, pruned_loss=0.04588, over 16785.00 frames. ], tot_loss[loss=0.1745, simple_loss=0.2593, pruned_loss=0.04487, over 2743843.84 frames. ], batch size: 102, lr: 3.42e-03, grad_scale: 1.0 2023-05-01 00:23:39,633 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=193211.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 00:24:22,157 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.7785, 3.8551, 2.3804, 4.5332, 3.0254, 4.4526, 2.4307, 3.1324], device='cuda:1'), covar=tensor([0.0332, 0.0449, 0.1600, 0.0305, 0.0905, 0.0517, 0.1615, 0.0789], device='cuda:1'), in_proj_covar=tensor([0.0167, 0.0175, 0.0193, 0.0157, 0.0175, 0.0212, 0.0202, 0.0177], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-05-01 00:24:26,487 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.8635, 2.1192, 2.3273, 3.1629, 2.1955, 2.3017, 2.3228, 2.2511], device='cuda:1'), covar=tensor([0.1289, 0.3158, 0.2594, 0.0743, 0.3924, 0.2382, 0.3017, 0.3292], device='cuda:1'), in_proj_covar=tensor([0.0392, 0.0435, 0.0361, 0.0320, 0.0432, 0.0498, 0.0406, 0.0507], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-01 00:24:37,894 INFO [train.py:904] (1/8) Epoch 20, batch 400, loss[loss=0.1998, simple_loss=0.2713, pruned_loss=0.06418, over 16888.00 frames. ], tot_loss[loss=0.1745, simple_loss=0.2587, pruned_loss=0.04516, over 2869487.42 frames. ], batch size: 116, lr: 3.42e-03, grad_scale: 2.0 2023-05-01 00:24:49,999 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([5.0488, 5.0670, 5.5277, 5.5238, 5.5053, 5.1668, 5.0886, 4.8802], device='cuda:1'), covar=tensor([0.0342, 0.0528, 0.0354, 0.0416, 0.0481, 0.0390, 0.0934, 0.0481], device='cuda:1'), in_proj_covar=tensor([0.0393, 0.0428, 0.0417, 0.0390, 0.0462, 0.0439, 0.0529, 0.0355], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:1') 2023-05-01 00:25:05,927 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=193272.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 00:25:06,032 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=193272.0, num_to_drop=1, layers_to_drop={1} 2023-05-01 00:25:06,669 INFO [optim.py:368] (1/8) 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 00:25:23,360 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 20, batch 450, loss[loss=0.1654, simple_loss=0.2467, pruned_loss=0.04206, over 16819.00 frames. ], tot_loss[loss=0.1725, simple_loss=0.2567, pruned_loss=0.04417, over 2967098.63 frames. ], batch size: 83, lr: 3.42e-03, grad_scale: 2.0 2023-05-01 00:26:01,954 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.0052, 4.2967, 4.3985, 3.3025, 3.6293, 4.3494, 3.9326, 2.6965], device='cuda:1'), covar=tensor([0.0440, 0.0064, 0.0035, 0.0296, 0.0153, 0.0083, 0.0086, 0.0422], device='cuda:1'), in_proj_covar=tensor([0.0136, 0.0081, 0.0080, 0.0134, 0.0095, 0.0106, 0.0092, 0.0129], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-05-01 00:26:13,644 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=193320.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 00:26:26,014 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=193329.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 00:26:48,785 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=193345.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 00:26:57,283 INFO [train.py:904] (1/8) Epoch 20, batch 500, loss[loss=0.1551, simple_loss=0.2354, pruned_loss=0.03736, over 16213.00 frames. ], tot_loss[loss=0.1705, simple_loss=0.255, pruned_loss=0.04297, over 3040227.63 frames. ], batch size: 165, lr: 3.42e-03, grad_scale: 2.0 2023-05-01 00:27:26,047 INFO [optim.py:368] (1/8) 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 00:27:31,763 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=193377.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 00:27:51,497 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=193391.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 00:27:59,425 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=193397.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 00:28:07,094 INFO [train.py:904] (1/8) Epoch 20, batch 550, loss[loss=0.1812, simple_loss=0.276, pruned_loss=0.04319, over 17141.00 frames. ], tot_loss[loss=0.1694, simple_loss=0.2545, pruned_loss=0.04211, over 3103935.28 frames. ], batch size: 49, lr: 3.42e-03, grad_scale: 2.0 2023-05-01 00:28:10,540 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([5.3181, 5.2704, 5.1712, 4.6071, 4.7089, 5.2044, 5.1705, 4.7770], device='cuda:1'), covar=tensor([0.0611, 0.0584, 0.0297, 0.0390, 0.1258, 0.0498, 0.0336, 0.0828], device='cuda:1'), in_proj_covar=tensor([0.0285, 0.0409, 0.0331, 0.0324, 0.0342, 0.0379, 0.0230, 0.0398], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-01 00:28:32,276 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=193421.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 00:28:56,783 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=193439.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 00:29:14,799 INFO [train.py:904] (1/8) Epoch 20, batch 600, loss[loss=0.1672, simple_loss=0.2593, pruned_loss=0.0376, over 16750.00 frames. ], tot_loss[loss=0.1692, simple_loss=0.2542, pruned_loss=0.04211, over 3159739.46 frames. ], batch size: 57, lr: 3.42e-03, grad_scale: 2.0 2023-05-01 00:29:18,103 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.0218, 3.8782, 4.0976, 4.2150, 4.2728, 3.8809, 4.1026, 4.2696], device='cuda:1'), covar=tensor([0.1612, 0.1258, 0.1259, 0.0724, 0.0596, 0.1600, 0.2208, 0.0935], device='cuda:1'), in_proj_covar=tensor([0.0626, 0.0777, 0.0909, 0.0794, 0.0593, 0.0620, 0.0642, 0.0741], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-01 00:29:18,138 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=193454.0, num_to_drop=1, layers_to_drop={0} 2023-05-01 00:29:38,082 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.20 vs. limit=2.0 2023-05-01 00:29:43,212 INFO [optim.py:368] (1/8) 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 00:29:55,303 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=193482.0, num_to_drop=1, layers_to_drop={2} 2023-05-01 00:30:03,338 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=193487.0, num_to_drop=1, layers_to_drop={1} 2023-05-01 00:30:21,024 INFO [train.py:904] (1/8) Epoch 20, batch 650, loss[loss=0.1505, simple_loss=0.2369, pruned_loss=0.03205, over 16831.00 frames. ], tot_loss[loss=0.1677, simple_loss=0.2524, pruned_loss=0.04153, over 3200503.78 frames. ], batch size: 42, lr: 3.42e-03, grad_scale: 2.0 2023-05-01 00:30:39,905 INFO [zipformer.py:625] (1/8) 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] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=193535.0, num_to_drop=1, layers_to_drop={0} 2023-05-01 00:31:29,636 INFO [train.py:904] (1/8) Epoch 20, batch 700, loss[loss=0.1712, simple_loss=0.2653, pruned_loss=0.03851, over 17092.00 frames. ], tot_loss[loss=0.1681, simple_loss=0.253, pruned_loss=0.04161, over 3231945.46 frames. ], batch size: 47, lr: 3.42e-03, grad_scale: 2.0 2023-05-01 00:31:42,796 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=3.40 vs. limit=5.0 2023-05-01 00:31:48,972 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=193567.0, num_to_drop=1, layers_to_drop={0} 2023-05-01 00:31:57,448 INFO [optim.py:368] (1/8) 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 00:32:08,585 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.0827, 2.1363, 2.6909, 3.0730, 2.9658, 3.5395, 2.4710, 3.4626], device='cuda:1'), covar=tensor([0.0208, 0.0513, 0.0301, 0.0287, 0.0273, 0.0201, 0.0437, 0.0170], device='cuda:1'), in_proj_covar=tensor([0.0181, 0.0190, 0.0175, 0.0178, 0.0191, 0.0149, 0.0191, 0.0142], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-01 00:32:35,677 INFO [train.py:904] (1/8) Epoch 20, batch 750, loss[loss=0.1829, simple_loss=0.256, pruned_loss=0.05487, over 16876.00 frames. ], tot_loss[loss=0.1676, simple_loss=0.2523, pruned_loss=0.04149, over 3246334.20 frames. ], batch size: 116, lr: 3.42e-03, grad_scale: 2.0 2023-05-01 00:32:54,605 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=193615.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 00:33:15,611 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([5.4251, 5.3669, 5.2434, 4.6893, 4.8567, 5.3149, 5.3012, 4.9105], device='cuda:1'), covar=tensor([0.0586, 0.0500, 0.0273, 0.0363, 0.1086, 0.0467, 0.0258, 0.0758], device='cuda:1'), in_proj_covar=tensor([0.0291, 0.0416, 0.0337, 0.0330, 0.0347, 0.0386, 0.0233, 0.0403], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002], device='cuda:1') 2023-05-01 00:33:25,014 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([5.1952, 5.1352, 4.9482, 4.4085, 5.0170, 1.9869, 4.7801, 4.8298], device='cuda:1'), covar=tensor([0.0075, 0.0087, 0.0225, 0.0386, 0.0107, 0.2668, 0.0147, 0.0213], device='cuda:1'), in_proj_covar=tensor([0.0159, 0.0148, 0.0192, 0.0171, 0.0168, 0.0204, 0.0181, 0.0169], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-01 00:33:28,437 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 20, batch 800, loss[loss=0.156, simple_loss=0.242, pruned_loss=0.03496, over 16851.00 frames. ], tot_loss[loss=0.1678, simple_loss=0.2523, pruned_loss=0.04168, over 3264794.65 frames. ], batch size: 42, lr: 3.42e-03, grad_scale: 4.0 2023-05-01 00:34:10,663 INFO [optim.py:368] (1/8) 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 00:34:16,595 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=193676.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 00:34:45,080 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 20, batch 850, loss[loss=0.1453, simple_loss=0.2326, pruned_loss=0.02905, over 16799.00 frames. ], tot_loss[loss=0.1672, simple_loss=0.2514, pruned_loss=0.0415, over 3258529.68 frames. ], batch size: 102, lr: 3.42e-03, grad_scale: 4.0 2023-05-01 00:35:02,533 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.6438, 3.5116, 3.8256, 2.0581, 4.0370, 4.0197, 3.2415, 3.0188], device='cuda:1'), covar=tensor([0.0736, 0.0251, 0.0210, 0.1122, 0.0087, 0.0161, 0.0359, 0.0440], device='cuda:1'), in_proj_covar=tensor([0.0146, 0.0107, 0.0095, 0.0138, 0.0078, 0.0121, 0.0126, 0.0130], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0004, 0.0003], device='cuda:1') 2023-05-01 00:35:20,271 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.7225, 1.8295, 2.3150, 2.5613, 2.7017, 2.6778, 2.0055, 2.8272], device='cuda:1'), covar=tensor([0.0192, 0.0502, 0.0326, 0.0312, 0.0294, 0.0291, 0.0504, 0.0169], device='cuda:1'), in_proj_covar=tensor([0.0182, 0.0191, 0.0176, 0.0178, 0.0192, 0.0149, 0.0192, 0.0143], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-01 00:35:50,808 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=193745.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 00:35:59,339 INFO [train.py:904] (1/8) Epoch 20, batch 900, loss[loss=0.1653, simple_loss=0.2513, pruned_loss=0.03969, over 16898.00 frames. ], tot_loss[loss=0.1667, simple_loss=0.2511, pruned_loss=0.04115, over 3254711.40 frames. ], batch size: 96, lr: 3.42e-03, grad_scale: 4.0 2023-05-01 00:35:59,946 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.17 vs. limit=2.0 2023-05-01 00:36:16,430 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.7945, 4.2664, 3.0825, 2.3097, 2.6526, 2.5843, 4.5795, 3.5324], device='cuda:1'), covar=tensor([0.2905, 0.0593, 0.1790, 0.2842, 0.2815, 0.1999, 0.0344, 0.1384], device='cuda:1'), in_proj_covar=tensor([0.0324, 0.0265, 0.0299, 0.0303, 0.0290, 0.0252, 0.0287, 0.0330], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:1') 2023-05-01 00:36:28,230 INFO [optim.py:368] (1/8) 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 00:36:33,730 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=193777.0, num_to_drop=1, layers_to_drop={1} 2023-05-01 00:36:35,676 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=4.70 vs. limit=5.0 2023-05-01 00:36:56,362 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.6787, 4.5067, 4.7206, 4.9166, 5.0505, 4.5516, 5.0884, 5.0783], device='cuda:1'), covar=tensor([0.2020, 0.1527, 0.1939, 0.0972, 0.0765, 0.1091, 0.0847, 0.0786], device='cuda:1'), in_proj_covar=tensor([0.0636, 0.0789, 0.0922, 0.0805, 0.0601, 0.0627, 0.0651, 0.0753], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-01 00:37:09,295 INFO [train.py:904] (1/8) Epoch 20, batch 950, loss[loss=0.1742, simple_loss=0.2735, pruned_loss=0.03749, over 17281.00 frames. ], tot_loss[loss=0.167, simple_loss=0.2511, pruned_loss=0.04147, over 3271154.72 frames. ], batch size: 52, lr: 3.42e-03, grad_scale: 4.0 2023-05-01 00:37:20,476 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=193810.0, num_to_drop=1, layers_to_drop={1} 2023-05-01 00:37:36,623 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.45 vs. limit=2.0 2023-05-01 00:38:17,860 INFO [train.py:904] (1/8) Epoch 20, batch 1000, loss[loss=0.1645, simple_loss=0.2381, pruned_loss=0.04549, over 16850.00 frames. ], tot_loss[loss=0.1666, simple_loss=0.2506, pruned_loss=0.04129, over 3282240.09 frames. ], batch size: 116, lr: 3.42e-03, grad_scale: 4.0 2023-05-01 00:38:39,225 INFO [zipformer.py:625] (1/8) 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] (1/8) 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 00:39:00,659 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=193884.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 00:39:24,781 INFO [train.py:904] (1/8) Epoch 20, batch 1050, loss[loss=0.161, simple_loss=0.2453, pruned_loss=0.03835, over 15859.00 frames. ], tot_loss[loss=0.1665, simple_loss=0.2509, pruned_loss=0.04105, over 3289583.80 frames. ], batch size: 35, lr: 3.42e-03, grad_scale: 4.0 2023-05-01 00:39:43,672 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=193915.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 00:40:18,183 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=193940.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 00:40:25,128 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=193945.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 00:40:35,456 INFO [train.py:904] (1/8) Epoch 20, batch 1100, loss[loss=0.1635, simple_loss=0.2423, pruned_loss=0.0424, over 16714.00 frames. ], tot_loss[loss=0.1664, simple_loss=0.2504, pruned_loss=0.04125, over 3291964.64 frames. ], batch size: 89, lr: 3.42e-03, grad_scale: 4.0 2023-05-01 00:40:54,799 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.5052, 3.5499, 4.0926, 2.2029, 3.1650, 2.3659, 3.8879, 3.7673], device='cuda:1'), covar=tensor([0.0245, 0.1015, 0.0486, 0.2035, 0.0840, 0.1046, 0.0625, 0.1088], device='cuda:1'), in_proj_covar=tensor([0.0155, 0.0160, 0.0166, 0.0152, 0.0144, 0.0128, 0.0144, 0.0172], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:1') 2023-05-01 00:41:01,679 INFO [zipformer.py:625] (1/8) 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] (1/8) 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 00:41:14,444 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.0824, 2.5888, 2.1272, 2.3630, 2.9544, 2.7403, 2.9645, 3.0893], device='cuda:1'), covar=tensor([0.0206, 0.0381, 0.0495, 0.0413, 0.0228, 0.0323, 0.0236, 0.0237], device='cuda:1'), in_proj_covar=tensor([0.0205, 0.0237, 0.0225, 0.0229, 0.0239, 0.0237, 0.0237, 0.0230], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-01 00:41:24,329 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=193988.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 00:41:30,455 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.6040, 4.6933, 5.0276, 5.0218, 5.0223, 4.7339, 4.7182, 4.5675], device='cuda:1'), covar=tensor([0.0379, 0.0694, 0.0444, 0.0397, 0.0479, 0.0451, 0.0880, 0.0494], device='cuda:1'), in_proj_covar=tensor([0.0404, 0.0444, 0.0429, 0.0401, 0.0480, 0.0451, 0.0544, 0.0365], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:1') 2023-05-01 00:41:46,949 INFO [train.py:904] (1/8) Epoch 20, batch 1150, loss[loss=0.1612, simple_loss=0.2539, pruned_loss=0.03421, over 17259.00 frames. ], tot_loss[loss=0.1654, simple_loss=0.2497, pruned_loss=0.04057, over 3297399.63 frames. ], batch size: 52, lr: 3.42e-03, grad_scale: 4.0 2023-05-01 00:41:57,250 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.9603, 5.0233, 5.4563, 5.4594, 5.4574, 5.1349, 5.0543, 4.8805], device='cuda:1'), covar=tensor([0.0345, 0.0622, 0.0485, 0.0485, 0.0511, 0.0395, 0.0932, 0.0472], device='cuda:1'), in_proj_covar=tensor([0.0404, 0.0444, 0.0429, 0.0402, 0.0481, 0.0451, 0.0545, 0.0365], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:1') 2023-05-01 00:42:13,039 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([5.0877, 5.1101, 5.5468, 5.5513, 5.5486, 5.2510, 5.1429, 4.9735], device='cuda:1'), covar=tensor([0.0325, 0.0576, 0.0410, 0.0412, 0.0454, 0.0360, 0.0907, 0.0440], device='cuda:1'), in_proj_covar=tensor([0.0405, 0.0445, 0.0430, 0.0402, 0.0482, 0.0452, 0.0546, 0.0366], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:1') 2023-05-01 00:42:56,045 INFO [train.py:904] (1/8) Epoch 20, batch 1200, loss[loss=0.1504, simple_loss=0.2259, pruned_loss=0.03746, over 11943.00 frames. ], tot_loss[loss=0.1643, simple_loss=0.2485, pruned_loss=0.04003, over 3305018.01 frames. ], batch size: 248, lr: 3.42e-03, grad_scale: 8.0 2023-05-01 00:43:09,510 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.7414, 3.7706, 2.1525, 4.4020, 2.9253, 4.3406, 2.0992, 3.0426], device='cuda:1'), covar=tensor([0.0301, 0.0394, 0.1936, 0.0312, 0.0846, 0.0438, 0.2049, 0.0807], device='cuda:1'), in_proj_covar=tensor([0.0171, 0.0179, 0.0197, 0.0163, 0.0179, 0.0218, 0.0205, 0.0180], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-05-01 00:43:25,035 INFO [zipformer.py:625] (1/8) 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] (1/8) 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 00:43:33,163 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=194077.0, num_to_drop=1, layers_to_drop={1} 2023-05-01 00:44:06,934 INFO [train.py:904] (1/8) Epoch 20, batch 1250, loss[loss=0.1542, simple_loss=0.2537, pruned_loss=0.02731, over 17107.00 frames. ], tot_loss[loss=0.1648, simple_loss=0.2489, pruned_loss=0.04034, over 3315156.43 frames. ], batch size: 49, lr: 3.42e-03, grad_scale: 8.0 2023-05-01 00:44:15,072 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.70 vs. limit=2.0 2023-05-01 00:44:17,006 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=194110.0, num_to_drop=1, layers_to_drop={2} 2023-05-01 00:44:23,733 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([5.1213, 5.4584, 5.2330, 5.2442, 4.9688, 4.8701, 4.9196, 5.5422], device='cuda:1'), covar=tensor([0.1220, 0.0946, 0.1039, 0.1015, 0.0912, 0.1024, 0.1239, 0.0984], device='cuda:1'), in_proj_covar=tensor([0.0662, 0.0816, 0.0664, 0.0606, 0.0511, 0.0518, 0.0682, 0.0625], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:1') 2023-05-01 00:44:38,137 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=194125.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 00:44:41,664 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.16 vs. limit=2.0 2023-05-01 00:44:49,780 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=194133.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 00:45:15,924 INFO [train.py:904] (1/8) Epoch 20, batch 1300, loss[loss=0.17, simple_loss=0.2582, pruned_loss=0.04093, over 16936.00 frames. ], tot_loss[loss=0.1646, simple_loss=0.2481, pruned_loss=0.04054, over 3302172.34 frames. ], batch size: 90, lr: 3.41e-03, grad_scale: 8.0 2023-05-01 00:45:26,571 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=194158.0, num_to_drop=1, layers_to_drop={0} 2023-05-01 00:45:26,731 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.2965, 4.1620, 4.4949, 2.4351, 4.8021, 4.8093, 3.4441, 3.7964], device='cuda:1'), covar=tensor([0.0661, 0.0206, 0.0183, 0.1123, 0.0061, 0.0142, 0.0385, 0.0360], device='cuda:1'), in_proj_covar=tensor([0.0148, 0.0108, 0.0097, 0.0140, 0.0079, 0.0124, 0.0127, 0.0131], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0004, 0.0004], device='cuda:1') 2023-05-01 00:45:46,646 INFO [optim.py:368] (1/8) Clipping_scale=2.0, grad-norm quartiles 1.297e+02 2.207e+02 2.541e+02 3.000e+02 4.873e+02, threshold=5.083e+02, percent-clipped=0.0 2023-05-01 00:46:22,969 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([5.0138, 5.0686, 5.4845, 5.5037, 5.4980, 5.1864, 5.0910, 4.8870], device='cuda:1'), covar=tensor([0.0335, 0.0549, 0.0403, 0.0375, 0.0404, 0.0359, 0.0960, 0.0459], device='cuda:1'), in_proj_covar=tensor([0.0405, 0.0445, 0.0430, 0.0403, 0.0481, 0.0452, 0.0546, 0.0365], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:1') 2023-05-01 00:46:27,312 INFO [train.py:904] (1/8) Epoch 20, batch 1350, loss[loss=0.1767, simple_loss=0.2748, pruned_loss=0.03927, over 17269.00 frames. ], tot_loss[loss=0.1648, simple_loss=0.2489, pruned_loss=0.0404, over 3297691.74 frames. ], batch size: 52, lr: 3.41e-03, grad_scale: 4.0 2023-05-01 00:47:20,518 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=194240.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 00:47:36,940 INFO [train.py:904] (1/8) Epoch 20, batch 1400, loss[loss=0.1594, simple_loss=0.2521, pruned_loss=0.03336, over 17179.00 frames. ], tot_loss[loss=0.1648, simple_loss=0.2486, pruned_loss=0.04044, over 3309100.35 frames. ], batch size: 46, lr: 3.41e-03, grad_scale: 4.0 2023-05-01 00:48:03,363 INFO [zipformer.py:625] (1/8) 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] (1/8) 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 00:48:30,354 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.8180, 4.8680, 5.2153, 5.2392, 5.2064, 4.9080, 4.8520, 4.7077], device='cuda:1'), covar=tensor([0.0391, 0.0639, 0.0485, 0.0431, 0.0519, 0.0416, 0.0951, 0.0526], device='cuda:1'), in_proj_covar=tensor([0.0405, 0.0444, 0.0428, 0.0402, 0.0480, 0.0452, 0.0544, 0.0364], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:1') 2023-05-01 00:48:44,299 INFO [train.py:904] (1/8) Epoch 20, batch 1450, loss[loss=0.167, simple_loss=0.24, pruned_loss=0.04699, over 16906.00 frames. ], tot_loss[loss=0.1649, simple_loss=0.2491, pruned_loss=0.04037, over 3320134.18 frames. ], batch size: 90, lr: 3.41e-03, grad_scale: 4.0 2023-05-01 00:49:03,353 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=4.75 vs. limit=5.0 2023-05-01 00:49:06,397 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=194318.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 00:49:07,985 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=194319.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 00:49:20,548 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.7866, 4.5962, 4.8119, 4.9860, 5.1417, 4.5135, 5.1379, 5.1543], device='cuda:1'), covar=tensor([0.1799, 0.1266, 0.1605, 0.0709, 0.0595, 0.1081, 0.0749, 0.0639], device='cuda:1'), in_proj_covar=tensor([0.0645, 0.0798, 0.0935, 0.0818, 0.0609, 0.0639, 0.0661, 0.0761], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-01 00:49:26,442 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([5.4506, 5.7986, 5.5535, 5.6393, 5.2421, 5.2059, 5.1198, 5.9130], device='cuda:1'), covar=tensor([0.1295, 0.1033, 0.1132, 0.0818, 0.0895, 0.0696, 0.1280, 0.1037], device='cuda:1'), in_proj_covar=tensor([0.0663, 0.0817, 0.0666, 0.0608, 0.0513, 0.0518, 0.0686, 0.0627], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:1') 2023-05-01 00:49:28,906 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([5.3505, 5.3063, 5.0901, 4.1703, 5.1251, 2.0594, 4.8811, 4.9621], device='cuda:1'), covar=tensor([0.0092, 0.0087, 0.0220, 0.0558, 0.0131, 0.2696, 0.0161, 0.0271], device='cuda:1'), in_proj_covar=tensor([0.0162, 0.0151, 0.0195, 0.0175, 0.0172, 0.0206, 0.0185, 0.0172], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-01 00:49:48,469 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.3046, 3.5258, 3.5776, 2.3051, 3.0904, 2.6652, 3.7796, 3.7263], device='cuda:1'), covar=tensor([0.0268, 0.0808, 0.0665, 0.1864, 0.0814, 0.0947, 0.0519, 0.1010], device='cuda:1'), in_proj_covar=tensor([0.0156, 0.0162, 0.0167, 0.0153, 0.0145, 0.0129, 0.0145, 0.0174], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-05-01 00:49:50,815 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.6703, 2.4601, 2.3617, 4.5129, 2.3154, 2.8612, 2.4710, 2.6133], device='cuda:1'), covar=tensor([0.1144, 0.3547, 0.3042, 0.0427, 0.4248, 0.2462, 0.3479, 0.3650], device='cuda:1'), in_proj_covar=tensor([0.0398, 0.0440, 0.0365, 0.0326, 0.0435, 0.0506, 0.0410, 0.0517], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-01 00:49:54,001 INFO [train.py:904] (1/8) Epoch 20, batch 1500, loss[loss=0.1689, simple_loss=0.2691, pruned_loss=0.0343, over 17056.00 frames. ], tot_loss[loss=0.1654, simple_loss=0.2495, pruned_loss=0.04059, over 3327108.25 frames. ], batch size: 50, lr: 3.41e-03, grad_scale: 4.0 2023-05-01 00:49:55,494 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=194353.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 00:50:07,446 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=194362.0, num_to_drop=1, layers_to_drop={1} 2023-05-01 00:50:24,391 INFO [optim.py:368] (1/8) 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 00:50:31,257 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=194379.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 00:51:03,653 INFO [train.py:904] (1/8) Epoch 20, batch 1550, loss[loss=0.1732, simple_loss=0.2695, pruned_loss=0.03844, over 17038.00 frames. ], tot_loss[loss=0.1668, simple_loss=0.2506, pruned_loss=0.04144, over 3328551.82 frames. ], batch size: 55, lr: 3.41e-03, grad_scale: 4.0 2023-05-01 00:51:20,852 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=194414.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 00:51:34,156 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=194423.0, num_to_drop=1, layers_to_drop={2} 2023-05-01 00:51:40,588 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 20, batch 1600, loss[loss=0.1592, simple_loss=0.2505, pruned_loss=0.03398, over 17253.00 frames. ], tot_loss[loss=0.1682, simple_loss=0.2522, pruned_loss=0.04214, over 3321961.90 frames. ], batch size: 45, lr: 3.41e-03, grad_scale: 8.0 2023-05-01 00:52:43,858 INFO [optim.py:368] (1/8) 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 00:53:22,707 INFO [train.py:904] (1/8) Epoch 20, batch 1650, loss[loss=0.2155, simple_loss=0.2956, pruned_loss=0.06773, over 12186.00 frames. ], tot_loss[loss=0.1687, simple_loss=0.2533, pruned_loss=0.04206, over 3318205.35 frames. ], batch size: 246, lr: 3.41e-03, grad_scale: 8.0 2023-05-01 00:53:29,012 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.5384, 3.5445, 3.7466, 2.6130, 3.4366, 3.7918, 3.5172, 2.1070], device='cuda:1'), covar=tensor([0.0475, 0.0145, 0.0055, 0.0373, 0.0117, 0.0105, 0.0103, 0.0478], device='cuda:1'), in_proj_covar=tensor([0.0136, 0.0082, 0.0082, 0.0135, 0.0096, 0.0108, 0.0093, 0.0130], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0002, 0.0004], device='cuda:1') 2023-05-01 00:54:16,608 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=194540.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 00:54:22,988 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.55 vs. limit=2.0 2023-05-01 00:54:33,460 INFO [train.py:904] (1/8) Epoch 20, batch 1700, loss[loss=0.1824, simple_loss=0.2642, pruned_loss=0.05032, over 16480.00 frames. ], tot_loss[loss=0.1707, simple_loss=0.2556, pruned_loss=0.04293, over 3301839.19 frames. ], batch size: 75, lr: 3.41e-03, grad_scale: 4.0 2023-05-01 00:55:05,697 INFO [optim.py:368] (1/8) 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 00:55:25,121 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 20, batch 1750, loss[loss=0.1748, simple_loss=0.2581, pruned_loss=0.04573, over 15465.00 frames. ], tot_loss[loss=0.1716, simple_loss=0.2571, pruned_loss=0.04301, over 3301058.30 frames. ], batch size: 191, lr: 3.41e-03, grad_scale: 4.0 2023-05-01 00:56:23,063 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.46 vs. limit=2.0 2023-05-01 00:56:27,185 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=194634.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 00:56:51,747 INFO [train.py:904] (1/8) Epoch 20, batch 1800, loss[loss=0.1821, simple_loss=0.269, pruned_loss=0.04755, over 16512.00 frames. ], tot_loss[loss=0.173, simple_loss=0.2588, pruned_loss=0.04362, over 3292780.68 frames. ], batch size: 68, lr: 3.41e-03, grad_scale: 4.0 2023-05-01 00:57:22,476 INFO [zipformer.py:625] (1/8) 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] (1/8) 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 00:57:50,490 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 20, batch 1850, loss[loss=0.1505, simple_loss=0.239, pruned_loss=0.03104, over 17180.00 frames. ], tot_loss[loss=0.1739, simple_loss=0.2596, pruned_loss=0.04412, over 3293496.85 frames. ], batch size: 46, lr: 3.41e-03, grad_scale: 4.0 2023-05-01 00:58:08,652 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=194709.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 00:58:13,991 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.9873, 3.7835, 4.1840, 2.0836, 4.4649, 4.5048, 3.1755, 3.4838], device='cuda:1'), covar=tensor([0.0706, 0.0241, 0.0232, 0.1184, 0.0070, 0.0185, 0.0465, 0.0383], device='cuda:1'), in_proj_covar=tensor([0.0149, 0.0109, 0.0098, 0.0140, 0.0079, 0.0125, 0.0128, 0.0132], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0004, 0.0004], device='cuda:1') 2023-05-01 00:58:21,440 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=194718.0, num_to_drop=1, layers_to_drop={1} 2023-05-01 00:58:33,797 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=194728.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 00:59:06,819 INFO [train.py:904] (1/8) Epoch 20, batch 1900, loss[loss=0.1683, simple_loss=0.2657, pruned_loss=0.03547, over 17126.00 frames. ], tot_loss[loss=0.1724, simple_loss=0.2587, pruned_loss=0.04308, over 3299598.57 frames. ], batch size: 47, lr: 3.41e-03, grad_scale: 4.0 2023-05-01 00:59:22,077 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([1.6903, 2.4363, 2.3120, 3.8148, 2.9700, 3.9615, 1.4510, 2.7571], device='cuda:1'), covar=tensor([0.1434, 0.0812, 0.1270, 0.0179, 0.0123, 0.0362, 0.1715, 0.0899], device='cuda:1'), in_proj_covar=tensor([0.0164, 0.0172, 0.0190, 0.0184, 0.0201, 0.0213, 0.0196, 0.0189], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-05-01 00:59:38,382 INFO [optim.py:368] (1/8) 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] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=194776.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 01:00:16,323 INFO [train.py:904] (1/8) Epoch 20, batch 1950, loss[loss=0.1694, simple_loss=0.2541, pruned_loss=0.04237, over 16777.00 frames. ], tot_loss[loss=0.1714, simple_loss=0.2581, pruned_loss=0.04237, over 3305991.02 frames. ], batch size: 83, lr: 3.41e-03, grad_scale: 4.0 2023-05-01 01:00:27,652 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.27 vs. limit=2.0 2023-05-01 01:00:42,816 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([5.1750, 4.9512, 5.2094, 5.4171, 5.6234, 4.8597, 5.5798, 5.5910], device='cuda:1'), covar=tensor([0.1989, 0.1334, 0.1755, 0.0739, 0.0509, 0.0871, 0.0519, 0.0546], device='cuda:1'), in_proj_covar=tensor([0.0653, 0.0809, 0.0946, 0.0828, 0.0619, 0.0649, 0.0666, 0.0771], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-01 01:01:17,086 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=194847.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 01:01:17,167 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.1612, 2.0939, 1.7179, 1.7909, 2.3164, 2.0295, 2.0845, 2.3886], device='cuda:1'), covar=tensor([0.0230, 0.0314, 0.0457, 0.0413, 0.0226, 0.0311, 0.0172, 0.0232], device='cuda:1'), in_proj_covar=tensor([0.0207, 0.0238, 0.0227, 0.0230, 0.0239, 0.0239, 0.0238, 0.0232], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-01 01:01:23,588 INFO [train.py:904] (1/8) Epoch 20, batch 2000, loss[loss=0.1798, simple_loss=0.2601, pruned_loss=0.04973, over 16805.00 frames. ], tot_loss[loss=0.1698, simple_loss=0.2565, pruned_loss=0.04152, over 3312438.44 frames. ], batch size: 102, lr: 3.41e-03, grad_scale: 8.0 2023-05-01 01:01:54,984 INFO [optim.py:368] (1/8) 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 01:02:09,567 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.5962, 2.4217, 2.4988, 4.5731, 2.3763, 2.8057, 2.5028, 2.6183], device='cuda:1'), covar=tensor([0.1146, 0.3596, 0.3017, 0.0413, 0.4180, 0.2703, 0.3468, 0.3786], device='cuda:1'), in_proj_covar=tensor([0.0397, 0.0439, 0.0363, 0.0326, 0.0433, 0.0505, 0.0408, 0.0515], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-01 01:02:32,415 INFO [train.py:904] (1/8) Epoch 20, batch 2050, loss[loss=0.1459, simple_loss=0.2431, pruned_loss=0.02433, over 17131.00 frames. ], tot_loss[loss=0.1691, simple_loss=0.2559, pruned_loss=0.04115, over 3319580.80 frames. ], batch size: 47, lr: 3.41e-03, grad_scale: 8.0 2023-05-01 01:02:41,381 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=194908.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 01:03:29,345 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.6519, 3.7379, 4.2650, 2.2872, 3.3795, 2.7759, 4.0417, 3.9681], device='cuda:1'), covar=tensor([0.0238, 0.0928, 0.0416, 0.1890, 0.0728, 0.0880, 0.0553, 0.1032], device='cuda:1'), in_proj_covar=tensor([0.0155, 0.0162, 0.0166, 0.0152, 0.0144, 0.0128, 0.0144, 0.0173], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:1') 2023-05-01 01:03:41,679 INFO [train.py:904] (1/8) Epoch 20, batch 2100, loss[loss=0.1907, simple_loss=0.2718, pruned_loss=0.05484, over 15579.00 frames. ], tot_loss[loss=0.1703, simple_loss=0.2567, pruned_loss=0.04197, over 3300731.76 frames. ], batch size: 191, lr: 3.41e-03, grad_scale: 8.0 2023-05-01 01:04:08,939 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.88 vs. limit=2.0 2023-05-01 01:04:12,818 INFO [zipformer.py:625] (1/8) 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] (1/8) 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 01:04:34,695 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=194990.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 01:04:50,814 INFO [train.py:904] (1/8) Epoch 20, batch 2150, loss[loss=0.158, simple_loss=0.2532, pruned_loss=0.03141, over 17116.00 frames. ], tot_loss[loss=0.1716, simple_loss=0.2578, pruned_loss=0.04272, over 3306239.80 frames. ], batch size: 47, lr: 3.41e-03, grad_scale: 8.0 2023-05-01 01:05:01,394 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=195009.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 01:05:03,695 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=195011.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 01:05:13,240 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=195018.0, num_to_drop=1, layers_to_drop={1} 2023-05-01 01:05:18,550 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=195022.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 01:05:49,676 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=195044.0, num_to_drop=1, layers_to_drop={1} 2023-05-01 01:06:01,907 INFO [train.py:904] (1/8) Epoch 20, batch 2200, loss[loss=0.2154, simple_loss=0.2922, pruned_loss=0.06925, over 11907.00 frames. ], tot_loss[loss=0.1725, simple_loss=0.2586, pruned_loss=0.04319, over 3308061.41 frames. ], batch size: 247, lr: 3.41e-03, grad_scale: 8.0 2023-05-01 01:06:09,189 INFO [zipformer.py:625] (1/8) 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] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=195066.0, num_to_drop=1, layers_to_drop={1} 2023-05-01 01:06:28,258 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=195071.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 01:06:30,103 INFO [zipformer.py:625] (1/8) 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] (1/8) 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 01:06:56,134 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([5.4254, 5.4478, 5.2319, 4.7687, 5.3339, 2.3669, 5.0496, 5.2128], device='cuda:1'), covar=tensor([0.0087, 0.0070, 0.0189, 0.0352, 0.0093, 0.2322, 0.0126, 0.0154], device='cuda:1'), in_proj_covar=tensor([0.0164, 0.0152, 0.0197, 0.0177, 0.0174, 0.0207, 0.0187, 0.0174], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-01 01:07:09,355 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=2.79 vs. limit=5.0 2023-05-01 01:07:10,872 INFO [train.py:904] (1/8) Epoch 20, batch 2250, loss[loss=0.2225, simple_loss=0.305, pruned_loss=0.06998, over 11760.00 frames. ], tot_loss[loss=0.1735, simple_loss=0.2597, pruned_loss=0.04364, over 3299440.75 frames. ], batch size: 246, lr: 3.41e-03, grad_scale: 8.0 2023-05-01 01:07:15,472 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=195105.0, num_to_drop=1, layers_to_drop={1} 2023-05-01 01:07:54,122 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=195132.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 01:08:21,144 INFO [train.py:904] (1/8) Epoch 20, batch 2300, loss[loss=0.1499, simple_loss=0.2332, pruned_loss=0.03334, over 16827.00 frames. ], tot_loss[loss=0.1736, simple_loss=0.26, pruned_loss=0.04357, over 3304502.24 frames. ], batch size: 39, lr: 3.41e-03, grad_scale: 8.0 2023-05-01 01:08:51,383 INFO [optim.py:368] (1/8) Clipping_scale=2.0, grad-norm quartiles 1.574e+02 2.280e+02 2.663e+02 3.175e+02 5.300e+02, threshold=5.327e+02, percent-clipped=0.0 2023-05-01 01:09:29,599 INFO [train.py:904] (1/8) Epoch 20, batch 2350, loss[loss=0.1836, simple_loss=0.277, pruned_loss=0.04505, over 16591.00 frames. ], tot_loss[loss=0.1739, simple_loss=0.26, pruned_loss=0.04394, over 3308133.54 frames. ], batch size: 62, lr: 3.41e-03, grad_scale: 8.0 2023-05-01 01:09:31,093 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=195203.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 01:09:54,974 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.27 vs. limit=2.0 2023-05-01 01:10:28,858 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.0387, 3.8647, 4.3328, 2.0521, 4.5802, 4.6023, 3.2805, 3.6532], device='cuda:1'), covar=tensor([0.0684, 0.0254, 0.0201, 0.1226, 0.0068, 0.0154, 0.0409, 0.0356], device='cuda:1'), in_proj_covar=tensor([0.0148, 0.0109, 0.0098, 0.0140, 0.0080, 0.0125, 0.0128, 0.0132], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0004, 0.0004], device='cuda:1') 2023-05-01 01:10:36,693 INFO [train.py:904] (1/8) Epoch 20, batch 2400, loss[loss=0.1994, simple_loss=0.2704, pruned_loss=0.06427, over 16915.00 frames. ], tot_loss[loss=0.1738, simple_loss=0.2601, pruned_loss=0.04374, over 3318734.21 frames. ], batch size: 109, lr: 3.41e-03, grad_scale: 8.0 2023-05-01 01:11:01,072 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.8506, 1.8759, 2.4804, 2.7970, 2.6415, 3.3650, 2.2611, 3.2948], device='cuda:1'), covar=tensor([0.0260, 0.0566, 0.0351, 0.0358, 0.0364, 0.0207, 0.0486, 0.0172], device='cuda:1'), in_proj_covar=tensor([0.0188, 0.0196, 0.0181, 0.0183, 0.0197, 0.0155, 0.0198, 0.0149], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-01 01:11:07,015 INFO [optim.py:368] (1/8) 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 01:11:29,107 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=195290.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 01:11:45,804 INFO [train.py:904] (1/8) Epoch 20, batch 2450, loss[loss=0.1797, simple_loss=0.2791, pruned_loss=0.04017, over 16698.00 frames. ], tot_loss[loss=0.1748, simple_loss=0.2617, pruned_loss=0.04397, over 3313590.53 frames. ], batch size: 62, lr: 3.40e-03, grad_scale: 8.0 2023-05-01 01:12:35,687 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 20, batch 2500, loss[loss=0.1829, simple_loss=0.2809, pruned_loss=0.04246, over 17099.00 frames. ], tot_loss[loss=0.1743, simple_loss=0.261, pruned_loss=0.04377, over 3319094.54 frames. ], batch size: 53, lr: 3.40e-03, grad_scale: 8.0 2023-05-01 01:13:09,155 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.33 vs. limit=2.0 2023-05-01 01:13:15,893 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=195367.0, num_to_drop=1, layers_to_drop={0} 2023-05-01 01:13:26,924 INFO [optim.py:368] (1/8) 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 01:13:33,676 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.8391, 2.7065, 2.3870, 2.6138, 3.0084, 2.8132, 3.4023, 3.2662], device='cuda:1'), covar=tensor([0.0143, 0.0443, 0.0533, 0.0440, 0.0313, 0.0410, 0.0278, 0.0274], device='cuda:1'), in_proj_covar=tensor([0.0209, 0.0241, 0.0227, 0.0230, 0.0241, 0.0240, 0.0241, 0.0234], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-01 01:13:50,894 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=195392.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 01:14:01,556 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=195400.0, num_to_drop=1, layers_to_drop={0} 2023-05-01 01:14:04,077 INFO [train.py:904] (1/8) Epoch 20, batch 2550, loss[loss=0.158, simple_loss=0.2382, pruned_loss=0.03892, over 16882.00 frames. ], tot_loss[loss=0.174, simple_loss=0.2607, pruned_loss=0.04361, over 3322705.21 frames. ], batch size: 109, lr: 3.40e-03, grad_scale: 8.0 2023-05-01 01:14:38,084 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=195427.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 01:14:47,500 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.6071, 4.5780, 4.9154, 4.8912, 4.9547, 4.6038, 4.6429, 4.3789], device='cuda:1'), covar=tensor([0.0363, 0.0642, 0.0444, 0.0467, 0.0558, 0.0464, 0.0927, 0.0631], device='cuda:1'), in_proj_covar=tensor([0.0405, 0.0444, 0.0430, 0.0404, 0.0478, 0.0455, 0.0544, 0.0365], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:1') 2023-05-01 01:14:47,968 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=4.63 vs. limit=5.0 2023-05-01 01:14:57,675 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.7344, 3.8141, 2.4303, 4.3927, 2.9447, 4.3067, 2.4101, 3.0438], device='cuda:1'), covar=tensor([0.0278, 0.0376, 0.1513, 0.0278, 0.0769, 0.0505, 0.1508, 0.0771], device='cuda:1'), in_proj_covar=tensor([0.0172, 0.0181, 0.0197, 0.0166, 0.0179, 0.0222, 0.0205, 0.0181], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-05-01 01:15:11,782 INFO [train.py:904] (1/8) Epoch 20, batch 2600, loss[loss=0.1788, simple_loss=0.2559, pruned_loss=0.05086, over 16853.00 frames. ], tot_loss[loss=0.1732, simple_loss=0.26, pruned_loss=0.04321, over 3332530.59 frames. ], batch size: 109, lr: 3.40e-03, grad_scale: 8.0 2023-05-01 01:15:13,323 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=195453.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 01:15:27,771 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.9236, 2.8365, 2.5509, 2.7219, 3.1766, 2.9721, 3.5180, 3.4523], device='cuda:1'), covar=tensor([0.0119, 0.0389, 0.0485, 0.0441, 0.0270, 0.0385, 0.0250, 0.0255], device='cuda:1'), in_proj_covar=tensor([0.0210, 0.0242, 0.0228, 0.0231, 0.0242, 0.0241, 0.0242, 0.0235], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-01 01:15:42,987 INFO [optim.py:368] (1/8) 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 01:16:20,745 INFO [train.py:904] (1/8) Epoch 20, batch 2650, loss[loss=0.1956, simple_loss=0.2777, pruned_loss=0.05673, over 16766.00 frames. ], tot_loss[loss=0.1734, simple_loss=0.2605, pruned_loss=0.04314, over 3331725.07 frames. ], batch size: 102, lr: 3.40e-03, grad_scale: 8.0 2023-05-01 01:16:22,303 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=195503.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 01:17:28,730 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=195551.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 01:17:29,636 INFO [train.py:904] (1/8) Epoch 20, batch 2700, loss[loss=0.168, simple_loss=0.2684, pruned_loss=0.03384, over 17122.00 frames. ], tot_loss[loss=0.1725, simple_loss=0.2603, pruned_loss=0.04238, over 3338844.71 frames. ], batch size: 49, lr: 3.40e-03, grad_scale: 8.0 2023-05-01 01:18:00,661 INFO [optim.py:368] (1/8) 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 01:18:02,891 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.15 vs. limit=2.0 2023-05-01 01:18:25,811 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([5.4000, 5.3642, 5.2548, 4.7621, 4.9122, 5.3242, 5.2556, 4.9138], device='cuda:1'), covar=tensor([0.0653, 0.0525, 0.0266, 0.0348, 0.0975, 0.0409, 0.0304, 0.0788], device='cuda:1'), in_proj_covar=tensor([0.0302, 0.0433, 0.0350, 0.0345, 0.0362, 0.0402, 0.0242, 0.0419], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-05-01 01:18:39,440 INFO [train.py:904] (1/8) Epoch 20, batch 2750, loss[loss=0.1712, simple_loss=0.2647, pruned_loss=0.03886, over 17231.00 frames. ], tot_loss[loss=0.1727, simple_loss=0.261, pruned_loss=0.04224, over 3341313.90 frames. ], batch size: 45, lr: 3.40e-03, grad_scale: 8.0 2023-05-01 01:19:47,489 INFO [train.py:904] (1/8) Epoch 20, batch 2800, loss[loss=0.1841, simple_loss=0.2674, pruned_loss=0.05034, over 16739.00 frames. ], tot_loss[loss=0.1719, simple_loss=0.2601, pruned_loss=0.04182, over 3348602.13 frames. ], batch size: 89, lr: 3.40e-03, grad_scale: 8.0 2023-05-01 01:20:06,707 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=195667.0, num_to_drop=1, layers_to_drop={0} 2023-05-01 01:20:18,615 INFO [optim.py:368] (1/8) 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 01:20:33,259 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.75 vs. limit=2.0 2023-05-01 01:20:52,387 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=195700.0, num_to_drop=1, layers_to_drop={1} 2023-05-01 01:20:54,302 INFO [train.py:904] (1/8) Epoch 20, batch 2850, loss[loss=0.1699, simple_loss=0.244, pruned_loss=0.04795, over 16407.00 frames. ], tot_loss[loss=0.1714, simple_loss=0.2594, pruned_loss=0.04169, over 3346542.61 frames. ], batch size: 146, lr: 3.40e-03, grad_scale: 8.0 2023-05-01 01:21:13,263 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=195715.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 01:21:29,769 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=195727.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 01:21:56,480 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=195748.0, num_to_drop=1, layers_to_drop={0} 2023-05-01 01:21:56,509 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=195748.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 01:22:02,476 INFO [train.py:904] (1/8) Epoch 20, batch 2900, loss[loss=0.2111, simple_loss=0.2806, pruned_loss=0.0708, over 16946.00 frames. ], tot_loss[loss=0.1718, simple_loss=0.2585, pruned_loss=0.0426, over 3345419.37 frames. ], batch size: 109, lr: 3.40e-03, grad_scale: 8.0 2023-05-01 01:22:33,088 INFO [optim.py:368] (1/8) 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 01:22:33,331 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=195775.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 01:23:10,992 INFO [train.py:904] (1/8) Epoch 20, batch 2950, loss[loss=0.1428, simple_loss=0.233, pruned_loss=0.0263, over 17214.00 frames. ], tot_loss[loss=0.172, simple_loss=0.2581, pruned_loss=0.043, over 3341090.57 frames. ], batch size: 45, lr: 3.40e-03, grad_scale: 8.0 2023-05-01 01:24:18,942 INFO [train.py:904] (1/8) Epoch 20, batch 3000, loss[loss=0.1745, simple_loss=0.2484, pruned_loss=0.05029, over 16899.00 frames. ], tot_loss[loss=0.1734, simple_loss=0.259, pruned_loss=0.04387, over 3324895.84 frames. ], batch size: 109, lr: 3.40e-03, grad_scale: 8.0 2023-05-01 01:24:18,942 INFO [train.py:929] (1/8) Computing validation loss 2023-05-01 01:24:27,132 INFO [train.py:938] (1/8) Epoch 20, validation: loss=0.1354, simple_loss=0.2409, pruned_loss=0.01492, over 944034.00 frames. 2023-05-01 01:24:27,133 INFO [train.py:939] (1/8) Maximum memory allocated so far is 18145MB 2023-05-01 01:24:58,716 INFO [optim.py:368] (1/8) 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 01:25:21,737 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([1.8385, 2.7005, 2.6496, 1.9209, 2.6670, 2.8230, 2.6067, 1.8617], device='cuda:1'), covar=tensor([0.0429, 0.0114, 0.0079, 0.0370, 0.0132, 0.0117, 0.0121, 0.0414], device='cuda:1'), in_proj_covar=tensor([0.0137, 0.0083, 0.0083, 0.0135, 0.0097, 0.0109, 0.0094, 0.0130], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0002, 0.0004], device='cuda:1') 2023-05-01 01:25:38,247 INFO [train.py:904] (1/8) Epoch 20, batch 3050, loss[loss=0.1625, simple_loss=0.2616, pruned_loss=0.03169, over 16997.00 frames. ], tot_loss[loss=0.1733, simple_loss=0.2588, pruned_loss=0.0439, over 3323196.17 frames. ], batch size: 53, lr: 3.40e-03, grad_scale: 8.0 2023-05-01 01:26:46,790 INFO [train.py:904] (1/8) Epoch 20, batch 3100, loss[loss=0.1532, simple_loss=0.2369, pruned_loss=0.03478, over 16763.00 frames. ], tot_loss[loss=0.1725, simple_loss=0.2578, pruned_loss=0.04361, over 3320691.65 frames. ], batch size: 83, lr: 3.40e-03, grad_scale: 8.0 2023-05-01 01:27:00,347 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.9761, 5.3862, 5.6178, 5.3289, 5.3448, 6.0065, 5.4892, 5.2059], device='cuda:1'), covar=tensor([0.1152, 0.2168, 0.2212, 0.2038, 0.3042, 0.1110, 0.1653, 0.2608], device='cuda:1'), in_proj_covar=tensor([0.0418, 0.0608, 0.0665, 0.0507, 0.0674, 0.0700, 0.0517, 0.0676], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-05-01 01:27:04,197 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([1.8496, 3.2258, 3.0464, 5.1350, 4.2426, 4.5501, 1.7156, 3.2747], device='cuda:1'), covar=tensor([0.1308, 0.0686, 0.1031, 0.0185, 0.0187, 0.0364, 0.1566, 0.0739], device='cuda:1'), in_proj_covar=tensor([0.0165, 0.0174, 0.0193, 0.0188, 0.0205, 0.0216, 0.0199, 0.0191], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-05-01 01:27:16,965 INFO [optim.py:368] (1/8) 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 01:27:55,506 INFO [train.py:904] (1/8) Epoch 20, batch 3150, loss[loss=0.1848, simple_loss=0.2738, pruned_loss=0.04792, over 15520.00 frames. ], tot_loss[loss=0.1717, simple_loss=0.2571, pruned_loss=0.04318, over 3327270.72 frames. ], batch size: 190, lr: 3.40e-03, grad_scale: 8.0 2023-05-01 01:28:57,095 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=196048.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 01:29:03,226 INFO [train.py:904] (1/8) Epoch 20, batch 3200, loss[loss=0.1511, simple_loss=0.245, pruned_loss=0.02864, over 17030.00 frames. ], tot_loss[loss=0.1707, simple_loss=0.2563, pruned_loss=0.04249, over 3331692.34 frames. ], batch size: 50, lr: 3.40e-03, grad_scale: 8.0 2023-05-01 01:29:35,518 INFO [optim.py:368] (1/8) Clipping_scale=2.0, grad-norm quartiles 1.399e+02 2.205e+02 2.543e+02 3.042e+02 4.560e+02, threshold=5.087e+02, percent-clipped=0.0 2023-05-01 01:30:04,481 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=196096.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 01:30:07,430 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.35 vs. limit=2.0 2023-05-01 01:30:11,570 INFO [train.py:904] (1/8) Epoch 20, batch 3250, loss[loss=0.2071, simple_loss=0.2868, pruned_loss=0.0637, over 16738.00 frames. ], tot_loss[loss=0.1705, simple_loss=0.2563, pruned_loss=0.04235, over 3325908.94 frames. ], batch size: 83, lr: 3.40e-03, grad_scale: 8.0 2023-05-01 01:30:23,013 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=4.65 vs. limit=5.0 2023-05-01 01:31:19,979 INFO [train.py:904] (1/8) Epoch 20, batch 3300, loss[loss=0.2119, simple_loss=0.2842, pruned_loss=0.06981, over 16870.00 frames. ], tot_loss[loss=0.1711, simple_loss=0.2569, pruned_loss=0.04271, over 3329516.97 frames. ], batch size: 109, lr: 3.40e-03, grad_scale: 8.0 2023-05-01 01:31:52,355 INFO [optim.py:368] (1/8) 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 01:31:56,803 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.38 vs. limit=2.0 2023-05-01 01:32:28,250 INFO [train.py:904] (1/8) Epoch 20, batch 3350, loss[loss=0.1589, simple_loss=0.2528, pruned_loss=0.03252, over 17269.00 frames. ], tot_loss[loss=0.1723, simple_loss=0.2582, pruned_loss=0.04313, over 3311109.40 frames. ], batch size: 52, lr: 3.40e-03, grad_scale: 8.0 2023-05-01 01:32:48,187 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.8815, 1.8705, 2.4480, 2.8368, 2.7171, 3.0079, 2.1339, 3.0785], device='cuda:1'), covar=tensor([0.0218, 0.0501, 0.0357, 0.0285, 0.0315, 0.0235, 0.0480, 0.0173], device='cuda:1'), in_proj_covar=tensor([0.0187, 0.0194, 0.0178, 0.0183, 0.0196, 0.0154, 0.0195, 0.0149], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-01 01:33:24,851 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.4329, 4.2207, 4.5507, 2.4053, 4.8422, 4.8171, 3.5537, 3.7845], device='cuda:1'), covar=tensor([0.0609, 0.0201, 0.0188, 0.1088, 0.0061, 0.0156, 0.0360, 0.0369], device='cuda:1'), in_proj_covar=tensor([0.0147, 0.0108, 0.0097, 0.0138, 0.0079, 0.0125, 0.0126, 0.0130], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0004, 0.0004], device='cuda:1') 2023-05-01 01:33:35,778 INFO [train.py:904] (1/8) Epoch 20, batch 3400, loss[loss=0.1574, simple_loss=0.2464, pruned_loss=0.03419, over 17196.00 frames. ], tot_loss[loss=0.1711, simple_loss=0.2574, pruned_loss=0.04235, over 3310221.58 frames. ], batch size: 44, lr: 3.40e-03, grad_scale: 8.0 2023-05-01 01:34:06,838 INFO [optim.py:368] (1/8) 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 01:34:44,339 INFO [train.py:904] (1/8) Epoch 20, batch 3450, loss[loss=0.1659, simple_loss=0.2414, pruned_loss=0.0452, over 16336.00 frames. ], tot_loss[loss=0.1699, simple_loss=0.2562, pruned_loss=0.0418, over 3314579.15 frames. ], batch size: 165, lr: 3.40e-03, grad_scale: 8.0 2023-05-01 01:35:10,338 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=196321.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 01:35:50,475 INFO [train.py:904] (1/8) Epoch 20, batch 3500, loss[loss=0.1771, simple_loss=0.2529, pruned_loss=0.05066, over 16876.00 frames. ], tot_loss[loss=0.1694, simple_loss=0.255, pruned_loss=0.04192, over 3313879.38 frames. ], batch size: 116, lr: 3.40e-03, grad_scale: 8.0 2023-05-01 01:36:03,243 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([5.7323, 6.1113, 5.7906, 5.9014, 5.5180, 5.4484, 5.4481, 6.2321], device='cuda:1'), covar=tensor([0.1362, 0.0967, 0.1160, 0.0932, 0.0899, 0.0700, 0.1271, 0.0958], device='cuda:1'), in_proj_covar=tensor([0.0676, 0.0835, 0.0680, 0.0623, 0.0522, 0.0525, 0.0697, 0.0638], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:1') 2023-05-01 01:36:03,376 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=196360.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 01:36:03,531 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.17 vs. limit=2.0 2023-05-01 01:36:23,663 INFO [optim.py:368] (1/8) 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 01:36:35,799 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=196382.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 01:36:44,244 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([5.2789, 5.2504, 5.0156, 4.5361, 5.1083, 2.0096, 4.8430, 4.8908], device='cuda:1'), covar=tensor([0.0079, 0.0073, 0.0196, 0.0357, 0.0102, 0.2682, 0.0138, 0.0201], device='cuda:1'), in_proj_covar=tensor([0.0165, 0.0153, 0.0199, 0.0179, 0.0176, 0.0207, 0.0189, 0.0175], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-01 01:37:01,668 INFO [train.py:904] (1/8) Epoch 20, batch 3550, loss[loss=0.1572, simple_loss=0.2455, pruned_loss=0.03442, over 16821.00 frames. ], tot_loss[loss=0.1684, simple_loss=0.2543, pruned_loss=0.04131, over 3309564.61 frames. ], batch size: 42, lr: 3.40e-03, grad_scale: 8.0 2023-05-01 01:37:27,882 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=196421.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 01:37:47,603 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.6320, 3.3599, 3.6847, 1.9912, 3.8321, 3.8611, 3.0498, 2.9259], device='cuda:1'), covar=tensor([0.0687, 0.0247, 0.0210, 0.1068, 0.0095, 0.0191, 0.0412, 0.0435], device='cuda:1'), in_proj_covar=tensor([0.0147, 0.0108, 0.0098, 0.0138, 0.0079, 0.0125, 0.0126, 0.0131], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0004, 0.0004], device='cuda:1') 2023-05-01 01:38:10,222 INFO [train.py:904] (1/8) Epoch 20, batch 3600, loss[loss=0.1497, simple_loss=0.2397, pruned_loss=0.02983, over 17200.00 frames. ], tot_loss[loss=0.1671, simple_loss=0.2533, pruned_loss=0.04049, over 3322872.60 frames. ], batch size: 46, lr: 3.39e-03, grad_scale: 8.0 2023-05-01 01:38:41,979 INFO [optim.py:368] (1/8) 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] (1/8) Epoch 20, batch 3650, loss[loss=0.1488, simple_loss=0.2231, pruned_loss=0.03727, over 16853.00 frames. ], tot_loss[loss=0.1668, simple_loss=0.2516, pruned_loss=0.04101, over 3311559.72 frames. ], batch size: 96, lr: 3.39e-03, grad_scale: 8.0 2023-05-01 01:39:50,807 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([5.2192, 5.6533, 5.8462, 5.5466, 5.6511, 6.2202, 5.7791, 5.4404], device='cuda:1'), covar=tensor([0.0817, 0.2219, 0.2041, 0.1849, 0.2310, 0.0881, 0.1252, 0.2247], device='cuda:1'), in_proj_covar=tensor([0.0415, 0.0603, 0.0662, 0.0503, 0.0666, 0.0697, 0.0513, 0.0671], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-05-01 01:40:16,351 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.0675, 4.0280, 4.0296, 3.3916, 3.9884, 1.8084, 3.7973, 3.4296], device='cuda:1'), covar=tensor([0.0149, 0.0119, 0.0187, 0.0278, 0.0090, 0.2781, 0.0131, 0.0254], device='cuda:1'), in_proj_covar=tensor([0.0163, 0.0152, 0.0198, 0.0178, 0.0175, 0.0206, 0.0187, 0.0174], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-01 01:40:23,567 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.9660, 2.1251, 2.5473, 2.8329, 2.8039, 2.9376, 2.1123, 3.1120], device='cuda:1'), covar=tensor([0.0173, 0.0415, 0.0316, 0.0251, 0.0284, 0.0266, 0.0498, 0.0153], device='cuda:1'), in_proj_covar=tensor([0.0186, 0.0192, 0.0177, 0.0182, 0.0195, 0.0153, 0.0194, 0.0148], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-01 01:40:32,643 INFO [train.py:904] (1/8) Epoch 20, batch 3700, loss[loss=0.1711, simple_loss=0.2423, pruned_loss=0.04997, over 16757.00 frames. ], tot_loss[loss=0.1684, simple_loss=0.2508, pruned_loss=0.04296, over 3287665.32 frames. ], batch size: 124, lr: 3.39e-03, grad_scale: 16.0 2023-05-01 01:41:07,135 INFO [optim.py:368] (1/8) 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] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=196578.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 01:41:24,712 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.31 vs. limit=2.0 2023-05-01 01:41:47,078 INFO [train.py:904] (1/8) Epoch 20, batch 3750, loss[loss=0.168, simple_loss=0.2552, pruned_loss=0.04036, over 16428.00 frames. ], tot_loss[loss=0.17, simple_loss=0.2516, pruned_loss=0.04422, over 3273755.35 frames. ], batch size: 68, lr: 3.39e-03, grad_scale: 16.0 2023-05-01 01:42:07,452 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.3176, 4.2768, 4.3002, 3.7573, 4.2953, 1.8413, 4.0777, 3.8244], device='cuda:1'), covar=tensor([0.0129, 0.0110, 0.0169, 0.0265, 0.0096, 0.2592, 0.0136, 0.0226], device='cuda:1'), in_proj_covar=tensor([0.0164, 0.0152, 0.0198, 0.0178, 0.0175, 0.0206, 0.0188, 0.0174], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-01 01:42:21,767 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.24 vs. limit=2.0 2023-05-01 01:42:38,985 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=196639.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 01:42:58,321 INFO [train.py:904] (1/8) Epoch 20, batch 3800, loss[loss=0.1717, simple_loss=0.2629, pruned_loss=0.04026, over 16802.00 frames. ], tot_loss[loss=0.1718, simple_loss=0.2526, pruned_loss=0.04554, over 3274306.76 frames. ], batch size: 102, lr: 3.39e-03, grad_scale: 16.0 2023-05-01 01:43:31,147 INFO [optim.py:368] (1/8) 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 01:43:35,323 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=196677.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 01:44:10,772 INFO [train.py:904] (1/8) Epoch 20, batch 3850, loss[loss=0.1783, simple_loss=0.2466, pruned_loss=0.05496, over 16871.00 frames. ], tot_loss[loss=0.1729, simple_loss=0.2531, pruned_loss=0.04631, over 3258455.57 frames. ], batch size: 90, lr: 3.39e-03, grad_scale: 16.0 2023-05-01 01:44:32,370 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=196716.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 01:44:44,892 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.0030, 2.1091, 2.6551, 2.9471, 2.9426, 3.5355, 2.1147, 3.3318], device='cuda:1'), covar=tensor([0.0207, 0.0498, 0.0288, 0.0331, 0.0293, 0.0128, 0.0520, 0.0120], device='cuda:1'), in_proj_covar=tensor([0.0185, 0.0192, 0.0178, 0.0182, 0.0195, 0.0153, 0.0194, 0.0147], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-01 01:45:00,213 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.9443, 4.2303, 4.0462, 4.1091, 3.7617, 3.7946, 3.8541, 4.2345], device='cuda:1'), covar=tensor([0.1294, 0.0996, 0.1124, 0.0803, 0.0861, 0.1824, 0.1037, 0.0961], device='cuda:1'), in_proj_covar=tensor([0.0672, 0.0826, 0.0675, 0.0619, 0.0520, 0.0524, 0.0694, 0.0638], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:1') 2023-05-01 01:45:20,019 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.2905, 4.2332, 4.2635, 3.6556, 4.2838, 1.6870, 4.0343, 3.7610], device='cuda:1'), covar=tensor([0.0132, 0.0108, 0.0173, 0.0332, 0.0100, 0.2783, 0.0145, 0.0275], device='cuda:1'), in_proj_covar=tensor([0.0163, 0.0151, 0.0198, 0.0178, 0.0175, 0.0205, 0.0187, 0.0174], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-01 01:45:24,168 INFO [train.py:904] (1/8) Epoch 20, batch 3900, loss[loss=0.1752, simple_loss=0.2568, pruned_loss=0.04678, over 16736.00 frames. ], tot_loss[loss=0.1729, simple_loss=0.2524, pruned_loss=0.04667, over 3268306.67 frames. ], batch size: 57, lr: 3.39e-03, grad_scale: 16.0 2023-05-01 01:45:57,454 INFO [optim.py:368] (1/8) 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 01:46:04,883 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=196779.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 01:46:36,466 INFO [train.py:904] (1/8) Epoch 20, batch 3950, loss[loss=0.1728, simple_loss=0.2629, pruned_loss=0.04137, over 17124.00 frames. ], tot_loss[loss=0.1732, simple_loss=0.2521, pruned_loss=0.04715, over 3279473.52 frames. ], batch size: 49, lr: 3.39e-03, grad_scale: 16.0 2023-05-01 01:46:56,959 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.2718, 3.6249, 3.7824, 2.6039, 3.4471, 3.8371, 3.5459, 2.0386], device='cuda:1'), covar=tensor([0.0506, 0.0113, 0.0055, 0.0360, 0.0094, 0.0102, 0.0093, 0.0478], device='cuda:1'), in_proj_covar=tensor([0.0135, 0.0082, 0.0081, 0.0132, 0.0096, 0.0107, 0.0093, 0.0128], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-05-01 01:47:31,558 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=196840.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 01:47:40,589 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.26 vs. limit=2.0 2023-05-01 01:47:48,960 INFO [train.py:904] (1/8) Epoch 20, batch 4000, loss[loss=0.1659, simple_loss=0.246, pruned_loss=0.04292, over 16735.00 frames. ], tot_loss[loss=0.1738, simple_loss=0.2524, pruned_loss=0.0476, over 3284785.91 frames. ], batch size: 89, lr: 3.39e-03, grad_scale: 16.0 2023-05-01 01:47:52,342 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=196854.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 01:47:55,379 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([1.8150, 2.7548, 2.5974, 4.5729, 3.5498, 4.1219, 1.6748, 2.9355], device='cuda:1'), covar=tensor([0.1262, 0.0749, 0.1198, 0.0153, 0.0339, 0.0354, 0.1541, 0.0902], device='cuda:1'), in_proj_covar=tensor([0.0164, 0.0172, 0.0191, 0.0186, 0.0205, 0.0214, 0.0198, 0.0190], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-05-01 01:48:14,270 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.3222, 2.2848, 2.2757, 4.0443, 2.2560, 2.6456, 2.3322, 2.4745], device='cuda:1'), covar=tensor([0.1275, 0.3469, 0.2893, 0.0537, 0.3827, 0.2257, 0.3543, 0.3130], device='cuda:1'), in_proj_covar=tensor([0.0402, 0.0445, 0.0367, 0.0329, 0.0436, 0.0514, 0.0414, 0.0522], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-01 01:48:21,943 INFO [optim.py:368] (1/8) 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 01:48:31,791 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 20, batch 4050, loss[loss=0.1768, simple_loss=0.2657, pruned_loss=0.04395, over 16761.00 frames. ], tot_loss[loss=0.1737, simple_loss=0.2533, pruned_loss=0.04699, over 3293117.07 frames. ], batch size: 83, lr: 3.39e-03, grad_scale: 16.0 2023-05-01 01:49:13,373 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.7354, 3.7287, 3.9143, 3.6508, 3.8093, 4.2535, 3.8662, 3.5279], device='cuda:1'), covar=tensor([0.2623, 0.2420, 0.2095, 0.2561, 0.2740, 0.1976, 0.1672, 0.2680], device='cuda:1'), in_proj_covar=tensor([0.0414, 0.0600, 0.0659, 0.0503, 0.0665, 0.0695, 0.0514, 0.0671], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-05-01 01:49:19,923 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=196915.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 01:49:47,740 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=196934.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 01:49:57,969 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 20, batch 4100, loss[loss=0.1831, simple_loss=0.2733, pruned_loss=0.04646, over 16631.00 frames. ], tot_loss[loss=0.1742, simple_loss=0.2555, pruned_loss=0.0465, over 3280393.73 frames. ], batch size: 57, lr: 3.39e-03, grad_scale: 16.0 2023-05-01 01:50:44,595 INFO [optim.py:368] (1/8) 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 01:50:47,906 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=196977.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 01:51:23,808 INFO [train.py:904] (1/8) Epoch 20, batch 4150, loss[loss=0.2599, simple_loss=0.3248, pruned_loss=0.09754, over 11447.00 frames. ], tot_loss[loss=0.1804, simple_loss=0.2627, pruned_loss=0.049, over 3241852.88 frames. ], batch size: 248, lr: 3.39e-03, grad_scale: 16.0 2023-05-01 01:51:45,587 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=197016.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 01:51:50,989 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.36 vs. limit=2.0 2023-05-01 01:52:00,049 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 20, batch 4200, loss[loss=0.2079, simple_loss=0.3012, pruned_loss=0.05723, over 16638.00 frames. ], tot_loss[loss=0.185, simple_loss=0.2693, pruned_loss=0.05037, over 3223075.16 frames. ], batch size: 57, lr: 3.39e-03, grad_scale: 16.0 2023-05-01 01:52:58,764 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=197064.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 01:53:12,969 INFO [optim.py:368] (1/8) 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 01:53:38,224 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([1.9828, 3.3240, 3.3026, 2.2443, 3.1069, 3.3417, 3.1573, 1.9836], device='cuda:1'), covar=tensor([0.0550, 0.0063, 0.0069, 0.0379, 0.0102, 0.0116, 0.0092, 0.0437], device='cuda:1'), in_proj_covar=tensor([0.0136, 0.0082, 0.0082, 0.0133, 0.0096, 0.0108, 0.0093, 0.0128], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-05-01 01:53:51,290 INFO [train.py:904] (1/8) Epoch 20, batch 4250, loss[loss=0.1732, simple_loss=0.2748, pruned_loss=0.03581, over 16858.00 frames. ], tot_loss[loss=0.1869, simple_loss=0.273, pruned_loss=0.05037, over 3200750.35 frames. ], batch size: 96, lr: 3.39e-03, grad_scale: 16.0 2023-05-01 01:54:31,634 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.6193, 4.4804, 4.3030, 3.1272, 3.7020, 4.3583, 3.7919, 2.3819], device='cuda:1'), covar=tensor([0.0472, 0.0026, 0.0046, 0.0312, 0.0092, 0.0089, 0.0088, 0.0423], device='cuda:1'), in_proj_covar=tensor([0.0135, 0.0082, 0.0081, 0.0132, 0.0096, 0.0107, 0.0093, 0.0127], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-05-01 01:54:39,852 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 20, batch 4300, loss[loss=0.215, simple_loss=0.297, pruned_loss=0.06652, over 16678.00 frames. ], tot_loss[loss=0.1865, simple_loss=0.274, pruned_loss=0.04948, over 3203987.34 frames. ], batch size: 57, lr: 3.39e-03, grad_scale: 16.0 2023-05-01 01:55:30,421 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.1167, 2.4714, 2.5829, 1.9052, 2.7297, 2.7994, 2.4044, 2.4204], device='cuda:1'), covar=tensor([0.0678, 0.0279, 0.0239, 0.0963, 0.0108, 0.0241, 0.0422, 0.0434], device='cuda:1'), in_proj_covar=tensor([0.0148, 0.0109, 0.0097, 0.0139, 0.0079, 0.0124, 0.0127, 0.0131], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0004, 0.0004], device='cuda:1') 2023-05-01 01:55:37,989 INFO [optim.py:368] (1/8) 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 01:56:09,660 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.3527, 3.2463, 2.6257, 2.1286, 2.1908, 2.2338, 3.3339, 2.9821], device='cuda:1'), covar=tensor([0.2935, 0.0749, 0.1793, 0.2442, 0.2421, 0.2087, 0.0529, 0.1247], device='cuda:1'), in_proj_covar=tensor([0.0322, 0.0268, 0.0302, 0.0306, 0.0297, 0.0253, 0.0292, 0.0333], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:1') 2023-05-01 01:56:18,918 INFO [train.py:904] (1/8) Epoch 20, batch 4350, loss[loss=0.1728, simple_loss=0.269, pruned_loss=0.0383, over 16849.00 frames. ], tot_loss[loss=0.1895, simple_loss=0.2773, pruned_loss=0.0508, over 3186644.11 frames. ], batch size: 96, lr: 3.39e-03, grad_scale: 16.0 2023-05-01 01:56:29,952 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=197210.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 01:56:52,712 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([5.1676, 5.2377, 5.5717, 5.5473, 5.6230, 5.2334, 5.1422, 4.9111], device='cuda:1'), covar=tensor([0.0275, 0.0367, 0.0278, 0.0341, 0.0445, 0.0308, 0.0963, 0.0432], device='cuda:1'), in_proj_covar=tensor([0.0402, 0.0444, 0.0427, 0.0400, 0.0474, 0.0453, 0.0542, 0.0362], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:1') 2023-05-01 01:56:55,340 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.7268, 4.8902, 4.6378, 4.3865, 4.1751, 4.7943, 4.5528, 4.3695], device='cuda:1'), covar=tensor([0.0640, 0.0382, 0.0326, 0.0288, 0.1024, 0.0404, 0.0404, 0.0604], device='cuda:1'), in_proj_covar=tensor([0.0293, 0.0421, 0.0341, 0.0336, 0.0352, 0.0389, 0.0234, 0.0407], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002], device='cuda:1') 2023-05-01 01:56:58,521 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([1.5168, 2.5327, 2.3686, 3.8721, 3.0124, 3.8130, 1.3596, 2.7137], device='cuda:1'), covar=tensor([0.1399, 0.0807, 0.1293, 0.0152, 0.0307, 0.0383, 0.1739, 0.0898], device='cuda:1'), in_proj_covar=tensor([0.0163, 0.0172, 0.0191, 0.0185, 0.0204, 0.0213, 0.0198, 0.0190], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-05-01 01:57:05,460 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=197234.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 01:57:09,109 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 20, batch 4400, loss[loss=0.1943, simple_loss=0.2776, pruned_loss=0.0555, over 17042.00 frames. ], tot_loss[loss=0.1915, simple_loss=0.2794, pruned_loss=0.05183, over 3201836.23 frames. ], batch size: 53, lr: 3.39e-03, grad_scale: 16.0 2023-05-01 01:58:04,302 INFO [optim.py:368] (1/8) 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 01:58:14,447 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=197282.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 01:58:42,794 INFO [train.py:904] (1/8) Epoch 20, batch 4450, loss[loss=0.2139, simple_loss=0.2895, pruned_loss=0.06917, over 11587.00 frames. ], tot_loss[loss=0.1945, simple_loss=0.2827, pruned_loss=0.0531, over 3201843.45 frames. ], batch size: 247, lr: 3.39e-03, grad_scale: 16.0 2023-05-01 01:59:16,217 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([5.5613, 5.5323, 5.4441, 5.1307, 5.1043, 5.5205, 5.3702, 5.1074], device='cuda:1'), covar=tensor([0.0515, 0.0252, 0.0208, 0.0219, 0.0823, 0.0247, 0.0214, 0.0516], device='cuda:1'), in_proj_covar=tensor([0.0291, 0.0417, 0.0338, 0.0334, 0.0349, 0.0385, 0.0232, 0.0404], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002], device='cuda:1') 2023-05-01 01:59:55,983 INFO [train.py:904] (1/8) Epoch 20, batch 4500, loss[loss=0.2063, simple_loss=0.2957, pruned_loss=0.05849, over 16727.00 frames. ], tot_loss[loss=0.1961, simple_loss=0.2835, pruned_loss=0.05432, over 3195662.17 frames. ], batch size: 124, lr: 3.39e-03, grad_scale: 8.0 2023-05-01 02:00:02,444 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.2755, 4.0088, 4.4593, 2.1712, 4.7573, 4.8233, 3.3706, 3.8642], device='cuda:1'), covar=tensor([0.0693, 0.0247, 0.0187, 0.1225, 0.0063, 0.0082, 0.0424, 0.0364], device='cuda:1'), in_proj_covar=tensor([0.0149, 0.0109, 0.0098, 0.0140, 0.0080, 0.0125, 0.0128, 0.0132], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0004, 0.0004], device='cuda:1') 2023-05-01 02:00:30,609 INFO [optim.py:368] (1/8) 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] (1/8) Epoch 20, batch 4550, loss[loss=0.2037, simple_loss=0.3015, pruned_loss=0.05298, over 16806.00 frames. ], tot_loss[loss=0.1973, simple_loss=0.2847, pruned_loss=0.05494, over 3219638.37 frames. ], batch size: 83, lr: 3.39e-03, grad_scale: 8.0 2023-05-01 02:01:27,163 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=197415.0, num_to_drop=1, layers_to_drop={1} 2023-05-01 02:01:56,157 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=197435.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 02:02:18,309 INFO [train.py:904] (1/8) Epoch 20, batch 4600, loss[loss=0.2092, simple_loss=0.2946, pruned_loss=0.06193, over 16482.00 frames. ], tot_loss[loss=0.1976, simple_loss=0.2855, pruned_loss=0.05488, over 3232624.40 frames. ], batch size: 68, lr: 3.39e-03, grad_scale: 8.0 2023-05-01 02:02:23,144 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.7956, 3.9360, 4.1262, 4.0847, 4.1096, 3.8613, 3.9121, 3.8461], device='cuda:1'), covar=tensor([0.0330, 0.0497, 0.0378, 0.0405, 0.0443, 0.0415, 0.0839, 0.0526], device='cuda:1'), in_proj_covar=tensor([0.0396, 0.0436, 0.0419, 0.0394, 0.0467, 0.0445, 0.0533, 0.0357], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:1') 2023-05-01 02:02:42,347 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.8015, 3.7951, 3.9087, 3.6454, 3.8087, 4.2500, 3.8907, 3.5525], device='cuda:1'), covar=tensor([0.2081, 0.2057, 0.1954, 0.2512, 0.2628, 0.1680, 0.1372, 0.2572], device='cuda:1'), in_proj_covar=tensor([0.0401, 0.0578, 0.0632, 0.0481, 0.0639, 0.0672, 0.0495, 0.0645], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-05-01 02:02:52,206 INFO [optim.py:368] (1/8) 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 02:02:52,802 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=197476.0, num_to_drop=1, layers_to_drop={1} 2023-05-01 02:03:02,410 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=197483.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 02:03:30,131 INFO [train.py:904] (1/8) Epoch 20, batch 4650, loss[loss=0.2179, simple_loss=0.2956, pruned_loss=0.07009, over 11610.00 frames. ], tot_loss[loss=0.1973, simple_loss=0.2844, pruned_loss=0.05512, over 3213193.99 frames. ], batch size: 246, lr: 3.39e-03, grad_scale: 8.0 2023-05-01 02:03:31,703 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.4146, 4.2935, 4.0785, 2.7655, 3.7413, 4.2173, 3.7499, 2.4113], device='cuda:1'), covar=tensor([0.0514, 0.0027, 0.0042, 0.0372, 0.0083, 0.0079, 0.0083, 0.0397], device='cuda:1'), in_proj_covar=tensor([0.0136, 0.0082, 0.0081, 0.0133, 0.0096, 0.0107, 0.0094, 0.0128], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-05-01 02:03:33,845 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.35 vs. limit=2.0 2023-05-01 02:03:41,281 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=197510.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 02:03:49,577 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=197516.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 02:04:14,810 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.8573, 4.8841, 4.7361, 4.3830, 4.4025, 4.8369, 4.5394, 4.4823], device='cuda:1'), covar=tensor([0.0480, 0.0388, 0.0225, 0.0257, 0.0770, 0.0328, 0.0413, 0.0528], device='cuda:1'), in_proj_covar=tensor([0.0287, 0.0411, 0.0333, 0.0330, 0.0345, 0.0379, 0.0230, 0.0397], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-01 02:04:20,838 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=197537.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 02:04:32,534 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=197545.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 02:04:42,155 INFO [train.py:904] (1/8) Epoch 20, batch 4700, loss[loss=0.1865, simple_loss=0.269, pruned_loss=0.05206, over 16828.00 frames. ], tot_loss[loss=0.1952, simple_loss=0.2821, pruned_loss=0.05413, over 3213108.86 frames. ], batch size: 39, lr: 3.39e-03, grad_scale: 8.0 2023-05-01 02:04:52,429 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=197558.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 02:04:52,637 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.1658, 2.0464, 1.7690, 1.7675, 2.2992, 1.9491, 1.9105, 2.3525], device='cuda:1'), covar=tensor([0.0167, 0.0411, 0.0510, 0.0493, 0.0245, 0.0355, 0.0207, 0.0280], device='cuda:1'), in_proj_covar=tensor([0.0204, 0.0233, 0.0225, 0.0226, 0.0235, 0.0234, 0.0236, 0.0231], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-01 02:05:18,235 INFO [optim.py:368] (1/8) 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 02:05:19,892 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=197577.0, num_to_drop=1, layers_to_drop={2} 2023-05-01 02:05:30,678 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=197585.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 02:05:41,283 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.77 vs. limit=2.0 2023-05-01 02:05:55,396 INFO [train.py:904] (1/8) Epoch 20, batch 4750, loss[loss=0.1928, simple_loss=0.2804, pruned_loss=0.05257, over 15460.00 frames. ], tot_loss[loss=0.1905, simple_loss=0.2776, pruned_loss=0.05168, over 3219762.67 frames. ], batch size: 190, lr: 3.38e-03, grad_scale: 8.0 2023-05-01 02:06:01,868 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=197606.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 02:06:03,428 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.46 vs. limit=2.0 2023-05-01 02:07:08,961 INFO [train.py:904] (1/8) Epoch 20, batch 4800, loss[loss=0.1747, simple_loss=0.2638, pruned_loss=0.04281, over 16918.00 frames. ], tot_loss[loss=0.1877, simple_loss=0.2749, pruned_loss=0.0502, over 3196138.95 frames. ], batch size: 109, lr: 3.38e-03, grad_scale: 8.0 2023-05-01 02:07:45,097 INFO [optim.py:368] (1/8) 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] (1/8) Epoch 20, batch 4850, loss[loss=0.2035, simple_loss=0.2867, pruned_loss=0.06012, over 12095.00 frames. ], tot_loss[loss=0.1874, simple_loss=0.2755, pruned_loss=0.04961, over 3180304.84 frames. ], batch size: 247, lr: 3.38e-03, grad_scale: 8.0 2023-05-01 02:08:32,233 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.7512, 4.8521, 4.6419, 4.2872, 4.2683, 4.7658, 4.5501, 4.4068], device='cuda:1'), covar=tensor([0.0659, 0.0419, 0.0295, 0.0310, 0.1060, 0.0441, 0.0430, 0.0665], device='cuda:1'), in_proj_covar=tensor([0.0287, 0.0411, 0.0333, 0.0329, 0.0344, 0.0380, 0.0229, 0.0398], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-01 02:08:40,862 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([5.2982, 5.3975, 5.7182, 5.6842, 5.7402, 5.3676, 5.3235, 4.9731], device='cuda:1'), covar=tensor([0.0245, 0.0347, 0.0265, 0.0292, 0.0433, 0.0264, 0.0904, 0.0385], device='cuda:1'), in_proj_covar=tensor([0.0393, 0.0433, 0.0416, 0.0391, 0.0465, 0.0442, 0.0531, 0.0353], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:1') 2023-05-01 02:09:40,078 INFO [train.py:904] (1/8) Epoch 20, batch 4900, loss[loss=0.1755, simple_loss=0.2681, pruned_loss=0.04146, over 16737.00 frames. ], tot_loss[loss=0.1849, simple_loss=0.2741, pruned_loss=0.04788, over 3178394.62 frames. ], batch size: 76, lr: 3.38e-03, grad_scale: 8.0 2023-05-01 02:10:08,660 INFO [zipformer.py:625] (1/8) 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] (1/8) 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 02:10:52,823 INFO [train.py:904] (1/8) Epoch 20, batch 4950, loss[loss=0.1938, simple_loss=0.2807, pruned_loss=0.05344, over 11913.00 frames. ], tot_loss[loss=0.1842, simple_loss=0.2737, pruned_loss=0.04732, over 3166279.04 frames. ], batch size: 246, lr: 3.38e-03, grad_scale: 8.0 2023-05-01 02:12:04,536 INFO [train.py:904] (1/8) Epoch 20, batch 5000, loss[loss=0.1788, simple_loss=0.2696, pruned_loss=0.04401, over 17135.00 frames. ], tot_loss[loss=0.185, simple_loss=0.2754, pruned_loss=0.04727, over 3181235.83 frames. ], batch size: 46, lr: 3.38e-03, grad_scale: 8.0 2023-05-01 02:12:32,781 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=197872.0, num_to_drop=1, layers_to_drop={2} 2023-05-01 02:12:38,859 INFO [optim.py:368] (1/8) 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 02:13:01,261 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=197892.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 02:13:13,655 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 20, batch 5050, loss[loss=0.1756, simple_loss=0.2728, pruned_loss=0.03926, over 16716.00 frames. ], tot_loss[loss=0.1853, simple_loss=0.2758, pruned_loss=0.04737, over 3187652.91 frames. ], batch size: 83, lr: 3.38e-03, grad_scale: 8.0 2023-05-01 02:13:25,513 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.7625, 2.5686, 2.0134, 2.4584, 3.0833, 2.7638, 3.2521, 3.3268], device='cuda:1'), covar=tensor([0.0092, 0.0445, 0.0629, 0.0483, 0.0249, 0.0382, 0.0220, 0.0279], device='cuda:1'), in_proj_covar=tensor([0.0201, 0.0230, 0.0223, 0.0223, 0.0233, 0.0232, 0.0233, 0.0228], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-01 02:13:54,799 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.4820, 3.4285, 2.7154, 2.1622, 2.4201, 2.3366, 3.7089, 3.1966], device='cuda:1'), covar=tensor([0.2793, 0.0833, 0.1766, 0.2616, 0.2302, 0.1902, 0.0499, 0.1249], device='cuda:1'), in_proj_covar=tensor([0.0324, 0.0269, 0.0302, 0.0307, 0.0296, 0.0253, 0.0293, 0.0333], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:1') 2023-05-01 02:14:25,041 INFO [train.py:904] (1/8) Epoch 20, batch 5100, loss[loss=0.1818, simple_loss=0.2712, pruned_loss=0.04624, over 16825.00 frames. ], tot_loss[loss=0.1838, simple_loss=0.2742, pruned_loss=0.04671, over 3191837.10 frames. ], batch size: 116, lr: 3.38e-03, grad_scale: 8.0 2023-05-01 02:14:27,234 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=197953.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 02:15:00,611 INFO [optim.py:368] (1/8) 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 02:15:41,439 INFO [train.py:904] (1/8) Epoch 20, batch 5150, loss[loss=0.2076, simple_loss=0.2897, pruned_loss=0.06274, over 11973.00 frames. ], tot_loss[loss=0.1829, simple_loss=0.2742, pruned_loss=0.04586, over 3189046.98 frames. ], batch size: 248, lr: 3.38e-03, grad_scale: 8.0 2023-05-01 02:16:36,160 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.5160, 4.5794, 4.4189, 4.0944, 4.0678, 4.5301, 4.3394, 4.2000], device='cuda:1'), covar=tensor([0.0680, 0.0565, 0.0340, 0.0336, 0.1034, 0.0514, 0.0438, 0.0711], device='cuda:1'), in_proj_covar=tensor([0.0287, 0.0413, 0.0334, 0.0330, 0.0345, 0.0383, 0.0230, 0.0400], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002], device='cuda:1') 2023-05-01 02:16:52,361 INFO [train.py:904] (1/8) Epoch 20, batch 5200, loss[loss=0.1658, simple_loss=0.2546, pruned_loss=0.03848, over 16531.00 frames. ], tot_loss[loss=0.1816, simple_loss=0.2725, pruned_loss=0.04534, over 3182382.60 frames. ], batch size: 75, lr: 3.38e-03, grad_scale: 8.0 2023-05-01 02:17:19,638 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=198071.0, num_to_drop=1, layers_to_drop={0} 2023-05-01 02:17:27,747 INFO [optim.py:368] (1/8) 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 02:17:34,742 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 20, batch 5250, loss[loss=0.1737, simple_loss=0.2655, pruned_loss=0.04094, over 16784.00 frames. ], tot_loss[loss=0.1801, simple_loss=0.2702, pruned_loss=0.04504, over 3182815.02 frames. ], batch size: 124, lr: 3.38e-03, grad_scale: 8.0 2023-05-01 02:18:28,862 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=198119.0, num_to_drop=1, layers_to_drop={0} 2023-05-01 02:18:32,451 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.8143, 3.7361, 3.8821, 3.9955, 4.0859, 3.7067, 4.0406, 4.1191], device='cuda:1'), covar=tensor([0.1526, 0.0990, 0.1210, 0.0633, 0.0512, 0.1675, 0.0677, 0.0621], device='cuda:1'), in_proj_covar=tensor([0.0621, 0.0767, 0.0894, 0.0783, 0.0585, 0.0614, 0.0628, 0.0726], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-01 02:19:01,624 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 20, batch 5300, loss[loss=0.1652, simple_loss=0.2531, pruned_loss=0.03869, over 16719.00 frames. ], tot_loss[loss=0.1775, simple_loss=0.267, pruned_loss=0.04397, over 3188287.86 frames. ], batch size: 76, lr: 3.38e-03, grad_scale: 8.0 2023-05-01 02:19:31,939 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.5186, 4.4551, 4.3342, 2.9463, 3.8106, 4.4018, 3.7875, 2.4340], device='cuda:1'), covar=tensor([0.0533, 0.0033, 0.0035, 0.0347, 0.0093, 0.0068, 0.0089, 0.0451], device='cuda:1'), in_proj_covar=tensor([0.0135, 0.0081, 0.0080, 0.0132, 0.0095, 0.0106, 0.0092, 0.0127], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-05-01 02:19:44,941 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=198172.0, num_to_drop=1, layers_to_drop={0} 2023-05-01 02:19:49,803 INFO [optim.py:368] (1/8) 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 02:20:26,951 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=198201.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 02:20:28,366 INFO [train.py:904] (1/8) Epoch 20, batch 5350, loss[loss=0.1835, simple_loss=0.2665, pruned_loss=0.0503, over 12375.00 frames. ], tot_loss[loss=0.1759, simple_loss=0.2654, pruned_loss=0.04323, over 3192997.92 frames. ], batch size: 247, lr: 3.38e-03, grad_scale: 8.0 2023-05-01 02:20:54,789 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=198220.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 02:21:35,326 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=198248.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 02:21:36,562 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=198249.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 02:21:40,980 INFO [train.py:904] (1/8) Epoch 20, batch 5400, loss[loss=0.1745, simple_loss=0.2684, pruned_loss=0.04032, over 17102.00 frames. ], tot_loss[loss=0.1779, simple_loss=0.2679, pruned_loss=0.04395, over 3182965.85 frames. ], batch size: 47, lr: 3.38e-03, grad_scale: 8.0 2023-05-01 02:21:57,709 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.2659, 4.3719, 4.4993, 4.2357, 4.3240, 4.8437, 4.3758, 4.0743], device='cuda:1'), covar=tensor([0.1681, 0.1830, 0.1647, 0.2055, 0.2537, 0.1037, 0.1387, 0.2625], device='cuda:1'), in_proj_covar=tensor([0.0398, 0.0570, 0.0622, 0.0474, 0.0633, 0.0662, 0.0490, 0.0639], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-05-01 02:22:15,985 INFO [optim.py:368] (1/8) 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] (1/8) Epoch 20, batch 5450, loss[loss=0.212, simple_loss=0.3058, pruned_loss=0.05911, over 16415.00 frames. ], tot_loss[loss=0.181, simple_loss=0.2712, pruned_loss=0.0454, over 3181478.65 frames. ], batch size: 146, lr: 3.38e-03, grad_scale: 8.0 2023-05-01 02:24:14,758 INFO [train.py:904] (1/8) Epoch 20, batch 5500, loss[loss=0.2009, simple_loss=0.2955, pruned_loss=0.05316, over 16303.00 frames. ], tot_loss[loss=0.1882, simple_loss=0.278, pruned_loss=0.04914, over 3160872.35 frames. ], batch size: 165, lr: 3.38e-03, grad_scale: 8.0 2023-05-01 02:24:17,158 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.5189, 3.5760, 3.3217, 3.0145, 3.1820, 3.4743, 3.3376, 3.2956], device='cuda:1'), covar=tensor([0.0608, 0.0686, 0.0262, 0.0271, 0.0528, 0.0484, 0.1248, 0.0468], device='cuda:1'), in_proj_covar=tensor([0.0290, 0.0421, 0.0339, 0.0334, 0.0350, 0.0389, 0.0233, 0.0405], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002], device='cuda:1') 2023-05-01 02:24:51,693 INFO [optim.py:368] (1/8) Clipping_scale=2.0, grad-norm quartiles 1.837e+02 2.873e+02 3.524e+02 4.449e+02 7.452e+02, threshold=7.049e+02, percent-clipped=24.0 2023-05-01 02:25:34,176 INFO [train.py:904] (1/8) Epoch 20, batch 5550, loss[loss=0.2796, simple_loss=0.3376, pruned_loss=0.1108, over 11067.00 frames. ], tot_loss[loss=0.197, simple_loss=0.2852, pruned_loss=0.05445, over 3144613.89 frames. ], batch size: 246, lr: 3.38e-03, grad_scale: 8.0 2023-05-01 02:26:01,481 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.0818, 2.4008, 2.4604, 2.6832, 1.9791, 3.1335, 1.9556, 2.7073], device='cuda:1'), covar=tensor([0.1072, 0.0540, 0.0995, 0.0208, 0.0151, 0.0417, 0.1359, 0.0687], device='cuda:1'), in_proj_covar=tensor([0.0164, 0.0174, 0.0194, 0.0185, 0.0205, 0.0214, 0.0200, 0.0192], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-05-01 02:26:30,782 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=198437.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 02:26:42,827 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.8833, 4.1619, 3.9889, 4.0434, 3.7285, 3.7641, 3.8503, 4.1433], device='cuda:1'), covar=tensor([0.1178, 0.0988, 0.1076, 0.0799, 0.0794, 0.1826, 0.0892, 0.1015], device='cuda:1'), in_proj_covar=tensor([0.0637, 0.0780, 0.0642, 0.0584, 0.0492, 0.0500, 0.0654, 0.0605], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:1') 2023-05-01 02:26:53,768 INFO [train.py:904] (1/8) Epoch 20, batch 5600, loss[loss=0.2553, simple_loss=0.3333, pruned_loss=0.08869, over 15275.00 frames. ], tot_loss[loss=0.2041, simple_loss=0.2902, pruned_loss=0.05902, over 3098256.15 frames. ], batch size: 191, lr: 3.38e-03, grad_scale: 8.0 2023-05-01 02:27:34,782 INFO [optim.py:368] (1/8) 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 02:27:35,217 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.7432, 5.0319, 5.1789, 4.9618, 5.0228, 5.5930, 5.0219, 4.8256], device='cuda:1'), covar=tensor([0.1166, 0.1912, 0.2170, 0.1827, 0.2152, 0.0883, 0.1636, 0.2345], device='cuda:1'), in_proj_covar=tensor([0.0397, 0.0572, 0.0624, 0.0475, 0.0633, 0.0663, 0.0490, 0.0640], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-05-01 02:28:14,190 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=198499.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 02:28:18,606 INFO [train.py:904] (1/8) Epoch 20, batch 5650, loss[loss=0.1904, simple_loss=0.2823, pruned_loss=0.04923, over 16611.00 frames. ], tot_loss[loss=0.2102, simple_loss=0.295, pruned_loss=0.06272, over 3081500.80 frames. ], batch size: 62, lr: 3.38e-03, grad_scale: 8.0 2023-05-01 02:29:02,351 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.1895, 4.2396, 4.0812, 3.8147, 3.7928, 4.1929, 3.8712, 3.9748], device='cuda:1'), covar=tensor([0.0617, 0.0533, 0.0289, 0.0308, 0.0773, 0.0495, 0.0866, 0.0561], device='cuda:1'), in_proj_covar=tensor([0.0287, 0.0414, 0.0334, 0.0330, 0.0346, 0.0384, 0.0230, 0.0400], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002], device='cuda:1') 2023-05-01 02:29:21,761 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.13 vs. limit=2.0 2023-05-01 02:29:31,862 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=198548.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 02:29:36,189 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.8143, 5.1638, 5.3361, 5.0689, 5.1554, 5.7157, 5.1495, 4.9326], device='cuda:1'), covar=tensor([0.1041, 0.1814, 0.1991, 0.1758, 0.2290, 0.0830, 0.1758, 0.2537], device='cuda:1'), in_proj_covar=tensor([0.0401, 0.0577, 0.0632, 0.0480, 0.0640, 0.0668, 0.0496, 0.0647], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-05-01 02:29:36,982 INFO [train.py:904] (1/8) Epoch 20, batch 5700, loss[loss=0.1895, simple_loss=0.2898, pruned_loss=0.04457, over 16770.00 frames. ], tot_loss[loss=0.2112, simple_loss=0.2954, pruned_loss=0.06345, over 3092654.31 frames. ], batch size: 83, lr: 3.38e-03, grad_scale: 8.0 2023-05-01 02:29:50,341 INFO [zipformer.py:625] (1/8) 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] (1/8) Clipping_scale=2.0, grad-norm quartiles 1.943e+02 3.352e+02 4.045e+02 5.075e+02 1.137e+03, threshold=8.090e+02, percent-clipped=5.0 2023-05-01 02:30:47,263 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=198596.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 02:30:55,554 INFO [train.py:904] (1/8) Epoch 20, batch 5750, loss[loss=0.2382, simple_loss=0.305, pruned_loss=0.08567, over 11760.00 frames. ], tot_loss[loss=0.214, simple_loss=0.2981, pruned_loss=0.065, over 3081250.65 frames. ], batch size: 246, lr: 3.38e-03, grad_scale: 8.0 2023-05-01 02:31:07,238 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.37 vs. limit=2.0 2023-05-01 02:31:17,280 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=198615.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 02:32:16,951 INFO [train.py:904] (1/8) Epoch 20, batch 5800, loss[loss=0.1913, simple_loss=0.2823, pruned_loss=0.05018, over 15270.00 frames. ], tot_loss[loss=0.2128, simple_loss=0.2978, pruned_loss=0.06386, over 3079265.91 frames. ], batch size: 190, lr: 3.38e-03, grad_scale: 8.0 2023-05-01 02:32:48,050 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.4783, 3.0546, 2.6283, 2.2273, 2.3491, 2.2468, 3.0295, 2.9308], device='cuda:1'), covar=tensor([0.2479, 0.0672, 0.1672, 0.2381, 0.2394, 0.2009, 0.0523, 0.1225], device='cuda:1'), in_proj_covar=tensor([0.0321, 0.0266, 0.0300, 0.0303, 0.0293, 0.0250, 0.0290, 0.0329], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:1') 2023-05-01 02:32:53,794 INFO [optim.py:368] (1/8) 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] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=198676.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 02:33:08,786 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=2.87 vs. limit=5.0 2023-05-01 02:33:34,960 INFO [train.py:904] (1/8) Epoch 20, batch 5850, loss[loss=0.1775, simple_loss=0.2664, pruned_loss=0.04428, over 16592.00 frames. ], tot_loss[loss=0.2106, simple_loss=0.2961, pruned_loss=0.0626, over 3081836.51 frames. ], batch size: 57, lr: 3.38e-03, grad_scale: 8.0 2023-05-01 02:34:06,644 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.5228, 3.4925, 3.4751, 2.7943, 3.3951, 2.1101, 3.1518, 2.8206], device='cuda:1'), covar=tensor([0.0150, 0.0133, 0.0190, 0.0238, 0.0103, 0.2225, 0.0137, 0.0256], device='cuda:1'), in_proj_covar=tensor([0.0158, 0.0147, 0.0191, 0.0173, 0.0168, 0.0201, 0.0181, 0.0167], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-01 02:34:30,972 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=198737.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 02:34:53,520 INFO [train.py:904] (1/8) Epoch 20, batch 5900, loss[loss=0.2039, simple_loss=0.2928, pruned_loss=0.05748, over 16443.00 frames. ], tot_loss[loss=0.21, simple_loss=0.2955, pruned_loss=0.0622, over 3100932.83 frames. ], batch size: 68, lr: 3.37e-03, grad_scale: 8.0 2023-05-01 02:35:18,060 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.4292, 2.5250, 2.1510, 2.3871, 3.0057, 2.6348, 3.0148, 3.1564], device='cuda:1'), covar=tensor([0.0121, 0.0409, 0.0537, 0.0415, 0.0235, 0.0344, 0.0212, 0.0230], device='cuda:1'), in_proj_covar=tensor([0.0198, 0.0227, 0.0220, 0.0220, 0.0230, 0.0227, 0.0228, 0.0225], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-01 02:35:19,265 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.4515, 4.2345, 4.2687, 2.8659, 3.8420, 4.3404, 3.9417, 2.3593], device='cuda:1'), covar=tensor([0.0564, 0.0043, 0.0047, 0.0381, 0.0085, 0.0105, 0.0072, 0.0460], device='cuda:1'), in_proj_covar=tensor([0.0135, 0.0082, 0.0081, 0.0133, 0.0096, 0.0107, 0.0093, 0.0128], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-05-01 02:35:34,248 INFO [optim.py:368] (1/8) Clipping_scale=2.0, grad-norm quartiles 1.541e+02 2.837e+02 3.258e+02 4.009e+02 8.301e+02, threshold=6.515e+02, percent-clipped=1.0 2023-05-01 02:35:48,535 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=198785.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 02:35:51,912 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.71 vs. limit=2.0 2023-05-01 02:36:14,567 INFO [train.py:904] (1/8) Epoch 20, batch 5950, loss[loss=0.2073, simple_loss=0.3087, pruned_loss=0.05298, over 16794.00 frames. ], tot_loss[loss=0.2092, simple_loss=0.2963, pruned_loss=0.06106, over 3098240.76 frames. ], batch size: 83, lr: 3.37e-03, grad_scale: 8.0 2023-05-01 02:37:31,021 INFO [train.py:904] (1/8) Epoch 20, batch 6000, loss[loss=0.1901, simple_loss=0.2803, pruned_loss=0.04998, over 16745.00 frames. ], tot_loss[loss=0.2087, simple_loss=0.2956, pruned_loss=0.06088, over 3088419.51 frames. ], batch size: 83, lr: 3.37e-03, grad_scale: 8.0 2023-05-01 02:37:31,021 INFO [train.py:929] (1/8) Computing validation loss 2023-05-01 02:37:41,832 INFO [train.py:938] (1/8) Epoch 20, validation: loss=0.1516, simple_loss=0.2644, pruned_loss=0.01942, over 944034.00 frames. 2023-05-01 02:37:41,833 INFO [train.py:939] (1/8) Maximum memory allocated so far is 18145MB 2023-05-01 02:37:42,266 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=198852.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 02:37:46,499 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=198855.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 02:37:57,193 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.8436, 3.9010, 2.5790, 4.7529, 3.1561, 4.6262, 2.6281, 3.1274], device='cuda:1'), covar=tensor([0.0305, 0.0404, 0.1629, 0.0220, 0.0778, 0.0519, 0.1541, 0.0839], device='cuda:1'), in_proj_covar=tensor([0.0166, 0.0175, 0.0193, 0.0158, 0.0175, 0.0213, 0.0201, 0.0176], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-05-01 02:38:05,652 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.6617, 3.0718, 3.1765, 1.9095, 2.7764, 2.1584, 3.2151, 3.3341], device='cuda:1'), covar=tensor([0.0277, 0.0786, 0.0628, 0.2140, 0.0859, 0.0986, 0.0702, 0.0872], device='cuda:1'), in_proj_covar=tensor([0.0154, 0.0162, 0.0166, 0.0151, 0.0143, 0.0129, 0.0144, 0.0172], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:1') 2023-05-01 02:38:17,162 INFO [optim.py:368] (1/8) 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 02:38:25,746 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.43 vs. limit=2.0 2023-05-01 02:38:28,019 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=198883.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 02:38:56,337 INFO [train.py:904] (1/8) Epoch 20, batch 6050, loss[loss=0.1913, simple_loss=0.29, pruned_loss=0.04631, over 16751.00 frames. ], tot_loss[loss=0.2064, simple_loss=0.2934, pruned_loss=0.05967, over 3112384.29 frames. ], batch size: 83, lr: 3.37e-03, grad_scale: 8.0 2023-05-01 02:39:14,677 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=198913.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 02:39:32,791 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.1510, 2.4501, 2.0094, 2.2022, 2.8920, 2.4773, 2.8605, 3.0241], device='cuda:1'), covar=tensor([0.0158, 0.0406, 0.0522, 0.0415, 0.0222, 0.0339, 0.0207, 0.0222], device='cuda:1'), in_proj_covar=tensor([0.0199, 0.0228, 0.0220, 0.0221, 0.0231, 0.0228, 0.0229, 0.0225], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-01 02:39:36,245 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=198927.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 02:39:51,397 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.3935, 4.4497, 4.7829, 4.7549, 4.7590, 4.4513, 4.4331, 4.3374], device='cuda:1'), covar=tensor([0.0333, 0.0599, 0.0392, 0.0423, 0.0446, 0.0419, 0.0968, 0.0516], device='cuda:1'), in_proj_covar=tensor([0.0400, 0.0442, 0.0426, 0.0397, 0.0475, 0.0450, 0.0540, 0.0361], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:1') 2023-05-01 02:40:05,150 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=198944.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 02:40:16,823 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.6057, 3.4337, 3.9179, 1.8582, 4.1132, 4.1831, 3.1486, 2.9840], device='cuda:1'), covar=tensor([0.0796, 0.0263, 0.0212, 0.1363, 0.0082, 0.0147, 0.0407, 0.0503], device='cuda:1'), in_proj_covar=tensor([0.0150, 0.0109, 0.0098, 0.0139, 0.0080, 0.0124, 0.0129, 0.0132], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0004, 0.0004], device='cuda:1') 2023-05-01 02:40:17,598 INFO [train.py:904] (1/8) Epoch 20, batch 6100, loss[loss=0.1879, simple_loss=0.2769, pruned_loss=0.0495, over 16505.00 frames. ], tot_loss[loss=0.205, simple_loss=0.2928, pruned_loss=0.05855, over 3126870.22 frames. ], batch size: 75, lr: 3.37e-03, grad_scale: 8.0 2023-05-01 02:40:31,217 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.6504, 3.6341, 2.8056, 2.2326, 2.4569, 2.4455, 3.8978, 3.3660], device='cuda:1'), covar=tensor([0.2789, 0.0662, 0.1695, 0.2464, 0.2475, 0.1883, 0.0432, 0.1181], device='cuda:1'), in_proj_covar=tensor([0.0323, 0.0267, 0.0301, 0.0305, 0.0294, 0.0252, 0.0291, 0.0330], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:1') 2023-05-01 02:40:46,238 INFO [zipformer.py:625] (1/8) 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] (1/8) 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 02:41:12,459 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 20, batch 6150, loss[loss=0.1585, simple_loss=0.2521, pruned_loss=0.03249, over 16932.00 frames. ], tot_loss[loss=0.2031, simple_loss=0.2904, pruned_loss=0.05791, over 3129008.26 frames. ], batch size: 96, lr: 3.37e-03, grad_scale: 8.0 2023-05-01 02:42:49,579 INFO [train.py:904] (1/8) Epoch 20, batch 6200, loss[loss=0.185, simple_loss=0.2744, pruned_loss=0.04785, over 16767.00 frames. ], tot_loss[loss=0.2019, simple_loss=0.2887, pruned_loss=0.05754, over 3126336.62 frames. ], batch size: 83, lr: 3.37e-03, grad_scale: 4.0 2023-05-01 02:43:28,024 INFO [optim.py:368] (1/8) 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 02:43:35,540 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.9956, 5.0214, 4.8407, 4.4897, 4.5072, 4.9256, 4.8598, 4.6036], device='cuda:1'), covar=tensor([0.0732, 0.0707, 0.0312, 0.0343, 0.0926, 0.0578, 0.0386, 0.0756], device='cuda:1'), in_proj_covar=tensor([0.0284, 0.0410, 0.0331, 0.0325, 0.0341, 0.0379, 0.0228, 0.0394], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-01 02:44:06,520 INFO [train.py:904] (1/8) Epoch 20, batch 6250, loss[loss=0.1943, simple_loss=0.2807, pruned_loss=0.05395, over 15414.00 frames. ], tot_loss[loss=0.2007, simple_loss=0.2877, pruned_loss=0.05681, over 3130842.04 frames. ], batch size: 190, lr: 3.37e-03, grad_scale: 4.0 2023-05-01 02:45:21,381 INFO [train.py:904] (1/8) Epoch 20, batch 6300, loss[loss=0.2211, simple_loss=0.2979, pruned_loss=0.07216, over 11610.00 frames. ], tot_loss[loss=0.1996, simple_loss=0.2874, pruned_loss=0.05592, over 3142689.89 frames. ], batch size: 247, lr: 3.37e-03, grad_scale: 4.0 2023-05-01 02:45:25,882 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=199155.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 02:45:38,298 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([1.8917, 3.3957, 3.3754, 2.1742, 3.1489, 3.4045, 3.1521, 2.0011], device='cuda:1'), covar=tensor([0.0618, 0.0061, 0.0062, 0.0441, 0.0114, 0.0126, 0.0108, 0.0461], device='cuda:1'), in_proj_covar=tensor([0.0134, 0.0081, 0.0080, 0.0132, 0.0095, 0.0107, 0.0092, 0.0127], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-05-01 02:45:59,422 INFO [zipformer.py:625] (1/8) 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] (1/8) 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 02:46:38,965 INFO [train.py:904] (1/8) Epoch 20, batch 6350, loss[loss=0.2142, simple_loss=0.3058, pruned_loss=0.06128, over 16730.00 frames. ], tot_loss[loss=0.2028, simple_loss=0.2893, pruned_loss=0.05811, over 3093775.27 frames. ], batch size: 89, lr: 3.37e-03, grad_scale: 4.0 2023-05-01 02:46:40,591 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=199203.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 02:46:47,429 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=199208.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 02:47:30,802 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=199237.0, num_to_drop=1, layers_to_drop={1} 2023-05-01 02:47:34,238 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=199239.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 02:47:52,851 INFO [train.py:904] (1/8) Epoch 20, batch 6400, loss[loss=0.2538, simple_loss=0.3234, pruned_loss=0.0921, over 11153.00 frames. ], tot_loss[loss=0.2049, simple_loss=0.2899, pruned_loss=0.05992, over 3090378.57 frames. ], batch size: 248, lr: 3.37e-03, grad_scale: 8.0 2023-05-01 02:48:21,656 INFO [zipformer.py:625] (1/8) 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] (1/8) 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 02:48:39,447 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=199283.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 02:48:59,069 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.76 vs. limit=2.0 2023-05-01 02:49:07,103 INFO [train.py:904] (1/8) Epoch 20, batch 6450, loss[loss=0.2164, simple_loss=0.3064, pruned_loss=0.06321, over 16627.00 frames. ], tot_loss[loss=0.2035, simple_loss=0.2896, pruned_loss=0.05876, over 3108507.42 frames. ], batch size: 62, lr: 3.37e-03, grad_scale: 8.0 2023-05-01 02:49:33,333 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=199319.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 02:50:24,669 INFO [train.py:904] (1/8) Epoch 20, batch 6500, loss[loss=0.1853, simple_loss=0.2751, pruned_loss=0.0478, over 16491.00 frames. ], tot_loss[loss=0.2017, simple_loss=0.2878, pruned_loss=0.05781, over 3112679.05 frames. ], batch size: 146, lr: 3.37e-03, grad_scale: 8.0 2023-05-01 02:51:02,008 INFO [optim.py:368] (1/8) 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 02:51:39,101 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=199400.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 02:51:41,757 INFO [train.py:904] (1/8) Epoch 20, batch 6550, loss[loss=0.212, simple_loss=0.3101, pruned_loss=0.05693, over 15308.00 frames. ], tot_loss[loss=0.2046, simple_loss=0.291, pruned_loss=0.05911, over 3097786.69 frames. ], batch size: 190, lr: 3.37e-03, grad_scale: 8.0 2023-05-01 02:52:13,178 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=4.69 vs. limit=5.0 2023-05-01 02:52:56,165 INFO [train.py:904] (1/8) Epoch 20, batch 6600, loss[loss=0.2132, simple_loss=0.3048, pruned_loss=0.06082, over 16488.00 frames. ], tot_loss[loss=0.2052, simple_loss=0.2924, pruned_loss=0.05902, over 3111135.90 frames. ], batch size: 68, lr: 3.37e-03, grad_scale: 8.0 2023-05-01 02:53:05,857 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=199458.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 02:53:10,113 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=199461.0, num_to_drop=1, layers_to_drop={3} 2023-05-01 02:53:33,376 INFO [optim.py:368] (1/8) 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 02:53:59,206 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=199494.0, num_to_drop=1, layers_to_drop={0} 2023-05-01 02:54:11,566 INFO [train.py:904] (1/8) Epoch 20, batch 6650, loss[loss=0.1928, simple_loss=0.2799, pruned_loss=0.05289, over 16732.00 frames. ], tot_loss[loss=0.2061, simple_loss=0.2926, pruned_loss=0.0598, over 3107155.28 frames. ], batch size: 76, lr: 3.37e-03, grad_scale: 8.0 2023-05-01 02:54:20,055 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=199508.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 02:54:37,188 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=199519.0, num_to_drop=1, layers_to_drop={0} 2023-05-01 02:54:54,135 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=2.02 vs. limit=2.0 2023-05-01 02:54:56,461 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=199532.0, num_to_drop=1, layers_to_drop={3} 2023-05-01 02:55:00,510 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=199535.0, num_to_drop=1, layers_to_drop={1} 2023-05-01 02:55:06,208 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 20, batch 6700, loss[loss=0.2113, simple_loss=0.3001, pruned_loss=0.06129, over 16931.00 frames. ], tot_loss[loss=0.2053, simple_loss=0.2913, pruned_loss=0.05962, over 3111309.76 frames. ], batch size: 116, lr: 3.37e-03, grad_scale: 8.0 2023-05-01 02:55:30,710 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=199555.0, num_to_drop=1, layers_to_drop={2} 2023-05-01 02:55:32,560 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=199556.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 02:55:52,845 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([1.8293, 1.3461, 1.7279, 1.6787, 1.7909, 1.9307, 1.6407, 1.7954], device='cuda:1'), covar=tensor([0.0242, 0.0378, 0.0207, 0.0271, 0.0243, 0.0161, 0.0389, 0.0130], device='cuda:1'), in_proj_covar=tensor([0.0183, 0.0191, 0.0176, 0.0181, 0.0192, 0.0150, 0.0193, 0.0145], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-01 02:56:02,759 INFO [optim.py:368] (1/8) 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 02:56:13,367 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=199583.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 02:56:18,210 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=199587.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 02:56:31,541 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=199596.0, num_to_drop=1, layers_to_drop={0} 2023-05-01 02:56:38,605 INFO [train.py:904] (1/8) Epoch 20, batch 6750, loss[loss=0.2489, simple_loss=0.3128, pruned_loss=0.0925, over 11867.00 frames. ], tot_loss[loss=0.2063, simple_loss=0.2913, pruned_loss=0.06066, over 3085034.12 frames. ], batch size: 248, lr: 3.37e-03, grad_scale: 8.0 2023-05-01 02:56:59,502 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.1627, 1.5720, 1.9747, 2.1258, 2.2297, 2.3893, 1.7380, 2.2713], device='cuda:1'), covar=tensor([0.0226, 0.0461, 0.0269, 0.0329, 0.0306, 0.0173, 0.0473, 0.0134], device='cuda:1'), in_proj_covar=tensor([0.0183, 0.0191, 0.0175, 0.0180, 0.0192, 0.0149, 0.0192, 0.0144], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-01 02:57:20,690 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=199631.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 02:57:28,264 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.6158, 3.8660, 2.8675, 2.3225, 2.6643, 2.5174, 4.1554, 3.4915], device='cuda:1'), covar=tensor([0.2873, 0.0632, 0.1793, 0.2631, 0.2628, 0.1938, 0.0459, 0.1178], device='cuda:1'), in_proj_covar=tensor([0.0326, 0.0268, 0.0303, 0.0307, 0.0296, 0.0254, 0.0293, 0.0332], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:1') 2023-05-01 02:57:53,091 INFO [train.py:904] (1/8) Epoch 20, batch 6800, loss[loss=0.2251, simple_loss=0.3104, pruned_loss=0.06988, over 16237.00 frames. ], tot_loss[loss=0.2061, simple_loss=0.2912, pruned_loss=0.06047, over 3087777.42 frames. ], batch size: 165, lr: 3.37e-03, grad_scale: 8.0 2023-05-01 02:58:29,266 INFO [optim.py:368] (1/8) 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 02:59:05,438 INFO [train.py:904] (1/8) Epoch 20, batch 6850, loss[loss=0.2171, simple_loss=0.3196, pruned_loss=0.05727, over 16086.00 frames. ], tot_loss[loss=0.2061, simple_loss=0.2919, pruned_loss=0.0602, over 3088511.85 frames. ], batch size: 35, lr: 3.37e-03, grad_scale: 8.0 2023-05-01 02:59:07,132 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.0725, 2.4278, 2.6030, 1.9495, 2.7356, 2.8231, 2.4346, 2.3808], device='cuda:1'), covar=tensor([0.0712, 0.0251, 0.0216, 0.0926, 0.0120, 0.0282, 0.0426, 0.0459], device='cuda:1'), in_proj_covar=tensor([0.0148, 0.0108, 0.0097, 0.0139, 0.0079, 0.0123, 0.0128, 0.0131], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0004, 0.0004], device='cuda:1') 2023-05-01 02:59:11,111 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.89 vs. limit=2.0 2023-05-01 02:59:51,962 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.23 vs. limit=2.0 2023-05-01 03:00:06,670 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.40 vs. limit=2.0 2023-05-01 03:00:20,622 INFO [train.py:904] (1/8) Epoch 20, batch 6900, loss[loss=0.2126, simple_loss=0.3074, pruned_loss=0.05892, over 16852.00 frames. ], tot_loss[loss=0.2066, simple_loss=0.2939, pruned_loss=0.05967, over 3107602.40 frames. ], batch size: 116, lr: 3.37e-03, grad_scale: 4.0 2023-05-01 03:00:26,873 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=199756.0, num_to_drop=1, layers_to_drop={3} 2023-05-01 03:00:59,797 INFO [optim.py:368] (1/8) 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 03:01:36,153 INFO [train.py:904] (1/8) Epoch 20, batch 6950, loss[loss=0.2273, simple_loss=0.3125, pruned_loss=0.07101, over 16535.00 frames. ], tot_loss[loss=0.2086, simple_loss=0.2953, pruned_loss=0.06096, over 3104887.76 frames. ], batch size: 75, lr: 3.37e-03, grad_scale: 4.0 2023-05-01 03:01:54,386 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=199814.0, num_to_drop=1, layers_to_drop={0} 2023-05-01 03:02:06,103 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.39 vs. limit=2.0 2023-05-01 03:02:16,046 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.9527, 3.9914, 4.3134, 4.2773, 4.2784, 4.0066, 4.0225, 3.9968], device='cuda:1'), covar=tensor([0.0385, 0.0615, 0.0387, 0.0394, 0.0498, 0.0438, 0.0936, 0.0509], device='cuda:1'), in_proj_covar=tensor([0.0398, 0.0438, 0.0424, 0.0396, 0.0473, 0.0449, 0.0539, 0.0358], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:1') 2023-05-01 03:02:19,445 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.4668, 2.1850, 1.8838, 1.9311, 2.4880, 2.1421, 2.2933, 2.5987], device='cuda:1'), covar=tensor([0.0202, 0.0409, 0.0514, 0.0456, 0.0239, 0.0357, 0.0197, 0.0260], device='cuda:1'), in_proj_covar=tensor([0.0197, 0.0226, 0.0219, 0.0220, 0.0228, 0.0227, 0.0226, 0.0224], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-01 03:02:22,464 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=199832.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 03:02:47,649 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=199850.0, num_to_drop=1, layers_to_drop={0} 2023-05-01 03:02:49,629 INFO [train.py:904] (1/8) Epoch 20, batch 7000, loss[loss=0.1974, simple_loss=0.2938, pruned_loss=0.05046, over 16901.00 frames. ], tot_loss[loss=0.2083, simple_loss=0.2956, pruned_loss=0.06054, over 3105927.64 frames. ], batch size: 116, lr: 3.37e-03, grad_scale: 4.0 2023-05-01 03:03:29,995 INFO [optim.py:368] (1/8) 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 03:03:32,319 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.77 vs. limit=2.0 2023-05-01 03:03:33,439 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=199880.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 03:03:49,864 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=199891.0, num_to_drop=1, layers_to_drop={1} 2023-05-01 03:04:07,202 INFO [train.py:904] (1/8) Epoch 20, batch 7050, loss[loss=0.2164, simple_loss=0.3112, pruned_loss=0.06083, over 16709.00 frames. ], tot_loss[loss=0.2089, simple_loss=0.2961, pruned_loss=0.06083, over 3075062.03 frames. ], batch size: 62, lr: 3.37e-03, grad_scale: 4.0 2023-05-01 03:04:17,915 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=199909.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 03:05:25,775 INFO [train.py:904] (1/8) Epoch 20, batch 7100, loss[loss=0.2631, simple_loss=0.3161, pruned_loss=0.105, over 11731.00 frames. ], tot_loss[loss=0.2082, simple_loss=0.2948, pruned_loss=0.0608, over 3078694.90 frames. ], batch size: 248, lr: 3.36e-03, grad_scale: 4.0 2023-05-01 03:05:54,813 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=199970.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 03:06:05,960 INFO [optim.py:368] (1/8) 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 03:06:45,340 INFO [train.py:904] (1/8) Epoch 20, batch 7150, loss[loss=0.253, simple_loss=0.3185, pruned_loss=0.09372, over 11300.00 frames. ], tot_loss[loss=0.2068, simple_loss=0.293, pruned_loss=0.06032, over 3096355.59 frames. ], batch size: 248, lr: 3.36e-03, grad_scale: 4.0 2023-05-01 03:07:59,654 INFO [train.py:904] (1/8) Epoch 20, batch 7200, loss[loss=0.1982, simple_loss=0.2859, pruned_loss=0.05524, over 15266.00 frames. ], tot_loss[loss=0.2038, simple_loss=0.2906, pruned_loss=0.05853, over 3105137.33 frames. ], batch size: 190, lr: 3.36e-03, grad_scale: 8.0 2023-05-01 03:08:06,649 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=200056.0, num_to_drop=1, layers_to_drop={1} 2023-05-01 03:08:40,490 INFO [optim.py:368] (1/8) Clipping_scale=2.0, grad-norm quartiles 1.812e+02 2.598e+02 3.072e+02 3.513e+02 6.350e+02, threshold=6.144e+02, percent-clipped=0.0 2023-05-01 03:09:19,963 INFO [train.py:904] (1/8) Epoch 20, batch 7250, loss[loss=0.1723, simple_loss=0.2648, pruned_loss=0.03986, over 16821.00 frames. ], tot_loss[loss=0.201, simple_loss=0.2879, pruned_loss=0.05708, over 3109263.08 frames. ], batch size: 102, lr: 3.36e-03, grad_scale: 8.0 2023-05-01 03:09:23,331 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=200104.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 03:09:37,503 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=200114.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 03:10:31,941 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 20, batch 7300, loss[loss=0.1906, simple_loss=0.2868, pruned_loss=0.04719, over 16330.00 frames. ], tot_loss[loss=0.2006, simple_loss=0.2874, pruned_loss=0.05689, over 3115909.22 frames. ], batch size: 146, lr: 3.36e-03, grad_scale: 8.0 2023-05-01 03:10:50,982 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=200162.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 03:11:14,627 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=200177.0, num_to_drop=1, layers_to_drop={1} 2023-05-01 03:11:15,880 INFO [optim.py:368] (1/8) Clipping_scale=2.0, grad-norm quartiles 1.776e+02 2.904e+02 3.413e+02 4.614e+02 7.836e+02, threshold=6.825e+02, percent-clipped=9.0 2023-05-01 03:11:37,041 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=200191.0, num_to_drop=1, layers_to_drop={2} 2023-05-01 03:11:47,182 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 20, batch 7350, loss[loss=0.1899, simple_loss=0.2819, pruned_loss=0.04899, over 16762.00 frames. ], tot_loss[loss=0.2026, simple_loss=0.2887, pruned_loss=0.05825, over 3089668.46 frames. ], batch size: 83, lr: 3.36e-03, grad_scale: 8.0 2023-05-01 03:12:51,299 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=200238.0, num_to_drop=1, layers_to_drop={2} 2023-05-01 03:12:52,206 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=200239.0, num_to_drop=1, layers_to_drop={0} 2023-05-01 03:13:12,411 INFO [train.py:904] (1/8) Epoch 20, batch 7400, loss[loss=0.2059, simple_loss=0.2934, pruned_loss=0.05923, over 16794.00 frames. ], tot_loss[loss=0.2035, simple_loss=0.2896, pruned_loss=0.05868, over 3093827.66 frames. ], batch size: 39, lr: 3.36e-03, grad_scale: 8.0 2023-05-01 03:13:33,058 INFO [zipformer.py:625] (1/8) 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] (1/8) 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 03:13:55,411 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 20, batch 7450, loss[loss=0.2107, simple_loss=0.3035, pruned_loss=0.05895, over 16925.00 frames. ], tot_loss[loss=0.2055, simple_loss=0.2914, pruned_loss=0.0598, over 3089283.40 frames. ], batch size: 109, lr: 3.36e-03, grad_scale: 8.0 2023-05-01 03:15:35,797 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=200340.0, num_to_drop=1, layers_to_drop={2} 2023-05-01 03:15:42,513 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.3148, 3.4820, 3.6187, 3.5786, 3.5983, 3.4177, 3.4340, 3.4892], device='cuda:1'), covar=tensor([0.0443, 0.0657, 0.0505, 0.0520, 0.0595, 0.0594, 0.0879, 0.0543], device='cuda:1'), in_proj_covar=tensor([0.0398, 0.0437, 0.0425, 0.0396, 0.0473, 0.0448, 0.0539, 0.0360], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:1') 2023-05-01 03:15:54,173 INFO [train.py:904] (1/8) Epoch 20, batch 7500, loss[loss=0.2429, simple_loss=0.3141, pruned_loss=0.08579, over 11683.00 frames. ], tot_loss[loss=0.2054, simple_loss=0.2916, pruned_loss=0.05963, over 3073930.87 frames. ], batch size: 248, lr: 3.36e-03, grad_scale: 8.0 2023-05-01 03:16:02,775 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.11 vs. limit=2.0 2023-05-01 03:16:03,840 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.6263, 2.5835, 1.8387, 2.7029, 2.1145, 2.7695, 2.0534, 2.3270], device='cuda:1'), covar=tensor([0.0329, 0.0341, 0.1243, 0.0263, 0.0621, 0.0473, 0.1254, 0.0611], device='cuda:1'), in_proj_covar=tensor([0.0167, 0.0174, 0.0194, 0.0157, 0.0174, 0.0213, 0.0201, 0.0176], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-05-01 03:16:34,099 INFO [optim.py:368] (1/8) 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 03:17:11,460 INFO [train.py:904] (1/8) Epoch 20, batch 7550, loss[loss=0.2253, simple_loss=0.2943, pruned_loss=0.07819, over 11392.00 frames. ], tot_loss[loss=0.2046, simple_loss=0.2904, pruned_loss=0.05945, over 3083419.02 frames. ], batch size: 247, lr: 3.36e-03, grad_scale: 8.0 2023-05-01 03:18:14,195 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.3395, 2.8984, 2.6566, 2.3063, 2.2963, 2.2848, 2.8450, 2.8416], device='cuda:1'), covar=tensor([0.2296, 0.0778, 0.1564, 0.2305, 0.2387, 0.2014, 0.0503, 0.1273], device='cuda:1'), in_proj_covar=tensor([0.0329, 0.0269, 0.0304, 0.0310, 0.0298, 0.0257, 0.0294, 0.0335], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:1') 2023-05-01 03:18:19,842 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.15 vs. limit=2.0 2023-05-01 03:18:26,295 INFO [train.py:904] (1/8) Epoch 20, batch 7600, loss[loss=0.1919, simple_loss=0.2807, pruned_loss=0.05148, over 16960.00 frames. ], tot_loss[loss=0.205, simple_loss=0.2902, pruned_loss=0.05995, over 3076688.81 frames. ], batch size: 55, lr: 3.36e-03, grad_scale: 8.0 2023-05-01 03:18:28,779 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.1875, 2.1672, 2.1900, 3.8538, 2.0394, 2.5450, 2.2264, 2.3359], device='cuda:1'), covar=tensor([0.1350, 0.3603, 0.2972, 0.0534, 0.4252, 0.2468, 0.3602, 0.3325], device='cuda:1'), in_proj_covar=tensor([0.0394, 0.0440, 0.0360, 0.0322, 0.0433, 0.0507, 0.0410, 0.0515], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-01 03:19:06,493 INFO [optim.py:368] (1/8) 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 03:19:44,057 INFO [train.py:904] (1/8) Epoch 20, batch 7650, loss[loss=0.2092, simple_loss=0.305, pruned_loss=0.05672, over 16835.00 frames. ], tot_loss[loss=0.2057, simple_loss=0.2905, pruned_loss=0.06039, over 3081592.98 frames. ], batch size: 90, lr: 3.36e-03, grad_scale: 8.0 2023-05-01 03:20:33,301 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=200533.0, num_to_drop=1, layers_to_drop={1} 2023-05-01 03:21:02,178 INFO [train.py:904] (1/8) Epoch 20, batch 7700, loss[loss=0.1851, simple_loss=0.2773, pruned_loss=0.0465, over 16857.00 frames. ], tot_loss[loss=0.2053, simple_loss=0.2901, pruned_loss=0.06023, over 3087099.18 frames. ], batch size: 102, lr: 3.36e-03, grad_scale: 8.0 2023-05-01 03:21:22,835 INFO [zipformer.py:625] (1/8) 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] (1/8) 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] (1/8) Epoch 20, batch 7750, loss[loss=0.2353, simple_loss=0.3004, pruned_loss=0.08512, over 11370.00 frames. ], tot_loss[loss=0.2057, simple_loss=0.2905, pruned_loss=0.06048, over 3077995.21 frames. ], batch size: 248, lr: 3.36e-03, grad_scale: 4.0 2023-05-01 03:22:38,651 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=200613.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 03:23:11,916 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=200635.0, num_to_drop=1, layers_to_drop={1} 2023-05-01 03:23:27,518 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=200645.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 03:23:36,970 INFO [train.py:904] (1/8) Epoch 20, batch 7800, loss[loss=0.1835, simple_loss=0.2732, pruned_loss=0.04691, over 16880.00 frames. ], tot_loss[loss=0.2068, simple_loss=0.291, pruned_loss=0.06126, over 3072914.92 frames. ], batch size: 96, lr: 3.36e-03, grad_scale: 4.0 2023-05-01 03:23:47,419 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.5749, 3.7003, 2.8022, 2.1615, 2.4107, 2.3519, 3.9661, 3.2253], device='cuda:1'), covar=tensor([0.2947, 0.0610, 0.1812, 0.2844, 0.2733, 0.2147, 0.0441, 0.1348], device='cuda:1'), in_proj_covar=tensor([0.0327, 0.0268, 0.0303, 0.0309, 0.0297, 0.0256, 0.0293, 0.0334], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:1') 2023-05-01 03:24:18,679 INFO [optim.py:368] (1/8) Clipping_scale=2.0, grad-norm quartiles 1.931e+02 2.831e+02 3.422e+02 4.035e+02 8.624e+02, threshold=6.845e+02, percent-clipped=1.0 2023-05-01 03:24:53,417 INFO [train.py:904] (1/8) Epoch 20, batch 7850, loss[loss=0.2023, simple_loss=0.2865, pruned_loss=0.05902, over 16740.00 frames. ], tot_loss[loss=0.2069, simple_loss=0.2921, pruned_loss=0.06086, over 3087942.07 frames. ], batch size: 124, lr: 3.36e-03, grad_scale: 4.0 2023-05-01 03:24:59,659 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=200706.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 03:25:12,328 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([5.1363, 5.5474, 5.7510, 5.4681, 5.5604, 6.1166, 5.5820, 5.3752], device='cuda:1'), covar=tensor([0.0895, 0.1779, 0.2227, 0.1948, 0.2246, 0.0917, 0.1525, 0.2257], device='cuda:1'), in_proj_covar=tensor([0.0397, 0.0580, 0.0636, 0.0478, 0.0636, 0.0665, 0.0500, 0.0645], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-05-01 03:25:27,503 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.36 vs. limit=2.0 2023-05-01 03:26:08,719 INFO [train.py:904] (1/8) Epoch 20, batch 7900, loss[loss=0.2655, simple_loss=0.3261, pruned_loss=0.1024, over 11721.00 frames. ], tot_loss[loss=0.2057, simple_loss=0.2913, pruned_loss=0.05999, over 3104162.11 frames. ], batch size: 248, lr: 3.36e-03, grad_scale: 4.0 2023-05-01 03:26:34,476 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.5127, 1.6447, 2.1370, 2.4619, 2.4848, 2.8774, 1.7871, 2.8360], device='cuda:1'), covar=tensor([0.0238, 0.0555, 0.0337, 0.0341, 0.0316, 0.0182, 0.0549, 0.0142], device='cuda:1'), in_proj_covar=tensor([0.0180, 0.0189, 0.0174, 0.0178, 0.0191, 0.0148, 0.0192, 0.0144], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-01 03:26:49,222 INFO [optim.py:368] (1/8) 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 03:27:27,140 INFO [train.py:904] (1/8) Epoch 20, batch 7950, loss[loss=0.1899, simple_loss=0.2769, pruned_loss=0.05144, over 16750.00 frames. ], tot_loss[loss=0.206, simple_loss=0.2917, pruned_loss=0.06019, over 3125046.76 frames. ], batch size: 83, lr: 3.36e-03, grad_scale: 4.0 2023-05-01 03:27:39,901 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.20 vs. limit=2.0 2023-05-01 03:28:15,508 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.2132, 4.3018, 4.4406, 4.2464, 4.3034, 4.8377, 4.3603, 4.1013], device='cuda:1'), covar=tensor([0.1742, 0.2101, 0.2416, 0.2019, 0.2636, 0.1110, 0.1648, 0.2459], device='cuda:1'), in_proj_covar=tensor([0.0399, 0.0581, 0.0639, 0.0479, 0.0637, 0.0668, 0.0502, 0.0648], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-05-01 03:28:15,569 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=200833.0, num_to_drop=1, layers_to_drop={1} 2023-05-01 03:28:44,604 INFO [train.py:904] (1/8) Epoch 20, batch 8000, loss[loss=0.1906, simple_loss=0.2864, pruned_loss=0.04741, over 16890.00 frames. ], tot_loss[loss=0.2081, simple_loss=0.2927, pruned_loss=0.06172, over 3091227.73 frames. ], batch size: 96, lr: 3.36e-03, grad_scale: 8.0 2023-05-01 03:29:24,818 INFO [optim.py:368] (1/8) 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 03:29:28,341 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=200881.0, num_to_drop=1, layers_to_drop={0} 2023-05-01 03:29:59,852 INFO [train.py:904] (1/8) Epoch 20, batch 8050, loss[loss=0.1878, simple_loss=0.2791, pruned_loss=0.04822, over 16442.00 frames. ], tot_loss[loss=0.2077, simple_loss=0.2929, pruned_loss=0.06127, over 3087999.53 frames. ], batch size: 75, lr: 3.36e-03, grad_scale: 8.0 2023-05-01 03:30:50,024 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 20, batch 8100, loss[loss=0.2198, simple_loss=0.306, pruned_loss=0.06685, over 16275.00 frames. ], tot_loss[loss=0.2068, simple_loss=0.2925, pruned_loss=0.06054, over 3094674.69 frames. ], batch size: 165, lr: 3.36e-03, grad_scale: 8.0 2023-05-01 03:31:57,146 INFO [optim.py:368] (1/8) 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 03:32:02,942 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.80 vs. limit=2.0 2023-05-01 03:32:04,949 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=200983.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 03:32:31,897 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=201001.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 03:32:33,329 INFO [train.py:904] (1/8) Epoch 20, batch 8150, loss[loss=0.2172, simple_loss=0.29, pruned_loss=0.07222, over 11912.00 frames. ], tot_loss[loss=0.2043, simple_loss=0.29, pruned_loss=0.0593, over 3107151.30 frames. ], batch size: 246, lr: 3.36e-03, grad_scale: 8.0 2023-05-01 03:33:23,362 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.46 vs. limit=2.0 2023-05-01 03:33:35,429 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.43 vs. limit=2.0 2023-05-01 03:33:52,112 INFO [train.py:904] (1/8) Epoch 20, batch 8200, loss[loss=0.1828, simple_loss=0.2697, pruned_loss=0.04795, over 16788.00 frames. ], tot_loss[loss=0.2022, simple_loss=0.2873, pruned_loss=0.05856, over 3111793.83 frames. ], batch size: 83, lr: 3.36e-03, grad_scale: 8.0 2023-05-01 03:34:24,295 INFO [zipformer.py:625] (1/8) 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] (1/8) 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 03:35:15,217 INFO [train.py:904] (1/8) Epoch 20, batch 8250, loss[loss=0.1853, simple_loss=0.2791, pruned_loss=0.04574, over 16883.00 frames. ], tot_loss[loss=0.1991, simple_loss=0.286, pruned_loss=0.05608, over 3099907.18 frames. ], batch size: 116, lr: 3.36e-03, grad_scale: 8.0 2023-05-01 03:36:05,985 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 20, batch 8300, loss[loss=0.1901, simple_loss=0.2833, pruned_loss=0.04845, over 16782.00 frames. ], tot_loss[loss=0.1946, simple_loss=0.2834, pruned_loss=0.0529, over 3103125.82 frames. ], batch size: 124, lr: 3.35e-03, grad_scale: 8.0 2023-05-01 03:37:22,448 INFO [optim.py:368] (1/8) 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 03:37:55,088 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.5270, 3.5923, 2.8668, 2.0521, 2.2320, 2.3027, 3.7668, 3.2706], device='cuda:1'), covar=tensor([0.2921, 0.0614, 0.1704, 0.3071, 0.2890, 0.2286, 0.0422, 0.1190], device='cuda:1'), in_proj_covar=tensor([0.0323, 0.0266, 0.0300, 0.0306, 0.0293, 0.0253, 0.0290, 0.0328], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:1') 2023-05-01 03:38:00,760 INFO [train.py:904] (1/8) Epoch 20, batch 8350, loss[loss=0.166, simple_loss=0.2716, pruned_loss=0.0302, over 16940.00 frames. ], tot_loss[loss=0.1925, simple_loss=0.2833, pruned_loss=0.05088, over 3112032.46 frames. ], batch size: 96, lr: 3.35e-03, grad_scale: 8.0 2023-05-01 03:39:23,231 INFO [train.py:904] (1/8) Epoch 20, batch 8400, loss[loss=0.1607, simple_loss=0.2609, pruned_loss=0.03021, over 16811.00 frames. ], tot_loss[loss=0.1891, simple_loss=0.2804, pruned_loss=0.04886, over 3096299.99 frames. ], batch size: 102, lr: 3.35e-03, grad_scale: 8.0 2023-05-01 03:39:25,951 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=201253.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 03:39:52,485 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([1.9786, 2.4012, 2.4110, 3.0285, 1.9251, 3.2623, 1.7855, 2.7954], device='cuda:1'), covar=tensor([0.1166, 0.0583, 0.0955, 0.0159, 0.0077, 0.0360, 0.1497, 0.0659], device='cuda:1'), in_proj_covar=tensor([0.0164, 0.0171, 0.0192, 0.0183, 0.0205, 0.0210, 0.0198, 0.0189], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-05-01 03:40:00,814 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=201274.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 03:40:00,843 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.6071, 2.1416, 1.8601, 1.8512, 2.4128, 2.0856, 2.1013, 2.5144], device='cuda:1'), covar=tensor([0.0194, 0.0372, 0.0503, 0.0453, 0.0242, 0.0378, 0.0235, 0.0278], device='cuda:1'), in_proj_covar=tensor([0.0193, 0.0224, 0.0217, 0.0217, 0.0225, 0.0225, 0.0223, 0.0220], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-01 03:40:08,502 INFO [optim.py:368] (1/8) 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 03:40:42,687 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.9593, 3.8253, 4.0470, 4.1313, 4.2614, 3.8726, 4.1845, 4.2760], device='cuda:1'), covar=tensor([0.1722, 0.1307, 0.1412, 0.0759, 0.0606, 0.1574, 0.0804, 0.0787], device='cuda:1'), in_proj_covar=tensor([0.0609, 0.0753, 0.0881, 0.0767, 0.0581, 0.0603, 0.0620, 0.0721], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-01 03:40:44,318 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=201301.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 03:40:45,033 INFO [train.py:904] (1/8) Epoch 20, batch 8450, loss[loss=0.1895, simple_loss=0.2685, pruned_loss=0.05527, over 12585.00 frames. ], tot_loss[loss=0.1861, simple_loss=0.2781, pruned_loss=0.04706, over 3096776.71 frames. ], batch size: 247, lr: 3.35e-03, grad_scale: 8.0 2023-05-01 03:41:05,339 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=201314.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 03:41:39,424 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=201335.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 03:41:54,231 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([1.9834, 3.3334, 2.9831, 5.0714, 3.8099, 4.5236, 1.7129, 3.1829], device='cuda:1'), covar=tensor([0.1282, 0.0630, 0.0997, 0.0130, 0.0162, 0.0318, 0.1665, 0.0709], device='cuda:1'), in_proj_covar=tensor([0.0164, 0.0170, 0.0191, 0.0183, 0.0204, 0.0210, 0.0198, 0.0189], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-05-01 03:42:02,465 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 20, batch 8500, loss[loss=0.1773, simple_loss=0.2623, pruned_loss=0.04616, over 16396.00 frames. ], tot_loss[loss=0.1824, simple_loss=0.2749, pruned_loss=0.04498, over 3104616.94 frames. ], batch size: 146, lr: 3.35e-03, grad_scale: 8.0 2023-05-01 03:42:32,312 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.8855, 2.7297, 2.9294, 2.1193, 2.7187, 2.1816, 2.7908, 2.9422], device='cuda:1'), covar=tensor([0.0254, 0.0896, 0.0440, 0.1808, 0.0732, 0.0935, 0.0570, 0.0755], device='cuda:1'), in_proj_covar=tensor([0.0150, 0.0157, 0.0161, 0.0148, 0.0140, 0.0126, 0.0140, 0.0167], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:1') 2023-05-01 03:42:49,499 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.0412, 2.3276, 2.4259, 2.9363, 1.9427, 3.2050, 1.8254, 2.7901], device='cuda:1'), covar=tensor([0.1138, 0.0622, 0.0914, 0.0150, 0.0085, 0.0365, 0.1486, 0.0644], device='cuda:1'), in_proj_covar=tensor([0.0164, 0.0171, 0.0192, 0.0183, 0.0204, 0.0210, 0.0198, 0.0189], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-05-01 03:42:50,125 INFO [optim.py:368] (1/8) 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 03:43:02,033 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.6894, 4.8743, 5.0332, 4.8256, 4.8726, 5.4327, 4.9850, 4.6775], device='cuda:1'), covar=tensor([0.1104, 0.2015, 0.2132, 0.2110, 0.2626, 0.1011, 0.1567, 0.2597], device='cuda:1'), in_proj_covar=tensor([0.0389, 0.0567, 0.0626, 0.0469, 0.0623, 0.0656, 0.0489, 0.0634], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-05-01 03:43:31,052 INFO [train.py:904] (1/8) Epoch 20, batch 8550, loss[loss=0.1655, simple_loss=0.265, pruned_loss=0.03304, over 16856.00 frames. ], tot_loss[loss=0.1808, simple_loss=0.2729, pruned_loss=0.04432, over 3081128.06 frames. ], batch size: 96, lr: 3.35e-03, grad_scale: 8.0 2023-05-01 03:44:18,623 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=201427.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 03:44:27,610 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.4478, 1.9555, 1.6930, 1.6834, 2.2533, 1.8970, 1.9226, 2.3368], device='cuda:1'), covar=tensor([0.0180, 0.0406, 0.0531, 0.0444, 0.0254, 0.0377, 0.0176, 0.0251], device='cuda:1'), in_proj_covar=tensor([0.0191, 0.0223, 0.0215, 0.0215, 0.0224, 0.0224, 0.0222, 0.0219], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-01 03:45:09,639 INFO [train.py:904] (1/8) Epoch 20, batch 8600, loss[loss=0.1839, simple_loss=0.2833, pruned_loss=0.04227, over 16411.00 frames. ], tot_loss[loss=0.1793, simple_loss=0.2723, pruned_loss=0.04314, over 3081380.76 frames. ], batch size: 68, lr: 3.35e-03, grad_scale: 8.0 2023-05-01 03:46:02,940 INFO [optim.py:368] (1/8) 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 03:46:39,391 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.9287, 2.1235, 2.3576, 3.1896, 2.1436, 2.2793, 2.2827, 2.1511], device='cuda:1'), covar=tensor([0.1250, 0.3560, 0.2745, 0.0696, 0.4594, 0.2838, 0.3544, 0.4010], device='cuda:1'), in_proj_covar=tensor([0.0385, 0.0431, 0.0355, 0.0314, 0.0424, 0.0494, 0.0401, 0.0503], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-01 03:46:48,554 INFO [train.py:904] (1/8) Epoch 20, batch 8650, loss[loss=0.1741, simple_loss=0.2638, pruned_loss=0.04216, over 12377.00 frames. ], tot_loss[loss=0.1766, simple_loss=0.27, pruned_loss=0.04163, over 3079861.49 frames. ], batch size: 246, lr: 3.35e-03, grad_scale: 4.0 2023-05-01 03:47:14,329 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.4025, 3.3541, 3.4822, 3.5136, 3.5619, 3.2691, 3.5327, 3.6055], device='cuda:1'), covar=tensor([0.1306, 0.0932, 0.0991, 0.0603, 0.0637, 0.2440, 0.0821, 0.0819], device='cuda:1'), in_proj_covar=tensor([0.0603, 0.0745, 0.0871, 0.0758, 0.0576, 0.0597, 0.0615, 0.0712], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-01 03:48:36,194 INFO [train.py:904] (1/8) Epoch 20, batch 8700, loss[loss=0.1928, simple_loss=0.2901, pruned_loss=0.04777, over 15305.00 frames. ], tot_loss[loss=0.1744, simple_loss=0.2677, pruned_loss=0.04056, over 3069147.19 frames. ], batch size: 190, lr: 3.35e-03, grad_scale: 4.0 2023-05-01 03:49:26,357 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.93 vs. limit=2.0 2023-05-01 03:49:28,968 INFO [optim.py:368] (1/8) 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 03:50:13,601 INFO [train.py:904] (1/8) Epoch 20, batch 8750, loss[loss=0.1722, simple_loss=0.2802, pruned_loss=0.0321, over 16845.00 frames. ], tot_loss[loss=0.1742, simple_loss=0.2674, pruned_loss=0.04047, over 3062866.60 frames. ], batch size: 102, lr: 3.35e-03, grad_scale: 4.0 2023-05-01 03:50:31,552 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=201609.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 03:51:21,442 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=201630.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 03:51:51,022 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 20, batch 8800, loss[loss=0.1697, simple_loss=0.2591, pruned_loss=0.04017, over 12677.00 frames. ], tot_loss[loss=0.1728, simple_loss=0.2663, pruned_loss=0.03966, over 3054249.01 frames. ], batch size: 247, lr: 3.35e-03, grad_scale: 4.0 2023-05-01 03:52:24,878 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=201661.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 03:52:34,180 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.71 vs. limit=2.0 2023-05-01 03:53:06,515 INFO [optim.py:368] (1/8) Clipping_scale=2.0, grad-norm quartiles 1.591e+02 2.243e+02 2.623e+02 3.212e+02 5.763e+02, threshold=5.246e+02, percent-clipped=4.0 2023-05-01 03:53:51,631 INFO [train.py:904] (1/8) Epoch 20, batch 8850, loss[loss=0.1718, simple_loss=0.2789, pruned_loss=0.03237, over 16757.00 frames. ], tot_loss[loss=0.1737, simple_loss=0.2689, pruned_loss=0.0392, over 3036894.86 frames. ], batch size: 124, lr: 3.35e-03, grad_scale: 4.0 2023-05-01 03:53:57,893 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=201705.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 03:53:59,698 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=201706.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 03:54:34,761 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=201722.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 03:54:45,906 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=201727.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 03:55:38,107 INFO [train.py:904] (1/8) Epoch 20, batch 8900, loss[loss=0.1721, simple_loss=0.2688, pruned_loss=0.03767, over 16248.00 frames. ], tot_loss[loss=0.1726, simple_loss=0.2687, pruned_loss=0.03822, over 3041161.08 frames. ], batch size: 166, lr: 3.35e-03, grad_scale: 2.0 2023-05-01 03:56:08,834 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=201767.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 03:56:27,847 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=201775.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 03:56:46,606 INFO [optim.py:368] (1/8) Clipping_scale=2.0, grad-norm quartiles 1.534e+02 2.216e+02 2.705e+02 3.264e+02 5.822e+02, threshold=5.410e+02, percent-clipped=1.0 2023-05-01 03:57:41,715 INFO [train.py:904] (1/8) Epoch 20, batch 8950, loss[loss=0.1639, simple_loss=0.2579, pruned_loss=0.03493, over 16230.00 frames. ], tot_loss[loss=0.1729, simple_loss=0.2686, pruned_loss=0.03853, over 3045847.39 frames. ], batch size: 165, lr: 3.35e-03, grad_scale: 2.0 2023-05-01 03:58:48,779 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.20 vs. limit=2.0 2023-05-01 03:59:24,358 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([1.7591, 2.6892, 2.5025, 4.1616, 2.6553, 3.9826, 1.4690, 3.0608], device='cuda:1'), covar=tensor([0.1288, 0.0731, 0.1138, 0.0135, 0.0136, 0.0368, 0.1660, 0.0633], device='cuda:1'), in_proj_covar=tensor([0.0163, 0.0169, 0.0190, 0.0180, 0.0201, 0.0208, 0.0197, 0.0188], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-05-01 03:59:30,696 INFO [train.py:904] (1/8) Epoch 20, batch 9000, loss[loss=0.1675, simple_loss=0.2581, pruned_loss=0.03846, over 16665.00 frames. ], tot_loss[loss=0.1701, simple_loss=0.2654, pruned_loss=0.03741, over 3051013.60 frames. ], batch size: 134, lr: 3.35e-03, grad_scale: 2.0 2023-05-01 03:59:30,696 INFO [train.py:929] (1/8) Computing validation loss 2023-05-01 03:59:41,114 INFO [train.py:938] (1/8) Epoch 20, validation: loss=0.1464, simple_loss=0.2502, pruned_loss=0.02125, over 944034.00 frames. 2023-05-01 03:59:41,115 INFO [train.py:939] (1/8) Maximum memory allocated so far is 18145MB 2023-05-01 04:00:41,793 INFO [optim.py:368] (1/8) Clipping_scale=2.0, grad-norm quartiles 1.381e+02 2.083e+02 2.552e+02 3.280e+02 1.556e+03, threshold=5.104e+02, percent-clipped=3.0 2023-05-01 04:01:12,770 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.18 vs. limit=2.0 2023-05-01 04:01:24,589 INFO [train.py:904] (1/8) Epoch 20, batch 9050, loss[loss=0.1778, simple_loss=0.2659, pruned_loss=0.04485, over 12517.00 frames. ], tot_loss[loss=0.1704, simple_loss=0.2659, pruned_loss=0.03742, over 3071782.47 frames. ], batch size: 248, lr: 3.35e-03, grad_scale: 2.0 2023-05-01 04:01:40,826 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=201909.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 04:02:21,257 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=201930.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 04:03:12,548 INFO [train.py:904] (1/8) Epoch 20, batch 9100, loss[loss=0.1885, simple_loss=0.29, pruned_loss=0.04345, over 16264.00 frames. ], tot_loss[loss=0.1708, simple_loss=0.2655, pruned_loss=0.03803, over 3065034.75 frames. ], batch size: 165, lr: 3.35e-03, grad_scale: 2.0 2023-05-01 04:03:22,655 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=201957.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 04:04:14,332 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=201978.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 04:04:21,616 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.4401, 3.5088, 2.7352, 2.1051, 2.2065, 2.4202, 3.6589, 3.1277], device='cuda:1'), covar=tensor([0.3147, 0.0719, 0.1836, 0.3124, 0.2946, 0.2106, 0.0461, 0.1418], device='cuda:1'), in_proj_covar=tensor([0.0318, 0.0262, 0.0296, 0.0300, 0.0284, 0.0250, 0.0284, 0.0322], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:1') 2023-05-01 04:04:23,056 INFO [optim.py:368] (1/8) 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 04:04:42,871 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=201990.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 04:05:11,349 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 20, batch 9150, loss[loss=0.1659, simple_loss=0.2606, pruned_loss=0.03564, over 15362.00 frames. ], tot_loss[loss=0.171, simple_loss=0.2662, pruned_loss=0.03785, over 3075649.19 frames. ], batch size: 191, lr: 3.35e-03, grad_scale: 2.0 2023-05-01 04:05:48,180 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=202017.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 04:06:58,616 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 20, batch 9200, loss[loss=0.177, simple_loss=0.2674, pruned_loss=0.04324, over 16854.00 frames. ], tot_loss[loss=0.168, simple_loss=0.2622, pruned_loss=0.03695, over 3068173.28 frames. ], batch size: 116, lr: 3.35e-03, grad_scale: 4.0 2023-05-01 04:07:07,635 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.8522, 4.0034, 4.2265, 4.1908, 4.2648, 4.0221, 3.9219, 3.9830], device='cuda:1'), covar=tensor([0.0554, 0.0865, 0.0697, 0.0764, 0.0643, 0.0629, 0.1297, 0.0572], device='cuda:1'), in_proj_covar=tensor([0.0389, 0.0427, 0.0418, 0.0387, 0.0461, 0.0436, 0.0521, 0.0350], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:1') 2023-05-01 04:07:19,865 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=202062.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 04:07:55,479 INFO [optim.py:368] (1/8) 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 04:08:37,608 INFO [train.py:904] (1/8) Epoch 20, batch 9250, loss[loss=0.1661, simple_loss=0.2591, pruned_loss=0.0366, over 16594.00 frames. ], tot_loss[loss=0.1686, simple_loss=0.2624, pruned_loss=0.03738, over 3058963.58 frames. ], batch size: 62, lr: 3.35e-03, grad_scale: 4.0 2023-05-01 04:09:37,498 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.4496, 2.3277, 2.3203, 4.2658, 2.2743, 2.7107, 2.4541, 2.4537], device='cuda:1'), covar=tensor([0.1096, 0.3627, 0.2945, 0.0394, 0.4116, 0.2439, 0.3436, 0.3505], device='cuda:1'), in_proj_covar=tensor([0.0386, 0.0432, 0.0357, 0.0315, 0.0425, 0.0495, 0.0403, 0.0504], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-01 04:10:29,970 INFO [train.py:904] (1/8) Epoch 20, batch 9300, loss[loss=0.1701, simple_loss=0.2505, pruned_loss=0.04483, over 12080.00 frames. ], tot_loss[loss=0.1674, simple_loss=0.261, pruned_loss=0.03687, over 3069793.39 frames. ], batch size: 248, lr: 3.35e-03, grad_scale: 4.0 2023-05-01 04:11:37,344 INFO [optim.py:368] (1/8) 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 04:12:16,136 INFO [train.py:904] (1/8) Epoch 20, batch 9350, loss[loss=0.1717, simple_loss=0.2627, pruned_loss=0.04036, over 16629.00 frames. ], tot_loss[loss=0.1675, simple_loss=0.261, pruned_loss=0.03702, over 3079437.35 frames. ], batch size: 62, lr: 3.35e-03, grad_scale: 4.0 2023-05-01 04:12:38,589 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.6152, 3.6574, 3.4906, 3.2465, 3.2969, 3.5799, 3.3245, 3.4527], device='cuda:1'), covar=tensor([0.0528, 0.0673, 0.0349, 0.0266, 0.0539, 0.0490, 0.1287, 0.0485], device='cuda:1'), in_proj_covar=tensor([0.0275, 0.0393, 0.0319, 0.0314, 0.0325, 0.0363, 0.0220, 0.0378], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-01 04:13:59,184 INFO [train.py:904] (1/8) Epoch 20, batch 9400, loss[loss=0.1839, simple_loss=0.2894, pruned_loss=0.0392, over 16671.00 frames. ], tot_loss[loss=0.1665, simple_loss=0.2597, pruned_loss=0.03665, over 3051136.04 frames. ], batch size: 134, lr: 3.35e-03, grad_scale: 4.0 2023-05-01 04:15:00,973 INFO [optim.py:368] (1/8) 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 04:15:39,389 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=202300.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 04:15:41,531 INFO [train.py:904] (1/8) Epoch 20, batch 9450, loss[loss=0.1743, simple_loss=0.266, pruned_loss=0.04131, over 12475.00 frames. ], tot_loss[loss=0.1677, simple_loss=0.2615, pruned_loss=0.03701, over 3033874.62 frames. ], batch size: 248, lr: 3.35e-03, grad_scale: 4.0 2023-05-01 04:16:12,933 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=202317.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 04:16:25,834 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.0316, 3.1057, 1.6362, 3.3031, 2.2817, 3.2410, 1.7851, 2.4640], device='cuda:1'), covar=tensor([0.0331, 0.0373, 0.1967, 0.0257, 0.0939, 0.0604, 0.2072, 0.0826], device='cuda:1'), in_proj_covar=tensor([0.0160, 0.0167, 0.0187, 0.0149, 0.0169, 0.0203, 0.0195, 0.0169], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0003, 0.0004, 0.0004, 0.0003], device='cuda:1') 2023-05-01 04:16:54,439 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=3.20 vs. limit=5.0 2023-05-01 04:17:14,533 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=202346.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 04:17:17,847 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=202348.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 04:17:25,743 INFO [train.py:904] (1/8) Epoch 20, batch 9500, loss[loss=0.1696, simple_loss=0.2682, pruned_loss=0.03553, over 16796.00 frames. ], tot_loss[loss=0.1669, simple_loss=0.2605, pruned_loss=0.03669, over 3037424.08 frames. ], batch size: 124, lr: 3.34e-03, grad_scale: 4.0 2023-05-01 04:17:49,552 INFO [zipformer.py:625] (1/8) 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] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=202365.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 04:18:26,629 INFO [optim.py:368] (1/8) 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 04:19:13,220 INFO [train.py:904] (1/8) Epoch 20, batch 9550, loss[loss=0.184, simple_loss=0.2862, pruned_loss=0.04083, over 15400.00 frames. ], tot_loss[loss=0.167, simple_loss=0.2605, pruned_loss=0.03677, over 3035520.15 frames. ], batch size: 191, lr: 3.34e-03, grad_scale: 4.0 2023-05-01 04:19:32,643 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=202410.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 04:20:28,261 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.23 vs. limit=2.0 2023-05-01 04:20:55,235 INFO [train.py:904] (1/8) Epoch 20, batch 9600, loss[loss=0.1958, simple_loss=0.3065, pruned_loss=0.04259, over 15369.00 frames. ], tot_loss[loss=0.1683, simple_loss=0.262, pruned_loss=0.03732, over 3039842.64 frames. ], batch size: 190, lr: 3.34e-03, grad_scale: 8.0 2023-05-01 04:21:53,895 INFO [optim.py:368] (1/8) 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] (1/8) Epoch 20, batch 9650, loss[loss=0.1638, simple_loss=0.2554, pruned_loss=0.03611, over 12500.00 frames. ], tot_loss[loss=0.1698, simple_loss=0.2642, pruned_loss=0.0377, over 3040909.26 frames. ], batch size: 248, lr: 3.34e-03, grad_scale: 8.0 2023-05-01 04:24:31,592 INFO [train.py:904] (1/8) Epoch 20, batch 9700, loss[loss=0.189, simple_loss=0.286, pruned_loss=0.04596, over 16758.00 frames. ], tot_loss[loss=0.1689, simple_loss=0.2627, pruned_loss=0.0375, over 3032384.10 frames. ], batch size: 124, lr: 3.34e-03, grad_scale: 4.0 2023-05-01 04:24:32,378 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.7751, 4.9851, 5.0942, 4.9247, 4.9843, 5.4996, 4.9978, 4.6860], device='cuda:1'), covar=tensor([0.1021, 0.2041, 0.2161, 0.1974, 0.2483, 0.0880, 0.1545, 0.2363], device='cuda:1'), in_proj_covar=tensor([0.0377, 0.0552, 0.0607, 0.0455, 0.0605, 0.0636, 0.0477, 0.0612], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-05-01 04:24:41,718 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.9429, 2.1026, 2.3711, 3.2083, 2.1720, 2.2888, 2.3025, 2.2179], device='cuda:1'), covar=tensor([0.1243, 0.3646, 0.2537, 0.0695, 0.4405, 0.2582, 0.3430, 0.3503], device='cuda:1'), in_proj_covar=tensor([0.0386, 0.0431, 0.0356, 0.0315, 0.0426, 0.0492, 0.0402, 0.0502], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-01 04:25:07,500 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.9750, 4.2738, 4.1035, 4.1380, 3.8018, 3.8551, 3.8138, 4.2735], device='cuda:1'), covar=tensor([0.1109, 0.0925, 0.0987, 0.0835, 0.0849, 0.1654, 0.0972, 0.0984], device='cuda:1'), in_proj_covar=tensor([0.0625, 0.0761, 0.0623, 0.0572, 0.0487, 0.0494, 0.0638, 0.0593], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:1') 2023-05-01 04:25:26,422 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=4.82 vs. limit=5.0 2023-05-01 04:25:37,517 INFO [optim.py:368] (1/8) 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 04:26:15,966 INFO [train.py:904] (1/8) Epoch 20, batch 9750, loss[loss=0.1655, simple_loss=0.2659, pruned_loss=0.03253, over 16393.00 frames. ], tot_loss[loss=0.1679, simple_loss=0.2615, pruned_loss=0.03716, over 3051736.91 frames. ], batch size: 146, lr: 3.34e-03, grad_scale: 4.0 2023-05-01 04:26:33,779 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.26 vs. limit=2.0 2023-05-01 04:26:58,184 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.8317, 5.0472, 5.2140, 5.0126, 5.1120, 5.6161, 5.1208, 4.7901], device='cuda:1'), covar=tensor([0.0945, 0.2162, 0.2467, 0.1781, 0.2069, 0.0879, 0.1639, 0.2362], device='cuda:1'), in_proj_covar=tensor([0.0376, 0.0551, 0.0607, 0.0455, 0.0605, 0.0635, 0.0476, 0.0612], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-05-01 04:27:47,502 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 20, batch 9800, loss[loss=0.1734, simple_loss=0.2785, pruned_loss=0.03408, over 15427.00 frames. ], tot_loss[loss=0.1678, simple_loss=0.2623, pruned_loss=0.03665, over 3063154.21 frames. ], batch size: 190, lr: 3.34e-03, grad_scale: 2.0 2023-05-01 04:28:00,983 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=2.95 vs. limit=5.0 2023-05-01 04:28:57,878 INFO [optim.py:368] (1/8) 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 04:29:04,024 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([1.7914, 3.0831, 2.8928, 5.0257, 3.8442, 4.4632, 1.7084, 3.1791], device='cuda:1'), covar=tensor([0.1385, 0.0724, 0.1067, 0.0126, 0.0184, 0.0323, 0.1634, 0.0738], device='cuda:1'), in_proj_covar=tensor([0.0164, 0.0169, 0.0190, 0.0179, 0.0197, 0.0208, 0.0197, 0.0187], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-05-01 04:29:26,527 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 20, batch 9850, loss[loss=0.1629, simple_loss=0.258, pruned_loss=0.03396, over 16094.00 frames. ], tot_loss[loss=0.1677, simple_loss=0.263, pruned_loss=0.03623, over 3065116.80 frames. ], batch size: 165, lr: 3.34e-03, grad_scale: 2.0 2023-05-01 04:31:34,887 INFO [train.py:904] (1/8) Epoch 20, batch 9900, loss[loss=0.1871, simple_loss=0.283, pruned_loss=0.0456, over 16916.00 frames. ], tot_loss[loss=0.1681, simple_loss=0.2637, pruned_loss=0.03626, over 3064323.75 frames. ], batch size: 116, lr: 3.34e-03, grad_scale: 2.0 2023-05-01 04:32:49,188 INFO [optim.py:368] (1/8) 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 04:33:30,080 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.6862, 4.2694, 3.0037, 2.4294, 2.7597, 2.7014, 4.4344, 3.6968], device='cuda:1'), covar=tensor([0.2815, 0.0433, 0.1796, 0.2717, 0.2292, 0.1766, 0.0357, 0.0967], device='cuda:1'), in_proj_covar=tensor([0.0316, 0.0260, 0.0295, 0.0299, 0.0281, 0.0248, 0.0282, 0.0321], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:1') 2023-05-01 04:33:30,682 INFO [train.py:904] (1/8) Epoch 20, batch 9950, loss[loss=0.1564, simple_loss=0.2513, pruned_loss=0.03072, over 16458.00 frames. ], tot_loss[loss=0.1692, simple_loss=0.2654, pruned_loss=0.03655, over 3071791.59 frames. ], batch size: 68, lr: 3.34e-03, grad_scale: 2.0 2023-05-01 04:34:41,346 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=202830.0, num_to_drop=1, layers_to_drop={1} 2023-05-01 04:34:45,230 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.7897, 5.0932, 4.9385, 4.9277, 4.6360, 4.6083, 4.5019, 5.1693], device='cuda:1'), covar=tensor([0.1117, 0.0828, 0.0758, 0.0743, 0.0720, 0.1013, 0.0989, 0.0777], device='cuda:1'), in_proj_covar=tensor([0.0621, 0.0759, 0.0620, 0.0570, 0.0485, 0.0492, 0.0637, 0.0591], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:1') 2023-05-01 04:35:32,300 INFO [train.py:904] (1/8) Epoch 20, batch 10000, loss[loss=0.2035, simple_loss=0.2928, pruned_loss=0.0571, over 12827.00 frames. ], tot_loss[loss=0.1682, simple_loss=0.2641, pruned_loss=0.0361, over 3104744.39 frames. ], batch size: 250, lr: 3.34e-03, grad_scale: 4.0 2023-05-01 04:36:36,180 INFO [optim.py:368] (1/8) 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 04:36:49,037 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.4437, 3.4687, 2.6069, 2.1387, 2.1815, 2.2497, 3.5933, 3.0443], device='cuda:1'), covar=tensor([0.2839, 0.0562, 0.1805, 0.2654, 0.2779, 0.2145, 0.0388, 0.1262], device='cuda:1'), in_proj_covar=tensor([0.0314, 0.0258, 0.0293, 0.0297, 0.0279, 0.0247, 0.0281, 0.0319], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:1') 2023-05-01 04:36:51,640 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=202891.0, num_to_drop=1, layers_to_drop={1} 2023-05-01 04:37:11,275 INFO [train.py:904] (1/8) Epoch 20, batch 10050, loss[loss=0.1804, simple_loss=0.2668, pruned_loss=0.04706, over 12044.00 frames. ], tot_loss[loss=0.1686, simple_loss=0.2647, pruned_loss=0.03626, over 3106379.41 frames. ], batch size: 246, lr: 3.34e-03, grad_scale: 4.0 2023-05-01 04:38:41,706 INFO [train.py:904] (1/8) Epoch 20, batch 10100, loss[loss=0.1472, simple_loss=0.2308, pruned_loss=0.03185, over 12762.00 frames. ], tot_loss[loss=0.1692, simple_loss=0.265, pruned_loss=0.03675, over 3075510.23 frames. ], batch size: 247, lr: 3.34e-03, grad_scale: 4.0 2023-05-01 04:39:40,380 INFO [optim.py:368] (1/8) 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 04:39:49,275 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=4.89 vs. limit=5.0 2023-05-01 04:40:23,118 INFO [train.py:904] (1/8) Epoch 21, batch 0, loss[loss=0.1796, simple_loss=0.2708, pruned_loss=0.04419, over 17212.00 frames. ], tot_loss[loss=0.1796, simple_loss=0.2708, pruned_loss=0.04419, over 17212.00 frames. ], batch size: 46, lr: 3.26e-03, grad_scale: 8.0 2023-05-01 04:40:23,119 INFO [train.py:929] (1/8) Computing validation loss 2023-05-01 04:40:30,890 INFO [train.py:938] (1/8) Epoch 21, validation: loss=0.1451, simple_loss=0.2491, pruned_loss=0.02058, over 944034.00 frames. 2023-05-01 04:40:30,890 INFO [train.py:939] (1/8) Maximum memory allocated so far is 18145MB 2023-05-01 04:40:49,570 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.4628, 4.6049, 4.7366, 4.4993, 4.5331, 5.1442, 4.7409, 4.4161], device='cuda:1'), covar=tensor([0.1623, 0.2145, 0.2314, 0.2355, 0.2738, 0.1213, 0.1574, 0.2544], device='cuda:1'), in_proj_covar=tensor([0.0373, 0.0549, 0.0604, 0.0456, 0.0604, 0.0637, 0.0474, 0.0610], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-05-01 04:41:11,098 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.7895, 3.7488, 3.9102, 3.6280, 3.8472, 4.2777, 3.9306, 3.5802], device='cuda:1'), covar=tensor([0.2011, 0.2643, 0.2509, 0.2780, 0.2847, 0.1633, 0.1659, 0.2793], device='cuda:1'), in_proj_covar=tensor([0.0375, 0.0551, 0.0608, 0.0458, 0.0608, 0.0640, 0.0476, 0.0613], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-05-01 04:41:19,662 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=203038.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 04:41:21,231 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.96 vs. limit=2.0 2023-05-01 04:41:36,911 INFO [train.py:904] (1/8) Epoch 21, batch 50, loss[loss=0.164, simple_loss=0.2595, pruned_loss=0.03424, over 17222.00 frames. ], tot_loss[loss=0.1847, simple_loss=0.2679, pruned_loss=0.05077, over 747964.37 frames. ], batch size: 46, lr: 3.26e-03, grad_scale: 2.0 2023-05-01 04:42:24,742 INFO [zipformer.py:625] (1/8) 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] (1/8) 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 04:42:43,822 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 21, batch 100, loss[loss=0.1821, simple_loss=0.269, pruned_loss=0.04756, over 16832.00 frames. ], tot_loss[loss=0.1834, simple_loss=0.2681, pruned_loss=0.04932, over 1311822.13 frames. ], batch size: 102, lr: 3.26e-03, grad_scale: 2.0 2023-05-01 04:43:40,152 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.3200, 3.0885, 3.3703, 1.8866, 3.4877, 3.4907, 2.7887, 2.6616], device='cuda:1'), covar=tensor([0.0796, 0.0244, 0.0194, 0.1112, 0.0107, 0.0204, 0.0450, 0.0463], device='cuda:1'), in_proj_covar=tensor([0.0145, 0.0106, 0.0093, 0.0137, 0.0077, 0.0119, 0.0125, 0.0128], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:1') 2023-05-01 04:43:50,274 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=203146.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 04:43:57,339 INFO [train.py:904] (1/8) Epoch 21, batch 150, loss[loss=0.2001, simple_loss=0.2787, pruned_loss=0.06077, over 16496.00 frames. ], tot_loss[loss=0.1812, simple_loss=0.2666, pruned_loss=0.04786, over 1763616.73 frames. ], batch size: 75, lr: 3.26e-03, grad_scale: 2.0 2023-05-01 04:44:03,666 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.8153, 4.5892, 4.8670, 4.9884, 5.1880, 4.6153, 5.1828, 5.1573], device='cuda:1'), covar=tensor([0.1812, 0.1206, 0.1577, 0.0773, 0.0535, 0.0917, 0.0608, 0.0612], device='cuda:1'), in_proj_covar=tensor([0.0605, 0.0740, 0.0863, 0.0764, 0.0574, 0.0593, 0.0614, 0.0710], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-01 04:44:44,682 INFO [optim.py:368] (1/8) 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] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=203186.0, num_to_drop=1, layers_to_drop={1} 2023-05-01 04:44:55,335 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.7794, 4.9098, 5.2679, 5.2712, 5.3431, 5.0121, 4.8380, 4.6980], device='cuda:1'), covar=tensor([0.0546, 0.0582, 0.0611, 0.0542, 0.0680, 0.0564, 0.1459, 0.0533], device='cuda:1'), in_proj_covar=tensor([0.0387, 0.0427, 0.0415, 0.0386, 0.0459, 0.0437, 0.0520, 0.0349], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:1') 2023-05-01 04:45:04,855 INFO [train.py:904] (1/8) Epoch 21, batch 200, loss[loss=0.1747, simple_loss=0.2657, pruned_loss=0.04181, over 16511.00 frames. ], tot_loss[loss=0.1798, simple_loss=0.2658, pruned_loss=0.04689, over 2110738.35 frames. ], batch size: 68, lr: 3.26e-03, grad_scale: 2.0 2023-05-01 04:46:15,294 INFO [train.py:904] (1/8) Epoch 21, batch 250, loss[loss=0.1464, simple_loss=0.2289, pruned_loss=0.0319, over 16883.00 frames. ], tot_loss[loss=0.1776, simple_loss=0.2633, pruned_loss=0.04599, over 2385312.56 frames. ], batch size: 42, lr: 3.26e-03, grad_scale: 2.0 2023-05-01 04:46:37,638 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.7725, 3.8985, 2.5505, 4.4330, 3.0270, 4.4097, 2.4988, 3.1792], device='cuda:1'), covar=tensor([0.0327, 0.0390, 0.1533, 0.0369, 0.0864, 0.0492, 0.1576, 0.0764], device='cuda:1'), in_proj_covar=tensor([0.0167, 0.0174, 0.0193, 0.0157, 0.0175, 0.0210, 0.0203, 0.0176], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-05-01 04:46:44,160 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.8548, 4.9809, 5.3840, 5.3738, 5.3774, 5.0495, 4.9649, 4.7292], device='cuda:1'), covar=tensor([0.0367, 0.0548, 0.0361, 0.0343, 0.0469, 0.0402, 0.0994, 0.0508], device='cuda:1'), in_proj_covar=tensor([0.0388, 0.0429, 0.0416, 0.0388, 0.0460, 0.0438, 0.0521, 0.0351], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:1') 2023-05-01 04:46:47,735 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.9504, 2.0693, 2.4831, 2.8555, 2.6947, 3.3219, 2.4524, 3.3508], device='cuda:1'), covar=tensor([0.0261, 0.0487, 0.0327, 0.0327, 0.0348, 0.0190, 0.0424, 0.0150], device='cuda:1'), in_proj_covar=tensor([0.0180, 0.0187, 0.0174, 0.0178, 0.0190, 0.0147, 0.0191, 0.0142], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-01 04:47:00,828 INFO [optim.py:368] (1/8) 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] (1/8) Epoch 21, batch 300, loss[loss=0.1879, simple_loss=0.2707, pruned_loss=0.05258, over 16094.00 frames. ], tot_loss[loss=0.176, simple_loss=0.2612, pruned_loss=0.04534, over 2596142.32 frames. ], batch size: 164, lr: 3.26e-03, grad_scale: 2.0 2023-05-01 04:47:45,392 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=3.37 vs. limit=5.0 2023-05-01 04:47:51,443 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.83 vs. limit=2.0 2023-05-01 04:48:10,327 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.37 vs. limit=2.0 2023-05-01 04:48:32,603 INFO [train.py:904] (1/8) Epoch 21, batch 350, loss[loss=0.2138, simple_loss=0.2818, pruned_loss=0.07288, over 16726.00 frames. ], tot_loss[loss=0.1736, simple_loss=0.2581, pruned_loss=0.04457, over 2741884.00 frames. ], batch size: 134, lr: 3.25e-03, grad_scale: 2.0 2023-05-01 04:48:41,976 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.8343, 4.2812, 3.0246, 2.3594, 2.7160, 2.5731, 4.6086, 3.5477], device='cuda:1'), covar=tensor([0.2845, 0.0588, 0.1768, 0.2955, 0.2795, 0.2052, 0.0350, 0.1384], device='cuda:1'), in_proj_covar=tensor([0.0322, 0.0266, 0.0300, 0.0306, 0.0288, 0.0254, 0.0289, 0.0330], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:1') 2023-05-01 04:49:17,934 INFO [optim.py:368] (1/8) 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 04:49:30,088 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=203394.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 04:49:41,954 INFO [train.py:904] (1/8) Epoch 21, batch 400, loss[loss=0.1748, simple_loss=0.2733, pruned_loss=0.03816, over 17244.00 frames. ], tot_loss[loss=0.1721, simple_loss=0.2563, pruned_loss=0.04397, over 2873174.76 frames. ], batch size: 52, lr: 3.25e-03, grad_scale: 4.0 2023-05-01 04:49:47,426 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.45 vs. limit=2.0 2023-05-01 04:49:48,200 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.20 vs. limit=2.0 2023-05-01 04:49:59,028 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([1.8405, 1.3131, 1.6209, 1.6605, 1.7032, 1.9365, 1.6202, 1.7501], device='cuda:1'), covar=tensor([0.0260, 0.0430, 0.0254, 0.0327, 0.0334, 0.0231, 0.0454, 0.0177], device='cuda:1'), in_proj_covar=tensor([0.0182, 0.0189, 0.0176, 0.0180, 0.0192, 0.0149, 0.0193, 0.0144], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-01 04:50:36,378 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=203441.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 04:50:47,447 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.8138, 5.0501, 5.2222, 5.0102, 5.0595, 5.6977, 5.1905, 4.8328], device='cuda:1'), covar=tensor([0.1270, 0.2176, 0.2246, 0.2281, 0.2867, 0.1188, 0.1963, 0.2716], device='cuda:1'), in_proj_covar=tensor([0.0395, 0.0578, 0.0636, 0.0477, 0.0637, 0.0669, 0.0500, 0.0639], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-05-01 04:50:50,734 INFO [train.py:904] (1/8) Epoch 21, batch 450, loss[loss=0.1666, simple_loss=0.256, pruned_loss=0.03861, over 16661.00 frames. ], tot_loss[loss=0.1699, simple_loss=0.2545, pruned_loss=0.04264, over 2972997.34 frames. ], batch size: 62, lr: 3.25e-03, grad_scale: 4.0 2023-05-01 04:50:55,500 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.6237, 3.7164, 2.9082, 2.1733, 2.3682, 2.3039, 3.7868, 3.2105], device='cuda:1'), covar=tensor([0.2746, 0.0595, 0.1630, 0.2843, 0.2708, 0.2155, 0.0500, 0.1522], device='cuda:1'), in_proj_covar=tensor([0.0324, 0.0267, 0.0303, 0.0308, 0.0291, 0.0256, 0.0291, 0.0333], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:1') 2023-05-01 04:51:11,011 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.39 vs. limit=2.0 2023-05-01 04:51:11,829 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=203467.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 04:51:37,694 INFO [optim.py:368] (1/8) 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 04:51:38,101 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=203486.0, num_to_drop=1, layers_to_drop={0} 2023-05-01 04:51:59,114 INFO [train.py:904] (1/8) Epoch 21, batch 500, loss[loss=0.1532, simple_loss=0.249, pruned_loss=0.02876, over 17047.00 frames. ], tot_loss[loss=0.1683, simple_loss=0.2527, pruned_loss=0.04192, over 3057128.72 frames. ], batch size: 50, lr: 3.25e-03, grad_scale: 2.0 2023-05-01 04:52:36,243 INFO [zipformer.py:625] (1/8) 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] (1/8) 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] (1/8) Epoch 21, batch 550, loss[loss=0.1331, simple_loss=0.2254, pruned_loss=0.0204, over 17238.00 frames. ], tot_loss[loss=0.1674, simple_loss=0.2521, pruned_loss=0.04139, over 3113539.75 frames. ], batch size: 44, lr: 3.25e-03, grad_scale: 2.0 2023-05-01 04:53:56,912 INFO [optim.py:368] (1/8) 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] (1/8) Epoch 21, batch 600, loss[loss=0.2023, simple_loss=0.2697, pruned_loss=0.06742, over 16719.00 frames. ], tot_loss[loss=0.1676, simple_loss=0.2519, pruned_loss=0.04165, over 3163639.85 frames. ], batch size: 134, lr: 3.25e-03, grad_scale: 2.0 2023-05-01 04:55:27,190 INFO [train.py:904] (1/8) Epoch 21, batch 650, loss[loss=0.1785, simple_loss=0.2705, pruned_loss=0.04327, over 16626.00 frames. ], tot_loss[loss=0.1664, simple_loss=0.2502, pruned_loss=0.04132, over 3194419.45 frames. ], batch size: 62, lr: 3.25e-03, grad_scale: 2.0 2023-05-01 04:55:50,905 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.38 vs. limit=2.0 2023-05-01 04:56:01,755 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([5.1408, 5.7054, 5.8863, 5.5358, 5.7169, 6.2546, 5.7950, 5.5117], device='cuda:1'), covar=tensor([0.0971, 0.1952, 0.2493, 0.2280, 0.2684, 0.0965, 0.1465, 0.2213], device='cuda:1'), in_proj_covar=tensor([0.0399, 0.0584, 0.0641, 0.0483, 0.0644, 0.0674, 0.0503, 0.0646], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-05-01 04:56:14,238 INFO [optim.py:368] (1/8) 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 04:56:22,661 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=4.60 vs. limit=5.0 2023-05-01 04:56:24,939 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 21, batch 700, loss[loss=0.2049, simple_loss=0.2742, pruned_loss=0.06783, over 12582.00 frames. ], tot_loss[loss=0.1663, simple_loss=0.2508, pruned_loss=0.0409, over 3223677.97 frames. ], batch size: 248, lr: 3.25e-03, grad_scale: 2.0 2023-05-01 04:57:30,219 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=203741.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 04:57:31,298 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=203742.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 04:57:37,629 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=203747.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 04:57:44,448 INFO [train.py:904] (1/8) Epoch 21, batch 750, loss[loss=0.1853, simple_loss=0.2565, pruned_loss=0.05706, over 16374.00 frames. ], tot_loss[loss=0.167, simple_loss=0.2512, pruned_loss=0.04143, over 3242936.64 frames. ], batch size: 145, lr: 3.25e-03, grad_scale: 2.0 2023-05-01 04:58:03,820 INFO [zipformer.py:625] (1/8) 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] (1/8) 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] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=203789.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 04:58:54,762 INFO [train.py:904] (1/8) Epoch 21, batch 800, loss[loss=0.1852, simple_loss=0.2586, pruned_loss=0.05596, over 16866.00 frames. ], tot_loss[loss=0.167, simple_loss=0.2512, pruned_loss=0.04139, over 3255965.82 frames. ], batch size: 109, lr: 3.25e-03, grad_scale: 4.0 2023-05-01 04:59:03,043 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=203808.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 04:59:24,696 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=203823.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 04:59:29,161 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=4.12 vs. limit=5.0 2023-05-01 04:59:30,023 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=203827.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 04:59:58,149 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=203847.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 05:00:04,930 INFO [train.py:904] (1/8) Epoch 21, batch 850, loss[loss=0.1657, simple_loss=0.2436, pruned_loss=0.04389, over 16905.00 frames. ], tot_loss[loss=0.1661, simple_loss=0.2503, pruned_loss=0.04093, over 3261169.94 frames. ], batch size: 116, lr: 3.25e-03, grad_scale: 4.0 2023-05-01 05:00:09,693 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([5.1407, 5.6746, 5.8335, 5.4841, 5.7115, 6.2291, 5.7931, 5.4734], device='cuda:1'), covar=tensor([0.0948, 0.2092, 0.2197, 0.1893, 0.2437, 0.0996, 0.1442, 0.2141], device='cuda:1'), in_proj_covar=tensor([0.0401, 0.0589, 0.0649, 0.0487, 0.0650, 0.0680, 0.0509, 0.0653], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-05-01 05:00:29,532 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.45 vs. limit=2.0 2023-05-01 05:00:53,310 INFO [optim.py:368] (1/8) 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 05:01:11,266 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=203900.0, num_to_drop=1, layers_to_drop={1} 2023-05-01 05:01:12,968 INFO [train.py:904] (1/8) Epoch 21, batch 900, loss[loss=0.1679, simple_loss=0.2439, pruned_loss=0.04599, over 16779.00 frames. ], tot_loss[loss=0.1649, simple_loss=0.2495, pruned_loss=0.04013, over 3283564.77 frames. ], batch size: 102, lr: 3.25e-03, grad_scale: 4.0 2023-05-01 05:01:21,664 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=203908.0, num_to_drop=1, layers_to_drop={1} 2023-05-01 05:01:29,591 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=3.44 vs. limit=5.0 2023-05-01 05:02:01,449 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.7860, 2.7860, 2.4795, 2.8301, 3.1489, 2.8827, 3.4415, 3.2888], device='cuda:1'), covar=tensor([0.0133, 0.0410, 0.0502, 0.0401, 0.0292, 0.0376, 0.0254, 0.0269], device='cuda:1'), in_proj_covar=tensor([0.0206, 0.0238, 0.0227, 0.0227, 0.0237, 0.0235, 0.0238, 0.0233], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-01 05:02:07,243 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=4.59 vs. limit=5.0 2023-05-01 05:02:21,433 INFO [train.py:904] (1/8) Epoch 21, batch 950, loss[loss=0.1774, simple_loss=0.2494, pruned_loss=0.05265, over 16280.00 frames. ], tot_loss[loss=0.1651, simple_loss=0.2495, pruned_loss=0.04034, over 3282399.77 frames. ], batch size: 145, lr: 3.25e-03, grad_scale: 4.0 2023-05-01 05:02:33,135 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.8956, 3.9439, 2.5133, 4.6354, 3.0848, 4.5892, 2.5724, 3.2497], device='cuda:1'), covar=tensor([0.0337, 0.0450, 0.1735, 0.0309, 0.0908, 0.0542, 0.1616, 0.0772], device='cuda:1'), in_proj_covar=tensor([0.0170, 0.0178, 0.0195, 0.0162, 0.0178, 0.0215, 0.0204, 0.0179], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-05-01 05:02:35,158 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=203961.0, num_to_drop=1, layers_to_drop={3} 2023-05-01 05:02:40,863 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([1.7966, 2.4959, 2.5064, 3.8457, 3.0597, 3.9459, 1.5655, 2.7810], device='cuda:1'), covar=tensor([0.1452, 0.0824, 0.1247, 0.0216, 0.0166, 0.0435, 0.1726, 0.0915], device='cuda:1'), in_proj_covar=tensor([0.0166, 0.0173, 0.0193, 0.0187, 0.0204, 0.0214, 0.0200, 0.0191], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-05-01 05:03:10,185 INFO [optim.py:368] (1/8) 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 05:03:34,201 INFO [train.py:904] (1/8) Epoch 21, batch 1000, loss[loss=0.1474, simple_loss=0.2223, pruned_loss=0.03627, over 16856.00 frames. ], tot_loss[loss=0.165, simple_loss=0.2495, pruned_loss=0.04024, over 3288580.91 frames. ], batch size: 39, lr: 3.25e-03, grad_scale: 4.0 2023-05-01 05:03:44,205 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.9226, 2.0852, 2.4254, 2.7051, 2.7839, 2.7340, 2.0948, 2.9639], device='cuda:1'), covar=tensor([0.0173, 0.0443, 0.0317, 0.0287, 0.0276, 0.0300, 0.0502, 0.0163], device='cuda:1'), in_proj_covar=tensor([0.0185, 0.0192, 0.0178, 0.0182, 0.0194, 0.0152, 0.0195, 0.0147], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-01 05:04:10,903 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.1678, 2.3037, 2.7325, 3.0892, 2.9078, 3.6266, 2.3373, 3.6461], device='cuda:1'), covar=tensor([0.0247, 0.0468, 0.0346, 0.0321, 0.0344, 0.0180, 0.0509, 0.0156], device='cuda:1'), in_proj_covar=tensor([0.0185, 0.0192, 0.0178, 0.0182, 0.0194, 0.0152, 0.0195, 0.0147], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-01 05:04:37,391 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([5.2941, 5.2281, 5.1143, 4.5792, 4.7699, 5.1303, 5.1597, 4.7662], device='cuda:1'), covar=tensor([0.0567, 0.0428, 0.0308, 0.0362, 0.1044, 0.0430, 0.0288, 0.0844], device='cuda:1'), in_proj_covar=tensor([0.0293, 0.0419, 0.0339, 0.0335, 0.0348, 0.0390, 0.0233, 0.0408], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002], device='cuda:1') 2023-05-01 05:04:43,042 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.1831, 4.1595, 2.7212, 4.7558, 3.4570, 4.7456, 2.6999, 3.4651], device='cuda:1'), covar=tensor([0.0262, 0.0330, 0.1548, 0.0310, 0.0672, 0.0418, 0.1667, 0.0642], device='cuda:1'), in_proj_covar=tensor([0.0170, 0.0177, 0.0195, 0.0162, 0.0177, 0.0214, 0.0203, 0.0178], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-05-01 05:04:43,737 INFO [train.py:904] (1/8) Epoch 21, batch 1050, loss[loss=0.1565, simple_loss=0.2413, pruned_loss=0.03586, over 15869.00 frames. ], tot_loss[loss=0.1645, simple_loss=0.2486, pruned_loss=0.0402, over 3301178.32 frames. ], batch size: 35, lr: 3.25e-03, grad_scale: 4.0 2023-05-01 05:05:09,393 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.54 vs. limit=2.0 2023-05-01 05:05:10,374 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.9517, 4.5072, 4.5219, 3.3245, 3.7692, 4.3984, 4.0227, 2.6825], device='cuda:1'), covar=tensor([0.0468, 0.0060, 0.0038, 0.0326, 0.0132, 0.0091, 0.0087, 0.0452], device='cuda:1'), in_proj_covar=tensor([0.0139, 0.0084, 0.0083, 0.0135, 0.0099, 0.0110, 0.0094, 0.0130], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0002, 0.0004], device='cuda:1') 2023-05-01 05:05:19,337 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=204077.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 05:05:34,276 INFO [optim.py:368] (1/8) 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 05:05:54,731 INFO [train.py:904] (1/8) Epoch 21, batch 1100, loss[loss=0.166, simple_loss=0.2619, pruned_loss=0.03507, over 16689.00 frames. ], tot_loss[loss=0.1636, simple_loss=0.2482, pruned_loss=0.03946, over 3299792.40 frames. ], batch size: 57, lr: 3.25e-03, grad_scale: 4.0 2023-05-01 05:05:56,312 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=204103.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 05:06:00,765 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.9700, 2.0485, 2.4167, 2.8123, 2.8033, 2.9777, 2.1151, 3.0654], device='cuda:1'), covar=tensor([0.0198, 0.0452, 0.0369, 0.0273, 0.0308, 0.0239, 0.0491, 0.0177], device='cuda:1'), in_proj_covar=tensor([0.0185, 0.0192, 0.0179, 0.0183, 0.0194, 0.0152, 0.0195, 0.0147], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-01 05:06:24,053 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=204122.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 05:06:25,277 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=204123.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 05:06:45,647 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=204138.0, num_to_drop=1, layers_to_drop={2} 2023-05-01 05:07:05,256 INFO [train.py:904] (1/8) Epoch 21, batch 1150, loss[loss=0.1705, simple_loss=0.269, pruned_loss=0.03599, over 17023.00 frames. ], tot_loss[loss=0.1633, simple_loss=0.2482, pruned_loss=0.03919, over 3300064.62 frames. ], batch size: 50, lr: 3.25e-03, grad_scale: 4.0 2023-05-01 05:07:20,315 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=204163.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 05:07:31,656 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=204171.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 05:07:53,760 INFO [optim.py:368] (1/8) Clipping_scale=2.0, grad-norm quartiles 1.488e+02 2.065e+02 2.500e+02 2.935e+02 5.015e+02, threshold=5.000e+02, percent-clipped=1.0 2023-05-01 05:08:07,130 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.8208, 5.1750, 4.9402, 4.9188, 4.7018, 4.6772, 4.6126, 5.2476], device='cuda:1'), covar=tensor([0.1317, 0.0930, 0.1121, 0.0944, 0.0832, 0.1212, 0.1259, 0.0950], device='cuda:1'), in_proj_covar=tensor([0.0678, 0.0826, 0.0679, 0.0623, 0.0526, 0.0533, 0.0696, 0.0642], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:1') 2023-05-01 05:08:13,780 INFO [train.py:904] (1/8) Epoch 21, batch 1200, loss[loss=0.137, simple_loss=0.2287, pruned_loss=0.02263, over 16861.00 frames. ], tot_loss[loss=0.1625, simple_loss=0.2474, pruned_loss=0.03882, over 3310643.05 frames. ], batch size: 42, lr: 3.25e-03, grad_scale: 8.0 2023-05-01 05:08:15,241 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=204203.0, num_to_drop=1, layers_to_drop={1} 2023-05-01 05:08:15,335 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.2194, 4.2975, 4.1505, 3.9426, 3.6366, 4.3640, 4.0978, 4.0437], device='cuda:1'), covar=tensor([0.1120, 0.1145, 0.0597, 0.0511, 0.1509, 0.0674, 0.0930, 0.0847], device='cuda:1'), in_proj_covar=tensor([0.0298, 0.0426, 0.0344, 0.0341, 0.0352, 0.0397, 0.0237, 0.0414], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002], device='cuda:1') 2023-05-01 05:08:28,566 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([5.1826, 5.6672, 5.8335, 5.5113, 5.6851, 6.2063, 5.7399, 5.4294], device='cuda:1'), covar=tensor([0.0894, 0.2184, 0.2490, 0.2109, 0.2847, 0.0920, 0.1514, 0.2179], device='cuda:1'), in_proj_covar=tensor([0.0404, 0.0594, 0.0654, 0.0491, 0.0656, 0.0683, 0.0515, 0.0659], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-05-01 05:08:44,939 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=204224.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 05:09:24,353 INFO [train.py:904] (1/8) Epoch 21, batch 1250, loss[loss=0.1674, simple_loss=0.2443, pruned_loss=0.0452, over 16803.00 frames. ], tot_loss[loss=0.1627, simple_loss=0.2476, pruned_loss=0.03892, over 3309274.44 frames. ], batch size: 96, lr: 3.25e-03, grad_scale: 8.0 2023-05-01 05:09:30,687 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=204256.0, num_to_drop=1, layers_to_drop={1} 2023-05-01 05:10:12,294 INFO [optim.py:368] (1/8) 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 05:10:33,931 INFO [train.py:904] (1/8) Epoch 21, batch 1300, loss[loss=0.1673, simple_loss=0.2498, pruned_loss=0.04238, over 16795.00 frames. ], tot_loss[loss=0.1629, simple_loss=0.2477, pruned_loss=0.03906, over 3308241.81 frames. ], batch size: 83, lr: 3.25e-03, grad_scale: 8.0 2023-05-01 05:11:17,974 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.5434, 3.7534, 4.0954, 2.2643, 3.3516, 2.5774, 3.9327, 3.8991], device='cuda:1'), covar=tensor([0.0256, 0.0832, 0.0452, 0.1873, 0.0745, 0.0916, 0.0578, 0.0908], device='cuda:1'), in_proj_covar=tensor([0.0156, 0.0163, 0.0167, 0.0153, 0.0145, 0.0130, 0.0145, 0.0175], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-05-01 05:11:42,899 INFO [train.py:904] (1/8) Epoch 21, batch 1350, loss[loss=0.1606, simple_loss=0.2407, pruned_loss=0.04022, over 15470.00 frames. ], tot_loss[loss=0.1634, simple_loss=0.2483, pruned_loss=0.03922, over 3312282.93 frames. ], batch size: 191, lr: 3.25e-03, grad_scale: 8.0 2023-05-01 05:11:48,278 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([5.0623, 4.8409, 5.0913, 5.2767, 5.5242, 4.8330, 5.4350, 5.4870], device='cuda:1'), covar=tensor([0.1933, 0.1282, 0.1759, 0.0804, 0.0571, 0.0830, 0.0570, 0.0580], device='cuda:1'), in_proj_covar=tensor([0.0650, 0.0795, 0.0928, 0.0818, 0.0611, 0.0635, 0.0657, 0.0759], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-01 05:12:17,677 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.29 vs. limit=2.0 2023-05-01 05:12:31,606 INFO [optim.py:368] (1/8) Clipping_scale=2.0, grad-norm quartiles 1.473e+02 2.188e+02 2.581e+02 2.996e+02 9.160e+02, threshold=5.162e+02, percent-clipped=1.0 2023-05-01 05:12:52,518 INFO [train.py:904] (1/8) Epoch 21, batch 1400, loss[loss=0.1407, simple_loss=0.2179, pruned_loss=0.03173, over 11947.00 frames. ], tot_loss[loss=0.1628, simple_loss=0.2481, pruned_loss=0.03879, over 3307800.68 frames. ], batch size: 246, lr: 3.25e-03, grad_scale: 8.0 2023-05-01 05:12:54,544 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=204403.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 05:13:20,714 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=204422.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 05:13:35,304 INFO [zipformer.py:625] (1/8) 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] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=204451.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 05:14:01,405 INFO [train.py:904] (1/8) Epoch 21, batch 1450, loss[loss=0.1665, simple_loss=0.24, pruned_loss=0.04646, over 16405.00 frames. ], tot_loss[loss=0.1635, simple_loss=0.2479, pruned_loss=0.03951, over 3313976.94 frames. ], batch size: 146, lr: 3.25e-03, grad_scale: 8.0 2023-05-01 05:14:26,854 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=204470.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 05:14:28,338 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([1.8524, 2.7088, 2.5384, 4.1308, 3.4249, 4.0477, 1.6399, 3.0176], device='cuda:1'), covar=tensor([0.1365, 0.0705, 0.1169, 0.0179, 0.0143, 0.0407, 0.1582, 0.0752], device='cuda:1'), in_proj_covar=tensor([0.0167, 0.0175, 0.0195, 0.0189, 0.0205, 0.0216, 0.0202, 0.0192], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-05-01 05:14:50,906 INFO [optim.py:368] (1/8) Clipping_scale=2.0, grad-norm quartiles 1.523e+02 2.087e+02 2.507e+02 3.018e+02 6.138e+02, threshold=5.015e+02, percent-clipped=2.0 2023-05-01 05:14:57,060 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([5.3982, 5.8044, 5.5524, 5.6557, 5.1937, 5.3276, 5.2241, 5.9368], device='cuda:1'), covar=tensor([0.1446, 0.1033, 0.1052, 0.0936, 0.0914, 0.0718, 0.1175, 0.0978], device='cuda:1'), in_proj_covar=tensor([0.0681, 0.0831, 0.0680, 0.0627, 0.0529, 0.0535, 0.0699, 0.0644], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:1') 2023-05-01 05:15:10,559 INFO [train.py:904] (1/8) Epoch 21, batch 1500, loss[loss=0.1983, simple_loss=0.265, pruned_loss=0.06587, over 16884.00 frames. ], tot_loss[loss=0.1644, simple_loss=0.2483, pruned_loss=0.04026, over 3308506.35 frames. ], batch size: 109, lr: 3.25e-03, grad_scale: 4.0 2023-05-01 05:15:12,776 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=204503.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 05:15:33,800 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=204519.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 05:16:17,197 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=204551.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 05:16:18,102 INFO [train.py:904] (1/8) Epoch 21, batch 1550, loss[loss=0.1853, simple_loss=0.2594, pruned_loss=0.05559, over 16534.00 frames. ], tot_loss[loss=0.1658, simple_loss=0.2496, pruned_loss=0.04102, over 3313289.17 frames. ], batch size: 146, lr: 3.25e-03, grad_scale: 4.0 2023-05-01 05:16:22,970 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=204556.0, num_to_drop=1, layers_to_drop={2} 2023-05-01 05:16:26,540 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=4.96 vs. limit=5.0 2023-05-01 05:17:07,000 INFO [optim.py:368] (1/8) Clipping_scale=2.0, grad-norm quartiles 1.626e+02 2.391e+02 2.852e+02 3.354e+02 7.624e+02, threshold=5.704e+02, percent-clipped=5.0 2023-05-01 05:17:08,281 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.23 vs. limit=2.0 2023-05-01 05:17:26,280 INFO [train.py:904] (1/8) Epoch 21, batch 1600, loss[loss=0.1827, simple_loss=0.2869, pruned_loss=0.03928, over 17114.00 frames. ], tot_loss[loss=0.1667, simple_loss=0.2508, pruned_loss=0.04135, over 3311185.83 frames. ], batch size: 49, lr: 3.24e-03, grad_scale: 8.0 2023-05-01 05:17:28,784 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=204604.0, num_to_drop=1, layers_to_drop={1} 2023-05-01 05:18:06,592 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([5.1136, 5.6831, 5.7625, 5.4241, 5.6372, 6.1619, 5.5871, 5.2740], device='cuda:1'), covar=tensor([0.0865, 0.2118, 0.2513, 0.1990, 0.2517, 0.0957, 0.1587, 0.2456], device='cuda:1'), in_proj_covar=tensor([0.0403, 0.0591, 0.0651, 0.0487, 0.0651, 0.0680, 0.0513, 0.0654], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-05-01 05:18:24,689 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.0960, 2.2877, 2.6177, 3.0376, 2.8054, 3.5085, 2.6296, 3.4203], device='cuda:1'), covar=tensor([0.0232, 0.0465, 0.0341, 0.0286, 0.0323, 0.0188, 0.0396, 0.0151], device='cuda:1'), in_proj_covar=tensor([0.0187, 0.0193, 0.0180, 0.0184, 0.0196, 0.0154, 0.0196, 0.0149], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-01 05:18:35,817 INFO [train.py:904] (1/8) Epoch 21, batch 1650, loss[loss=0.1791, simple_loss=0.2691, pruned_loss=0.04457, over 16689.00 frames. ], tot_loss[loss=0.1677, simple_loss=0.2524, pruned_loss=0.04148, over 3317080.54 frames. ], batch size: 57, lr: 3.24e-03, grad_scale: 8.0 2023-05-01 05:19:25,764 INFO [optim.py:368] (1/8) Clipping_scale=2.0, grad-norm quartiles 1.742e+02 2.251e+02 2.625e+02 3.143e+02 5.667e+02, threshold=5.251e+02, percent-clipped=0.0 2023-05-01 05:19:38,366 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=3.45 vs. limit=5.0 2023-05-01 05:19:45,536 INFO [train.py:904] (1/8) Epoch 21, batch 1700, loss[loss=0.1749, simple_loss=0.2491, pruned_loss=0.05039, over 16918.00 frames. ], tot_loss[loss=0.1695, simple_loss=0.2544, pruned_loss=0.04229, over 3297048.65 frames. ], batch size: 109, lr: 3.24e-03, grad_scale: 8.0 2023-05-01 05:20:29,905 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=204733.0, num_to_drop=1, layers_to_drop={1} 2023-05-01 05:20:46,423 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.4969, 3.9702, 4.0461, 2.3992, 3.2458, 2.7631, 3.8238, 4.0719], device='cuda:1'), covar=tensor([0.0279, 0.0723, 0.0466, 0.1775, 0.0792, 0.0837, 0.0635, 0.0959], device='cuda:1'), in_proj_covar=tensor([0.0156, 0.0163, 0.0167, 0.0152, 0.0144, 0.0129, 0.0144, 0.0174], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:1') 2023-05-01 05:20:55,678 INFO [train.py:904] (1/8) Epoch 21, batch 1750, loss[loss=0.1833, simple_loss=0.2538, pruned_loss=0.05637, over 16826.00 frames. ], tot_loss[loss=0.1694, simple_loss=0.2549, pruned_loss=0.04198, over 3306396.05 frames. ], batch size: 109, lr: 3.24e-03, grad_scale: 8.0 2023-05-01 05:21:37,075 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=204781.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 05:21:46,293 INFO [optim.py:368] (1/8) 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 05:22:06,988 INFO [train.py:904] (1/8) Epoch 21, batch 1800, loss[loss=0.1525, simple_loss=0.237, pruned_loss=0.03401, over 17191.00 frames. ], tot_loss[loss=0.1687, simple_loss=0.2547, pruned_loss=0.04138, over 3312359.57 frames. ], batch size: 44, lr: 3.24e-03, grad_scale: 8.0 2023-05-01 05:22:30,017 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=204819.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 05:22:32,937 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=204821.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 05:22:33,979 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.6773, 4.8603, 4.9947, 4.7794, 4.8232, 5.4503, 4.9292, 4.5747], device='cuda:1'), covar=tensor([0.1447, 0.2118, 0.2480, 0.2361, 0.2963, 0.1093, 0.1790, 0.2797], device='cuda:1'), in_proj_covar=tensor([0.0409, 0.0599, 0.0660, 0.0495, 0.0661, 0.0689, 0.0521, 0.0664], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-05-01 05:23:03,729 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([5.5053, 5.9174, 5.6581, 5.7187, 5.3632, 5.3107, 5.3088, 6.0203], device='cuda:1'), covar=tensor([0.1445, 0.0932, 0.1147, 0.0925, 0.0908, 0.0701, 0.1216, 0.1032], device='cuda:1'), in_proj_covar=tensor([0.0678, 0.0826, 0.0679, 0.0624, 0.0526, 0.0531, 0.0697, 0.0640], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:1') 2023-05-01 05:23:15,229 INFO [train.py:904] (1/8) Epoch 21, batch 1850, loss[loss=0.1664, simple_loss=0.2473, pruned_loss=0.04268, over 16887.00 frames. ], tot_loss[loss=0.1697, simple_loss=0.2555, pruned_loss=0.04192, over 3313669.48 frames. ], batch size: 96, lr: 3.24e-03, grad_scale: 8.0 2023-05-01 05:23:37,455 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=204867.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 05:23:57,947 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=204882.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 05:24:05,752 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.17 vs. limit=2.0 2023-05-01 05:24:06,221 INFO [optim.py:368] (1/8) 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 05:24:26,171 INFO [train.py:904] (1/8) Epoch 21, batch 1900, loss[loss=0.1503, simple_loss=0.238, pruned_loss=0.03132, over 16824.00 frames. ], tot_loss[loss=0.1684, simple_loss=0.2547, pruned_loss=0.04105, over 3322589.49 frames. ], batch size: 42, lr: 3.24e-03, grad_scale: 8.0 2023-05-01 05:25:35,963 INFO [train.py:904] (1/8) Epoch 21, batch 1950, loss[loss=0.1247, simple_loss=0.2083, pruned_loss=0.02053, over 16819.00 frames. ], tot_loss[loss=0.1684, simple_loss=0.255, pruned_loss=0.04089, over 3329419.60 frames. ], batch size: 39, lr: 3.24e-03, grad_scale: 8.0 2023-05-01 05:25:36,463 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.7518, 4.4175, 3.0383, 2.2151, 2.7201, 2.4426, 4.7559, 3.6019], device='cuda:1'), covar=tensor([0.3077, 0.0588, 0.1823, 0.3087, 0.3108, 0.2188, 0.0330, 0.1369], device='cuda:1'), in_proj_covar=tensor([0.0328, 0.0272, 0.0306, 0.0311, 0.0297, 0.0258, 0.0295, 0.0338], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:1') 2023-05-01 05:26:26,411 INFO [optim.py:368] (1/8) 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 05:26:28,037 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.2292, 4.5346, 4.4044, 3.2269, 3.8973, 4.3511, 3.9928, 2.4467], device='cuda:1'), covar=tensor([0.0493, 0.0074, 0.0065, 0.0431, 0.0150, 0.0161, 0.0113, 0.0682], device='cuda:1'), in_proj_covar=tensor([0.0138, 0.0084, 0.0083, 0.0135, 0.0099, 0.0110, 0.0094, 0.0129], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0002, 0.0004], device='cuda:1') 2023-05-01 05:26:36,172 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.5382, 3.5453, 2.1524, 3.7596, 2.7595, 3.7162, 2.3714, 2.8216], device='cuda:1'), covar=tensor([0.0246, 0.0407, 0.1425, 0.0271, 0.0697, 0.0688, 0.1307, 0.0682], device='cuda:1'), in_proj_covar=tensor([0.0172, 0.0179, 0.0196, 0.0166, 0.0178, 0.0219, 0.0205, 0.0180], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-05-01 05:26:44,684 INFO [train.py:904] (1/8) Epoch 21, batch 2000, loss[loss=0.1653, simple_loss=0.2408, pruned_loss=0.04489, over 16809.00 frames. ], tot_loss[loss=0.1681, simple_loss=0.2547, pruned_loss=0.04072, over 3322522.21 frames. ], batch size: 102, lr: 3.24e-03, grad_scale: 8.0 2023-05-01 05:27:55,521 INFO [train.py:904] (1/8) Epoch 21, batch 2050, loss[loss=0.1708, simple_loss=0.2684, pruned_loss=0.03654, over 17286.00 frames. ], tot_loss[loss=0.1688, simple_loss=0.2552, pruned_loss=0.0412, over 3325468.70 frames. ], batch size: 52, lr: 3.24e-03, grad_scale: 8.0 2023-05-01 05:28:18,247 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([5.0625, 4.8392, 5.0825, 5.2782, 5.5035, 4.7862, 5.4944, 5.4683], device='cuda:1'), covar=tensor([0.1794, 0.1285, 0.1722, 0.0780, 0.0524, 0.0925, 0.0465, 0.0565], device='cuda:1'), in_proj_covar=tensor([0.0656, 0.0804, 0.0944, 0.0828, 0.0617, 0.0647, 0.0667, 0.0770], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-01 05:28:44,310 INFO [optim.py:368] (1/8) Clipping_scale=2.0, grad-norm quartiles 1.595e+02 2.164e+02 2.478e+02 3.098e+02 5.896e+02, threshold=4.956e+02, percent-clipped=1.0 2023-05-01 05:28:50,282 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.96 vs. limit=2.0 2023-05-01 05:29:04,144 INFO [train.py:904] (1/8) Epoch 21, batch 2100, loss[loss=0.1673, simple_loss=0.2549, pruned_loss=0.03988, over 11847.00 frames. ], tot_loss[loss=0.1705, simple_loss=0.2566, pruned_loss=0.0422, over 3313377.47 frames. ], batch size: 246, lr: 3.24e-03, grad_scale: 8.0 2023-05-01 05:29:11,771 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.4215, 1.6738, 2.1433, 2.2345, 2.3959, 2.3291, 1.6275, 2.4305], device='cuda:1'), covar=tensor([0.0226, 0.0536, 0.0311, 0.0320, 0.0302, 0.0314, 0.0593, 0.0208], device='cuda:1'), in_proj_covar=tensor([0.0189, 0.0195, 0.0180, 0.0186, 0.0198, 0.0155, 0.0197, 0.0150], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-01 05:29:14,563 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=205109.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 05:30:14,923 INFO [train.py:904] (1/8) Epoch 21, batch 2150, loss[loss=0.1854, simple_loss=0.2784, pruned_loss=0.04625, over 17079.00 frames. ], tot_loss[loss=0.1715, simple_loss=0.2577, pruned_loss=0.04264, over 3316369.23 frames. ], batch size: 53, lr: 3.24e-03, grad_scale: 8.0 2023-05-01 05:30:39,949 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=205170.0, num_to_drop=1, layers_to_drop={0} 2023-05-01 05:30:49,797 INFO [zipformer.py:625] (1/8) 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] (1/8) 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 05:31:19,697 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=205198.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 05:31:24,882 INFO [train.py:904] (1/8) Epoch 21, batch 2200, loss[loss=0.1724, simple_loss=0.2538, pruned_loss=0.0455, over 16432.00 frames. ], tot_loss[loss=0.1716, simple_loss=0.2582, pruned_loss=0.04247, over 3312408.58 frames. ], batch size: 146, lr: 3.24e-03, grad_scale: 8.0 2023-05-01 05:31:44,712 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.90 vs. limit=2.0 2023-05-01 05:32:34,216 INFO [train.py:904] (1/8) Epoch 21, batch 2250, loss[loss=0.1531, simple_loss=0.2334, pruned_loss=0.03634, over 15757.00 frames. ], tot_loss[loss=0.1719, simple_loss=0.2582, pruned_loss=0.04282, over 3308361.77 frames. ], batch size: 35, lr: 3.24e-03, grad_scale: 8.0 2023-05-01 05:32:45,040 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=205259.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 05:32:48,399 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.9437, 5.2844, 5.4577, 5.1387, 5.2201, 5.8494, 5.3526, 5.0171], device='cuda:1'), covar=tensor([0.1217, 0.2003, 0.2490, 0.2082, 0.2677, 0.1059, 0.1905, 0.2760], device='cuda:1'), in_proj_covar=tensor([0.0411, 0.0601, 0.0662, 0.0499, 0.0663, 0.0692, 0.0521, 0.0668], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-05-01 05:33:18,676 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.9023, 1.9560, 2.4304, 2.7388, 2.6812, 2.7916, 2.0111, 2.9960], device='cuda:1'), covar=tensor([0.0195, 0.0482, 0.0342, 0.0300, 0.0330, 0.0310, 0.0534, 0.0168], device='cuda:1'), in_proj_covar=tensor([0.0189, 0.0194, 0.0180, 0.0186, 0.0198, 0.0155, 0.0197, 0.0150], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-01 05:33:23,477 INFO [optim.py:368] (1/8) Clipping_scale=2.0, grad-norm quartiles 1.631e+02 2.149e+02 2.507e+02 2.983e+02 5.754e+02, threshold=5.014e+02, percent-clipped=1.0 2023-05-01 05:33:44,283 INFO [train.py:904] (1/8) Epoch 21, batch 2300, loss[loss=0.1704, simple_loss=0.2503, pruned_loss=0.0453, over 16795.00 frames. ], tot_loss[loss=0.1721, simple_loss=0.2582, pruned_loss=0.04304, over 3313936.63 frames. ], batch size: 102, lr: 3.24e-03, grad_scale: 8.0 2023-05-01 05:34:53,188 INFO [train.py:904] (1/8) Epoch 21, batch 2350, loss[loss=0.1842, simple_loss=0.275, pruned_loss=0.04666, over 16736.00 frames. ], tot_loss[loss=0.1729, simple_loss=0.2588, pruned_loss=0.04351, over 3303071.76 frames. ], batch size: 62, lr: 3.24e-03, grad_scale: 8.0 2023-05-01 05:35:18,397 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.0901, 4.6050, 3.1090, 2.4875, 2.9601, 2.6553, 4.8486, 3.7766], device='cuda:1'), covar=tensor([0.2804, 0.0560, 0.1897, 0.2798, 0.2978, 0.2086, 0.0443, 0.1355], device='cuda:1'), in_proj_covar=tensor([0.0327, 0.0270, 0.0303, 0.0309, 0.0296, 0.0257, 0.0294, 0.0336], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:1') 2023-05-01 05:35:42,800 INFO [optim.py:368] (1/8) Clipping_scale=2.0, grad-norm quartiles 1.569e+02 2.163e+02 2.457e+02 2.980e+02 4.846e+02, threshold=4.914e+02, percent-clipped=0.0 2023-05-01 05:35:55,175 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.77 vs. limit=2.0 2023-05-01 05:36:02,959 INFO [train.py:904] (1/8) Epoch 21, batch 2400, loss[loss=0.194, simple_loss=0.2746, pruned_loss=0.05675, over 16466.00 frames. ], tot_loss[loss=0.1737, simple_loss=0.2598, pruned_loss=0.0438, over 3306641.02 frames. ], batch size: 146, lr: 3.24e-03, grad_scale: 8.0 2023-05-01 05:36:10,867 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.20 vs. limit=2.0 2023-05-01 05:36:50,389 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.83 vs. limit=2.0 2023-05-01 05:37:10,842 INFO [train.py:904] (1/8) Epoch 21, batch 2450, loss[loss=0.1879, simple_loss=0.2818, pruned_loss=0.04697, over 16645.00 frames. ], tot_loss[loss=0.1743, simple_loss=0.2608, pruned_loss=0.04389, over 3301276.69 frames. ], batch size: 62, lr: 3.24e-03, grad_scale: 8.0 2023-05-01 05:37:24,965 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([1.8745, 2.7055, 2.6263, 1.9439, 2.5719, 2.7856, 2.6056, 1.8952], device='cuda:1'), covar=tensor([0.0475, 0.0124, 0.0091, 0.0400, 0.0148, 0.0122, 0.0129, 0.0402], device='cuda:1'), in_proj_covar=tensor([0.0137, 0.0084, 0.0083, 0.0134, 0.0099, 0.0109, 0.0094, 0.0128], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-05-01 05:37:29,556 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=205465.0, num_to_drop=1, layers_to_drop={2} 2023-05-01 05:37:45,981 INFO [zipformer.py:625] (1/8) 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] (1/8) Clipping_scale=2.0, grad-norm quartiles 1.425e+02 2.346e+02 2.648e+02 3.284e+02 5.369e+02, threshold=5.296e+02, percent-clipped=1.0 2023-05-01 05:38:16,925 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.8862, 2.1464, 2.5504, 2.9123, 2.7084, 3.3765, 2.3524, 3.3386], device='cuda:1'), covar=tensor([0.0260, 0.0515, 0.0321, 0.0302, 0.0335, 0.0190, 0.0479, 0.0154], device='cuda:1'), in_proj_covar=tensor([0.0189, 0.0195, 0.0180, 0.0186, 0.0198, 0.0155, 0.0198, 0.0151], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-01 05:38:21,260 INFO [train.py:904] (1/8) Epoch 21, batch 2500, loss[loss=0.2315, simple_loss=0.307, pruned_loss=0.078, over 11920.00 frames. ], tot_loss[loss=0.1747, simple_loss=0.2613, pruned_loss=0.04408, over 3291112.05 frames. ], batch size: 246, lr: 3.24e-03, grad_scale: 8.0 2023-05-01 05:38:40,334 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.0452, 4.5313, 4.5781, 3.3034, 3.8103, 4.4899, 3.9753, 2.6679], device='cuda:1'), covar=tensor([0.0448, 0.0057, 0.0036, 0.0330, 0.0127, 0.0092, 0.0082, 0.0439], device='cuda:1'), in_proj_covar=tensor([0.0137, 0.0084, 0.0083, 0.0134, 0.0099, 0.0109, 0.0094, 0.0128], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-05-01 05:38:53,422 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 21, batch 2550, loss[loss=0.1578, simple_loss=0.2453, pruned_loss=0.03521, over 17194.00 frames. ], tot_loss[loss=0.1745, simple_loss=0.2611, pruned_loss=0.04396, over 3291545.34 frames. ], batch size: 44, lr: 3.24e-03, grad_scale: 8.0 2023-05-01 05:39:33,575 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=205554.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 05:40:16,808 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.1153, 3.9814, 4.1861, 4.2956, 4.3853, 3.9595, 4.1763, 4.3558], device='cuda:1'), covar=tensor([0.1553, 0.1178, 0.1187, 0.0721, 0.0605, 0.1534, 0.2295, 0.0874], device='cuda:1'), in_proj_covar=tensor([0.0660, 0.0810, 0.0949, 0.0832, 0.0619, 0.0649, 0.0670, 0.0776], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-01 05:40:19,332 INFO [optim.py:368] (1/8) Clipping_scale=2.0, grad-norm quartiles 1.360e+02 2.328e+02 2.679e+02 3.297e+02 1.189e+03, threshold=5.358e+02, percent-clipped=3.0 2023-05-01 05:40:21,000 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.0897, 4.5557, 3.3644, 2.4754, 2.9802, 2.8399, 4.9365, 3.9163], device='cuda:1'), covar=tensor([0.2582, 0.0572, 0.1694, 0.2914, 0.2864, 0.1936, 0.0357, 0.1237], device='cuda:1'), in_proj_covar=tensor([0.0327, 0.0270, 0.0304, 0.0308, 0.0296, 0.0257, 0.0294, 0.0336], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:1') 2023-05-01 05:40:38,679 INFO [train.py:904] (1/8) Epoch 21, batch 2600, loss[loss=0.1802, simple_loss=0.2776, pruned_loss=0.04136, over 17025.00 frames. ], tot_loss[loss=0.1736, simple_loss=0.2605, pruned_loss=0.04338, over 3307657.76 frames. ], batch size: 55, lr: 3.24e-03, grad_scale: 8.0 2023-05-01 05:41:49,802 INFO [train.py:904] (1/8) Epoch 21, batch 2650, loss[loss=0.1701, simple_loss=0.2688, pruned_loss=0.0357, over 17030.00 frames. ], tot_loss[loss=0.1729, simple_loss=0.2606, pruned_loss=0.04262, over 3309433.06 frames. ], batch size: 55, lr: 3.24e-03, grad_scale: 4.0 2023-05-01 05:41:54,618 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=205655.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 05:42:40,191 INFO [optim.py:368] (1/8) 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 05:43:00,011 INFO [train.py:904] (1/8) Epoch 21, batch 2700, loss[loss=0.2092, simple_loss=0.2845, pruned_loss=0.06698, over 12648.00 frames. ], tot_loss[loss=0.1725, simple_loss=0.2605, pruned_loss=0.04224, over 3309806.39 frames. ], batch size: 246, lr: 3.24e-03, grad_scale: 4.0 2023-05-01 05:43:18,677 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=205716.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 05:43:59,603 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=205744.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 05:44:09,882 INFO [train.py:904] (1/8) Epoch 21, batch 2750, loss[loss=0.165, simple_loss=0.2535, pruned_loss=0.0383, over 16815.00 frames. ], tot_loss[loss=0.1724, simple_loss=0.261, pruned_loss=0.04189, over 3313802.91 frames. ], batch size: 102, lr: 3.24e-03, grad_scale: 4.0 2023-05-01 05:44:13,710 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.4616, 3.2921, 2.7338, 2.1461, 2.1735, 2.2326, 3.4307, 2.9456], device='cuda:1'), covar=tensor([0.2713, 0.0737, 0.1677, 0.2705, 0.2589, 0.2157, 0.0545, 0.1664], device='cuda:1'), in_proj_covar=tensor([0.0327, 0.0270, 0.0304, 0.0309, 0.0297, 0.0257, 0.0294, 0.0337], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:1') 2023-05-01 05:44:29,198 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=205765.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 05:45:02,029 INFO [optim.py:368] (1/8) 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] (1/8) Epoch 21, batch 2800, loss[loss=0.1576, simple_loss=0.252, pruned_loss=0.03162, over 17026.00 frames. ], tot_loss[loss=0.1716, simple_loss=0.26, pruned_loss=0.04155, over 3319674.98 frames. ], batch size: 55, lr: 3.24e-03, grad_scale: 8.0 2023-05-01 05:45:22,622 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.0608, 4.0339, 3.9755, 3.3028, 3.9616, 1.7492, 3.7586, 3.4374], device='cuda:1'), covar=tensor([0.0117, 0.0112, 0.0182, 0.0249, 0.0097, 0.2887, 0.0133, 0.0242], device='cuda:1'), in_proj_covar=tensor([0.0166, 0.0157, 0.0199, 0.0180, 0.0178, 0.0210, 0.0190, 0.0176], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-01 05:45:23,822 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=205805.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 05:45:35,279 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=205813.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 05:46:29,087 INFO [train.py:904] (1/8) Epoch 21, batch 2850, loss[loss=0.1453, simple_loss=0.2343, pruned_loss=0.02815, over 16775.00 frames. ], tot_loss[loss=0.1711, simple_loss=0.2593, pruned_loss=0.04143, over 3319673.72 frames. ], batch size: 39, lr: 3.23e-03, grad_scale: 8.0 2023-05-01 05:46:31,729 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=205854.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 05:47:17,820 INFO [optim.py:368] (1/8) 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 05:47:19,177 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.7862, 4.8755, 4.9857, 4.8161, 4.8147, 5.4513, 4.8960, 4.5755], device='cuda:1'), covar=tensor([0.1332, 0.2058, 0.2331, 0.1978, 0.2686, 0.0959, 0.1737, 0.2470], device='cuda:1'), in_proj_covar=tensor([0.0417, 0.0611, 0.0669, 0.0505, 0.0669, 0.0699, 0.0526, 0.0673], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-05-01 05:47:19,281 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=205890.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 05:47:36,338 INFO [train.py:904] (1/8) Epoch 21, batch 2900, loss[loss=0.1902, simple_loss=0.2597, pruned_loss=0.06032, over 15476.00 frames. ], tot_loss[loss=0.1703, simple_loss=0.2576, pruned_loss=0.04151, over 3323361.37 frames. ], batch size: 190, lr: 3.23e-03, grad_scale: 8.0 2023-05-01 05:47:36,635 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=205902.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 05:48:44,784 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 21, batch 2950, loss[loss=0.1991, simple_loss=0.2779, pruned_loss=0.06012, over 16308.00 frames. ], tot_loss[loss=0.1705, simple_loss=0.257, pruned_loss=0.04198, over 3316839.27 frames. ], batch size: 165, lr: 3.23e-03, grad_scale: 8.0 2023-05-01 05:49:12,166 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.3719, 2.4378, 2.3813, 4.1869, 2.3222, 2.7936, 2.4707, 2.5711], device='cuda:1'), covar=tensor([0.1333, 0.3357, 0.2953, 0.0609, 0.3827, 0.2449, 0.3462, 0.3146], device='cuda:1'), in_proj_covar=tensor([0.0402, 0.0449, 0.0369, 0.0331, 0.0435, 0.0515, 0.0418, 0.0525], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-01 05:49:28,991 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([5.0695, 5.0780, 4.9355, 4.5121, 4.6004, 5.0150, 4.9280, 4.6844], device='cuda:1'), covar=tensor([0.0627, 0.0577, 0.0295, 0.0309, 0.0989, 0.0430, 0.0358, 0.0665], device='cuda:1'), in_proj_covar=tensor([0.0305, 0.0437, 0.0355, 0.0350, 0.0363, 0.0407, 0.0243, 0.0427], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-05-01 05:49:36,947 INFO [optim.py:368] (1/8) 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] (1/8) Epoch 21, batch 3000, loss[loss=0.1676, simple_loss=0.2517, pruned_loss=0.04173, over 16800.00 frames. ], tot_loss[loss=0.1715, simple_loss=0.2575, pruned_loss=0.04277, over 3322132.87 frames. ], batch size: 102, lr: 3.23e-03, grad_scale: 8.0 2023-05-01 05:49:58,057 INFO [train.py:929] (1/8) Computing validation loss 2023-05-01 05:50:06,481 INFO [train.py:938] (1/8) Epoch 21, validation: loss=0.1356, simple_loss=0.2408, pruned_loss=0.01521, over 944034.00 frames. 2023-05-01 05:50:06,482 INFO [train.py:939] (1/8) Maximum memory allocated so far is 18145MB 2023-05-01 05:50:06,932 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([1.7461, 2.4891, 2.4560, 3.8549, 3.1667, 3.9384, 1.6034, 2.8521], device='cuda:1'), covar=tensor([0.1380, 0.0733, 0.1163, 0.0181, 0.0143, 0.0359, 0.1570, 0.0838], device='cuda:1'), in_proj_covar=tensor([0.0166, 0.0175, 0.0195, 0.0192, 0.0207, 0.0217, 0.0202, 0.0193], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-05-01 05:50:18,500 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=206011.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 05:51:14,522 INFO [train.py:904] (1/8) Epoch 21, batch 3050, loss[loss=0.194, simple_loss=0.2736, pruned_loss=0.05721, over 16555.00 frames. ], tot_loss[loss=0.1718, simple_loss=0.2578, pruned_loss=0.0429, over 3311645.16 frames. ], batch size: 75, lr: 3.23e-03, grad_scale: 8.0 2023-05-01 05:51:21,059 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=206056.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 05:52:05,513 INFO [optim.py:368] (1/8) 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] (1/8) 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] (1/8) Epoch 21, batch 3100, loss[loss=0.1413, simple_loss=0.2374, pruned_loss=0.02263, over 17264.00 frames. ], tot_loss[loss=0.171, simple_loss=0.2573, pruned_loss=0.04234, over 3306301.30 frames. ], batch size: 52, lr: 3.23e-03, grad_scale: 8.0 2023-05-01 05:52:43,677 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=206117.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 05:53:30,989 INFO [train.py:904] (1/8) Epoch 21, batch 3150, loss[loss=0.1888, simple_loss=0.2623, pruned_loss=0.05766, over 16929.00 frames. ], tot_loss[loss=0.1718, simple_loss=0.2575, pruned_loss=0.04304, over 3297053.84 frames. ], batch size: 96, lr: 3.23e-03, grad_scale: 8.0 2023-05-01 05:53:51,705 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.5555, 3.6331, 3.3694, 3.0477, 3.2464, 3.4968, 3.3068, 3.3319], device='cuda:1'), covar=tensor([0.0638, 0.0566, 0.0304, 0.0275, 0.0536, 0.0427, 0.1587, 0.0465], device='cuda:1'), in_proj_covar=tensor([0.0308, 0.0440, 0.0358, 0.0352, 0.0367, 0.0409, 0.0245, 0.0430], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-05-01 05:54:22,972 INFO [optim.py:368] (1/8) 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 05:54:34,401 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=206197.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 05:54:41,550 INFO [train.py:904] (1/8) Epoch 21, batch 3200, loss[loss=0.1771, simple_loss=0.2714, pruned_loss=0.04142, over 16621.00 frames. ], tot_loss[loss=0.1704, simple_loss=0.2561, pruned_loss=0.04233, over 3300913.47 frames. ], batch size: 62, lr: 3.23e-03, grad_scale: 8.0 2023-05-01 05:55:42,472 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 21, batch 3250, loss[loss=0.1837, simple_loss=0.2724, pruned_loss=0.04757, over 16831.00 frames. ], tot_loss[loss=0.1706, simple_loss=0.2565, pruned_loss=0.04233, over 3304451.71 frames. ], batch size: 96, lr: 3.23e-03, grad_scale: 8.0 2023-05-01 05:55:59,429 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=206258.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 05:56:42,635 INFO [optim.py:368] (1/8) 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 05:57:00,589 INFO [train.py:904] (1/8) Epoch 21, batch 3300, loss[loss=0.1493, simple_loss=0.2428, pruned_loss=0.02793, over 17200.00 frames. ], tot_loss[loss=0.1709, simple_loss=0.2572, pruned_loss=0.0423, over 3309387.78 frames. ], batch size: 46, lr: 3.23e-03, grad_scale: 8.0 2023-05-01 05:57:13,834 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=206311.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 05:57:20,379 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.47 vs. limit=2.0 2023-05-01 05:58:09,897 INFO [train.py:904] (1/8) Epoch 21, batch 3350, loss[loss=0.163, simple_loss=0.2681, pruned_loss=0.02891, over 17060.00 frames. ], tot_loss[loss=0.1709, simple_loss=0.2573, pruned_loss=0.04226, over 3313435.58 frames. ], batch size: 50, lr: 3.23e-03, grad_scale: 8.0 2023-05-01 05:58:20,156 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=206359.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 05:59:00,092 INFO [optim.py:368] (1/8) 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 05:59:16,477 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 21, batch 3400, loss[loss=0.1759, simple_loss=0.2711, pruned_loss=0.0404, over 17009.00 frames. ], tot_loss[loss=0.1704, simple_loss=0.257, pruned_loss=0.04185, over 3316986.90 frames. ], batch size: 50, lr: 3.23e-03, grad_scale: 8.0 2023-05-01 05:59:32,318 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=206412.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 06:00:02,286 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.4906, 3.5994, 3.3500, 2.9895, 3.1763, 3.4881, 3.2820, 3.2919], device='cuda:1'), covar=tensor([0.0656, 0.0534, 0.0341, 0.0324, 0.0656, 0.0452, 0.1496, 0.0529], device='cuda:1'), in_proj_covar=tensor([0.0307, 0.0442, 0.0359, 0.0353, 0.0366, 0.0409, 0.0245, 0.0431], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-05-01 06:00:06,535 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.7263, 3.7698, 2.2784, 4.2591, 2.8870, 4.2185, 2.5032, 3.1088], device='cuda:1'), covar=tensor([0.0293, 0.0407, 0.1567, 0.0411, 0.0766, 0.0582, 0.1413, 0.0724], device='cuda:1'), in_proj_covar=tensor([0.0171, 0.0179, 0.0195, 0.0167, 0.0178, 0.0220, 0.0204, 0.0180], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-05-01 06:00:21,655 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 21, batch 3450, loss[loss=0.1786, simple_loss=0.2484, pruned_loss=0.05443, over 16752.00 frames. ], tot_loss[loss=0.1697, simple_loss=0.2557, pruned_loss=0.04183, over 3305615.96 frames. ], batch size: 124, lr: 3.23e-03, grad_scale: 8.0 2023-05-01 06:00:37,827 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.7346, 3.7814, 2.8830, 2.2663, 2.4421, 2.4283, 3.8903, 3.3725], device='cuda:1'), covar=tensor([0.2751, 0.0619, 0.1734, 0.3079, 0.2911, 0.2138, 0.0559, 0.1498], device='cuda:1'), in_proj_covar=tensor([0.0327, 0.0271, 0.0305, 0.0310, 0.0298, 0.0259, 0.0295, 0.0339], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:1') 2023-05-01 06:00:51,717 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.3427, 4.2699, 4.2743, 3.9662, 4.0756, 4.3510, 4.0450, 4.1214], device='cuda:1'), covar=tensor([0.0675, 0.0860, 0.0320, 0.0302, 0.0712, 0.0527, 0.0653, 0.0610], device='cuda:1'), in_proj_covar=tensor([0.0308, 0.0443, 0.0359, 0.0353, 0.0367, 0.0410, 0.0245, 0.0431], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-05-01 06:01:17,175 INFO [optim.py:368] (1/8) 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 06:01:29,388 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.5823, 3.4511, 3.7480, 1.9738, 3.9011, 3.9312, 3.0823, 2.8872], device='cuda:1'), covar=tensor([0.0736, 0.0224, 0.0185, 0.1173, 0.0093, 0.0167, 0.0393, 0.0446], device='cuda:1'), in_proj_covar=tensor([0.0148, 0.0109, 0.0098, 0.0139, 0.0080, 0.0126, 0.0129, 0.0131], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0004, 0.0004], device='cuda:1') 2023-05-01 06:01:36,850 INFO [train.py:904] (1/8) Epoch 21, batch 3500, loss[loss=0.1987, simple_loss=0.2806, pruned_loss=0.0584, over 16720.00 frames. ], tot_loss[loss=0.1682, simple_loss=0.2543, pruned_loss=0.04108, over 3307942.39 frames. ], batch size: 124, lr: 3.23e-03, grad_scale: 8.0 2023-05-01 06:02:37,248 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=206546.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 06:02:44,993 INFO [train.py:904] (1/8) Epoch 21, batch 3550, loss[loss=0.2027, simple_loss=0.2752, pruned_loss=0.06514, over 12040.00 frames. ], tot_loss[loss=0.1678, simple_loss=0.254, pruned_loss=0.04077, over 3312524.73 frames. ], batch size: 246, lr: 3.23e-03, grad_scale: 8.0 2023-05-01 06:02:47,116 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=206553.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 06:03:37,041 INFO [optim.py:368] (1/8) 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 06:03:44,159 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=206594.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 06:03:55,931 INFO [train.py:904] (1/8) Epoch 21, batch 3600, loss[loss=0.1394, simple_loss=0.2208, pruned_loss=0.02905, over 16679.00 frames. ], tot_loss[loss=0.1672, simple_loss=0.253, pruned_loss=0.04069, over 3310379.56 frames. ], batch size: 89, lr: 3.23e-03, grad_scale: 8.0 2023-05-01 06:04:02,343 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.5923, 4.7238, 4.8469, 4.6958, 4.6992, 5.3179, 4.8354, 4.4891], device='cuda:1'), covar=tensor([0.1471, 0.2016, 0.2137, 0.1989, 0.2702, 0.1010, 0.1588, 0.2748], device='cuda:1'), in_proj_covar=tensor([0.0414, 0.0608, 0.0667, 0.0503, 0.0670, 0.0697, 0.0522, 0.0669], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-05-01 06:04:28,671 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([5.5193, 5.8895, 5.6854, 5.7446, 5.2929, 5.3493, 5.2751, 6.0413], device='cuda:1'), covar=tensor([0.1364, 0.0953, 0.1065, 0.0776, 0.0986, 0.0754, 0.1194, 0.0820], device='cuda:1'), in_proj_covar=tensor([0.0681, 0.0840, 0.0690, 0.0632, 0.0533, 0.0538, 0.0703, 0.0651], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:1') 2023-05-01 06:05:06,976 INFO [train.py:904] (1/8) Epoch 21, batch 3650, loss[loss=0.1542, simple_loss=0.2336, pruned_loss=0.03738, over 16798.00 frames. ], tot_loss[loss=0.1661, simple_loss=0.2508, pruned_loss=0.04073, over 3316127.50 frames. ], batch size: 102, lr: 3.23e-03, grad_scale: 8.0 2023-05-01 06:05:59,786 INFO [optim.py:368] (1/8) 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] (1/8) Epoch 21, batch 3700, loss[loss=0.1581, simple_loss=0.2297, pruned_loss=0.04324, over 16714.00 frames. ], tot_loss[loss=0.1671, simple_loss=0.2498, pruned_loss=0.0422, over 3311020.99 frames. ], batch size: 89, lr: 3.23e-03, grad_scale: 8.0 2023-05-01 06:06:31,521 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.7659, 3.5165, 3.8859, 2.1253, 4.0073, 4.0067, 3.2020, 2.9731], device='cuda:1'), covar=tensor([0.0715, 0.0236, 0.0158, 0.1119, 0.0092, 0.0158, 0.0355, 0.0428], device='cuda:1'), in_proj_covar=tensor([0.0148, 0.0109, 0.0098, 0.0139, 0.0080, 0.0126, 0.0129, 0.0131], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0004, 0.0004], device='cuda:1') 2023-05-01 06:06:34,574 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=206712.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 06:06:58,375 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.7471, 3.5745, 3.9422, 2.1698, 4.0618, 4.0695, 3.2083, 2.9618], device='cuda:1'), covar=tensor([0.0748, 0.0231, 0.0154, 0.1164, 0.0087, 0.0169, 0.0379, 0.0466], device='cuda:1'), in_proj_covar=tensor([0.0149, 0.0109, 0.0098, 0.0139, 0.0080, 0.0126, 0.0129, 0.0131], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0004, 0.0004], device='cuda:1') 2023-05-01 06:07:26,565 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.4407, 4.3861, 4.3539, 3.8143, 4.3937, 1.7176, 4.1445, 4.0049], device='cuda:1'), covar=tensor([0.0114, 0.0107, 0.0180, 0.0283, 0.0096, 0.2874, 0.0143, 0.0213], device='cuda:1'), in_proj_covar=tensor([0.0167, 0.0157, 0.0200, 0.0180, 0.0179, 0.0209, 0.0190, 0.0176], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-01 06:07:32,563 INFO [train.py:904] (1/8) Epoch 21, batch 3750, loss[loss=0.1805, simple_loss=0.2644, pruned_loss=0.04827, over 11567.00 frames. ], tot_loss[loss=0.169, simple_loss=0.2509, pruned_loss=0.04355, over 3299199.25 frames. ], batch size: 248, lr: 3.23e-03, grad_scale: 8.0 2023-05-01 06:07:45,100 INFO [zipformer.py:625] (1/8) 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] (1/8) 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 06:08:27,637 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=3.11 vs. limit=5.0 2023-05-01 06:08:35,312 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.0424, 2.1507, 2.6147, 2.9863, 2.8808, 3.0639, 2.0360, 3.2169], device='cuda:1'), covar=tensor([0.0184, 0.0448, 0.0318, 0.0247, 0.0285, 0.0203, 0.0573, 0.0143], device='cuda:1'), in_proj_covar=tensor([0.0190, 0.0194, 0.0180, 0.0186, 0.0198, 0.0156, 0.0197, 0.0152], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-01 06:08:45,460 INFO [train.py:904] (1/8) Epoch 21, batch 3800, loss[loss=0.1914, simple_loss=0.2611, pruned_loss=0.06086, over 16712.00 frames. ], tot_loss[loss=0.1708, simple_loss=0.2519, pruned_loss=0.04485, over 3286695.84 frames. ], batch size: 134, lr: 3.23e-03, grad_scale: 8.0 2023-05-01 06:08:54,081 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.4701, 2.8319, 2.2549, 2.5609, 3.1593, 2.8238, 3.2106, 3.2813], device='cuda:1'), covar=tensor([0.0137, 0.0348, 0.0520, 0.0388, 0.0225, 0.0319, 0.0198, 0.0222], device='cuda:1'), in_proj_covar=tensor([0.0213, 0.0238, 0.0226, 0.0227, 0.0238, 0.0237, 0.0241, 0.0236], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-01 06:09:24,393 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.9896, 2.6663, 2.1763, 2.3521, 2.9803, 2.7668, 3.0122, 3.1145], device='cuda:1'), covar=tensor([0.0221, 0.0396, 0.0533, 0.0479, 0.0252, 0.0323, 0.0255, 0.0267], device='cuda:1'), in_proj_covar=tensor([0.0212, 0.0237, 0.0226, 0.0227, 0.0238, 0.0237, 0.0241, 0.0235], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-01 06:09:32,821 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=206834.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 06:09:58,314 INFO [train.py:904] (1/8) Epoch 21, batch 3850, loss[loss=0.1529, simple_loss=0.2309, pruned_loss=0.03742, over 16435.00 frames. ], tot_loss[loss=0.1714, simple_loss=0.252, pruned_loss=0.04537, over 3297972.48 frames. ], batch size: 68, lr: 3.23e-03, grad_scale: 8.0 2023-05-01 06:10:00,705 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=206853.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 06:10:52,983 INFO [optim.py:368] (1/8) 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 06:11:00,976 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=206895.0, num_to_drop=1, layers_to_drop={0} 2023-05-01 06:11:09,940 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 21, batch 3900, loss[loss=0.1688, simple_loss=0.2468, pruned_loss=0.04538, over 16426.00 frames. ], tot_loss[loss=0.1717, simple_loss=0.2516, pruned_loss=0.04591, over 3305719.55 frames. ], batch size: 75, lr: 3.23e-03, grad_scale: 8.0 2023-05-01 06:12:03,155 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.49 vs. limit=2.0 2023-05-01 06:12:24,778 INFO [train.py:904] (1/8) Epoch 21, batch 3950, loss[loss=0.1744, simple_loss=0.2474, pruned_loss=0.0507, over 16678.00 frames. ], tot_loss[loss=0.1723, simple_loss=0.2517, pruned_loss=0.04651, over 3310864.00 frames. ], batch size: 89, lr: 3.23e-03, grad_scale: 8.0 2023-05-01 06:13:16,330 INFO [optim.py:368] (1/8) 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] (1/8) Epoch 21, batch 4000, loss[loss=0.144, simple_loss=0.2297, pruned_loss=0.02918, over 16447.00 frames. ], tot_loss[loss=0.1732, simple_loss=0.2521, pruned_loss=0.0471, over 3310198.33 frames. ], batch size: 68, lr: 3.23e-03, grad_scale: 8.0 2023-05-01 06:13:42,713 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.7256, 4.8313, 4.5525, 3.0593, 3.9299, 4.6517, 3.8478, 2.8474], device='cuda:1'), covar=tensor([0.0498, 0.0022, 0.0035, 0.0364, 0.0088, 0.0064, 0.0107, 0.0371], device='cuda:1'), in_proj_covar=tensor([0.0137, 0.0084, 0.0083, 0.0134, 0.0098, 0.0110, 0.0095, 0.0129], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:1') 2023-05-01 06:14:09,106 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.1599, 3.3560, 3.5045, 2.0058, 3.0279, 2.4172, 3.5292, 3.7213], device='cuda:1'), covar=tensor([0.0220, 0.0791, 0.0569, 0.2102, 0.0838, 0.0930, 0.0543, 0.0863], device='cuda:1'), in_proj_covar=tensor([0.0156, 0.0165, 0.0167, 0.0152, 0.0145, 0.0130, 0.0144, 0.0177], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-05-01 06:14:09,500 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=5.01 vs. limit=5.0 2023-05-01 06:14:21,190 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.7783, 1.7440, 2.3818, 2.6578, 2.6677, 2.9684, 1.6413, 2.9420], device='cuda:1'), covar=tensor([0.0170, 0.0575, 0.0298, 0.0283, 0.0295, 0.0173, 0.0746, 0.0118], device='cuda:1'), in_proj_covar=tensor([0.0190, 0.0195, 0.0181, 0.0186, 0.0199, 0.0156, 0.0199, 0.0152], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-01 06:14:45,530 INFO [train.py:904] (1/8) Epoch 21, batch 4050, loss[loss=0.1813, simple_loss=0.2603, pruned_loss=0.05114, over 16362.00 frames. ], tot_loss[loss=0.1729, simple_loss=0.2528, pruned_loss=0.04646, over 3300824.60 frames. ], batch size: 35, lr: 3.23e-03, grad_scale: 8.0 2023-05-01 06:15:05,399 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.48 vs. limit=2.0 2023-05-01 06:15:37,538 INFO [optim.py:368] (1/8) 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 06:15:55,999 INFO [train.py:904] (1/8) Epoch 21, batch 4100, loss[loss=0.1663, simple_loss=0.2475, pruned_loss=0.04258, over 16820.00 frames. ], tot_loss[loss=0.1733, simple_loss=0.2546, pruned_loss=0.04597, over 3295345.90 frames. ], batch size: 42, lr: 3.23e-03, grad_scale: 8.0 2023-05-01 06:17:10,031 INFO [train.py:904] (1/8) Epoch 21, batch 4150, loss[loss=0.2141, simple_loss=0.3023, pruned_loss=0.063, over 16875.00 frames. ], tot_loss[loss=0.1789, simple_loss=0.2618, pruned_loss=0.04797, over 3271427.15 frames. ], batch size: 109, lr: 3.22e-03, grad_scale: 8.0 2023-05-01 06:17:23,196 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.9532, 4.9368, 4.6459, 3.1767, 4.0335, 4.7215, 3.9505, 3.0312], device='cuda:1'), covar=tensor([0.0494, 0.0021, 0.0039, 0.0384, 0.0083, 0.0084, 0.0085, 0.0368], device='cuda:1'), in_proj_covar=tensor([0.0136, 0.0084, 0.0083, 0.0134, 0.0098, 0.0109, 0.0094, 0.0128], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-05-01 06:17:45,672 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=207176.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 06:18:04,516 INFO [optim.py:368] (1/8) 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 06:18:06,792 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=207190.0, num_to_drop=1, layers_to_drop={1} 2023-05-01 06:18:24,667 INFO [train.py:904] (1/8) Epoch 21, batch 4200, loss[loss=0.2181, simple_loss=0.3019, pruned_loss=0.06716, over 16844.00 frames. ], tot_loss[loss=0.184, simple_loss=0.269, pruned_loss=0.04952, over 3256186.87 frames. ], batch size: 116, lr: 3.22e-03, grad_scale: 8.0 2023-05-01 06:18:56,507 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.6177, 2.7296, 2.2790, 2.4194, 3.0428, 2.6563, 3.2034, 3.1977], device='cuda:1'), covar=tensor([0.0140, 0.0393, 0.0547, 0.0512, 0.0267, 0.0415, 0.0239, 0.0288], device='cuda:1'), in_proj_covar=tensor([0.0211, 0.0237, 0.0226, 0.0228, 0.0238, 0.0236, 0.0240, 0.0235], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-01 06:18:57,629 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.9510, 2.3626, 1.9771, 2.0578, 2.6691, 2.2281, 2.6306, 2.8089], device='cuda:1'), covar=tensor([0.0204, 0.0440, 0.0542, 0.0522, 0.0294, 0.0453, 0.0187, 0.0294], device='cuda:1'), in_proj_covar=tensor([0.0211, 0.0237, 0.0226, 0.0228, 0.0238, 0.0236, 0.0240, 0.0235], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-01 06:19:04,380 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.0345, 3.4732, 3.4878, 1.7998, 3.6656, 3.8255, 2.9655, 2.7155], device='cuda:1'), covar=tensor([0.1238, 0.0207, 0.0215, 0.1342, 0.0097, 0.0169, 0.0429, 0.0608], device='cuda:1'), in_proj_covar=tensor([0.0149, 0.0109, 0.0097, 0.0139, 0.0080, 0.0125, 0.0129, 0.0131], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0004, 0.0004], device='cuda:1') 2023-05-01 06:19:18,473 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=207237.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 06:19:40,047 INFO [train.py:904] (1/8) Epoch 21, batch 4250, loss[loss=0.1777, simple_loss=0.2691, pruned_loss=0.04322, over 15309.00 frames. ], tot_loss[loss=0.1855, simple_loss=0.2724, pruned_loss=0.0493, over 3237763.94 frames. ], batch size: 190, lr: 3.22e-03, grad_scale: 8.0 2023-05-01 06:19:51,551 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.6390, 2.5981, 1.8476, 2.7720, 2.0806, 2.7837, 2.1328, 2.3991], device='cuda:1'), covar=tensor([0.0288, 0.0339, 0.1314, 0.0291, 0.0681, 0.0378, 0.1185, 0.0580], device='cuda:1'), in_proj_covar=tensor([0.0169, 0.0176, 0.0193, 0.0162, 0.0177, 0.0216, 0.0200, 0.0177], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-05-01 06:20:35,820 INFO [optim.py:368] (1/8) 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 06:20:55,788 INFO [train.py:904] (1/8) Epoch 21, batch 4300, loss[loss=0.1982, simple_loss=0.2828, pruned_loss=0.05682, over 11842.00 frames. ], tot_loss[loss=0.1853, simple_loss=0.2734, pruned_loss=0.04863, over 3216074.09 frames. ], batch size: 246, lr: 3.22e-03, grad_scale: 8.0 2023-05-01 06:21:18,666 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=4.63 vs. limit=5.0 2023-05-01 06:21:42,201 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.0489, 2.1824, 2.6329, 3.0705, 2.9069, 3.4813, 2.1720, 3.3824], device='cuda:1'), covar=tensor([0.0227, 0.0521, 0.0329, 0.0304, 0.0313, 0.0144, 0.0544, 0.0120], device='cuda:1'), in_proj_covar=tensor([0.0189, 0.0194, 0.0179, 0.0186, 0.0198, 0.0154, 0.0197, 0.0151], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-01 06:22:07,507 INFO [train.py:904] (1/8) Epoch 21, batch 4350, loss[loss=0.1995, simple_loss=0.2923, pruned_loss=0.05334, over 17098.00 frames. ], tot_loss[loss=0.1889, simple_loss=0.2775, pruned_loss=0.05016, over 3226300.84 frames. ], batch size: 47, lr: 3.22e-03, grad_scale: 8.0 2023-05-01 06:22:21,865 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.5622, 3.7544, 2.2839, 4.4110, 2.9012, 4.3063, 2.5037, 2.9570], device='cuda:1'), covar=tensor([0.0319, 0.0320, 0.1677, 0.0127, 0.0789, 0.0442, 0.1440, 0.0800], device='cuda:1'), in_proj_covar=tensor([0.0170, 0.0177, 0.0194, 0.0163, 0.0177, 0.0217, 0.0200, 0.0179], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-05-01 06:22:34,577 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.8491, 3.4889, 4.0179, 1.9680, 4.2394, 4.2659, 3.0857, 3.1031], device='cuda:1'), covar=tensor([0.0722, 0.0285, 0.0207, 0.1220, 0.0058, 0.0105, 0.0464, 0.0451], device='cuda:1'), in_proj_covar=tensor([0.0149, 0.0109, 0.0097, 0.0139, 0.0080, 0.0125, 0.0129, 0.0131], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0004, 0.0004], device='cuda:1') 2023-05-01 06:23:02,765 INFO [optim.py:368] (1/8) 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] (1/8) Epoch 21, batch 4400, loss[loss=0.2142, simple_loss=0.3017, pruned_loss=0.06341, over 16395.00 frames. ], tot_loss[loss=0.1912, simple_loss=0.2796, pruned_loss=0.05135, over 3197901.06 frames. ], batch size: 35, lr: 3.22e-03, grad_scale: 8.0 2023-05-01 06:23:30,448 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.3102, 3.4828, 3.6085, 3.5804, 3.5890, 3.4245, 3.4530, 3.4823], device='cuda:1'), covar=tensor([0.0355, 0.0523, 0.0393, 0.0416, 0.0506, 0.0439, 0.0703, 0.0500], device='cuda:1'), in_proj_covar=tensor([0.0408, 0.0454, 0.0440, 0.0408, 0.0484, 0.0461, 0.0549, 0.0370], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:1') 2023-05-01 06:24:02,541 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.3807, 1.6841, 2.1002, 2.3283, 2.4058, 2.6509, 1.7460, 2.4732], device='cuda:1'), covar=tensor([0.0243, 0.0554, 0.0298, 0.0342, 0.0298, 0.0202, 0.0580, 0.0159], device='cuda:1'), in_proj_covar=tensor([0.0188, 0.0193, 0.0178, 0.0185, 0.0196, 0.0153, 0.0197, 0.0150], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-01 06:24:37,197 INFO [train.py:904] (1/8) Epoch 21, batch 4450, loss[loss=0.2029, simple_loss=0.2825, pruned_loss=0.06161, over 16461.00 frames. ], tot_loss[loss=0.1938, simple_loss=0.2826, pruned_loss=0.05247, over 3210183.42 frames. ], batch size: 35, lr: 3.22e-03, grad_scale: 8.0 2023-05-01 06:25:31,509 INFO [optim.py:368] (1/8) 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 06:25:33,766 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=207490.0, num_to_drop=1, layers_to_drop={0} 2023-05-01 06:25:35,165 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([1.9591, 2.3430, 2.4050, 2.4901, 1.9744, 3.1737, 1.9228, 2.7024], device='cuda:1'), covar=tensor([0.1101, 0.0661, 0.0973, 0.0169, 0.0132, 0.0312, 0.1318, 0.0691], device='cuda:1'), in_proj_covar=tensor([0.0166, 0.0175, 0.0194, 0.0191, 0.0208, 0.0215, 0.0201, 0.0193], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-05-01 06:25:50,497 INFO [train.py:904] (1/8) Epoch 21, batch 4500, loss[loss=0.1815, simple_loss=0.2697, pruned_loss=0.04664, over 16462.00 frames. ], tot_loss[loss=0.195, simple_loss=0.2834, pruned_loss=0.05326, over 3204928.10 frames. ], batch size: 146, lr: 3.22e-03, grad_scale: 8.0 2023-05-01 06:26:35,639 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=207532.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 06:26:43,762 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=207538.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 06:27:03,813 INFO [train.py:904] (1/8) Epoch 21, batch 4550, loss[loss=0.2035, simple_loss=0.2856, pruned_loss=0.0607, over 16680.00 frames. ], tot_loss[loss=0.1956, simple_loss=0.2836, pruned_loss=0.05379, over 3213333.62 frames. ], batch size: 124, lr: 3.22e-03, grad_scale: 8.0 2023-05-01 06:27:05,678 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.0348, 3.0786, 1.8526, 3.2944, 2.3710, 3.3345, 2.0787, 2.5380], device='cuda:1'), covar=tensor([0.0304, 0.0364, 0.1638, 0.0163, 0.0794, 0.0462, 0.1507, 0.0750], device='cuda:1'), in_proj_covar=tensor([0.0170, 0.0177, 0.0194, 0.0163, 0.0177, 0.0216, 0.0200, 0.0179], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-05-01 06:27:40,967 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.18 vs. limit=2.0 2023-05-01 06:27:57,547 INFO [optim.py:368] (1/8) 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 06:28:16,260 INFO [train.py:904] (1/8) Epoch 21, batch 4600, loss[loss=0.1978, simple_loss=0.2814, pruned_loss=0.05707, over 16431.00 frames. ], tot_loss[loss=0.1963, simple_loss=0.2843, pruned_loss=0.0542, over 3194393.87 frames. ], batch size: 35, lr: 3.22e-03, grad_scale: 8.0 2023-05-01 06:29:02,862 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=2.76 vs. limit=5.0 2023-05-01 06:29:07,206 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.9130, 5.2970, 5.4836, 5.1773, 5.2584, 5.8508, 5.3771, 5.1368], device='cuda:1'), covar=tensor([0.1005, 0.1757, 0.1933, 0.1977, 0.2555, 0.0888, 0.1358, 0.2282], device='cuda:1'), in_proj_covar=tensor([0.0402, 0.0584, 0.0637, 0.0486, 0.0646, 0.0672, 0.0498, 0.0648], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-05-01 06:29:29,137 INFO [train.py:904] (1/8) Epoch 21, batch 4650, loss[loss=0.1852, simple_loss=0.2738, pruned_loss=0.04829, over 16864.00 frames. ], tot_loss[loss=0.1963, simple_loss=0.2837, pruned_loss=0.05448, over 3200746.59 frames. ], batch size: 96, lr: 3.22e-03, grad_scale: 16.0 2023-05-01 06:30:23,479 INFO [optim.py:368] (1/8) 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] (1/8) Epoch 21, batch 4700, loss[loss=0.1832, simple_loss=0.2681, pruned_loss=0.04913, over 11620.00 frames. ], tot_loss[loss=0.1933, simple_loss=0.2806, pruned_loss=0.05297, over 3211445.63 frames. ], batch size: 247, lr: 3.22e-03, grad_scale: 16.0 2023-05-01 06:31:56,745 INFO [train.py:904] (1/8) Epoch 21, batch 4750, loss[loss=0.1683, simple_loss=0.2449, pruned_loss=0.04591, over 16328.00 frames. ], tot_loss[loss=0.1882, simple_loss=0.2757, pruned_loss=0.05034, over 3218493.81 frames. ], batch size: 35, lr: 3.22e-03, grad_scale: 16.0 2023-05-01 06:32:03,200 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=4.42 vs. limit=5.0 2023-05-01 06:32:25,324 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.82 vs. limit=2.0 2023-05-01 06:32:46,941 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([5.3815, 5.4098, 5.2339, 4.8040, 4.8221, 5.2680, 5.2950, 4.9312], device='cuda:1'), covar=tensor([0.0642, 0.0425, 0.0309, 0.0305, 0.1168, 0.0619, 0.0231, 0.0733], device='cuda:1'), in_proj_covar=tensor([0.0290, 0.0417, 0.0339, 0.0333, 0.0348, 0.0388, 0.0231, 0.0407], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002], device='cuda:1') 2023-05-01 06:32:50,130 INFO [optim.py:368] (1/8) 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 06:32:54,809 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.15 vs. limit=2.0 2023-05-01 06:32:58,448 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=4.75 vs. limit=5.0 2023-05-01 06:33:11,166 INFO [train.py:904] (1/8) Epoch 21, batch 4800, loss[loss=0.1906, simple_loss=0.2712, pruned_loss=0.05499, over 11858.00 frames. ], tot_loss[loss=0.184, simple_loss=0.2718, pruned_loss=0.0481, over 3223473.61 frames. ], batch size: 246, lr: 3.22e-03, grad_scale: 16.0 2023-05-01 06:33:14,436 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.48 vs. limit=2.0 2023-05-01 06:33:55,469 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=207832.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 06:34:01,225 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.4659, 3.6637, 3.6386, 2.1645, 2.9849, 2.3907, 3.7692, 3.8244], device='cuda:1'), covar=tensor([0.0237, 0.0743, 0.0631, 0.1953, 0.0876, 0.0955, 0.0611, 0.0998], device='cuda:1'), in_proj_covar=tensor([0.0154, 0.0164, 0.0166, 0.0151, 0.0144, 0.0129, 0.0143, 0.0175], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:1') 2023-05-01 06:34:24,539 INFO [train.py:904] (1/8) Epoch 21, batch 4850, loss[loss=0.1793, simple_loss=0.2612, pruned_loss=0.04868, over 16825.00 frames. ], tot_loss[loss=0.1833, simple_loss=0.2722, pruned_loss=0.04714, over 3214801.63 frames. ], batch size: 39, lr: 3.22e-03, grad_scale: 8.0 2023-05-01 06:35:08,162 INFO [zipformer.py:625] (1/8) 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] (1/8) 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] (1/8) Epoch 21, batch 4900, loss[loss=0.1627, simple_loss=0.2609, pruned_loss=0.03226, over 16857.00 frames. ], tot_loss[loss=0.1823, simple_loss=0.2721, pruned_loss=0.04623, over 3193507.29 frames. ], batch size: 96, lr: 3.22e-03, grad_scale: 8.0 2023-05-01 06:35:47,656 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.0116, 4.1154, 3.9228, 3.6632, 3.6907, 3.9925, 3.6987, 3.7980], device='cuda:1'), covar=tensor([0.0624, 0.0488, 0.0298, 0.0269, 0.0716, 0.0476, 0.1050, 0.0505], device='cuda:1'), in_proj_covar=tensor([0.0286, 0.0413, 0.0335, 0.0330, 0.0344, 0.0384, 0.0229, 0.0402], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-01 06:36:52,788 INFO [train.py:904] (1/8) Epoch 21, batch 4950, loss[loss=0.1758, simple_loss=0.2711, pruned_loss=0.04029, over 16230.00 frames. ], tot_loss[loss=0.1821, simple_loss=0.2722, pruned_loss=0.04596, over 3192787.07 frames. ], batch size: 165, lr: 3.22e-03, grad_scale: 8.0 2023-05-01 06:36:57,455 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.1133, 1.5598, 1.8886, 2.1003, 2.2789, 2.4257, 1.7802, 2.2818], device='cuda:1'), covar=tensor([0.0231, 0.0476, 0.0294, 0.0384, 0.0293, 0.0190, 0.0501, 0.0129], device='cuda:1'), in_proj_covar=tensor([0.0187, 0.0192, 0.0178, 0.0184, 0.0195, 0.0152, 0.0197, 0.0149], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-01 06:37:47,806 INFO [optim.py:368] (1/8) Clipping_scale=2.0, grad-norm quartiles 1.452e+02 1.964e+02 2.322e+02 2.960e+02 4.980e+02, threshold=4.644e+02, percent-clipped=1.0 2023-05-01 06:38:08,406 INFO [train.py:904] (1/8) Epoch 21, batch 5000, loss[loss=0.1669, simple_loss=0.2564, pruned_loss=0.03871, over 16798.00 frames. ], tot_loss[loss=0.1838, simple_loss=0.2748, pruned_loss=0.04641, over 3194797.77 frames. ], batch size: 39, lr: 3.22e-03, grad_scale: 8.0 2023-05-01 06:38:52,042 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=208032.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 06:39:21,547 INFO [train.py:904] (1/8) Epoch 21, batch 5050, loss[loss=0.1838, simple_loss=0.2665, pruned_loss=0.05052, over 17151.00 frames. ], tot_loss[loss=0.1833, simple_loss=0.2745, pruned_loss=0.04601, over 3203120.81 frames. ], batch size: 46, lr: 3.22e-03, grad_scale: 8.0 2023-05-01 06:40:18,551 INFO [optim.py:368] (1/8) 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 06:40:23,269 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=208093.0, num_to_drop=1, layers_to_drop={2} 2023-05-01 06:40:35,283 INFO [train.py:904] (1/8) Epoch 21, batch 5100, loss[loss=0.1555, simple_loss=0.2511, pruned_loss=0.02998, over 16675.00 frames. ], tot_loss[loss=0.1812, simple_loss=0.2722, pruned_loss=0.04515, over 3220591.29 frames. ], batch size: 89, lr: 3.22e-03, grad_scale: 8.0 2023-05-01 06:40:48,252 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.3174, 2.2775, 2.3258, 4.0172, 2.2244, 2.5716, 2.3236, 2.4393], device='cuda:1'), covar=tensor([0.1308, 0.3763, 0.2860, 0.0538, 0.4088, 0.2754, 0.3592, 0.3314], device='cuda:1'), in_proj_covar=tensor([0.0400, 0.0448, 0.0365, 0.0328, 0.0435, 0.0515, 0.0416, 0.0522], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-01 06:41:48,626 INFO [train.py:904] (1/8) Epoch 21, batch 5150, loss[loss=0.1841, simple_loss=0.2797, pruned_loss=0.04423, over 16452.00 frames. ], tot_loss[loss=0.181, simple_loss=0.2726, pruned_loss=0.04465, over 3215531.90 frames. ], batch size: 68, lr: 3.22e-03, grad_scale: 8.0 2023-05-01 06:41:57,834 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.8069, 3.9684, 4.1434, 4.1215, 4.1074, 3.9279, 3.9069, 3.9218], device='cuda:1'), covar=tensor([0.0345, 0.0552, 0.0380, 0.0396, 0.0492, 0.0389, 0.0777, 0.0446], device='cuda:1'), in_proj_covar=tensor([0.0400, 0.0444, 0.0430, 0.0400, 0.0475, 0.0453, 0.0539, 0.0363], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:1') 2023-05-01 06:42:36,094 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.8304, 5.0868, 5.2596, 5.0190, 5.1215, 5.6735, 5.1364, 4.8521], device='cuda:1'), covar=tensor([0.0956, 0.1851, 0.2032, 0.1998, 0.2374, 0.0831, 0.1568, 0.2433], device='cuda:1'), in_proj_covar=tensor([0.0399, 0.0574, 0.0626, 0.0481, 0.0639, 0.0661, 0.0493, 0.0642], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-05-01 06:42:43,675 INFO [optim.py:368] (1/8) 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 06:43:01,060 INFO [train.py:904] (1/8) Epoch 21, batch 5200, loss[loss=0.1769, simple_loss=0.2717, pruned_loss=0.04101, over 15519.00 frames. ], tot_loss[loss=0.1796, simple_loss=0.2711, pruned_loss=0.044, over 3206427.34 frames. ], batch size: 190, lr: 3.22e-03, grad_scale: 8.0 2023-05-01 06:44:00,528 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.0891, 4.9249, 4.7638, 3.2485, 4.0940, 4.7983, 3.9820, 2.7374], device='cuda:1'), covar=tensor([0.0427, 0.0027, 0.0029, 0.0326, 0.0078, 0.0072, 0.0092, 0.0394], device='cuda:1'), in_proj_covar=tensor([0.0134, 0.0082, 0.0082, 0.0132, 0.0097, 0.0107, 0.0093, 0.0126], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-05-01 06:44:11,872 INFO [train.py:904] (1/8) Epoch 21, batch 5250, loss[loss=0.1586, simple_loss=0.2515, pruned_loss=0.0328, over 16586.00 frames. ], tot_loss[loss=0.1775, simple_loss=0.2681, pruned_loss=0.04347, over 3219417.57 frames. ], batch size: 62, lr: 3.22e-03, grad_scale: 8.0 2023-05-01 06:44:31,839 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=208265.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 06:44:45,235 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.4558, 2.9120, 3.0011, 2.0453, 2.6670, 2.1387, 3.0856, 3.1280], device='cuda:1'), covar=tensor([0.0290, 0.0770, 0.0647, 0.1929, 0.0912, 0.0991, 0.0586, 0.0805], device='cuda:1'), in_proj_covar=tensor([0.0155, 0.0163, 0.0167, 0.0152, 0.0145, 0.0130, 0.0144, 0.0175], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-05-01 06:45:07,260 INFO [optim.py:368] (1/8) 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] (1/8) Epoch 21, batch 5300, loss[loss=0.1669, simple_loss=0.2564, pruned_loss=0.0387, over 16652.00 frames. ], tot_loss[loss=0.1752, simple_loss=0.2652, pruned_loss=0.04255, over 3223311.25 frames. ], batch size: 134, lr: 3.22e-03, grad_scale: 8.0 2023-05-01 06:45:36,045 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.1645, 2.2058, 2.2785, 3.8323, 2.1190, 2.5899, 2.3134, 2.4285], device='cuda:1'), covar=tensor([0.1421, 0.3536, 0.3043, 0.0556, 0.4093, 0.2418, 0.3545, 0.3203], device='cuda:1'), in_proj_covar=tensor([0.0400, 0.0445, 0.0365, 0.0327, 0.0434, 0.0514, 0.0415, 0.0520], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-01 06:46:00,445 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=208326.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 06:46:00,802 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.30 vs. limit=2.0 2023-05-01 06:46:05,932 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([5.3928, 5.4113, 5.2615, 4.8472, 4.8820, 5.2861, 5.3085, 4.9865], device='cuda:1'), covar=tensor([0.0658, 0.0405, 0.0289, 0.0273, 0.1148, 0.0484, 0.0215, 0.0610], device='cuda:1'), in_proj_covar=tensor([0.0288, 0.0416, 0.0337, 0.0332, 0.0346, 0.0388, 0.0230, 0.0404], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002], device='cuda:1') 2023-05-01 06:46:27,536 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.1620, 4.2630, 4.0739, 3.7625, 3.7465, 4.1629, 3.9147, 3.9359], device='cuda:1'), covar=tensor([0.0660, 0.0582, 0.0322, 0.0299, 0.0892, 0.0539, 0.0775, 0.0588], device='cuda:1'), in_proj_covar=tensor([0.0288, 0.0416, 0.0338, 0.0333, 0.0347, 0.0389, 0.0231, 0.0405], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002], device='cuda:1') 2023-05-01 06:46:37,966 INFO [train.py:904] (1/8) Epoch 21, batch 5350, loss[loss=0.1553, simple_loss=0.2525, pruned_loss=0.02903, over 16772.00 frames. ], tot_loss[loss=0.1736, simple_loss=0.2637, pruned_loss=0.04178, over 3230533.86 frames. ], batch size: 102, lr: 3.22e-03, grad_scale: 8.0 2023-05-01 06:46:56,646 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([1.6754, 2.4367, 2.3087, 3.5644, 2.2554, 3.7386, 1.3791, 2.7662], device='cuda:1'), covar=tensor([0.1337, 0.0827, 0.1291, 0.0170, 0.0152, 0.0358, 0.1779, 0.0846], device='cuda:1'), in_proj_covar=tensor([0.0166, 0.0175, 0.0194, 0.0190, 0.0208, 0.0214, 0.0201, 0.0193], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-05-01 06:47:32,375 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=208388.0, num_to_drop=1, layers_to_drop={0} 2023-05-01 06:47:34,961 INFO [optim.py:368] (1/8) 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 06:47:53,665 INFO [train.py:904] (1/8) Epoch 21, batch 5400, loss[loss=0.2006, simple_loss=0.2983, pruned_loss=0.05145, over 16194.00 frames. ], tot_loss[loss=0.1755, simple_loss=0.2664, pruned_loss=0.04226, over 3242317.03 frames. ], batch size: 165, lr: 3.22e-03, grad_scale: 8.0 2023-05-01 06:49:10,972 INFO [train.py:904] (1/8) Epoch 21, batch 5450, loss[loss=0.1736, simple_loss=0.268, pruned_loss=0.03956, over 16763.00 frames. ], tot_loss[loss=0.1787, simple_loss=0.2694, pruned_loss=0.04399, over 3219676.62 frames. ], batch size: 83, lr: 3.21e-03, grad_scale: 8.0 2023-05-01 06:50:05,002 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.6591, 2.5936, 1.8468, 2.7710, 2.1849, 2.8022, 2.1180, 2.4128], device='cuda:1'), covar=tensor([0.0289, 0.0357, 0.1250, 0.0252, 0.0611, 0.0463, 0.1183, 0.0560], device='cuda:1'), in_proj_covar=tensor([0.0167, 0.0175, 0.0192, 0.0160, 0.0175, 0.0214, 0.0200, 0.0177], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-05-01 06:50:09,143 INFO [optim.py:368] (1/8) 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] (1/8) Epoch 21, batch 5500, loss[loss=0.2309, simple_loss=0.3131, pruned_loss=0.07433, over 15373.00 frames. ], tot_loss[loss=0.1872, simple_loss=0.277, pruned_loss=0.04871, over 3192595.43 frames. ], batch size: 191, lr: 3.21e-03, grad_scale: 8.0 2023-05-01 06:51:09,378 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.4477, 4.5033, 4.7400, 4.5160, 4.6930, 5.1757, 4.6673, 4.4208], device='cuda:1'), covar=tensor([0.1452, 0.2144, 0.2444, 0.2092, 0.2450, 0.0976, 0.1667, 0.2528], device='cuda:1'), in_proj_covar=tensor([0.0400, 0.0578, 0.0630, 0.0483, 0.0642, 0.0666, 0.0496, 0.0645], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-05-01 06:51:47,648 INFO [train.py:904] (1/8) Epoch 21, batch 5550, loss[loss=0.1948, simple_loss=0.2844, pruned_loss=0.05262, over 16854.00 frames. ], tot_loss[loss=0.1951, simple_loss=0.2837, pruned_loss=0.05329, over 3175767.51 frames. ], batch size: 42, lr: 3.21e-03, grad_scale: 8.0 2023-05-01 06:52:25,674 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.70 vs. limit=2.0 2023-05-01 06:52:27,832 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=208576.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 06:52:49,284 INFO [optim.py:368] (1/8) 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 06:52:54,487 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([1.5974, 2.4960, 2.3577, 3.6393, 2.5619, 3.8135, 1.4430, 2.7445], device='cuda:1'), covar=tensor([0.1500, 0.0892, 0.1391, 0.0239, 0.0243, 0.0425, 0.1902, 0.0911], device='cuda:1'), in_proj_covar=tensor([0.0166, 0.0175, 0.0194, 0.0189, 0.0207, 0.0214, 0.0201, 0.0193], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-05-01 06:53:07,711 INFO [train.py:904] (1/8) Epoch 21, batch 5600, loss[loss=0.1982, simple_loss=0.2852, pruned_loss=0.05557, over 17188.00 frames. ], tot_loss[loss=0.2028, simple_loss=0.289, pruned_loss=0.05833, over 3130713.77 frames. ], batch size: 45, lr: 3.21e-03, grad_scale: 8.0 2023-05-01 06:53:39,436 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=208621.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 06:54:06,485 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=208637.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 06:54:18,188 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.1266, 2.1740, 2.2224, 3.8495, 2.0843, 2.5255, 2.2926, 2.3357], device='cuda:1'), covar=tensor([0.1360, 0.3239, 0.2747, 0.0537, 0.3913, 0.2334, 0.3116, 0.3197], device='cuda:1'), in_proj_covar=tensor([0.0397, 0.0441, 0.0361, 0.0323, 0.0429, 0.0507, 0.0410, 0.0514], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-01 06:54:29,858 INFO [train.py:904] (1/8) Epoch 21, batch 5650, loss[loss=0.2149, simple_loss=0.3011, pruned_loss=0.06435, over 16253.00 frames. ], tot_loss[loss=0.2095, simple_loss=0.2943, pruned_loss=0.06236, over 3105526.40 frames. ], batch size: 35, lr: 3.21e-03, grad_scale: 8.0 2023-05-01 06:55:28,104 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=208688.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 06:55:30,677 INFO [optim.py:368] (1/8) Clipping_scale=2.0, grad-norm quartiles 2.148e+02 3.160e+02 3.845e+02 4.644e+02 8.553e+02, threshold=7.690e+02, percent-clipped=3.0 2023-05-01 06:55:50,854 INFO [train.py:904] (1/8) Epoch 21, batch 5700, loss[loss=0.2566, simple_loss=0.3174, pruned_loss=0.09783, over 11341.00 frames. ], tot_loss[loss=0.2137, simple_loss=0.297, pruned_loss=0.06522, over 3056886.97 frames. ], batch size: 248, lr: 3.21e-03, grad_scale: 8.0 2023-05-01 06:56:12,120 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.33 vs. limit=2.0 2023-05-01 06:56:45,710 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=208736.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 06:57:08,933 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.0689, 4.0576, 3.9572, 3.1335, 3.9865, 1.8806, 3.7868, 3.4766], device='cuda:1'), covar=tensor([0.0114, 0.0096, 0.0182, 0.0291, 0.0085, 0.2735, 0.0124, 0.0248], device='cuda:1'), in_proj_covar=tensor([0.0160, 0.0150, 0.0191, 0.0173, 0.0170, 0.0201, 0.0181, 0.0168], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-01 06:57:10,969 INFO [train.py:904] (1/8) Epoch 21, batch 5750, loss[loss=0.203, simple_loss=0.2921, pruned_loss=0.05697, over 16828.00 frames. ], tot_loss[loss=0.2164, simple_loss=0.2994, pruned_loss=0.06667, over 3029109.73 frames. ], batch size: 116, lr: 3.21e-03, grad_scale: 8.0 2023-05-01 06:58:13,920 INFO [optim.py:368] (1/8) 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 06:58:33,851 INFO [train.py:904] (1/8) Epoch 21, batch 5800, loss[loss=0.1874, simple_loss=0.2829, pruned_loss=0.04601, over 16398.00 frames. ], tot_loss[loss=0.2138, simple_loss=0.2986, pruned_loss=0.06451, over 3043193.37 frames. ], batch size: 146, lr: 3.21e-03, grad_scale: 8.0 2023-05-01 06:59:11,507 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([1.6932, 2.2933, 2.1664, 3.2242, 2.1628, 3.5419, 1.5057, 2.6753], device='cuda:1'), covar=tensor([0.1434, 0.0868, 0.1433, 0.0212, 0.0185, 0.0431, 0.1845, 0.0901], device='cuda:1'), in_proj_covar=tensor([0.0165, 0.0174, 0.0194, 0.0189, 0.0206, 0.0213, 0.0201, 0.0192], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-05-01 06:59:52,593 INFO [train.py:904] (1/8) Epoch 21, batch 5850, loss[loss=0.2072, simple_loss=0.2988, pruned_loss=0.05776, over 16581.00 frames. ], tot_loss[loss=0.2111, simple_loss=0.2962, pruned_loss=0.06293, over 3017867.98 frames. ], batch size: 62, lr: 3.21e-03, grad_scale: 8.0 2023-05-01 07:00:00,760 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.1801, 3.2251, 1.8874, 3.4516, 2.4434, 3.4981, 2.1030, 2.6052], device='cuda:1'), covar=tensor([0.0306, 0.0379, 0.1681, 0.0221, 0.0840, 0.0569, 0.1508, 0.0743], device='cuda:1'), in_proj_covar=tensor([0.0167, 0.0175, 0.0192, 0.0160, 0.0175, 0.0213, 0.0199, 0.0176], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-05-01 07:00:41,372 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.2826, 3.4430, 3.5987, 3.5683, 3.5860, 3.3604, 3.4192, 3.4757], device='cuda:1'), covar=tensor([0.0450, 0.0809, 0.0515, 0.0497, 0.0565, 0.0668, 0.0877, 0.0613], device='cuda:1'), in_proj_covar=tensor([0.0402, 0.0447, 0.0434, 0.0402, 0.0476, 0.0457, 0.0543, 0.0366], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:1') 2023-05-01 07:00:53,525 INFO [optim.py:368] (1/8) 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 07:01:07,929 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([1.7676, 2.7227, 2.5065, 4.5059, 3.3427, 4.0216, 1.6527, 3.0789], device='cuda:1'), covar=tensor([0.1350, 0.0763, 0.1291, 0.0145, 0.0265, 0.0389, 0.1638, 0.0822], device='cuda:1'), in_proj_covar=tensor([0.0166, 0.0174, 0.0194, 0.0189, 0.0207, 0.0214, 0.0201, 0.0192], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-05-01 07:01:12,718 INFO [train.py:904] (1/8) Epoch 21, batch 5900, loss[loss=0.1855, simple_loss=0.2804, pruned_loss=0.04528, over 16273.00 frames. ], tot_loss[loss=0.2119, simple_loss=0.2962, pruned_loss=0.06381, over 2997036.19 frames. ], batch size: 35, lr: 3.21e-03, grad_scale: 8.0 2023-05-01 07:01:43,442 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.83 vs. limit=2.0 2023-05-01 07:01:48,476 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=208921.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 07:01:54,327 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.75 vs. limit=2.0 2023-05-01 07:02:04,488 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=208932.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 07:02:11,294 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.35 vs. limit=2.0 2023-05-01 07:02:36,220 INFO [train.py:904] (1/8) Epoch 21, batch 5950, loss[loss=0.1875, simple_loss=0.2782, pruned_loss=0.0484, over 17215.00 frames. ], tot_loss[loss=0.2104, simple_loss=0.2965, pruned_loss=0.06216, over 3021425.72 frames. ], batch size: 44, lr: 3.21e-03, grad_scale: 8.0 2023-05-01 07:03:02,963 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=208969.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 07:03:03,220 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.3803, 3.2956, 2.6175, 2.1702, 2.2780, 2.2410, 3.3956, 3.0725], device='cuda:1'), covar=tensor([0.2999, 0.0690, 0.1810, 0.2665, 0.2467, 0.2178, 0.0502, 0.1271], device='cuda:1'), in_proj_covar=tensor([0.0324, 0.0267, 0.0301, 0.0307, 0.0293, 0.0254, 0.0293, 0.0331], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:1') 2023-05-01 07:03:36,545 INFO [optim.py:368] (1/8) 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 07:03:56,981 INFO [train.py:904] (1/8) Epoch 21, batch 6000, loss[loss=0.1898, simple_loss=0.2776, pruned_loss=0.05097, over 16710.00 frames. ], tot_loss[loss=0.2079, simple_loss=0.2944, pruned_loss=0.06068, over 3049426.67 frames. ], batch size: 62, lr: 3.21e-03, grad_scale: 8.0 2023-05-01 07:03:56,981 INFO [train.py:929] (1/8) Computing validation loss 2023-05-01 07:04:08,270 INFO [train.py:938] (1/8) Epoch 21, validation: loss=0.1512, simple_loss=0.2639, pruned_loss=0.01924, over 944034.00 frames. 2023-05-01 07:04:08,270 INFO [train.py:939] (1/8) Maximum memory allocated so far is 18145MB 2023-05-01 07:04:21,626 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.46 vs. limit=2.0 2023-05-01 07:05:25,680 INFO [train.py:904] (1/8) Epoch 21, batch 6050, loss[loss=0.2059, simple_loss=0.3035, pruned_loss=0.0542, over 16726.00 frames. ], tot_loss[loss=0.2059, simple_loss=0.2927, pruned_loss=0.05955, over 3076977.81 frames. ], batch size: 57, lr: 3.21e-03, grad_scale: 8.0 2023-05-01 07:05:34,140 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([1.7366, 2.7658, 2.6130, 4.8085, 3.5065, 4.1730, 1.5936, 3.1162], device='cuda:1'), covar=tensor([0.1394, 0.0795, 0.1264, 0.0182, 0.0332, 0.0426, 0.1720, 0.0835], device='cuda:1'), in_proj_covar=tensor([0.0165, 0.0174, 0.0194, 0.0189, 0.0206, 0.0214, 0.0201, 0.0192], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-05-01 07:05:43,584 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.9336, 2.0646, 2.2504, 3.3923, 2.0203, 2.3251, 2.1769, 2.1906], device='cuda:1'), covar=tensor([0.1478, 0.3586, 0.2981, 0.0677, 0.4347, 0.2660, 0.3662, 0.3506], device='cuda:1'), in_proj_covar=tensor([0.0399, 0.0444, 0.0363, 0.0325, 0.0433, 0.0510, 0.0413, 0.0518], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-01 07:06:14,784 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.7444, 4.5535, 4.7454, 4.9272, 5.0822, 4.5660, 5.0772, 5.0807], device='cuda:1'), covar=tensor([0.1912, 0.1344, 0.1639, 0.0755, 0.0679, 0.1023, 0.0735, 0.0698], device='cuda:1'), in_proj_covar=tensor([0.0621, 0.0765, 0.0893, 0.0780, 0.0586, 0.0616, 0.0634, 0.0734], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-01 07:06:26,989 INFO [optim.py:368] (1/8) 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 07:06:45,918 INFO [train.py:904] (1/8) Epoch 21, batch 6100, loss[loss=0.1931, simple_loss=0.2782, pruned_loss=0.05401, over 15520.00 frames. ], tot_loss[loss=0.2041, simple_loss=0.2918, pruned_loss=0.05824, over 3077464.45 frames. ], batch size: 191, lr: 3.21e-03, grad_scale: 8.0 2023-05-01 07:07:54,287 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([1.7784, 2.3922, 2.1760, 3.1694, 2.1389, 3.5528, 1.6059, 2.6177], device='cuda:1'), covar=tensor([0.1242, 0.0782, 0.1378, 0.0211, 0.0160, 0.0473, 0.1655, 0.0899], device='cuda:1'), in_proj_covar=tensor([0.0166, 0.0174, 0.0194, 0.0189, 0.0206, 0.0214, 0.0201, 0.0192], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-05-01 07:08:04,716 INFO [train.py:904] (1/8) Epoch 21, batch 6150, loss[loss=0.2138, simple_loss=0.3027, pruned_loss=0.06243, over 16736.00 frames. ], tot_loss[loss=0.2031, simple_loss=0.2903, pruned_loss=0.05795, over 3082794.38 frames. ], batch size: 124, lr: 3.21e-03, grad_scale: 8.0 2023-05-01 07:08:42,667 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.84 vs. limit=2.0 2023-05-01 07:08:51,087 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=4.37 vs. limit=5.0 2023-05-01 07:09:04,297 INFO [optim.py:368] (1/8) 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 07:09:23,294 INFO [train.py:904] (1/8) Epoch 21, batch 6200, loss[loss=0.2251, simple_loss=0.2962, pruned_loss=0.07699, over 11827.00 frames. ], tot_loss[loss=0.202, simple_loss=0.2884, pruned_loss=0.05779, over 3088902.36 frames. ], batch size: 248, lr: 3.21e-03, grad_scale: 8.0 2023-05-01 07:09:23,899 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=209202.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 07:09:41,751 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.51 vs. limit=2.0 2023-05-01 07:09:43,668 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.9001, 4.1696, 4.0029, 4.0173, 3.7153, 3.7618, 3.8365, 4.1694], device='cuda:1'), covar=tensor([0.1103, 0.0867, 0.0931, 0.0854, 0.0801, 0.1835, 0.0933, 0.1026], device='cuda:1'), in_proj_covar=tensor([0.0653, 0.0793, 0.0661, 0.0603, 0.0504, 0.0512, 0.0666, 0.0620], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:1') 2023-05-01 07:10:07,072 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.4996, 3.5662, 3.3345, 3.0631, 3.1738, 3.4539, 3.3183, 3.2584], device='cuda:1'), covar=tensor([0.0601, 0.0625, 0.0269, 0.0257, 0.0451, 0.0487, 0.1090, 0.0453], device='cuda:1'), in_proj_covar=tensor([0.0287, 0.0415, 0.0335, 0.0331, 0.0343, 0.0385, 0.0230, 0.0402], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-01 07:10:09,495 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=209232.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 07:10:28,263 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=4.45 vs. limit=5.0 2023-05-01 07:10:35,202 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.2459, 2.2894, 2.3606, 3.9853, 2.1832, 2.6314, 2.3516, 2.4840], device='cuda:1'), covar=tensor([0.1407, 0.3618, 0.2983, 0.0550, 0.4070, 0.2532, 0.3506, 0.3207], device='cuda:1'), in_proj_covar=tensor([0.0398, 0.0444, 0.0364, 0.0325, 0.0434, 0.0511, 0.0414, 0.0519], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-01 07:10:39,911 INFO [train.py:904] (1/8) Epoch 21, batch 6250, loss[loss=0.1895, simple_loss=0.2811, pruned_loss=0.04899, over 16179.00 frames. ], tot_loss[loss=0.2021, simple_loss=0.2885, pruned_loss=0.05787, over 3075076.34 frames. ], batch size: 165, lr: 3.21e-03, grad_scale: 8.0 2023-05-01 07:10:45,590 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=209255.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 07:10:57,584 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=209263.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 07:11:22,154 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=209280.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 07:11:32,672 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.71 vs. limit=2.0 2023-05-01 07:11:35,896 INFO [optim.py:368] (1/8) 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 07:11:36,499 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([5.1784, 4.9923, 5.1984, 5.3929, 5.5586, 4.8518, 5.5863, 5.5754], device='cuda:1'), covar=tensor([0.1959, 0.1224, 0.1755, 0.0792, 0.0618, 0.0855, 0.0622, 0.0546], device='cuda:1'), in_proj_covar=tensor([0.0622, 0.0766, 0.0894, 0.0780, 0.0585, 0.0616, 0.0635, 0.0736], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-01 07:11:49,869 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=209298.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 07:11:54,987 INFO [train.py:904] (1/8) Epoch 21, batch 6300, loss[loss=0.2061, simple_loss=0.2944, pruned_loss=0.05885, over 16679.00 frames. ], tot_loss[loss=0.2023, simple_loss=0.2887, pruned_loss=0.05793, over 3064915.36 frames. ], batch size: 124, lr: 3.21e-03, grad_scale: 8.0 2023-05-01 07:12:17,952 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 21, batch 6350, loss[loss=0.2504, simple_loss=0.3161, pruned_loss=0.09241, over 11634.00 frames. ], tot_loss[loss=0.2039, simple_loss=0.2898, pruned_loss=0.059, over 3064938.61 frames. ], batch size: 246, lr: 3.21e-03, grad_scale: 8.0 2023-05-01 07:13:24,899 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=209359.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 07:14:13,987 INFO [optim.py:368] (1/8) 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] (1/8) Epoch 21, batch 6400, loss[loss=0.1885, simple_loss=0.2767, pruned_loss=0.05015, over 16892.00 frames. ], tot_loss[loss=0.204, simple_loss=0.2892, pruned_loss=0.05942, over 3065614.73 frames. ], batch size: 90, lr: 3.21e-03, grad_scale: 8.0 2023-05-01 07:15:02,636 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([1.8387, 1.3663, 1.6803, 1.6652, 1.8008, 1.8983, 1.6786, 1.8065], device='cuda:1'), covar=tensor([0.0239, 0.0377, 0.0210, 0.0297, 0.0265, 0.0161, 0.0400, 0.0152], device='cuda:1'), in_proj_covar=tensor([0.0187, 0.0194, 0.0178, 0.0184, 0.0196, 0.0152, 0.0196, 0.0150], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-01 07:15:47,430 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.4117, 2.4081, 2.3324, 4.2052, 2.2640, 2.8356, 2.4630, 2.6060], device='cuda:1'), covar=tensor([0.1295, 0.3611, 0.2953, 0.0492, 0.4103, 0.2322, 0.3362, 0.3290], device='cuda:1'), in_proj_covar=tensor([0.0398, 0.0443, 0.0363, 0.0324, 0.0432, 0.0510, 0.0412, 0.0518], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-01 07:15:48,007 INFO [train.py:904] (1/8) Epoch 21, batch 6450, loss[loss=0.2024, simple_loss=0.2896, pruned_loss=0.05761, over 17265.00 frames. ], tot_loss[loss=0.203, simple_loss=0.2889, pruned_loss=0.0586, over 3089333.76 frames. ], batch size: 52, lr: 3.21e-03, grad_scale: 4.0 2023-05-01 07:16:10,788 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=209466.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 07:16:52,760 INFO [optim.py:368] (1/8) 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 07:17:06,155 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=3.88 vs. limit=5.0 2023-05-01 07:17:08,455 INFO [train.py:904] (1/8) Epoch 21, batch 6500, loss[loss=0.1907, simple_loss=0.2722, pruned_loss=0.05463, over 16384.00 frames. ], tot_loss[loss=0.2013, simple_loss=0.287, pruned_loss=0.05783, over 3108788.54 frames. ], batch size: 68, lr: 3.21e-03, grad_scale: 4.0 2023-05-01 07:17:20,758 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=209510.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 07:17:33,831 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.9170, 2.0712, 2.1029, 3.5144, 2.0190, 2.4266, 2.1885, 2.2040], device='cuda:1'), covar=tensor([0.1442, 0.3595, 0.2975, 0.0607, 0.4265, 0.2486, 0.3435, 0.3527], device='cuda:1'), in_proj_covar=tensor([0.0397, 0.0442, 0.0362, 0.0323, 0.0432, 0.0509, 0.0412, 0.0517], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-01 07:17:48,013 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=209527.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 07:18:29,349 INFO [train.py:904] (1/8) Epoch 21, batch 6550, loss[loss=0.2822, simple_loss=0.3447, pruned_loss=0.1098, over 11600.00 frames. ], tot_loss[loss=0.204, simple_loss=0.2897, pruned_loss=0.05914, over 3083542.64 frames. ], batch size: 248, lr: 3.21e-03, grad_scale: 4.0 2023-05-01 07:18:40,052 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=209558.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 07:19:00,586 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=209571.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 07:19:22,021 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.27 vs. limit=2.0 2023-05-01 07:19:33,742 INFO [optim.py:368] (1/8) 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 07:19:49,516 INFO [train.py:904] (1/8) Epoch 21, batch 6600, loss[loss=0.2103, simple_loss=0.3009, pruned_loss=0.05979, over 16922.00 frames. ], tot_loss[loss=0.2064, simple_loss=0.2926, pruned_loss=0.06013, over 3076337.24 frames. ], batch size: 109, lr: 3.21e-03, grad_scale: 4.0 2023-05-01 07:19:56,061 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=209606.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 07:20:03,684 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=209611.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 07:21:07,640 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.36 vs. limit=2.0 2023-05-01 07:21:07,995 INFO [train.py:904] (1/8) Epoch 21, batch 6650, loss[loss=0.1883, simple_loss=0.2772, pruned_loss=0.04974, over 16724.00 frames. ], tot_loss[loss=0.2073, simple_loss=0.2926, pruned_loss=0.06094, over 3067338.67 frames. ], batch size: 134, lr: 3.21e-03, grad_scale: 2.0 2023-05-01 07:21:12,193 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=209654.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 07:21:32,350 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=209667.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 07:22:04,962 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=209688.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 07:22:11,940 INFO [optim.py:368] (1/8) 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 07:22:25,654 INFO [train.py:904] (1/8) Epoch 21, batch 6700, loss[loss=0.2048, simple_loss=0.2893, pruned_loss=0.0601, over 15335.00 frames. ], tot_loss[loss=0.2071, simple_loss=0.2918, pruned_loss=0.06122, over 3053622.22 frames. ], batch size: 190, lr: 3.21e-03, grad_scale: 2.0 2023-05-01 07:22:31,780 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=4.93 vs. limit=5.0 2023-05-01 07:23:10,896 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([1.6128, 2.4425, 2.2967, 3.5698, 2.5758, 3.7882, 1.4324, 2.8284], device='cuda:1'), covar=tensor([0.1427, 0.0817, 0.1356, 0.0168, 0.0205, 0.0391, 0.1792, 0.0825], device='cuda:1'), in_proj_covar=tensor([0.0166, 0.0174, 0.0195, 0.0188, 0.0206, 0.0214, 0.0201, 0.0193], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-05-01 07:23:40,239 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=209749.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 07:23:44,112 INFO [train.py:904] (1/8) Epoch 21, batch 6750, loss[loss=0.2322, simple_loss=0.3095, pruned_loss=0.07748, over 11787.00 frames. ], tot_loss[loss=0.2063, simple_loss=0.2907, pruned_loss=0.06095, over 3061278.72 frames. ], batch size: 248, lr: 3.20e-03, grad_scale: 2.0 2023-05-01 07:24:47,247 INFO [optim.py:368] (1/8) 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 07:25:01,545 INFO [train.py:904] (1/8) Epoch 21, batch 6800, loss[loss=0.2051, simple_loss=0.2833, pruned_loss=0.06344, over 11148.00 frames. ], tot_loss[loss=0.2067, simple_loss=0.2913, pruned_loss=0.06101, over 3066998.01 frames. ], batch size: 247, lr: 3.20e-03, grad_scale: 4.0 2023-05-01 07:25:33,297 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=209822.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 07:26:19,716 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.1045, 4.2261, 2.5948, 4.7740, 3.1264, 4.6669, 2.7522, 3.2534], device='cuda:1'), covar=tensor([0.0231, 0.0308, 0.1608, 0.0209, 0.0774, 0.0564, 0.1393, 0.0761], device='cuda:1'), in_proj_covar=tensor([0.0168, 0.0176, 0.0194, 0.0161, 0.0176, 0.0215, 0.0201, 0.0178], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-05-01 07:26:20,407 INFO [train.py:904] (1/8) Epoch 21, batch 6850, loss[loss=0.1857, simple_loss=0.2831, pruned_loss=0.04416, over 16826.00 frames. ], tot_loss[loss=0.207, simple_loss=0.2921, pruned_loss=0.06092, over 3085930.40 frames. ], batch size: 42, lr: 3.20e-03, grad_scale: 4.0 2023-05-01 07:26:28,966 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=209858.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 07:26:29,607 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=4.82 vs. limit=5.0 2023-05-01 07:26:41,682 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=209866.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 07:27:01,874 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.0019, 1.9813, 1.6983, 1.7247, 2.2845, 1.9891, 1.9151, 2.3720], device='cuda:1'), covar=tensor([0.0244, 0.0440, 0.0557, 0.0505, 0.0266, 0.0357, 0.0209, 0.0293], device='cuda:1'), in_proj_covar=tensor([0.0203, 0.0231, 0.0223, 0.0224, 0.0232, 0.0230, 0.0231, 0.0227], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-01 07:27:21,614 INFO [optim.py:368] (1/8) Clipping_scale=2.0, grad-norm quartiles 1.924e+02 2.648e+02 3.249e+02 3.781e+02 8.526e+02, threshold=6.499e+02, percent-clipped=1.0 2023-05-01 07:27:34,652 INFO [train.py:904] (1/8) Epoch 21, batch 6900, loss[loss=0.2043, simple_loss=0.2948, pruned_loss=0.05691, over 16891.00 frames. ], tot_loss[loss=0.2082, simple_loss=0.2945, pruned_loss=0.06093, over 3076950.29 frames. ], batch size: 102, lr: 3.20e-03, grad_scale: 4.0 2023-05-01 07:27:41,303 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=209906.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 07:27:48,367 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=209911.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 07:27:50,382 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.5289, 3.4671, 2.6866, 2.1580, 2.3453, 2.3123, 3.6325, 3.1549], device='cuda:1'), covar=tensor([0.2759, 0.0663, 0.1791, 0.2773, 0.2418, 0.2064, 0.0509, 0.1284], device='cuda:1'), in_proj_covar=tensor([0.0327, 0.0271, 0.0303, 0.0311, 0.0296, 0.0257, 0.0294, 0.0334], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:1') 2023-05-01 07:27:50,562 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.83 vs. limit=2.0 2023-05-01 07:28:51,083 INFO [train.py:904] (1/8) Epoch 21, batch 6950, loss[loss=0.2211, simple_loss=0.3024, pruned_loss=0.0699, over 15229.00 frames. ], tot_loss[loss=0.2107, simple_loss=0.2965, pruned_loss=0.06249, over 3070097.42 frames. ], batch size: 190, lr: 3.20e-03, grad_scale: 4.0 2023-05-01 07:28:53,960 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=209954.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 07:29:02,106 INFO [zipformer.py:625] (1/8) 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] (1/8) 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] (1/8) 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 07:30:09,078 INFO [train.py:904] (1/8) Epoch 21, batch 7000, loss[loss=0.1838, simple_loss=0.2848, pruned_loss=0.04135, over 16854.00 frames. ], tot_loss[loss=0.209, simple_loss=0.2958, pruned_loss=0.06106, over 3097101.05 frames. ], batch size: 83, lr: 3.20e-03, grad_scale: 4.0 2023-05-01 07:30:09,335 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=210002.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 07:30:49,938 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.1227, 2.1860, 2.2416, 3.7880, 2.1637, 2.5922, 2.2809, 2.3415], device='cuda:1'), covar=tensor([0.1349, 0.3578, 0.2857, 0.0517, 0.3991, 0.2464, 0.3484, 0.3328], device='cuda:1'), in_proj_covar=tensor([0.0398, 0.0444, 0.0364, 0.0324, 0.0434, 0.0510, 0.0414, 0.0518], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-01 07:31:11,211 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 21, batch 7050, loss[loss=0.1919, simple_loss=0.2867, pruned_loss=0.04856, over 16263.00 frames. ], tot_loss[loss=0.2085, simple_loss=0.2961, pruned_loss=0.06045, over 3101253.99 frames. ], batch size: 165, lr: 3.20e-03, grad_scale: 4.0 2023-05-01 07:31:30,569 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=210057.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 07:32:24,157 INFO [optim.py:368] (1/8) 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 07:32:37,544 INFO [train.py:904] (1/8) Epoch 21, batch 7100, loss[loss=0.2359, simple_loss=0.2994, pruned_loss=0.0862, over 11322.00 frames. ], tot_loss[loss=0.2078, simple_loss=0.2949, pruned_loss=0.06038, over 3103332.23 frames. ], batch size: 248, lr: 3.20e-03, grad_scale: 4.0 2023-05-01 07:33:03,287 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=210118.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 07:33:08,772 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.83 vs. limit=2.0 2023-05-01 07:33:09,186 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=210122.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 07:33:11,669 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.23 vs. limit=2.0 2023-05-01 07:33:55,428 INFO [train.py:904] (1/8) Epoch 21, batch 7150, loss[loss=0.2015, simple_loss=0.2956, pruned_loss=0.0537, over 16738.00 frames. ], tot_loss[loss=0.2063, simple_loss=0.2926, pruned_loss=0.05999, over 3109856.82 frames. ], batch size: 83, lr: 3.20e-03, grad_scale: 4.0 2023-05-01 07:34:16,451 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=210166.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 07:34:21,259 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=210170.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 07:34:47,393 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.83 vs. limit=2.0 2023-05-01 07:34:53,945 INFO [optim.py:368] (1/8) 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] (1/8) Epoch 21, batch 7200, loss[loss=0.1894, simple_loss=0.281, pruned_loss=0.04892, over 16679.00 frames. ], tot_loss[loss=0.2045, simple_loss=0.2909, pruned_loss=0.05905, over 3084096.16 frames. ], batch size: 134, lr: 3.20e-03, grad_scale: 8.0 2023-05-01 07:35:13,841 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.3077, 3.4807, 3.6215, 3.5982, 3.5908, 3.4314, 3.4436, 3.4683], device='cuda:1'), covar=tensor([0.0412, 0.0642, 0.0437, 0.0411, 0.0501, 0.0506, 0.0833, 0.0553], device='cuda:1'), in_proj_covar=tensor([0.0400, 0.0444, 0.0433, 0.0402, 0.0480, 0.0455, 0.0540, 0.0364], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:1') 2023-05-01 07:35:26,243 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=210214.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 07:35:36,278 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([5.2423, 5.5488, 5.3077, 5.3197, 4.9883, 4.9703, 4.9600, 5.6584], device='cuda:1'), covar=tensor([0.1177, 0.0815, 0.1057, 0.0840, 0.0791, 0.0871, 0.1114, 0.0805], device='cuda:1'), in_proj_covar=tensor([0.0662, 0.0800, 0.0670, 0.0610, 0.0510, 0.0520, 0.0674, 0.0629], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:1') 2023-05-01 07:36:28,284 INFO [train.py:904] (1/8) Epoch 21, batch 7250, loss[loss=0.2001, simple_loss=0.2844, pruned_loss=0.05788, over 16729.00 frames. ], tot_loss[loss=0.2017, simple_loss=0.2882, pruned_loss=0.05762, over 3089753.22 frames. ], batch size: 124, lr: 3.20e-03, grad_scale: 8.0 2023-05-01 07:36:43,150 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.84 vs. limit=2.0 2023-05-01 07:36:43,759 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=210262.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 07:37:31,771 INFO [optim.py:368] (1/8) Clipping_scale=2.0, grad-norm quartiles 1.805e+02 2.676e+02 3.072e+02 3.848e+02 7.717e+02, threshold=6.145e+02, percent-clipped=2.0 2023-05-01 07:37:44,652 INFO [train.py:904] (1/8) Epoch 21, batch 7300, loss[loss=0.1961, simple_loss=0.2856, pruned_loss=0.05332, over 16988.00 frames. ], tot_loss[loss=0.2013, simple_loss=0.2877, pruned_loss=0.05751, over 3088487.53 frames. ], batch size: 55, lr: 3.20e-03, grad_scale: 8.0 2023-05-01 07:37:50,471 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.70 vs. limit=2.0 2023-05-01 07:37:58,791 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=210310.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 07:38:44,684 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([1.8842, 2.7940, 2.5398, 4.5449, 3.3690, 4.0416, 1.6729, 2.8556], device='cuda:1'), covar=tensor([0.1156, 0.0752, 0.1196, 0.0151, 0.0280, 0.0435, 0.1497, 0.0862], device='cuda:1'), in_proj_covar=tensor([0.0166, 0.0174, 0.0196, 0.0189, 0.0207, 0.0215, 0.0202, 0.0194], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-05-01 07:38:48,827 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 21, batch 7350, loss[loss=0.2078, simple_loss=0.2993, pruned_loss=0.05815, over 16827.00 frames. ], tot_loss[loss=0.2035, simple_loss=0.2894, pruned_loss=0.05875, over 3077057.30 frames. ], batch size: 102, lr: 3.20e-03, grad_scale: 8.0 2023-05-01 07:40:00,129 INFO [zipformer.py:625] (1/8) 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] (1/8) 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 07:40:11,063 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.7148, 4.5556, 4.7703, 4.9202, 5.0875, 4.5642, 5.0804, 5.0959], device='cuda:1'), covar=tensor([0.1792, 0.1228, 0.1521, 0.0728, 0.0626, 0.0981, 0.0725, 0.0581], device='cuda:1'), in_proj_covar=tensor([0.0618, 0.0762, 0.0882, 0.0777, 0.0584, 0.0612, 0.0632, 0.0727], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-01 07:40:14,854 INFO [train.py:904] (1/8) Epoch 21, batch 7400, loss[loss=0.1986, simple_loss=0.2853, pruned_loss=0.05591, over 16872.00 frames. ], tot_loss[loss=0.2041, simple_loss=0.2901, pruned_loss=0.05907, over 3079500.14 frames. ], batch size: 42, lr: 3.20e-03, grad_scale: 4.0 2023-05-01 07:40:26,993 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.4694, 4.5294, 4.3735, 4.0891, 4.0726, 4.4753, 4.2134, 4.1814], device='cuda:1'), covar=tensor([0.0656, 0.0649, 0.0300, 0.0297, 0.0872, 0.0523, 0.0600, 0.0666], device='cuda:1'), in_proj_covar=tensor([0.0280, 0.0406, 0.0327, 0.0322, 0.0337, 0.0376, 0.0226, 0.0392], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-01 07:40:32,233 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=210413.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 07:40:44,352 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.8501, 2.1082, 2.4335, 3.1116, 2.2014, 2.3434, 2.3505, 2.2598], device='cuda:1'), covar=tensor([0.1353, 0.3180, 0.2472, 0.0711, 0.4062, 0.2215, 0.2822, 0.3099], device='cuda:1'), in_proj_covar=tensor([0.0397, 0.0445, 0.0362, 0.0324, 0.0435, 0.0512, 0.0414, 0.0518], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-01 07:41:02,793 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=3.05 vs. limit=5.0 2023-05-01 07:41:32,224 INFO [train.py:904] (1/8) Epoch 21, batch 7450, loss[loss=0.2099, simple_loss=0.2994, pruned_loss=0.06026, over 16496.00 frames. ], tot_loss[loss=0.2062, simple_loss=0.2917, pruned_loss=0.06039, over 3077713.34 frames. ], batch size: 75, lr: 3.20e-03, grad_scale: 4.0 2023-05-01 07:41:57,712 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.3884, 2.8499, 3.0310, 1.9374, 2.6910, 2.0437, 3.0084, 3.0623], device='cuda:1'), covar=tensor([0.0340, 0.0813, 0.0630, 0.2126, 0.0887, 0.1089, 0.0762, 0.0864], device='cuda:1'), in_proj_covar=tensor([0.0154, 0.0163, 0.0167, 0.0152, 0.0144, 0.0129, 0.0143, 0.0174], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:1') 2023-05-01 07:42:42,641 INFO [optim.py:368] (1/8) 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 07:42:48,265 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.5203, 3.5079, 3.4876, 2.7620, 3.3794, 2.0622, 3.1801, 2.8940], device='cuda:1'), covar=tensor([0.0162, 0.0137, 0.0197, 0.0240, 0.0119, 0.2392, 0.0143, 0.0248], device='cuda:1'), in_proj_covar=tensor([0.0161, 0.0149, 0.0192, 0.0173, 0.0170, 0.0202, 0.0181, 0.0167], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-01 07:42:53,253 INFO [train.py:904] (1/8) Epoch 21, batch 7500, loss[loss=0.2133, simple_loss=0.2983, pruned_loss=0.06413, over 16377.00 frames. ], tot_loss[loss=0.2062, simple_loss=0.2921, pruned_loss=0.06013, over 3065211.86 frames. ], batch size: 146, lr: 3.20e-03, grad_scale: 4.0 2023-05-01 07:43:25,816 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.76 vs. limit=2.0 2023-05-01 07:43:47,081 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.2944, 3.6673, 3.6592, 2.4686, 3.4194, 3.7044, 3.3478, 2.0214], device='cuda:1'), covar=tensor([0.0510, 0.0061, 0.0060, 0.0405, 0.0102, 0.0115, 0.0105, 0.0492], device='cuda:1'), in_proj_covar=tensor([0.0134, 0.0082, 0.0083, 0.0134, 0.0096, 0.0107, 0.0093, 0.0127], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-05-01 07:44:09,156 INFO [train.py:904] (1/8) Epoch 21, batch 7550, loss[loss=0.1688, simple_loss=0.2566, pruned_loss=0.04047, over 16484.00 frames. ], tot_loss[loss=0.2059, simple_loss=0.2912, pruned_loss=0.0603, over 3064630.56 frames. ], batch size: 68, lr: 3.20e-03, grad_scale: 4.0 2023-05-01 07:45:11,347 INFO [optim.py:368] (1/8) Clipping_scale=2.0, grad-norm quartiles 1.786e+02 2.775e+02 3.392e+02 4.108e+02 6.854e+02, threshold=6.785e+02, percent-clipped=0.0 2023-05-01 07:45:23,226 INFO [train.py:904] (1/8) Epoch 21, batch 7600, loss[loss=0.2027, simple_loss=0.2833, pruned_loss=0.06107, over 16944.00 frames. ], tot_loss[loss=0.2056, simple_loss=0.2903, pruned_loss=0.06043, over 3060171.97 frames. ], batch size: 109, lr: 3.20e-03, grad_scale: 8.0 2023-05-01 07:45:47,407 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=2.96 vs. limit=5.0 2023-05-01 07:46:02,448 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.2161, 1.5771, 1.9386, 2.0676, 2.2586, 2.3716, 1.7511, 2.3119], device='cuda:1'), covar=tensor([0.0231, 0.0465, 0.0278, 0.0310, 0.0284, 0.0197, 0.0479, 0.0157], device='cuda:1'), in_proj_covar=tensor([0.0185, 0.0191, 0.0176, 0.0182, 0.0194, 0.0150, 0.0194, 0.0148], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-01 07:46:08,262 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=210632.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 07:46:37,230 INFO [train.py:904] (1/8) Epoch 21, batch 7650, loss[loss=0.2902, simple_loss=0.3406, pruned_loss=0.1199, over 11583.00 frames. ], tot_loss[loss=0.2066, simple_loss=0.291, pruned_loss=0.06106, over 3062564.83 frames. ], batch size: 248, lr: 3.20e-03, grad_scale: 8.0 2023-05-01 07:46:59,505 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=210666.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 07:47:25,381 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=210682.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 07:47:42,141 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=210693.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 07:47:42,756 INFO [optim.py:368] (1/8) 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 07:47:44,159 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([1.8410, 2.7839, 2.4962, 4.7282, 3.6854, 4.1894, 1.6737, 3.1641], device='cuda:1'), covar=tensor([0.1311, 0.0821, 0.1351, 0.0197, 0.0336, 0.0351, 0.1673, 0.0817], device='cuda:1'), in_proj_covar=tensor([0.0167, 0.0175, 0.0196, 0.0190, 0.0208, 0.0216, 0.0202, 0.0195], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-05-01 07:47:55,351 INFO [train.py:904] (1/8) Epoch 21, batch 7700, loss[loss=0.192, simple_loss=0.2799, pruned_loss=0.05211, over 15308.00 frames. ], tot_loss[loss=0.2068, simple_loss=0.291, pruned_loss=0.0613, over 3062167.60 frames. ], batch size: 190, lr: 3.20e-03, grad_scale: 8.0 2023-05-01 07:48:12,271 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=210713.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 07:48:34,120 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=210727.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 07:48:58,944 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=210743.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 07:49:12,087 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=3.04 vs. limit=5.0 2023-05-01 07:49:12,532 INFO [train.py:904] (1/8) Epoch 21, batch 7750, loss[loss=0.1844, simple_loss=0.2789, pruned_loss=0.04497, over 16695.00 frames. ], tot_loss[loss=0.2069, simple_loss=0.2916, pruned_loss=0.0611, over 3060129.96 frames. ], batch size: 134, lr: 3.20e-03, grad_scale: 8.0 2023-05-01 07:49:27,315 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=210761.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 07:49:32,423 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=210764.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 07:49:36,154 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([5.6060, 5.6528, 5.4721, 5.1052, 5.0994, 5.5372, 5.4720, 5.2228], device='cuda:1'), covar=tensor([0.0768, 0.0700, 0.0314, 0.0346, 0.1039, 0.0577, 0.0334, 0.0869], device='cuda:1'), in_proj_covar=tensor([0.0279, 0.0405, 0.0326, 0.0320, 0.0335, 0.0375, 0.0226, 0.0391], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-01 07:49:36,747 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.86 vs. limit=2.0 2023-05-01 07:50:18,620 INFO [optim.py:368] (1/8) Clipping_scale=2.0, grad-norm quartiles 1.950e+02 2.795e+02 3.327e+02 4.093e+02 7.547e+02, threshold=6.654e+02, percent-clipped=2.0 2023-05-01 07:50:31,005 INFO [train.py:904] (1/8) Epoch 21, batch 7800, loss[loss=0.2459, simple_loss=0.3121, pruned_loss=0.08989, over 11470.00 frames. ], tot_loss[loss=0.2078, simple_loss=0.2922, pruned_loss=0.06175, over 3061924.70 frames. ], batch size: 246, lr: 3.20e-03, grad_scale: 8.0 2023-05-01 07:51:07,548 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 21, batch 7850, loss[loss=0.1905, simple_loss=0.2797, pruned_loss=0.05065, over 16785.00 frames. ], tot_loss[loss=0.2067, simple_loss=0.2923, pruned_loss=0.06052, over 3079573.92 frames. ], batch size: 124, lr: 3.20e-03, grad_scale: 8.0 2023-05-01 07:52:54,098 INFO [optim.py:368] (1/8) Clipping_scale=2.0, grad-norm quartiles 2.040e+02 2.722e+02 3.190e+02 3.938e+02 6.813e+02, threshold=6.379e+02, percent-clipped=1.0 2023-05-01 07:53:05,678 INFO [train.py:904] (1/8) Epoch 21, batch 7900, loss[loss=0.196, simple_loss=0.2935, pruned_loss=0.04924, over 16721.00 frames. ], tot_loss[loss=0.2046, simple_loss=0.2907, pruned_loss=0.05923, over 3099590.54 frames. ], batch size: 89, lr: 3.20e-03, grad_scale: 8.0 2023-05-01 07:53:59,031 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.14 vs. limit=2.0 2023-05-01 07:54:03,461 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=4.34 vs. limit=5.0 2023-05-01 07:54:24,386 INFO [train.py:904] (1/8) Epoch 21, batch 7950, loss[loss=0.201, simple_loss=0.2848, pruned_loss=0.05859, over 16546.00 frames. ], tot_loss[loss=0.2053, simple_loss=0.2912, pruned_loss=0.05965, over 3108406.38 frames. ], batch size: 75, lr: 3.20e-03, grad_scale: 8.0 2023-05-01 07:54:33,143 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.9596, 4.0260, 4.2962, 4.2646, 4.2978, 4.0247, 4.0232, 4.0246], device='cuda:1'), covar=tensor([0.0337, 0.0660, 0.0421, 0.0477, 0.0475, 0.0458, 0.0904, 0.0516], device='cuda:1'), in_proj_covar=tensor([0.0399, 0.0443, 0.0431, 0.0401, 0.0478, 0.0453, 0.0539, 0.0361], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:1') 2023-05-01 07:55:14,441 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=210984.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 07:55:20,264 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=210988.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 07:55:29,110 INFO [optim.py:368] (1/8) 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] (1/8) Epoch 21, batch 8000, loss[loss=0.1994, simple_loss=0.287, pruned_loss=0.05592, over 16233.00 frames. ], tot_loss[loss=0.2072, simple_loss=0.2923, pruned_loss=0.06109, over 3087109.37 frames. ], batch size: 165, lr: 3.20e-03, grad_scale: 8.0 2023-05-01 07:56:12,595 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=211022.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 07:56:35,352 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=211038.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 07:56:45,496 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=211045.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 07:56:56,538 INFO [train.py:904] (1/8) Epoch 21, batch 8050, loss[loss=0.2119, simple_loss=0.3, pruned_loss=0.06184, over 16179.00 frames. ], tot_loss[loss=0.2067, simple_loss=0.292, pruned_loss=0.06069, over 3080816.01 frames. ], batch size: 165, lr: 3.19e-03, grad_scale: 8.0 2023-05-01 07:57:56,761 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.2272, 2.9122, 3.0959, 1.8083, 3.2674, 3.3238, 2.6932, 2.5847], device='cuda:1'), covar=tensor([0.0800, 0.0281, 0.0211, 0.1199, 0.0093, 0.0220, 0.0486, 0.0494], device='cuda:1'), in_proj_covar=tensor([0.0148, 0.0107, 0.0096, 0.0137, 0.0079, 0.0123, 0.0128, 0.0128], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0004, 0.0003], device='cuda:1') 2023-05-01 07:57:59,279 INFO [optim.py:368] (1/8) 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 07:58:08,125 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.6246, 2.5575, 1.9374, 2.6871, 2.1504, 2.7623, 2.1216, 2.3711], device='cuda:1'), covar=tensor([0.0336, 0.0420, 0.1231, 0.0240, 0.0655, 0.0578, 0.1154, 0.0542], device='cuda:1'), in_proj_covar=tensor([0.0167, 0.0174, 0.0192, 0.0158, 0.0174, 0.0214, 0.0199, 0.0177], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-05-01 07:58:10,500 INFO [train.py:904] (1/8) Epoch 21, batch 8100, loss[loss=0.2288, simple_loss=0.3062, pruned_loss=0.07574, over 11687.00 frames. ], tot_loss[loss=0.2062, simple_loss=0.2916, pruned_loss=0.06037, over 3062874.49 frames. ], batch size: 248, lr: 3.19e-03, grad_scale: 8.0 2023-05-01 07:58:38,222 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=211120.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 07:59:05,207 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([1.8620, 2.7599, 2.6696, 1.8685, 2.5728, 2.7124, 2.5767, 1.8859], device='cuda:1'), covar=tensor([0.0493, 0.0091, 0.0095, 0.0440, 0.0143, 0.0149, 0.0131, 0.0469], device='cuda:1'), in_proj_covar=tensor([0.0135, 0.0082, 0.0083, 0.0134, 0.0096, 0.0108, 0.0093, 0.0127], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-05-01 07:59:22,861 INFO [train.py:904] (1/8) Epoch 21, batch 8150, loss[loss=0.1805, simple_loss=0.2647, pruned_loss=0.04814, over 16854.00 frames. ], tot_loss[loss=0.2028, simple_loss=0.2883, pruned_loss=0.05866, over 3084362.34 frames. ], batch size: 102, lr: 3.19e-03, grad_scale: 8.0 2023-05-01 07:59:36,406 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.23 vs. limit=2.0 2023-05-01 08:00:19,529 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.1888, 4.0436, 4.2607, 4.4030, 4.5463, 4.1090, 4.4498, 4.5499], device='cuda:1'), covar=tensor([0.1858, 0.1218, 0.1467, 0.0768, 0.0603, 0.1340, 0.0873, 0.0694], device='cuda:1'), in_proj_covar=tensor([0.0616, 0.0759, 0.0879, 0.0772, 0.0581, 0.0611, 0.0631, 0.0728], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-01 08:00:27,469 INFO [optim.py:368] (1/8) 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 08:00:34,028 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.5243, 3.4887, 3.4581, 2.7690, 3.3490, 2.1034, 3.1651, 2.8269], device='cuda:1'), covar=tensor([0.0178, 0.0128, 0.0192, 0.0250, 0.0105, 0.2256, 0.0151, 0.0252], device='cuda:1'), in_proj_covar=tensor([0.0163, 0.0151, 0.0195, 0.0175, 0.0172, 0.0204, 0.0182, 0.0169], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-01 08:00:40,751 INFO [train.py:904] (1/8) Epoch 21, batch 8200, loss[loss=0.1812, simple_loss=0.2677, pruned_loss=0.04733, over 16821.00 frames. ], tot_loss[loss=0.2006, simple_loss=0.2853, pruned_loss=0.05792, over 3090545.42 frames. ], batch size: 102, lr: 3.19e-03, grad_scale: 8.0 2023-05-01 08:01:59,682 INFO [train.py:904] (1/8) Epoch 21, batch 8250, loss[loss=0.1677, simple_loss=0.2598, pruned_loss=0.03781, over 16613.00 frames. ], tot_loss[loss=0.1975, simple_loss=0.2844, pruned_loss=0.05524, over 3085423.00 frames. ], batch size: 62, lr: 3.19e-03, grad_scale: 8.0 2023-05-01 08:02:37,476 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.37 vs. limit=2.0 2023-05-01 08:02:56,782 INFO [zipformer.py:625] (1/8) 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] (1/8) 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 08:03:18,673 INFO [train.py:904] (1/8) Epoch 21, batch 8300, loss[loss=0.1746, simple_loss=0.2684, pruned_loss=0.04035, over 16584.00 frames. ], tot_loss[loss=0.1937, simple_loss=0.2822, pruned_loss=0.0526, over 3063577.31 frames. ], batch size: 75, lr: 3.19e-03, grad_scale: 8.0 2023-05-01 08:03:51,787 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=211322.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 08:04:12,350 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=211336.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 08:04:15,511 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=211338.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 08:04:18,454 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=211340.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 08:04:24,867 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([5.0141, 5.3302, 5.0842, 5.1211, 4.9114, 4.8641, 4.7007, 5.4073], device='cuda:1'), covar=tensor([0.1189, 0.0849, 0.0963, 0.0852, 0.0808, 0.0895, 0.1319, 0.0813], device='cuda:1'), in_proj_covar=tensor([0.0657, 0.0792, 0.0662, 0.0605, 0.0504, 0.0517, 0.0669, 0.0623], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:1') 2023-05-01 08:04:36,926 INFO [train.py:904] (1/8) Epoch 21, batch 8350, loss[loss=0.2227, simple_loss=0.2959, pruned_loss=0.07475, over 12183.00 frames. ], tot_loss[loss=0.1918, simple_loss=0.2816, pruned_loss=0.05105, over 3048892.94 frames. ], batch size: 247, lr: 3.19e-03, grad_scale: 8.0 2023-05-01 08:05:05,461 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=211370.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 08:05:30,340 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=211386.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 08:05:30,430 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=211386.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 08:05:43,485 INFO [optim.py:368] (1/8) 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] (1/8) Epoch 21, batch 8400, loss[loss=0.1653, simple_loss=0.26, pruned_loss=0.03533, over 15223.00 frames. ], tot_loss[loss=0.1885, simple_loss=0.2791, pruned_loss=0.04896, over 3040059.38 frames. ], batch size: 191, lr: 3.19e-03, grad_scale: 8.0 2023-05-01 08:06:10,920 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=211412.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 08:06:22,408 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=211420.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 08:07:03,758 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=211447.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 08:07:10,964 INFO [train.py:904] (1/8) Epoch 21, batch 8450, loss[loss=0.1877, simple_loss=0.28, pruned_loss=0.04771, over 16705.00 frames. ], tot_loss[loss=0.1857, simple_loss=0.2772, pruned_loss=0.04715, over 3048760.00 frames. ], batch size: 134, lr: 3.19e-03, grad_scale: 8.0 2023-05-01 08:07:34,369 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.4900, 3.5352, 2.2291, 3.9567, 2.7150, 3.9783, 2.4168, 3.0026], device='cuda:1'), covar=tensor([0.0279, 0.0340, 0.1382, 0.0239, 0.0767, 0.0414, 0.1298, 0.0588], device='cuda:1'), in_proj_covar=tensor([0.0165, 0.0171, 0.0189, 0.0156, 0.0172, 0.0210, 0.0198, 0.0174], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0003, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-05-01 08:07:36,102 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=211468.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 08:07:44,873 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=211473.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 08:08:10,667 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.5896, 2.3659, 2.2940, 4.3260, 2.2451, 2.7576, 2.4403, 2.5529], device='cuda:1'), covar=tensor([0.1061, 0.3737, 0.3104, 0.0422, 0.4327, 0.2524, 0.3607, 0.3472], device='cuda:1'), in_proj_covar=tensor([0.0390, 0.0437, 0.0359, 0.0317, 0.0429, 0.0500, 0.0407, 0.0510], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-01 08:08:16,382 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.2173, 2.1448, 2.1201, 3.8502, 2.1065, 2.5096, 2.2924, 2.3156], device='cuda:1'), covar=tensor([0.1218, 0.3979, 0.3245, 0.0522, 0.4476, 0.2680, 0.3786, 0.3456], device='cuda:1'), in_proj_covar=tensor([0.0390, 0.0438, 0.0359, 0.0317, 0.0429, 0.0500, 0.0407, 0.0510], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-01 08:08:18,591 INFO [optim.py:368] (1/8) 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 08:08:29,988 INFO [train.py:904] (1/8) Epoch 21, batch 8500, loss[loss=0.1728, simple_loss=0.2656, pruned_loss=0.03999, over 16257.00 frames. ], tot_loss[loss=0.1815, simple_loss=0.2733, pruned_loss=0.04481, over 3056306.22 frames. ], batch size: 35, lr: 3.19e-03, grad_scale: 8.0 2023-05-01 08:09:04,444 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.84 vs. limit=2.0 2023-05-01 08:09:42,450 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.7244, 3.7497, 2.3316, 4.2485, 2.8557, 4.2350, 2.6738, 3.2508], device='cuda:1'), covar=tensor([0.0267, 0.0315, 0.1480, 0.0222, 0.0813, 0.0404, 0.1248, 0.0558], device='cuda:1'), in_proj_covar=tensor([0.0165, 0.0171, 0.0189, 0.0155, 0.0172, 0.0210, 0.0197, 0.0174], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0003, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-05-01 08:09:54,385 INFO [train.py:904] (1/8) Epoch 21, batch 8550, loss[loss=0.1688, simple_loss=0.2512, pruned_loss=0.04318, over 11748.00 frames. ], tot_loss[loss=0.1794, simple_loss=0.2707, pruned_loss=0.04401, over 3014519.34 frames. ], batch size: 247, lr: 3.19e-03, grad_scale: 8.0 2023-05-01 08:09:58,571 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.82 vs. limit=2.0 2023-05-01 08:11:18,416 INFO [optim.py:368] (1/8) 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 08:11:32,298 INFO [train.py:904] (1/8) Epoch 21, batch 8600, loss[loss=0.1785, simple_loss=0.2766, pruned_loss=0.04025, over 15273.00 frames. ], tot_loss[loss=0.1792, simple_loss=0.2716, pruned_loss=0.0434, over 3026566.19 frames. ], batch size: 190, lr: 3.19e-03, grad_scale: 8.0 2023-05-01 08:12:09,003 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.2496, 3.6149, 3.9139, 2.1970, 3.2745, 2.6043, 3.6888, 3.7577], device='cuda:1'), covar=tensor([0.0243, 0.0690, 0.0459, 0.1903, 0.0696, 0.0861, 0.0573, 0.0861], device='cuda:1'), in_proj_covar=tensor([0.0151, 0.0159, 0.0164, 0.0150, 0.0142, 0.0127, 0.0140, 0.0169], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:1') 2023-05-01 08:12:11,582 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.3730, 3.4198, 3.6362, 3.6324, 3.6393, 3.4538, 3.4746, 3.5302], device='cuda:1'), covar=tensor([0.0410, 0.0725, 0.0513, 0.0478, 0.0499, 0.0519, 0.0808, 0.0504], device='cuda:1'), in_proj_covar=tensor([0.0394, 0.0439, 0.0425, 0.0394, 0.0472, 0.0444, 0.0530, 0.0356], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:1') 2023-05-01 08:12:48,716 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=211640.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 08:12:51,570 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.57 vs. limit=2.0 2023-05-01 08:13:09,686 INFO [train.py:904] (1/8) Epoch 21, batch 8650, loss[loss=0.1669, simple_loss=0.2586, pruned_loss=0.0376, over 12307.00 frames. ], tot_loss[loss=0.176, simple_loss=0.2691, pruned_loss=0.04147, over 3017280.09 frames. ], batch size: 250, lr: 3.19e-03, grad_scale: 8.0 2023-05-01 08:14:18,154 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.40 vs. limit=2.0 2023-05-01 08:14:30,915 INFO [zipformer.py:625] (1/8) 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] (1/8) Clipping_scale=2.0, grad-norm quartiles 1.495e+02 2.069e+02 2.535e+02 3.155e+02 6.979e+02, threshold=5.069e+02, percent-clipped=2.0 2023-05-01 08:14:57,319 INFO [train.py:904] (1/8) Epoch 21, batch 8700, loss[loss=0.1801, simple_loss=0.2733, pruned_loss=0.04341, over 15350.00 frames. ], tot_loss[loss=0.1739, simple_loss=0.2669, pruned_loss=0.04049, over 3034044.21 frames. ], batch size: 190, lr: 3.19e-03, grad_scale: 8.0 2023-05-01 08:16:09,883 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 21, batch 8750, loss[loss=0.1795, simple_loss=0.2757, pruned_loss=0.04162, over 16177.00 frames. ], tot_loss[loss=0.1738, simple_loss=0.2671, pruned_loss=0.04024, over 3035207.99 frames. ], batch size: 165, lr: 3.19e-03, grad_scale: 8.0 2023-05-01 08:17:06,238 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.37 vs. limit=2.0 2023-05-01 08:17:13,748 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=211768.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 08:18:09,821 INFO [optim.py:368] (1/8) 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] (1/8) Epoch 21, batch 8800, loss[loss=0.1802, simple_loss=0.2726, pruned_loss=0.04383, over 16298.00 frames. ], tot_loss[loss=0.1721, simple_loss=0.2656, pruned_loss=0.03926, over 3047771.97 frames. ], batch size: 165, lr: 3.19e-03, grad_scale: 8.0 2023-05-01 08:19:10,017 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([1.8781, 2.2439, 2.2802, 3.0602, 1.7291, 3.2002, 1.6888, 2.6314], device='cuda:1'), covar=tensor([0.1440, 0.0783, 0.1263, 0.0187, 0.0085, 0.0396, 0.1692, 0.0843], device='cuda:1'), in_proj_covar=tensor([0.0164, 0.0170, 0.0190, 0.0183, 0.0200, 0.0210, 0.0198, 0.0190], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-05-01 08:20:07,833 INFO [train.py:904] (1/8) Epoch 21, batch 8850, loss[loss=0.1693, simple_loss=0.2736, pruned_loss=0.03254, over 15516.00 frames. ], tot_loss[loss=0.1722, simple_loss=0.2677, pruned_loss=0.03837, over 3046327.25 frames. ], batch size: 191, lr: 3.19e-03, grad_scale: 8.0 2023-05-01 08:21:38,908 INFO [optim.py:368] (1/8) 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 08:21:54,320 INFO [train.py:904] (1/8) Epoch 21, batch 8900, loss[loss=0.1744, simple_loss=0.2564, pruned_loss=0.04623, over 12341.00 frames. ], tot_loss[loss=0.1719, simple_loss=0.2679, pruned_loss=0.03795, over 3041903.03 frames. ], batch size: 249, lr: 3.19e-03, grad_scale: 8.0 2023-05-01 08:23:57,812 INFO [train.py:904] (1/8) Epoch 21, batch 8950, loss[loss=0.1675, simple_loss=0.2565, pruned_loss=0.03923, over 16672.00 frames. ], tot_loss[loss=0.1723, simple_loss=0.2677, pruned_loss=0.03845, over 3064490.60 frames. ], batch size: 134, lr: 3.19e-03, grad_scale: 8.0 2023-05-01 08:24:11,370 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.1373, 2.5399, 2.6439, 1.8476, 2.8126, 2.8880, 2.5525, 2.4955], device='cuda:1'), covar=tensor([0.0635, 0.0237, 0.0202, 0.0993, 0.0108, 0.0236, 0.0428, 0.0409], device='cuda:1'), in_proj_covar=tensor([0.0143, 0.0104, 0.0092, 0.0134, 0.0076, 0.0118, 0.0124, 0.0124], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:1') 2023-05-01 08:25:05,256 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.30 vs. limit=2.0 2023-05-01 08:25:26,558 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.3408, 3.5694, 3.7524, 2.0845, 3.2323, 2.4872, 3.7684, 3.7638], device='cuda:1'), covar=tensor([0.0218, 0.0760, 0.0503, 0.2023, 0.0720, 0.0906, 0.0534, 0.0928], device='cuda:1'), in_proj_covar=tensor([0.0150, 0.0157, 0.0162, 0.0149, 0.0141, 0.0126, 0.0139, 0.0167], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:1') 2023-05-01 08:25:29,293 INFO [optim.py:368] (1/8) 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] (1/8) Epoch 21, batch 9000, loss[loss=0.1653, simple_loss=0.2566, pruned_loss=0.037, over 15405.00 frames. ], tot_loss[loss=0.1691, simple_loss=0.2644, pruned_loss=0.03696, over 3072349.80 frames. ], batch size: 190, lr: 3.19e-03, grad_scale: 8.0 2023-05-01 08:25:46,900 INFO [train.py:929] (1/8) Computing validation loss 2023-05-01 08:25:57,432 INFO [train.py:938] (1/8) Epoch 21, validation: loss=0.1457, simple_loss=0.2498, pruned_loss=0.02077, over 944034.00 frames. 2023-05-01 08:25:57,432 INFO [train.py:939] (1/8) Maximum memory allocated so far is 18145MB 2023-05-01 08:27:19,755 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=212042.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 08:27:39,216 INFO [train.py:904] (1/8) Epoch 21, batch 9050, loss[loss=0.1623, simple_loss=0.2545, pruned_loss=0.03504, over 16766.00 frames. ], tot_loss[loss=0.1705, simple_loss=0.2657, pruned_loss=0.03762, over 3079329.19 frames. ], batch size: 90, lr: 3.19e-03, grad_scale: 8.0 2023-05-01 08:27:47,072 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.8947, 5.2030, 4.8916, 4.6229, 4.2038, 5.1207, 4.9568, 4.7052], device='cuda:1'), covar=tensor([0.0999, 0.0707, 0.0477, 0.0411, 0.1889, 0.0543, 0.0354, 0.0800], device='cuda:1'), in_proj_covar=tensor([0.0277, 0.0402, 0.0324, 0.0320, 0.0330, 0.0372, 0.0225, 0.0388], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-01 08:28:14,811 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=212068.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 08:28:24,091 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.6086, 3.5771, 3.5427, 2.8720, 3.4929, 2.0404, 3.2682, 2.9839], device='cuda:1'), covar=tensor([0.0180, 0.0142, 0.0196, 0.0216, 0.0134, 0.2414, 0.0151, 0.0266], device='cuda:1'), in_proj_covar=tensor([0.0158, 0.0147, 0.0188, 0.0168, 0.0167, 0.0200, 0.0177, 0.0162], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-01 08:28:58,636 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=212090.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 08:29:10,425 INFO [optim.py:368] (1/8) 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 08:29:26,248 INFO [train.py:904] (1/8) Epoch 21, batch 9100, loss[loss=0.169, simple_loss=0.2675, pruned_loss=0.03524, over 16882.00 frames. ], tot_loss[loss=0.1702, simple_loss=0.2649, pruned_loss=0.03777, over 3094292.40 frames. ], batch size: 90, lr: 3.19e-03, grad_scale: 8.0 2023-05-01 08:29:30,888 INFO [zipformer.py:625] (1/8) 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] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=212116.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 08:30:50,922 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=2.95 vs. limit=5.0 2023-05-01 08:31:15,077 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.0450, 4.7180, 4.7774, 3.3946, 4.0149, 4.6929, 4.1968, 2.7138], device='cuda:1'), covar=tensor([0.0454, 0.0041, 0.0033, 0.0316, 0.0099, 0.0074, 0.0066, 0.0435], device='cuda:1'), in_proj_covar=tensor([0.0133, 0.0080, 0.0081, 0.0132, 0.0095, 0.0105, 0.0092, 0.0126], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-05-01 08:31:22,746 INFO [train.py:904] (1/8) Epoch 21, batch 9150, loss[loss=0.176, simple_loss=0.2797, pruned_loss=0.03612, over 17127.00 frames. ], tot_loss[loss=0.1695, simple_loss=0.2648, pruned_loss=0.0371, over 3083783.44 frames. ], batch size: 49, lr: 3.19e-03, grad_scale: 8.0 2023-05-01 08:31:52,290 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=212165.0, num_to_drop=1, layers_to_drop={1} 2023-05-01 08:31:57,186 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([5.0738, 5.3450, 5.1165, 5.1392, 4.8839, 4.8725, 4.6871, 5.4155], device='cuda:1'), covar=tensor([0.1073, 0.0866, 0.0892, 0.0804, 0.0737, 0.0915, 0.1171, 0.0840], device='cuda:1'), in_proj_covar=tensor([0.0643, 0.0781, 0.0649, 0.0594, 0.0497, 0.0507, 0.0655, 0.0612], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:1') 2023-05-01 08:31:57,360 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([1.7883, 2.9685, 2.4478, 4.3910, 2.8400, 4.1144, 1.5409, 3.1235], device='cuda:1'), covar=tensor([0.1314, 0.0681, 0.1218, 0.0139, 0.0115, 0.0328, 0.1677, 0.0637], device='cuda:1'), in_proj_covar=tensor([0.0166, 0.0171, 0.0192, 0.0184, 0.0200, 0.0211, 0.0200, 0.0191], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-05-01 08:32:13,809 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.7302, 4.7164, 4.4865, 3.9588, 4.6282, 1.8341, 4.3934, 4.3029], device='cuda:1'), covar=tensor([0.0160, 0.0139, 0.0224, 0.0306, 0.0163, 0.2600, 0.0127, 0.0240], device='cuda:1'), in_proj_covar=tensor([0.0158, 0.0147, 0.0189, 0.0168, 0.0167, 0.0200, 0.0177, 0.0163], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-01 08:32:49,426 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=3.24 vs. limit=5.0 2023-05-01 08:32:57,277 INFO [optim.py:368] (1/8) 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 08:33:02,161 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.4408, 3.3995, 3.5125, 3.5950, 3.6318, 3.3392, 3.6210, 3.6828], device='cuda:1'), covar=tensor([0.1284, 0.0946, 0.1059, 0.0613, 0.0589, 0.2126, 0.0798, 0.0800], device='cuda:1'), in_proj_covar=tensor([0.0594, 0.0734, 0.0853, 0.0748, 0.0565, 0.0594, 0.0610, 0.0707], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-01 08:33:09,950 INFO [train.py:904] (1/8) Epoch 21, batch 9200, loss[loss=0.153, simple_loss=0.2374, pruned_loss=0.03432, over 11666.00 frames. ], tot_loss[loss=0.1668, simple_loss=0.261, pruned_loss=0.03632, over 3088803.49 frames. ], batch size: 249, lr: 3.19e-03, grad_scale: 8.0 2023-05-01 08:34:13,805 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.6162, 3.9163, 2.9000, 2.2196, 2.4206, 2.4822, 4.1980, 3.2316], device='cuda:1'), covar=tensor([0.2847, 0.0584, 0.1755, 0.2873, 0.3070, 0.2070, 0.0334, 0.1395], device='cuda:1'), in_proj_covar=tensor([0.0319, 0.0262, 0.0298, 0.0304, 0.0285, 0.0252, 0.0285, 0.0325], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:1') 2023-05-01 08:34:48,820 INFO [train.py:904] (1/8) Epoch 21, batch 9250, loss[loss=0.1581, simple_loss=0.2575, pruned_loss=0.02931, over 16955.00 frames. ], tot_loss[loss=0.1666, simple_loss=0.2608, pruned_loss=0.03615, over 3100693.35 frames. ], batch size: 96, lr: 3.19e-03, grad_scale: 8.0 2023-05-01 08:34:51,546 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=212253.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 08:36:25,634 INFO [optim.py:368] (1/8) 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 08:36:39,744 INFO [train.py:904] (1/8) Epoch 21, batch 9300, loss[loss=0.1369, simple_loss=0.2292, pruned_loss=0.02226, over 16592.00 frames. ], tot_loss[loss=0.1656, simple_loss=0.2596, pruned_loss=0.03576, over 3098535.97 frames. ], batch size: 62, lr: 3.19e-03, grad_scale: 8.0 2023-05-01 08:37:07,730 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=212314.0, num_to_drop=1, layers_to_drop={2} 2023-05-01 08:38:15,110 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.11 vs. limit=2.0 2023-05-01 08:38:22,833 INFO [train.py:904] (1/8) Epoch 21, batch 9350, loss[loss=0.1613, simple_loss=0.2565, pruned_loss=0.03309, over 12065.00 frames. ], tot_loss[loss=0.1658, simple_loss=0.2598, pruned_loss=0.03593, over 3095813.67 frames. ], batch size: 247, lr: 3.19e-03, grad_scale: 8.0 2023-05-01 08:39:47,852 INFO [optim.py:368] (1/8) 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 08:40:02,993 INFO [train.py:904] (1/8) Epoch 21, batch 9400, loss[loss=0.142, simple_loss=0.2302, pruned_loss=0.02688, over 12245.00 frames. ], tot_loss[loss=0.166, simple_loss=0.26, pruned_loss=0.03596, over 3087482.59 frames. ], batch size: 248, lr: 3.18e-03, grad_scale: 8.0 2023-05-01 08:41:08,328 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([1.9885, 2.1910, 2.3759, 2.9008, 1.8872, 3.1816, 1.8174, 2.7999], device='cuda:1'), covar=tensor([0.1260, 0.0717, 0.1055, 0.0165, 0.0081, 0.0333, 0.1444, 0.0682], device='cuda:1'), in_proj_covar=tensor([0.0165, 0.0170, 0.0191, 0.0182, 0.0197, 0.0210, 0.0199, 0.0190], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-05-01 08:41:18,580 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=212439.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 08:41:42,918 INFO [train.py:904] (1/8) Epoch 21, batch 9450, loss[loss=0.1738, simple_loss=0.268, pruned_loss=0.0398, over 16984.00 frames. ], tot_loss[loss=0.1671, simple_loss=0.2619, pruned_loss=0.03615, over 3080518.85 frames. ], batch size: 109, lr: 3.18e-03, grad_scale: 8.0 2023-05-01 08:41:58,670 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=212460.0, num_to_drop=1, layers_to_drop={3} 2023-05-01 08:42:06,077 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.72 vs. limit=2.0 2023-05-01 08:43:11,247 INFO [optim.py:368] (1/8) Clipping_scale=2.0, grad-norm quartiles 1.570e+02 1.974e+02 2.463e+02 3.006e+02 5.675e+02, threshold=4.927e+02, percent-clipped=1.0 2023-05-01 08:43:21,163 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 21, batch 9500, loss[loss=0.1585, simple_loss=0.2376, pruned_loss=0.03967, over 12644.00 frames. ], tot_loss[loss=0.166, simple_loss=0.2606, pruned_loss=0.03573, over 3072528.94 frames. ], batch size: 246, lr: 3.18e-03, grad_scale: 8.0 2023-05-01 08:45:02,016 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([1.9661, 1.8315, 1.6708, 1.5549, 2.0057, 1.7297, 1.5484, 1.9882], device='cuda:1'), covar=tensor([0.0175, 0.0343, 0.0445, 0.0399, 0.0227, 0.0300, 0.0178, 0.0219], device='cuda:1'), in_proj_covar=tensor([0.0195, 0.0225, 0.0218, 0.0217, 0.0225, 0.0224, 0.0221, 0.0218], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-01 08:45:12,071 INFO [train.py:904] (1/8) Epoch 21, batch 9550, loss[loss=0.168, simple_loss=0.2566, pruned_loss=0.03973, over 12487.00 frames. ], tot_loss[loss=0.1661, simple_loss=0.2604, pruned_loss=0.03588, over 3067356.01 frames. ], batch size: 248, lr: 3.18e-03, grad_scale: 8.0 2023-05-01 08:45:19,558 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.6171, 3.5768, 3.5480, 2.8558, 3.5015, 1.9611, 3.2900, 2.9160], device='cuda:1'), covar=tensor([0.0126, 0.0120, 0.0175, 0.0216, 0.0105, 0.2549, 0.0129, 0.0301], device='cuda:1'), in_proj_covar=tensor([0.0156, 0.0145, 0.0185, 0.0164, 0.0164, 0.0197, 0.0174, 0.0160], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-01 08:46:39,906 INFO [optim.py:368] (1/8) 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] (1/8) Epoch 21, batch 9600, loss[loss=0.1909, simple_loss=0.2908, pruned_loss=0.04555, over 15231.00 frames. ], tot_loss[loss=0.1674, simple_loss=0.2618, pruned_loss=0.03653, over 3063467.44 frames. ], batch size: 190, lr: 3.18e-03, grad_scale: 8.0 2023-05-01 08:47:06,245 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=212609.0, num_to_drop=1, layers_to_drop={2} 2023-05-01 08:48:38,602 INFO [train.py:904] (1/8) Epoch 21, batch 9650, loss[loss=0.1698, simple_loss=0.2595, pruned_loss=0.04005, over 12347.00 frames. ], tot_loss[loss=0.1691, simple_loss=0.2641, pruned_loss=0.03698, over 3071141.60 frames. ], batch size: 248, lr: 3.18e-03, grad_scale: 8.0 2023-05-01 08:48:54,728 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.9958, 2.7808, 2.9446, 2.0696, 2.7192, 2.0924, 2.7466, 2.8783], device='cuda:1'), covar=tensor([0.0279, 0.0814, 0.0442, 0.1760, 0.0768, 0.0932, 0.0635, 0.0818], device='cuda:1'), in_proj_covar=tensor([0.0149, 0.0154, 0.0161, 0.0147, 0.0139, 0.0125, 0.0138, 0.0165], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:1') 2023-05-01 08:49:24,753 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.4462, 3.6417, 4.0441, 2.2234, 3.3381, 2.5840, 3.8814, 3.6866], device='cuda:1'), covar=tensor([0.0206, 0.0781, 0.0425, 0.2033, 0.0693, 0.0887, 0.0590, 0.0988], device='cuda:1'), in_proj_covar=tensor([0.0149, 0.0154, 0.0161, 0.0147, 0.0139, 0.0125, 0.0138, 0.0165], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:1') 2023-05-01 08:49:47,546 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.2949, 4.1539, 4.3597, 4.4593, 4.6109, 4.1456, 4.6351, 4.6258], device='cuda:1'), covar=tensor([0.1817, 0.1194, 0.1502, 0.0747, 0.0530, 0.1180, 0.0550, 0.0684], device='cuda:1'), in_proj_covar=tensor([0.0593, 0.0728, 0.0849, 0.0744, 0.0564, 0.0591, 0.0609, 0.0704], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-01 08:50:11,919 INFO [optim.py:368] (1/8) 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 08:50:20,711 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 21, batch 9700, loss[loss=0.171, simple_loss=0.2663, pruned_loss=0.0378, over 15379.00 frames. ], tot_loss[loss=0.1682, simple_loss=0.2629, pruned_loss=0.03671, over 3070048.81 frames. ], batch size: 190, lr: 3.18e-03, grad_scale: 8.0 2023-05-01 08:50:30,771 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.4261, 1.7165, 2.1081, 2.4298, 2.4586, 2.7095, 1.8750, 2.5901], device='cuda:1'), covar=tensor([0.0267, 0.0579, 0.0354, 0.0338, 0.0336, 0.0218, 0.0549, 0.0150], device='cuda:1'), in_proj_covar=tensor([0.0181, 0.0188, 0.0172, 0.0176, 0.0190, 0.0147, 0.0190, 0.0143], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-01 08:50:46,012 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=212711.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 08:52:10,271 INFO [train.py:904] (1/8) Epoch 21, batch 9750, loss[loss=0.1742, simple_loss=0.2712, pruned_loss=0.03854, over 16638.00 frames. ], tot_loss[loss=0.1676, simple_loss=0.2617, pruned_loss=0.03678, over 3066064.96 frames. ], batch size: 134, lr: 3.18e-03, grad_scale: 8.0 2023-05-01 08:52:24,885 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=212759.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 08:52:26,937 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=212760.0, num_to_drop=1, layers_to_drop={2} 2023-05-01 08:52:29,819 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.62 vs. limit=2.0 2023-05-01 08:52:49,220 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=212772.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 08:52:53,149 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=212774.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 08:53:36,934 INFO [optim.py:368] (1/8) 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 08:53:37,560 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=212795.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 08:53:46,240 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.7956, 3.6081, 4.0800, 1.8612, 4.2362, 4.3458, 3.2527, 3.2621], device='cuda:1'), covar=tensor([0.0702, 0.0254, 0.0159, 0.1317, 0.0055, 0.0103, 0.0361, 0.0411], device='cuda:1'), in_proj_covar=tensor([0.0142, 0.0103, 0.0091, 0.0134, 0.0076, 0.0117, 0.0123, 0.0124], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:1') 2023-05-01 08:53:48,983 INFO [train.py:904] (1/8) Epoch 21, batch 9800, loss[loss=0.1857, simple_loss=0.2901, pruned_loss=0.04069, over 16251.00 frames. ], tot_loss[loss=0.1669, simple_loss=0.262, pruned_loss=0.03586, over 3076286.11 frames. ], batch size: 165, lr: 3.18e-03, grad_scale: 8.0 2023-05-01 08:54:01,430 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=212808.0, num_to_drop=1, layers_to_drop={1} 2023-05-01 08:54:05,569 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.93 vs. limit=2.0 2023-05-01 08:54:10,345 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.7762, 3.7942, 4.0024, 3.7203, 3.9503, 4.3256, 3.9471, 3.6559], device='cuda:1'), covar=tensor([0.2206, 0.2501, 0.2252, 0.2314, 0.2528, 0.1614, 0.1619, 0.2517], device='cuda:1'), in_proj_covar=tensor([0.0380, 0.0556, 0.0614, 0.0463, 0.0613, 0.0645, 0.0485, 0.0620], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-05-01 08:54:35,152 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.43 vs. limit=2.0 2023-05-01 08:54:51,517 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 21, batch 9850, loss[loss=0.1678, simple_loss=0.263, pruned_loss=0.03625, over 15355.00 frames. ], tot_loss[loss=0.1672, simple_loss=0.2629, pruned_loss=0.03575, over 3059562.62 frames. ], batch size: 191, lr: 3.18e-03, grad_scale: 8.0 2023-05-01 08:56:28,028 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.7070, 4.0156, 2.9590, 2.2366, 2.5298, 2.5982, 4.2946, 3.3920], device='cuda:1'), covar=tensor([0.2879, 0.0527, 0.1822, 0.3062, 0.2785, 0.2067, 0.0344, 0.1302], device='cuda:1'), in_proj_covar=tensor([0.0316, 0.0259, 0.0296, 0.0301, 0.0282, 0.0249, 0.0282, 0.0320], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:1') 2023-05-01 08:57:07,910 INFO [optim.py:368] (1/8) 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 08:57:09,116 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.9448, 3.8726, 4.2307, 1.9125, 4.4238, 4.5964, 3.3330, 3.4460], device='cuda:1'), covar=tensor([0.0741, 0.0261, 0.0220, 0.1403, 0.0085, 0.0123, 0.0401, 0.0450], device='cuda:1'), in_proj_covar=tensor([0.0141, 0.0103, 0.0090, 0.0133, 0.0075, 0.0117, 0.0123, 0.0123], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:1') 2023-05-01 08:57:21,370 INFO [train.py:904] (1/8) Epoch 21, batch 9900, loss[loss=0.1743, simple_loss=0.2592, pruned_loss=0.0447, over 12330.00 frames. ], tot_loss[loss=0.1675, simple_loss=0.2633, pruned_loss=0.0358, over 3071978.86 frames. ], batch size: 248, lr: 3.18e-03, grad_scale: 8.0 2023-05-01 08:57:38,693 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=212909.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 08:58:16,452 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.28 vs. limit=2.0 2023-05-01 08:59:17,458 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=4.79 vs. limit=5.0 2023-05-01 08:59:17,572 INFO [train.py:904] (1/8) Epoch 21, batch 9950, loss[loss=0.1754, simple_loss=0.2778, pruned_loss=0.03649, over 16557.00 frames. ], tot_loss[loss=0.1687, simple_loss=0.2655, pruned_loss=0.03596, over 3085877.23 frames. ], batch size: 68, lr: 3.18e-03, grad_scale: 8.0 2023-05-01 08:59:29,580 INFO [zipformer.py:625] (1/8) 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] (1/8) 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] (1/8) Epoch 21, batch 10000, loss[loss=0.1601, simple_loss=0.2595, pruned_loss=0.03032, over 15544.00 frames. ], tot_loss[loss=0.1677, simple_loss=0.264, pruned_loss=0.0357, over 3091037.97 frames. ], batch size: 192, lr: 3.18e-03, grad_scale: 8.0 2023-05-01 09:01:51,277 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.4827, 3.8121, 2.7606, 2.1177, 2.3580, 2.4388, 4.0476, 3.1846], device='cuda:1'), covar=tensor([0.3064, 0.0482, 0.1834, 0.2956, 0.2787, 0.2063, 0.0325, 0.1282], device='cuda:1'), in_proj_covar=tensor([0.0315, 0.0258, 0.0295, 0.0300, 0.0280, 0.0249, 0.0282, 0.0319], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:1') 2023-05-01 09:02:58,011 INFO [train.py:904] (1/8) Epoch 21, batch 10050, loss[loss=0.1493, simple_loss=0.2502, pruned_loss=0.02416, over 16496.00 frames. ], tot_loss[loss=0.1674, simple_loss=0.2636, pruned_loss=0.03557, over 3089507.67 frames. ], batch size: 68, lr: 3.18e-03, grad_scale: 8.0 2023-05-01 09:03:02,065 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=213054.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 09:03:28,731 INFO [zipformer.py:625] (1/8) 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] (1/8) 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 09:04:21,589 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 21, batch 10100, loss[loss=0.165, simple_loss=0.2547, pruned_loss=0.0377, over 16235.00 frames. ], tot_loss[loss=0.1672, simple_loss=0.2635, pruned_loss=0.0355, over 3095339.03 frames. ], batch size: 165, lr: 3.18e-03, grad_scale: 8.0 2023-05-01 09:05:27,294 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=213130.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 09:05:42,671 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=213143.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 09:05:51,218 INFO [train.py:904] (1/8) Epoch 21, batch 10150, loss[loss=0.1516, simple_loss=0.2376, pruned_loss=0.03283, over 12351.00 frames. ], tot_loss[loss=0.1664, simple_loss=0.2618, pruned_loss=0.03552, over 3060361.97 frames. ], batch size: 247, lr: 3.18e-03, grad_scale: 8.0 2023-05-01 09:06:16,172 INFO [train.py:904] (1/8) Epoch 22, batch 0, loss[loss=0.2366, simple_loss=0.3064, pruned_loss=0.08341, over 16596.00 frames. ], tot_loss[loss=0.2366, simple_loss=0.3064, pruned_loss=0.08341, over 16596.00 frames. ], batch size: 68, lr: 3.10e-03, grad_scale: 8.0 2023-05-01 09:06:16,172 INFO [train.py:929] (1/8) Computing validation loss 2023-05-01 09:06:23,632 INFO [train.py:938] (1/8) Epoch 22, validation: loss=0.1457, simple_loss=0.2489, pruned_loss=0.0212, over 944034.00 frames. 2023-05-01 09:06:23,632 INFO [train.py:939] (1/8) Maximum memory allocated so far is 18145MB 2023-05-01 09:06:32,012 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=4.22 vs. limit=5.0 2023-05-01 09:07:26,339 INFO [optim.py:368] (1/8) 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] (1/8) Epoch 22, batch 50, loss[loss=0.1504, simple_loss=0.2334, pruned_loss=0.03372, over 16807.00 frames. ], tot_loss[loss=0.1845, simple_loss=0.2715, pruned_loss=0.04879, over 748747.13 frames. ], batch size: 39, lr: 3.10e-03, grad_scale: 1.0 2023-05-01 09:08:03,071 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=4.90 vs. limit=5.0 2023-05-01 09:08:41,847 INFO [train.py:904] (1/8) Epoch 22, batch 100, loss[loss=0.1724, simple_loss=0.2563, pruned_loss=0.04426, over 16790.00 frames. ], tot_loss[loss=0.1813, simple_loss=0.2669, pruned_loss=0.04786, over 1321853.82 frames. ], batch size: 39, lr: 3.10e-03, grad_scale: 1.0 2023-05-01 09:09:28,918 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.6487, 3.7216, 2.6135, 2.2474, 2.3504, 2.1617, 3.7474, 3.1690], device='cuda:1'), covar=tensor([0.2989, 0.0640, 0.1979, 0.3082, 0.2893, 0.2460, 0.0648, 0.1624], device='cuda:1'), in_proj_covar=tensor([0.0320, 0.0262, 0.0299, 0.0305, 0.0285, 0.0253, 0.0286, 0.0325], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:1') 2023-05-01 09:09:44,707 INFO [optim.py:368] (1/8) 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 09:09:51,953 INFO [train.py:904] (1/8) Epoch 22, batch 150, loss[loss=0.1509, simple_loss=0.2434, pruned_loss=0.02922, over 17206.00 frames. ], tot_loss[loss=0.1793, simple_loss=0.2651, pruned_loss=0.04682, over 1758829.84 frames. ], batch size: 45, lr: 3.10e-03, grad_scale: 1.0 2023-05-01 09:11:00,679 INFO [train.py:904] (1/8) Epoch 22, batch 200, loss[loss=0.1418, simple_loss=0.233, pruned_loss=0.02526, over 16839.00 frames. ], tot_loss[loss=0.1788, simple_loss=0.2652, pruned_loss=0.04619, over 2113276.81 frames. ], batch size: 42, lr: 3.10e-03, grad_scale: 1.0 2023-05-01 09:11:02,264 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=213354.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 09:11:20,367 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=213367.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 09:12:00,918 INFO [optim.py:368] (1/8) 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 09:12:07,195 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=213402.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 09:12:08,108 INFO [train.py:904] (1/8) Epoch 22, batch 250, loss[loss=0.184, simple_loss=0.2677, pruned_loss=0.05011, over 12236.00 frames. ], tot_loss[loss=0.1795, simple_loss=0.2651, pruned_loss=0.04692, over 2356971.29 frames. ], batch size: 246, lr: 3.10e-03, grad_scale: 1.0 2023-05-01 09:12:17,461 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.3730, 2.1508, 1.7519, 1.9478, 2.4855, 2.2793, 2.3654, 2.5835], device='cuda:1'), covar=tensor([0.0264, 0.0457, 0.0609, 0.0498, 0.0250, 0.0358, 0.0226, 0.0315], device='cuda:1'), in_proj_covar=tensor([0.0205, 0.0234, 0.0226, 0.0226, 0.0234, 0.0233, 0.0232, 0.0228], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-01 09:12:25,570 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=213415.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 09:12:46,317 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=213430.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 09:13:17,487 INFO [train.py:904] (1/8) Epoch 22, batch 300, loss[loss=0.1749, simple_loss=0.2553, pruned_loss=0.04722, over 16529.00 frames. ], tot_loss[loss=0.1763, simple_loss=0.2613, pruned_loss=0.04561, over 2558177.89 frames. ], batch size: 68, lr: 3.10e-03, grad_scale: 1.0 2023-05-01 09:13:25,552 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=3.55 vs. limit=5.0 2023-05-01 09:13:31,295 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.44 vs. limit=2.0 2023-05-01 09:13:52,066 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=213478.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 09:14:19,525 INFO [optim.py:368] (1/8) 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] (1/8) Epoch 22, batch 350, loss[loss=0.1559, simple_loss=0.2328, pruned_loss=0.03949, over 16170.00 frames. ], tot_loss[loss=0.1728, simple_loss=0.2577, pruned_loss=0.04393, over 2735036.00 frames. ], batch size: 164, lr: 3.10e-03, grad_scale: 1.0 2023-05-01 09:15:31,681 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.4732, 3.0263, 2.5775, 2.3073, 2.2491, 2.2335, 3.0979, 2.7818], device='cuda:1'), covar=tensor([0.2742, 0.0810, 0.1852, 0.2512, 0.2657, 0.2271, 0.0570, 0.1448], device='cuda:1'), in_proj_covar=tensor([0.0324, 0.0266, 0.0303, 0.0309, 0.0290, 0.0257, 0.0290, 0.0332], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:1') 2023-05-01 09:15:34,171 INFO [train.py:904] (1/8) Epoch 22, batch 400, loss[loss=0.2009, simple_loss=0.2727, pruned_loss=0.06448, over 16854.00 frames. ], tot_loss[loss=0.1717, simple_loss=0.2563, pruned_loss=0.04354, over 2863974.63 frames. ], batch size: 96, lr: 3.10e-03, grad_scale: 2.0 2023-05-01 09:15:52,091 INFO [zipformer.py:625] (1/8) 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] (1/8) 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] (1/8) Epoch 22, batch 450, loss[loss=0.1456, simple_loss=0.2226, pruned_loss=0.03435, over 16786.00 frames. ], tot_loss[loss=0.1696, simple_loss=0.2548, pruned_loss=0.04218, over 2967851.72 frames. ], batch size: 83, lr: 3.10e-03, grad_scale: 2.0 2023-05-01 09:17:16,835 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=213627.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 09:17:47,102 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([1.6856, 2.8558, 2.6176, 4.9567, 4.0359, 4.4056, 1.6237, 3.2548], device='cuda:1'), covar=tensor([0.1410, 0.0818, 0.1288, 0.0183, 0.0201, 0.0364, 0.1661, 0.0745], device='cuda:1'), in_proj_covar=tensor([0.0167, 0.0173, 0.0193, 0.0186, 0.0201, 0.0213, 0.0202, 0.0192], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-05-01 09:17:48,785 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.90 vs. limit=2.0 2023-05-01 09:17:53,028 INFO [train.py:904] (1/8) Epoch 22, batch 500, loss[loss=0.1777, simple_loss=0.2523, pruned_loss=0.05149, over 16496.00 frames. ], tot_loss[loss=0.1678, simple_loss=0.2538, pruned_loss=0.04096, over 3048291.39 frames. ], batch size: 146, lr: 3.10e-03, grad_scale: 2.0 2023-05-01 09:18:11,682 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.0874, 2.0586, 2.6623, 3.0092, 2.8759, 3.4325, 2.0471, 3.4776], device='cuda:1'), covar=tensor([0.0245, 0.0619, 0.0314, 0.0300, 0.0320, 0.0203, 0.0706, 0.0164], device='cuda:1'), in_proj_covar=tensor([0.0186, 0.0192, 0.0178, 0.0182, 0.0196, 0.0151, 0.0195, 0.0147], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-01 09:18:54,886 INFO [optim.py:368] (1/8) 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 09:19:01,584 INFO [train.py:904] (1/8) Epoch 22, batch 550, loss[loss=0.1813, simple_loss=0.2633, pruned_loss=0.04962, over 16494.00 frames. ], tot_loss[loss=0.1668, simple_loss=0.2529, pruned_loss=0.04039, over 3114807.07 frames. ], batch size: 75, lr: 3.10e-03, grad_scale: 2.0 2023-05-01 09:19:45,388 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.17 vs. limit=2.0 2023-05-01 09:20:10,673 INFO [train.py:904] (1/8) Epoch 22, batch 600, loss[loss=0.1558, simple_loss=0.2537, pruned_loss=0.02897, over 17117.00 frames. ], tot_loss[loss=0.1666, simple_loss=0.2519, pruned_loss=0.04064, over 3155897.94 frames. ], batch size: 47, lr: 3.10e-03, grad_scale: 2.0 2023-05-01 09:20:26,396 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.9645, 5.3573, 5.1123, 5.0960, 4.8626, 4.7730, 4.7394, 5.4323], device='cuda:1'), covar=tensor([0.1345, 0.0876, 0.1061, 0.0931, 0.0892, 0.1206, 0.1273, 0.0977], device='cuda:1'), in_proj_covar=tensor([0.0674, 0.0820, 0.0678, 0.0625, 0.0524, 0.0530, 0.0689, 0.0643], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:1') 2023-05-01 09:20:45,178 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.74 vs. limit=2.0 2023-05-01 09:21:13,515 INFO [optim.py:368] (1/8) 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] (1/8) Epoch 22, batch 650, loss[loss=0.169, simple_loss=0.2481, pruned_loss=0.04492, over 16928.00 frames. ], tot_loss[loss=0.166, simple_loss=0.2508, pruned_loss=0.04058, over 3196821.90 frames. ], batch size: 109, lr: 3.10e-03, grad_scale: 2.0 2023-05-01 09:21:30,662 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.3069, 4.1362, 4.5945, 2.6664, 4.8220, 4.8846, 3.4649, 3.9795], device='cuda:1'), covar=tensor([0.0616, 0.0220, 0.0219, 0.0942, 0.0072, 0.0156, 0.0407, 0.0308], device='cuda:1'), in_proj_covar=tensor([0.0148, 0.0108, 0.0097, 0.0140, 0.0080, 0.0125, 0.0129, 0.0129], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0004, 0.0003], device='cuda:1') 2023-05-01 09:21:38,406 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.8844, 2.8725, 2.6042, 2.8187, 3.2215, 2.9616, 3.5232, 3.3584], device='cuda:1'), covar=tensor([0.0135, 0.0424, 0.0494, 0.0448, 0.0297, 0.0402, 0.0233, 0.0323], device='cuda:1'), in_proj_covar=tensor([0.0210, 0.0236, 0.0227, 0.0228, 0.0237, 0.0236, 0.0236, 0.0231], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-01 09:22:16,700 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=2.81 vs. limit=5.0 2023-05-01 09:22:30,249 INFO [train.py:904] (1/8) Epoch 22, batch 700, loss[loss=0.206, simple_loss=0.2764, pruned_loss=0.06777, over 16449.00 frames. ], tot_loss[loss=0.165, simple_loss=0.2501, pruned_loss=0.03998, over 3224030.33 frames. ], batch size: 146, lr: 3.10e-03, grad_scale: 2.0 2023-05-01 09:22:31,837 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.9825, 5.0218, 5.4290, 5.4335, 5.4501, 5.1250, 5.0347, 4.8489], device='cuda:1'), covar=tensor([0.0352, 0.0553, 0.0426, 0.0425, 0.0498, 0.0412, 0.0904, 0.0423], device='cuda:1'), in_proj_covar=tensor([0.0404, 0.0450, 0.0437, 0.0405, 0.0485, 0.0459, 0.0541, 0.0369], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:1') 2023-05-01 09:23:17,062 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([5.2894, 5.6119, 5.4191, 5.4396, 5.1336, 5.1099, 5.0593, 5.7625], device='cuda:1'), covar=tensor([0.1323, 0.1011, 0.1030, 0.0844, 0.0885, 0.0778, 0.1138, 0.0948], device='cuda:1'), in_proj_covar=tensor([0.0675, 0.0823, 0.0678, 0.0626, 0.0525, 0.0532, 0.0690, 0.0645], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:1') 2023-05-01 09:23:35,483 INFO [optim.py:368] (1/8) 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] (1/8) Epoch 22, batch 750, loss[loss=0.1717, simple_loss=0.2691, pruned_loss=0.0371, over 17092.00 frames. ], tot_loss[loss=0.1652, simple_loss=0.2507, pruned_loss=0.03984, over 3239578.55 frames. ], batch size: 55, lr: 3.10e-03, grad_scale: 2.0 2023-05-01 09:23:43,460 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.9342, 4.5221, 4.5179, 3.2428, 3.7484, 4.4053, 4.0682, 2.7655], device='cuda:1'), covar=tensor([0.0506, 0.0052, 0.0040, 0.0357, 0.0134, 0.0081, 0.0079, 0.0432], device='cuda:1'), in_proj_covar=tensor([0.0136, 0.0083, 0.0084, 0.0135, 0.0098, 0.0108, 0.0094, 0.0129], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-05-01 09:24:08,343 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=213922.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 09:24:53,400 INFO [train.py:904] (1/8) Epoch 22, batch 800, loss[loss=0.1585, simple_loss=0.2434, pruned_loss=0.03681, over 16620.00 frames. ], tot_loss[loss=0.1649, simple_loss=0.2503, pruned_loss=0.03977, over 3257993.60 frames. ], batch size: 68, lr: 3.10e-03, grad_scale: 4.0 2023-05-01 09:25:35,100 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.7895, 1.9983, 2.3535, 2.6702, 2.7090, 2.7102, 1.9595, 2.9159], device='cuda:1'), covar=tensor([0.0203, 0.0489, 0.0367, 0.0299, 0.0312, 0.0315, 0.0583, 0.0172], device='cuda:1'), in_proj_covar=tensor([0.0188, 0.0192, 0.0177, 0.0182, 0.0196, 0.0152, 0.0194, 0.0148], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-01 09:25:56,892 INFO [optim.py:368] (1/8) 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 09:26:06,621 INFO [train.py:904] (1/8) Epoch 22, batch 850, loss[loss=0.1639, simple_loss=0.2498, pruned_loss=0.03905, over 16800.00 frames. ], tot_loss[loss=0.1642, simple_loss=0.2498, pruned_loss=0.03936, over 3271838.97 frames. ], batch size: 102, lr: 3.10e-03, grad_scale: 4.0 2023-05-01 09:26:23,349 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.74 vs. limit=2.0 2023-05-01 09:27:17,240 INFO [train.py:904] (1/8) Epoch 22, batch 900, loss[loss=0.1768, simple_loss=0.2673, pruned_loss=0.04317, over 17169.00 frames. ], tot_loss[loss=0.1637, simple_loss=0.249, pruned_loss=0.03927, over 3278527.74 frames. ], batch size: 47, lr: 3.10e-03, grad_scale: 4.0 2023-05-01 09:27:57,101 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.5343, 4.5226, 4.4366, 3.6733, 4.4811, 1.6961, 4.2095, 4.0308], device='cuda:1'), covar=tensor([0.0186, 0.0161, 0.0258, 0.0508, 0.0165, 0.3093, 0.0232, 0.0345], device='cuda:1'), in_proj_covar=tensor([0.0165, 0.0154, 0.0196, 0.0174, 0.0174, 0.0207, 0.0185, 0.0171], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-01 09:28:19,798 INFO [optim.py:368] (1/8) 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 09:28:27,527 INFO [train.py:904] (1/8) Epoch 22, batch 950, loss[loss=0.1927, simple_loss=0.2732, pruned_loss=0.05611, over 17131.00 frames. ], tot_loss[loss=0.1635, simple_loss=0.2487, pruned_loss=0.03913, over 3290448.50 frames. ], batch size: 49, lr: 3.10e-03, grad_scale: 4.0 2023-05-01 09:29:03,267 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=214129.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 09:29:35,585 INFO [train.py:904] (1/8) Epoch 22, batch 1000, loss[loss=0.1565, simple_loss=0.2307, pruned_loss=0.04119, over 16703.00 frames. ], tot_loss[loss=0.1628, simple_loss=0.2474, pruned_loss=0.03905, over 3299045.55 frames. ], batch size: 89, lr: 3.10e-03, grad_scale: 4.0 2023-05-01 09:30:29,079 INFO [zipformer.py:625] (1/8) 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] (1/8) 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 09:30:46,635 INFO [train.py:904] (1/8) Epoch 22, batch 1050, loss[loss=0.181, simple_loss=0.2535, pruned_loss=0.05428, over 16756.00 frames. ], tot_loss[loss=0.1625, simple_loss=0.2466, pruned_loss=0.03923, over 3302706.72 frames. ], batch size: 134, lr: 3.10e-03, grad_scale: 4.0 2023-05-01 09:31:07,775 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=214218.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 09:31:13,729 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=214222.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 09:31:56,092 INFO [train.py:904] (1/8) Epoch 22, batch 1100, loss[loss=0.1659, simple_loss=0.2395, pruned_loss=0.04608, over 16762.00 frames. ], tot_loss[loss=0.1624, simple_loss=0.2463, pruned_loss=0.03924, over 3300391.29 frames. ], batch size: 124, lr: 3.10e-03, grad_scale: 4.0 2023-05-01 09:32:04,083 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.74 vs. limit=2.0 2023-05-01 09:32:19,670 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=214270.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 09:32:31,193 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=214279.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 09:32:57,750 INFO [optim.py:368] (1/8) 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] (1/8) Epoch 22, batch 1150, loss[loss=0.1532, simple_loss=0.2289, pruned_loss=0.0387, over 16946.00 frames. ], tot_loss[loss=0.1619, simple_loss=0.2462, pruned_loss=0.03878, over 3309954.25 frames. ], batch size: 41, lr: 3.10e-03, grad_scale: 4.0 2023-05-01 09:33:20,807 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.9096, 1.9936, 2.2208, 3.5118, 2.0316, 2.1496, 2.1030, 2.1118], device='cuda:1'), covar=tensor([0.1881, 0.4665, 0.3333, 0.0873, 0.5013, 0.3425, 0.4422, 0.4072], device='cuda:1'), in_proj_covar=tensor([0.0403, 0.0449, 0.0370, 0.0330, 0.0439, 0.0514, 0.0421, 0.0527], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-01 09:34:15,635 INFO [train.py:904] (1/8) Epoch 22, batch 1200, loss[loss=0.165, simple_loss=0.2535, pruned_loss=0.03831, over 16784.00 frames. ], tot_loss[loss=0.1614, simple_loss=0.2459, pruned_loss=0.03842, over 3320027.35 frames. ], batch size: 57, lr: 3.10e-03, grad_scale: 8.0 2023-05-01 09:35:18,107 INFO [optim.py:368] (1/8) 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 09:35:25,088 INFO [train.py:904] (1/8) Epoch 22, batch 1250, loss[loss=0.1686, simple_loss=0.2482, pruned_loss=0.0445, over 16844.00 frames. ], tot_loss[loss=0.1616, simple_loss=0.2459, pruned_loss=0.03867, over 3327093.57 frames. ], batch size: 102, lr: 3.10e-03, grad_scale: 8.0 2023-05-01 09:36:35,051 INFO [train.py:904] (1/8) Epoch 22, batch 1300, loss[loss=0.1561, simple_loss=0.2357, pruned_loss=0.03825, over 16654.00 frames. ], tot_loss[loss=0.1615, simple_loss=0.246, pruned_loss=0.03845, over 3327723.17 frames. ], batch size: 89, lr: 3.10e-03, grad_scale: 8.0 2023-05-01 09:37:11,705 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([5.2337, 5.2532, 4.9723, 4.4169, 5.0218, 1.9947, 4.7526, 4.7853], device='cuda:1'), covar=tensor([0.0100, 0.0084, 0.0223, 0.0426, 0.0117, 0.2843, 0.0162, 0.0265], device='cuda:1'), in_proj_covar=tensor([0.0167, 0.0156, 0.0199, 0.0176, 0.0177, 0.0208, 0.0188, 0.0173], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-01 09:37:18,479 INFO [zipformer.py:625] (1/8) 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] (1/8) 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] (1/8) Epoch 22, batch 1350, loss[loss=0.1716, simple_loss=0.2639, pruned_loss=0.03965, over 16696.00 frames. ], tot_loss[loss=0.1624, simple_loss=0.2473, pruned_loss=0.03876, over 3321440.89 frames. ], batch size: 57, lr: 3.09e-03, grad_scale: 8.0 2023-05-01 09:38:48,216 INFO [train.py:904] (1/8) Epoch 22, batch 1400, loss[loss=0.1576, simple_loss=0.2393, pruned_loss=0.03799, over 16847.00 frames. ], tot_loss[loss=0.1625, simple_loss=0.2479, pruned_loss=0.0386, over 3317804.57 frames. ], batch size: 102, lr: 3.09e-03, grad_scale: 8.0 2023-05-01 09:39:17,560 INFO [zipformer.py:625] (1/8) 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] (1/8) 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] (1/8) Epoch 22, batch 1450, loss[loss=0.154, simple_loss=0.2323, pruned_loss=0.03785, over 16280.00 frames. ], tot_loss[loss=0.1627, simple_loss=0.2473, pruned_loss=0.03909, over 3323695.13 frames. ], batch size: 165, lr: 3.09e-03, grad_scale: 8.0 2023-05-01 09:41:07,195 INFO [train.py:904] (1/8) Epoch 22, batch 1500, loss[loss=0.1777, simple_loss=0.2604, pruned_loss=0.04755, over 16173.00 frames. ], tot_loss[loss=0.163, simple_loss=0.2473, pruned_loss=0.03937, over 3325170.09 frames. ], batch size: 164, lr: 3.09e-03, grad_scale: 4.0 2023-05-01 09:41:09,971 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.3632, 4.3422, 4.3326, 3.8033, 4.3301, 1.8608, 4.1077, 3.9970], device='cuda:1'), covar=tensor([0.0156, 0.0115, 0.0175, 0.0342, 0.0116, 0.2741, 0.0163, 0.0237], device='cuda:1'), in_proj_covar=tensor([0.0168, 0.0156, 0.0199, 0.0177, 0.0177, 0.0209, 0.0188, 0.0174], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-01 09:41:48,003 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.26 vs. limit=2.0 2023-05-01 09:42:09,909 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.86 vs. limit=2.0 2023-05-01 09:42:10,190 INFO [optim.py:368] (1/8) 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] (1/8) Epoch 22, batch 1550, loss[loss=0.2108, simple_loss=0.2921, pruned_loss=0.06471, over 12164.00 frames. ], tot_loss[loss=0.1648, simple_loss=0.2487, pruned_loss=0.04043, over 3311148.72 frames. ], batch size: 246, lr: 3.09e-03, grad_scale: 4.0 2023-05-01 09:43:27,996 INFO [train.py:904] (1/8) Epoch 22, batch 1600, loss[loss=0.1782, simple_loss=0.2735, pruned_loss=0.04147, over 16012.00 frames. ], tot_loss[loss=0.1659, simple_loss=0.25, pruned_loss=0.04089, over 3312176.67 frames. ], batch size: 35, lr: 3.09e-03, grad_scale: 8.0 2023-05-01 09:44:12,530 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=214785.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 09:44:32,064 INFO [optim.py:368] (1/8) 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 09:44:37,870 INFO [train.py:904] (1/8) Epoch 22, batch 1650, loss[loss=0.1815, simple_loss=0.2726, pruned_loss=0.0452, over 17060.00 frames. ], tot_loss[loss=0.1671, simple_loss=0.2516, pruned_loss=0.0413, over 3316344.98 frames. ], batch size: 55, lr: 3.09e-03, grad_scale: 8.0 2023-05-01 09:45:20,146 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=214833.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 09:45:35,037 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.29 vs. limit=2.0 2023-05-01 09:45:47,647 INFO [train.py:904] (1/8) Epoch 22, batch 1700, loss[loss=0.176, simple_loss=0.2489, pruned_loss=0.05155, over 16865.00 frames. ], tot_loss[loss=0.1684, simple_loss=0.2533, pruned_loss=0.04181, over 3315630.39 frames. ], batch size: 116, lr: 3.09e-03, grad_scale: 8.0 2023-05-01 09:46:18,506 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=214874.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 09:46:23,099 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=4.69 vs. limit=5.0 2023-05-01 09:46:30,326 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.4444, 4.3056, 4.3512, 4.0539, 4.1364, 4.4248, 4.1227, 4.1709], device='cuda:1'), covar=tensor([0.0678, 0.0791, 0.0328, 0.0314, 0.0736, 0.0510, 0.0623, 0.0604], device='cuda:1'), in_proj_covar=tensor([0.0302, 0.0438, 0.0353, 0.0351, 0.0362, 0.0406, 0.0241, 0.0422], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002], device='cuda:1') 2023-05-01 09:46:39,722 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([5.3747, 5.3000, 5.1957, 4.7015, 4.8516, 5.2288, 5.2013, 4.7951], device='cuda:1'), covar=tensor([0.0591, 0.0530, 0.0309, 0.0365, 0.1153, 0.0518, 0.0290, 0.0804], device='cuda:1'), in_proj_covar=tensor([0.0301, 0.0438, 0.0353, 0.0350, 0.0361, 0.0406, 0.0241, 0.0421], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002], device='cuda:1') 2023-05-01 09:46:53,103 INFO [optim.py:368] (1/8) 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] (1/8) Epoch 22, batch 1750, loss[loss=0.1824, simple_loss=0.2618, pruned_loss=0.05143, over 16307.00 frames. ], tot_loss[loss=0.1693, simple_loss=0.2548, pruned_loss=0.04195, over 3319762.28 frames. ], batch size: 165, lr: 3.09e-03, grad_scale: 8.0 2023-05-01 09:47:00,407 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.31 vs. limit=2.0 2023-05-01 09:47:25,706 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 22, batch 1800, loss[loss=0.1728, simple_loss=0.2638, pruned_loss=0.0409, over 16487.00 frames. ], tot_loss[loss=0.1689, simple_loss=0.2552, pruned_loss=0.04132, over 3327657.43 frames. ], batch size: 75, lr: 3.09e-03, grad_scale: 8.0 2023-05-01 09:49:13,316 INFO [optim.py:368] (1/8) 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 09:49:18,204 INFO [train.py:904] (1/8) Epoch 22, batch 1850, loss[loss=0.1562, simple_loss=0.2414, pruned_loss=0.03553, over 16852.00 frames. ], tot_loss[loss=0.1698, simple_loss=0.2562, pruned_loss=0.04169, over 3317798.77 frames. ], batch size: 102, lr: 3.09e-03, grad_scale: 4.0 2023-05-01 09:49:28,188 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.6562, 3.8355, 2.0990, 4.4077, 2.7988, 4.3151, 2.2549, 3.0917], device='cuda:1'), covar=tensor([0.0368, 0.0390, 0.1997, 0.0310, 0.0973, 0.0426, 0.1924, 0.0787], device='cuda:1'), in_proj_covar=tensor([0.0174, 0.0180, 0.0198, 0.0168, 0.0180, 0.0222, 0.0206, 0.0182], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-05-01 09:50:27,231 INFO [train.py:904] (1/8) Epoch 22, batch 1900, loss[loss=0.174, simple_loss=0.265, pruned_loss=0.04151, over 15354.00 frames. ], tot_loss[loss=0.1679, simple_loss=0.2545, pruned_loss=0.04066, over 3321960.40 frames. ], batch size: 190, lr: 3.09e-03, grad_scale: 4.0 2023-05-01 09:51:31,794 INFO [optim.py:368] (1/8) 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] (1/8) Epoch 22, batch 1950, loss[loss=0.1764, simple_loss=0.269, pruned_loss=0.04184, over 17034.00 frames. ], tot_loss[loss=0.1686, simple_loss=0.2555, pruned_loss=0.04086, over 3319483.27 frames. ], batch size: 53, lr: 3.09e-03, grad_scale: 4.0 2023-05-01 09:52:41,441 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.86 vs. limit=2.0 2023-05-01 09:52:44,782 INFO [train.py:904] (1/8) Epoch 22, batch 2000, loss[loss=0.1953, simple_loss=0.2605, pruned_loss=0.06507, over 16837.00 frames. ], tot_loss[loss=0.1688, simple_loss=0.2553, pruned_loss=0.04119, over 3306685.56 frames. ], batch size: 109, lr: 3.09e-03, grad_scale: 8.0 2023-05-01 09:53:21,982 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.2253, 3.2828, 3.3481, 2.3268, 3.0556, 3.4240, 3.1313, 2.0845], device='cuda:1'), covar=tensor([0.0462, 0.0120, 0.0064, 0.0402, 0.0136, 0.0107, 0.0110, 0.0448], device='cuda:1'), in_proj_covar=tensor([0.0135, 0.0083, 0.0084, 0.0134, 0.0098, 0.0109, 0.0095, 0.0128], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-05-01 09:53:46,014 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.37 vs. limit=2.0 2023-05-01 09:53:48,736 INFO [optim.py:368] (1/8) Clipping_scale=2.0, grad-norm quartiles 1.559e+02 2.173e+02 2.667e+02 3.241e+02 6.066e+02, threshold=5.334e+02, percent-clipped=3.0 2023-05-01 09:53:53,918 INFO [train.py:904] (1/8) Epoch 22, batch 2050, loss[loss=0.141, simple_loss=0.2251, pruned_loss=0.02842, over 16745.00 frames. ], tot_loss[loss=0.1694, simple_loss=0.2554, pruned_loss=0.04173, over 3303819.38 frames. ], batch size: 39, lr: 3.09e-03, grad_scale: 8.0 2023-05-01 09:54:17,876 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([5.2015, 5.7277, 5.8406, 5.5405, 5.6077, 6.1863, 5.6501, 5.3881], device='cuda:1'), covar=tensor([0.0852, 0.1818, 0.2361, 0.2041, 0.2817, 0.1008, 0.1457, 0.2436], device='cuda:1'), in_proj_covar=tensor([0.0420, 0.0616, 0.0674, 0.0510, 0.0679, 0.0708, 0.0532, 0.0680], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-05-01 09:54:31,452 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=215230.0, num_to_drop=1, layers_to_drop={1} 2023-05-01 09:55:01,198 INFO [train.py:904] (1/8) Epoch 22, batch 2100, loss[loss=0.1479, simple_loss=0.2324, pruned_loss=0.03166, over 17017.00 frames. ], tot_loss[loss=0.1695, simple_loss=0.2559, pruned_loss=0.0416, over 3317301.57 frames. ], batch size: 41, lr: 3.09e-03, grad_scale: 8.0 2023-05-01 09:55:54,400 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=215291.0, num_to_drop=1, layers_to_drop={0} 2023-05-01 09:56:04,635 INFO [optim.py:368] (1/8) 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 09:56:09,031 INFO [train.py:904] (1/8) Epoch 22, batch 2150, loss[loss=0.1877, simple_loss=0.2718, pruned_loss=0.05181, over 16866.00 frames. ], tot_loss[loss=0.1705, simple_loss=0.2569, pruned_loss=0.04202, over 3323936.30 frames. ], batch size: 109, lr: 3.09e-03, grad_scale: 8.0 2023-05-01 09:56:39,627 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=215325.0, num_to_drop=1, layers_to_drop={1} 2023-05-01 09:56:51,098 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 22, batch 2200, loss[loss=0.1641, simple_loss=0.2567, pruned_loss=0.03579, over 17186.00 frames. ], tot_loss[loss=0.1704, simple_loss=0.2569, pruned_loss=0.04198, over 3328341.01 frames. ], batch size: 46, lr: 3.09e-03, grad_scale: 8.0 2023-05-01 09:58:02,288 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=215386.0, num_to_drop=1, layers_to_drop={3} 2023-05-01 09:58:11,523 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.7503, 2.6786, 2.2433, 2.4542, 3.0072, 2.7490, 3.3767, 3.2333], device='cuda:1'), covar=tensor([0.0150, 0.0416, 0.0534, 0.0483, 0.0284, 0.0396, 0.0225, 0.0259], device='cuda:1'), in_proj_covar=tensor([0.0218, 0.0241, 0.0231, 0.0231, 0.0242, 0.0241, 0.0243, 0.0237], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-01 09:58:14,359 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=215394.0, num_to_drop=1, layers_to_drop={1} 2023-05-01 09:58:20,772 INFO [optim.py:368] (1/8) 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 09:58:24,659 INFO [train.py:904] (1/8) Epoch 22, batch 2250, loss[loss=0.1873, simple_loss=0.2619, pruned_loss=0.05636, over 16856.00 frames. ], tot_loss[loss=0.1717, simple_loss=0.2582, pruned_loss=0.04262, over 3327731.98 frames. ], batch size: 116, lr: 3.09e-03, grad_scale: 8.0 2023-05-01 09:58:35,540 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.77 vs. limit=2.0 2023-05-01 09:59:36,632 INFO [train.py:904] (1/8) Epoch 22, batch 2300, loss[loss=0.1639, simple_loss=0.2562, pruned_loss=0.03581, over 17235.00 frames. ], tot_loss[loss=0.1712, simple_loss=0.258, pruned_loss=0.04221, over 3331385.08 frames. ], batch size: 52, lr: 3.09e-03, grad_scale: 8.0 2023-05-01 10:00:27,111 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([5.3458, 5.3285, 5.1786, 4.7011, 4.8668, 5.2624, 5.1626, 4.8443], device='cuda:1'), covar=tensor([0.0567, 0.0439, 0.0332, 0.0358, 0.1071, 0.0421, 0.0316, 0.0769], device='cuda:1'), in_proj_covar=tensor([0.0306, 0.0443, 0.0356, 0.0355, 0.0365, 0.0410, 0.0244, 0.0427], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-05-01 10:00:27,277 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.5211, 2.4615, 2.4709, 4.4322, 2.3162, 2.8432, 2.5303, 2.6724], device='cuda:1'), covar=tensor([0.1182, 0.3405, 0.2995, 0.0469, 0.4166, 0.2496, 0.3501, 0.3550], device='cuda:1'), in_proj_covar=tensor([0.0405, 0.0452, 0.0372, 0.0331, 0.0439, 0.0519, 0.0423, 0.0530], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-01 10:00:42,878 INFO [optim.py:368] (1/8) 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] (1/8) Epoch 22, batch 2350, loss[loss=0.1666, simple_loss=0.2475, pruned_loss=0.0429, over 16809.00 frames. ], tot_loss[loss=0.1715, simple_loss=0.2581, pruned_loss=0.0425, over 3318942.12 frames. ], batch size: 102, lr: 3.09e-03, grad_scale: 8.0 2023-05-01 10:00:46,957 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=215503.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 10:01:41,198 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.88 vs. limit=2.0 2023-05-01 10:01:48,858 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=215548.0, num_to_drop=1, layers_to_drop={0} 2023-05-01 10:01:55,002 INFO [train.py:904] (1/8) Epoch 22, batch 2400, loss[loss=0.1749, simple_loss=0.2712, pruned_loss=0.03935, over 16751.00 frames. ], tot_loss[loss=0.1721, simple_loss=0.2587, pruned_loss=0.04275, over 3326564.99 frames. ], batch size: 62, lr: 3.09e-03, grad_scale: 8.0 2023-05-01 10:02:11,580 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=215564.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 10:02:12,672 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=215565.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 10:02:41,693 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=215586.0, num_to_drop=1, layers_to_drop={2} 2023-05-01 10:02:59,709 INFO [optim.py:368] (1/8) 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] (1/8) Epoch 22, batch 2450, loss[loss=0.16, simple_loss=0.2496, pruned_loss=0.03518, over 17120.00 frames. ], tot_loss[loss=0.1725, simple_loss=0.2594, pruned_loss=0.04281, over 3319529.36 frames. ], batch size: 47, lr: 3.09e-03, grad_scale: 8.0 2023-05-01 10:03:08,795 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.91 vs. limit=2.0 2023-05-01 10:03:11,937 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=215609.0, num_to_drop=1, layers_to_drop={2} 2023-05-01 10:03:35,724 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=215626.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 10:04:11,703 INFO [train.py:904] (1/8) Epoch 22, batch 2500, loss[loss=0.1778, simple_loss=0.2629, pruned_loss=0.04631, over 16486.00 frames. ], tot_loss[loss=0.1722, simple_loss=0.2593, pruned_loss=0.04254, over 3328696.28 frames. ], batch size: 146, lr: 3.09e-03, grad_scale: 8.0 2023-05-01 10:04:51,431 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=215681.0, num_to_drop=1, layers_to_drop={2} 2023-05-01 10:05:02,668 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=215689.0, num_to_drop=1, layers_to_drop={1} 2023-05-01 10:05:02,824 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.3697, 3.3869, 2.1304, 3.5750, 2.6630, 3.5541, 2.2715, 2.7946], device='cuda:1'), covar=tensor([0.0301, 0.0406, 0.1473, 0.0348, 0.0832, 0.0799, 0.1380, 0.0728], device='cuda:1'), in_proj_covar=tensor([0.0174, 0.0180, 0.0197, 0.0168, 0.0180, 0.0222, 0.0205, 0.0181], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-05-01 10:05:15,511 INFO [optim.py:368] (1/8) 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] (1/8) Epoch 22, batch 2550, loss[loss=0.1549, simple_loss=0.2419, pruned_loss=0.03399, over 17236.00 frames. ], tot_loss[loss=0.1721, simple_loss=0.2592, pruned_loss=0.04251, over 3323803.21 frames. ], batch size: 45, lr: 3.09e-03, grad_scale: 8.0 2023-05-01 10:06:00,857 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.83 vs. limit=2.0 2023-05-01 10:06:11,028 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.5627, 4.9764, 5.2668, 5.2325, 5.2803, 4.9561, 4.6213, 4.7265], device='cuda:1'), covar=tensor([0.0735, 0.0695, 0.0644, 0.0715, 0.0715, 0.0720, 0.1586, 0.0580], device='cuda:1'), in_proj_covar=tensor([0.0416, 0.0461, 0.0450, 0.0416, 0.0496, 0.0473, 0.0555, 0.0376], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:1') 2023-05-01 10:06:30,390 INFO [train.py:904] (1/8) Epoch 22, batch 2600, loss[loss=0.1836, simple_loss=0.2634, pruned_loss=0.05192, over 16838.00 frames. ], tot_loss[loss=0.1715, simple_loss=0.2588, pruned_loss=0.04209, over 3322957.42 frames. ], batch size: 116, lr: 3.09e-03, grad_scale: 8.0 2023-05-01 10:07:02,346 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.16 vs. limit=2.0 2023-05-01 10:07:03,711 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.18 vs. limit=2.0 2023-05-01 10:07:33,111 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.9264, 4.8801, 4.6992, 4.2145, 4.8363, 2.0000, 4.5782, 4.5046], device='cuda:1'), covar=tensor([0.0113, 0.0102, 0.0233, 0.0374, 0.0105, 0.2734, 0.0149, 0.0245], device='cuda:1'), in_proj_covar=tensor([0.0170, 0.0158, 0.0202, 0.0179, 0.0180, 0.0211, 0.0191, 0.0176], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-01 10:07:36,116 INFO [optim.py:368] (1/8) 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 10:07:39,997 INFO [train.py:904] (1/8) Epoch 22, batch 2650, loss[loss=0.1946, simple_loss=0.2717, pruned_loss=0.05874, over 16354.00 frames. ], tot_loss[loss=0.1717, simple_loss=0.2595, pruned_loss=0.04198, over 3323385.61 frames. ], batch size: 165, lr: 3.09e-03, grad_scale: 8.0 2023-05-01 10:07:41,705 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.15 vs. limit=2.0 2023-05-01 10:08:12,981 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.9706, 4.9265, 4.7293, 4.2489, 4.8829, 1.9519, 4.6455, 4.6047], device='cuda:1'), covar=tensor([0.0106, 0.0089, 0.0209, 0.0337, 0.0093, 0.2715, 0.0124, 0.0203], device='cuda:1'), in_proj_covar=tensor([0.0169, 0.0158, 0.0201, 0.0179, 0.0179, 0.0210, 0.0190, 0.0176], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-01 10:08:15,759 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=215829.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 10:08:48,041 INFO [train.py:904] (1/8) Epoch 22, batch 2700, loss[loss=0.1892, simple_loss=0.2694, pruned_loss=0.05451, over 16649.00 frames. ], tot_loss[loss=0.1719, simple_loss=0.2601, pruned_loss=0.04188, over 3327610.11 frames. ], batch size: 134, lr: 3.09e-03, grad_scale: 8.0 2023-05-01 10:08:57,507 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=215859.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 10:09:09,789 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.5387, 2.4016, 2.3631, 4.4442, 2.3881, 2.8016, 2.4943, 2.6467], device='cuda:1'), covar=tensor([0.1259, 0.3747, 0.3092, 0.0493, 0.4114, 0.2643, 0.3514, 0.3703], device='cuda:1'), in_proj_covar=tensor([0.0405, 0.0452, 0.0373, 0.0332, 0.0440, 0.0521, 0.0423, 0.0531], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-01 10:09:34,738 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=215886.0, num_to_drop=1, layers_to_drop={2} 2023-05-01 10:09:39,301 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=215890.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 10:09:53,921 INFO [optim.py:368] (1/8) 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 10:09:57,373 INFO [train.py:904] (1/8) Epoch 22, batch 2750, loss[loss=0.1745, simple_loss=0.2637, pruned_loss=0.04268, over 16877.00 frames. ], tot_loss[loss=0.1715, simple_loss=0.26, pruned_loss=0.04145, over 3333881.52 frames. ], batch size: 96, lr: 3.08e-03, grad_scale: 8.0 2023-05-01 10:09:59,452 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=215904.0, num_to_drop=1, layers_to_drop={1} 2023-05-01 10:10:10,353 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.0904, 2.1884, 2.6739, 3.0444, 2.8722, 3.5698, 2.2292, 3.5077], device='cuda:1'), covar=tensor([0.0278, 0.0556, 0.0373, 0.0330, 0.0358, 0.0194, 0.0593, 0.0187], device='cuda:1'), in_proj_covar=tensor([0.0194, 0.0197, 0.0183, 0.0188, 0.0201, 0.0157, 0.0199, 0.0154], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-01 10:10:21,932 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=215921.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 10:10:38,796 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=215934.0, num_to_drop=1, layers_to_drop={1} 2023-05-01 10:11:05,009 INFO [train.py:904] (1/8) Epoch 22, batch 2800, loss[loss=0.1971, simple_loss=0.2736, pruned_loss=0.06033, over 16904.00 frames. ], tot_loss[loss=0.1716, simple_loss=0.26, pruned_loss=0.04157, over 3329833.00 frames. ], batch size: 96, lr: 3.08e-03, grad_scale: 8.0 2023-05-01 10:11:13,987 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.9183, 4.0332, 2.7532, 4.7932, 3.1237, 4.7004, 2.7485, 3.4064], device='cuda:1'), covar=tensor([0.0363, 0.0402, 0.1495, 0.0248, 0.0894, 0.0460, 0.1503, 0.0709], device='cuda:1'), in_proj_covar=tensor([0.0173, 0.0180, 0.0197, 0.0168, 0.0180, 0.0223, 0.0205, 0.0181], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-05-01 10:11:42,360 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=215981.0, num_to_drop=1, layers_to_drop={0} 2023-05-01 10:11:53,898 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=215989.0, num_to_drop=1, layers_to_drop={0} 2023-05-01 10:11:58,043 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.9217, 3.7008, 4.0328, 2.4368, 4.1247, 4.1431, 3.3060, 3.2254], device='cuda:1'), covar=tensor([0.0764, 0.0229, 0.0185, 0.1042, 0.0083, 0.0196, 0.0382, 0.0420], device='cuda:1'), in_proj_covar=tensor([0.0149, 0.0109, 0.0099, 0.0140, 0.0081, 0.0129, 0.0131, 0.0131], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0004, 0.0004], device='cuda:1') 2023-05-01 10:12:07,418 INFO [optim.py:368] (1/8) 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] (1/8) Epoch 22, batch 2850, loss[loss=0.1361, simple_loss=0.2359, pruned_loss=0.01814, over 17155.00 frames. ], tot_loss[loss=0.1706, simple_loss=0.2588, pruned_loss=0.04119, over 3327577.12 frames. ], batch size: 46, lr: 3.08e-03, grad_scale: 8.0 2023-05-01 10:12:51,497 INFO [zipformer.py:625] (1/8) 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] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=216029.0, num_to_drop=1, layers_to_drop={1} 2023-05-01 10:12:59,486 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.41 vs. limit=2.0 2023-05-01 10:13:03,974 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=216037.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 10:13:26,222 INFO [train.py:904] (1/8) Epoch 22, batch 2900, loss[loss=0.1916, simple_loss=0.2611, pruned_loss=0.06101, over 16793.00 frames. ], tot_loss[loss=0.1697, simple_loss=0.2574, pruned_loss=0.04099, over 3327993.51 frames. ], batch size: 124, lr: 3.08e-03, grad_scale: 8.0 2023-05-01 10:13:42,519 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.7087, 3.8244, 3.9002, 3.6146, 3.6997, 4.2378, 3.8927, 3.6068], device='cuda:1'), covar=tensor([0.2053, 0.2431, 0.2699, 0.2843, 0.3343, 0.2024, 0.2005, 0.2944], device='cuda:1'), in_proj_covar=tensor([0.0423, 0.0615, 0.0676, 0.0513, 0.0682, 0.0712, 0.0533, 0.0681], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-05-01 10:13:56,561 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.1407, 3.6624, 3.8111, 2.1949, 3.1342, 2.4965, 3.5149, 3.7818], device='cuda:1'), covar=tensor([0.0407, 0.0862, 0.0491, 0.1998, 0.0882, 0.0977, 0.0846, 0.1022], device='cuda:1'), in_proj_covar=tensor([0.0158, 0.0165, 0.0168, 0.0155, 0.0146, 0.0131, 0.0145, 0.0178], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-05-01 10:14:16,958 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=216089.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 10:14:24,306 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.6798, 4.5997, 4.6043, 4.2864, 4.2996, 4.6370, 4.4670, 4.3930], device='cuda:1'), covar=tensor([0.0669, 0.0824, 0.0322, 0.0319, 0.0914, 0.0606, 0.0491, 0.0674], device='cuda:1'), in_proj_covar=tensor([0.0308, 0.0448, 0.0358, 0.0359, 0.0370, 0.0415, 0.0246, 0.0433], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-05-01 10:14:31,569 INFO [optim.py:368] (1/8) 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 10:14:35,745 INFO [train.py:904] (1/8) Epoch 22, batch 2950, loss[loss=0.1644, simple_loss=0.2479, pruned_loss=0.04048, over 16652.00 frames. ], tot_loss[loss=0.1703, simple_loss=0.2568, pruned_loss=0.04184, over 3310997.40 frames. ], batch size: 89, lr: 3.08e-03, grad_scale: 8.0 2023-05-01 10:15:45,385 INFO [train.py:904] (1/8) Epoch 22, batch 3000, loss[loss=0.1458, simple_loss=0.2384, pruned_loss=0.0266, over 17155.00 frames. ], tot_loss[loss=0.171, simple_loss=0.2576, pruned_loss=0.04218, over 3313764.62 frames. ], batch size: 48, lr: 3.08e-03, grad_scale: 8.0 2023-05-01 10:15:45,385 INFO [train.py:929] (1/8) Computing validation loss 2023-05-01 10:15:52,849 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.9073, 5.2771, 5.0845, 5.1377, 4.9063, 4.9503, 4.6281, 5.3594], device='cuda:1'), covar=tensor([0.1198, 0.0923, 0.0859, 0.0853, 0.0807, 0.0461, 0.1242, 0.0732], device='cuda:1'), in_proj_covar=tensor([0.0695, 0.0854, 0.0704, 0.0649, 0.0540, 0.0546, 0.0713, 0.0668], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-01 10:15:54,104 INFO [train.py:938] (1/8) Epoch 22, validation: loss=0.1347, simple_loss=0.2399, pruned_loss=0.0148, over 944034.00 frames. 2023-05-01 10:15:54,105 INFO [train.py:939] (1/8) Maximum memory allocated so far is 18145MB 2023-05-01 10:16:02,689 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=216159.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 10:16:38,730 INFO [zipformer.py:625] (1/8) 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] (1/8) 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 10:17:00,608 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([5.1040, 4.8600, 5.1458, 5.3495, 5.5842, 4.8290, 5.4960, 5.5212], device='cuda:1'), covar=tensor([0.2095, 0.1359, 0.1830, 0.0771, 0.0552, 0.0919, 0.0570, 0.0655], device='cuda:1'), in_proj_covar=tensor([0.0668, 0.0829, 0.0964, 0.0837, 0.0634, 0.0666, 0.0685, 0.0796], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-01 10:17:03,725 INFO [train.py:904] (1/8) Epoch 22, batch 3050, loss[loss=0.1803, simple_loss=0.2754, pruned_loss=0.04265, over 17039.00 frames. ], tot_loss[loss=0.1705, simple_loss=0.257, pruned_loss=0.042, over 3316352.87 frames. ], batch size: 55, lr: 3.08e-03, grad_scale: 8.0 2023-05-01 10:17:05,181 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=216204.0, num_to_drop=1, layers_to_drop={1} 2023-05-01 10:17:09,668 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=216207.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 10:17:29,278 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=216221.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 10:17:39,741 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.7624, 4.8220, 5.2102, 5.2022, 5.2087, 4.8799, 4.8257, 4.7228], device='cuda:1'), covar=tensor([0.0344, 0.0591, 0.0389, 0.0424, 0.0437, 0.0431, 0.0853, 0.0457], device='cuda:1'), in_proj_covar=tensor([0.0420, 0.0467, 0.0454, 0.0421, 0.0501, 0.0478, 0.0562, 0.0381], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:1') 2023-05-01 10:18:10,762 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=216252.0, num_to_drop=1, layers_to_drop={0} 2023-05-01 10:18:12,224 INFO [train.py:904] (1/8) Epoch 22, batch 3100, loss[loss=0.1517, simple_loss=0.2457, pruned_loss=0.02879, over 17180.00 frames. ], tot_loss[loss=0.1699, simple_loss=0.2568, pruned_loss=0.04148, over 3316108.51 frames. ], batch size: 46, lr: 3.08e-03, grad_scale: 8.0 2023-05-01 10:18:33,893 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=216269.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 10:19:16,926 INFO [optim.py:368] (1/8) 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 10:19:21,072 INFO [train.py:904] (1/8) Epoch 22, batch 3150, loss[loss=0.1728, simple_loss=0.2678, pruned_loss=0.03891, over 17065.00 frames. ], tot_loss[loss=0.1689, simple_loss=0.2557, pruned_loss=0.0411, over 3330168.78 frames. ], batch size: 55, lr: 3.08e-03, grad_scale: 8.0 2023-05-01 10:19:26,841 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=2.72 vs. limit=5.0 2023-05-01 10:20:29,459 INFO [train.py:904] (1/8) Epoch 22, batch 3200, loss[loss=0.1605, simple_loss=0.2413, pruned_loss=0.03991, over 16783.00 frames. ], tot_loss[loss=0.1676, simple_loss=0.2542, pruned_loss=0.04054, over 3324867.00 frames. ], batch size: 102, lr: 3.08e-03, grad_scale: 8.0 2023-05-01 10:21:13,882 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=216384.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 10:21:36,410 INFO [optim.py:368] (1/8) 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 10:21:40,318 INFO [train.py:904] (1/8) Epoch 22, batch 3250, loss[loss=0.1455, simple_loss=0.238, pruned_loss=0.02651, over 17163.00 frames. ], tot_loss[loss=0.167, simple_loss=0.2538, pruned_loss=0.04007, over 3331309.14 frames. ], batch size: 46, lr: 3.08e-03, grad_scale: 8.0 2023-05-01 10:22:52,220 INFO [train.py:904] (1/8) Epoch 22, batch 3300, loss[loss=0.1792, simple_loss=0.2591, pruned_loss=0.04963, over 16158.00 frames. ], tot_loss[loss=0.1677, simple_loss=0.2551, pruned_loss=0.04016, over 3324309.12 frames. ], batch size: 165, lr: 3.08e-03, grad_scale: 8.0 2023-05-01 10:23:35,851 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=216485.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 10:23:54,330 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=3.08 vs. limit=5.0 2023-05-01 10:23:56,628 INFO [optim.py:368] (1/8) 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] (1/8) Epoch 22, batch 3350, loss[loss=0.1488, simple_loss=0.2473, pruned_loss=0.02513, over 16633.00 frames. ], tot_loss[loss=0.1677, simple_loss=0.2553, pruned_loss=0.04009, over 3317964.01 frames. ], batch size: 62, lr: 3.08e-03, grad_scale: 8.0 2023-05-01 10:24:05,855 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([5.3286, 5.7141, 5.4529, 5.5103, 5.1294, 5.0139, 5.0609, 5.8200], device='cuda:1'), covar=tensor([0.1324, 0.0885, 0.0859, 0.0788, 0.0912, 0.0775, 0.1179, 0.0837], device='cuda:1'), in_proj_covar=tensor([0.0694, 0.0851, 0.0701, 0.0645, 0.0538, 0.0544, 0.0712, 0.0666], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:1') 2023-05-01 10:24:42,589 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=216533.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 10:25:10,652 INFO [train.py:904] (1/8) Epoch 22, batch 3400, loss[loss=0.1376, simple_loss=0.225, pruned_loss=0.02504, over 15820.00 frames. ], tot_loss[loss=0.1668, simple_loss=0.2545, pruned_loss=0.03956, over 3322673.92 frames. ], batch size: 35, lr: 3.08e-03, grad_scale: 8.0 2023-05-01 10:26:02,837 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=216591.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 10:26:07,076 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.6601, 2.6776, 2.3034, 2.4728, 2.9678, 2.7216, 3.2295, 3.2071], device='cuda:1'), covar=tensor([0.0198, 0.0455, 0.0571, 0.0536, 0.0331, 0.0438, 0.0310, 0.0303], device='cuda:1'), in_proj_covar=tensor([0.0220, 0.0241, 0.0231, 0.0231, 0.0241, 0.0241, 0.0245, 0.0238], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-01 10:26:15,739 INFO [optim.py:368] (1/8) 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 10:26:19,666 INFO [train.py:904] (1/8) Epoch 22, batch 3450, loss[loss=0.1439, simple_loss=0.2303, pruned_loss=0.02882, over 17224.00 frames. ], tot_loss[loss=0.1661, simple_loss=0.2532, pruned_loss=0.03953, over 3325693.12 frames. ], batch size: 44, lr: 3.08e-03, grad_scale: 8.0 2023-05-01 10:27:29,235 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 22, batch 3500, loss[loss=0.1548, simple_loss=0.2518, pruned_loss=0.02885, over 17040.00 frames. ], tot_loss[loss=0.1662, simple_loss=0.2529, pruned_loss=0.03974, over 3326004.89 frames. ], batch size: 50, lr: 3.08e-03, grad_scale: 8.0 2023-05-01 10:27:32,748 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=216655.0, num_to_drop=1, layers_to_drop={1} 2023-05-01 10:27:57,442 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.9911, 3.6768, 4.1023, 1.9821, 4.2246, 4.4550, 3.3217, 3.4342], device='cuda:1'), covar=tensor([0.0772, 0.0299, 0.0287, 0.1358, 0.0132, 0.0205, 0.0463, 0.0431], device='cuda:1'), in_proj_covar=tensor([0.0150, 0.0109, 0.0100, 0.0140, 0.0082, 0.0130, 0.0131, 0.0131], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-05-01 10:28:12,631 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=216684.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 10:28:35,763 INFO [optim.py:368] (1/8) 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] (1/8) Epoch 22, batch 3550, loss[loss=0.154, simple_loss=0.2367, pruned_loss=0.03564, over 15437.00 frames. ], tot_loss[loss=0.1661, simple_loss=0.2526, pruned_loss=0.03976, over 3318510.91 frames. ], batch size: 190, lr: 3.08e-03, grad_scale: 4.0 2023-05-01 10:28:42,566 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=216705.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 10:28:56,687 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=216716.0, num_to_drop=1, layers_to_drop={3} 2023-05-01 10:29:20,498 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=216732.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 10:29:49,493 INFO [train.py:904] (1/8) Epoch 22, batch 3600, loss[loss=0.15, simple_loss=0.2274, pruned_loss=0.0363, over 16551.00 frames. ], tot_loss[loss=0.1657, simple_loss=0.2517, pruned_loss=0.03982, over 3313585.44 frames. ], batch size: 75, lr: 3.08e-03, grad_scale: 8.0 2023-05-01 10:30:08,445 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=216766.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 10:30:24,179 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.8218, 2.8322, 2.4647, 2.7460, 3.1756, 3.0012, 3.5004, 3.3256], device='cuda:1'), covar=tensor([0.0158, 0.0408, 0.0507, 0.0445, 0.0290, 0.0377, 0.0231, 0.0289], device='cuda:1'), in_proj_covar=tensor([0.0221, 0.0242, 0.0232, 0.0232, 0.0242, 0.0242, 0.0246, 0.0240], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-01 10:31:00,614 INFO [optim.py:368] (1/8) 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] (1/8) Epoch 22, batch 3650, loss[loss=0.1744, simple_loss=0.2459, pruned_loss=0.05142, over 16888.00 frames. ], tot_loss[loss=0.1661, simple_loss=0.2512, pruned_loss=0.04045, over 3296677.98 frames. ], batch size: 109, lr: 3.08e-03, grad_scale: 8.0 2023-05-01 10:31:32,885 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.1705, 2.2341, 2.3286, 3.8124, 2.2627, 2.5267, 2.3019, 2.3812], device='cuda:1'), covar=tensor([0.1520, 0.3661, 0.2952, 0.0640, 0.3803, 0.2562, 0.3844, 0.3163], device='cuda:1'), in_proj_covar=tensor([0.0408, 0.0453, 0.0374, 0.0335, 0.0442, 0.0524, 0.0426, 0.0533], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-01 10:32:08,747 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.6801, 2.9683, 3.1664, 1.9974, 2.7455, 2.0682, 3.3672, 3.3241], device='cuda:1'), covar=tensor([0.0240, 0.0960, 0.0603, 0.1989, 0.0900, 0.1076, 0.0544, 0.0840], device='cuda:1'), in_proj_covar=tensor([0.0159, 0.0166, 0.0167, 0.0155, 0.0146, 0.0131, 0.0145, 0.0178], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-05-01 10:32:18,427 INFO [train.py:904] (1/8) Epoch 22, batch 3700, loss[loss=0.193, simple_loss=0.2602, pruned_loss=0.06285, over 16869.00 frames. ], tot_loss[loss=0.1669, simple_loss=0.25, pruned_loss=0.04187, over 3287062.31 frames. ], batch size: 116, lr: 3.08e-03, grad_scale: 4.0 2023-05-01 10:32:31,646 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.18 vs. limit=2.0 2023-05-01 10:33:19,755 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.5898, 3.2587, 3.6249, 2.0090, 3.7116, 3.7372, 3.1077, 2.7091], device='cuda:1'), covar=tensor([0.0750, 0.0268, 0.0197, 0.1116, 0.0117, 0.0188, 0.0371, 0.0462], device='cuda:1'), in_proj_covar=tensor([0.0150, 0.0110, 0.0100, 0.0140, 0.0082, 0.0129, 0.0131, 0.0131], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-05-01 10:33:31,443 INFO [optim.py:368] (1/8) 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 10:33:32,646 INFO [train.py:904] (1/8) Epoch 22, batch 3750, loss[loss=0.1768, simple_loss=0.2461, pruned_loss=0.05376, over 16910.00 frames. ], tot_loss[loss=0.1685, simple_loss=0.2503, pruned_loss=0.04331, over 3290014.87 frames. ], batch size: 109, lr: 3.08e-03, grad_scale: 4.0 2023-05-01 10:33:35,169 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.78 vs. limit=2.0 2023-05-01 10:34:36,639 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 22, batch 3800, loss[loss=0.1823, simple_loss=0.2643, pruned_loss=0.05013, over 16410.00 frames. ], tot_loss[loss=0.1703, simple_loss=0.2517, pruned_loss=0.04445, over 3282348.29 frames. ], batch size: 146, lr: 3.08e-03, grad_scale: 4.0 2023-05-01 10:35:03,013 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=216965.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 10:35:55,033 INFO [optim.py:368] (1/8) 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 10:35:56,826 INFO [train.py:904] (1/8) Epoch 22, batch 3850, loss[loss=0.1793, simple_loss=0.2651, pruned_loss=0.04675, over 17063.00 frames. ], tot_loss[loss=0.1709, simple_loss=0.2518, pruned_loss=0.04494, over 3282114.13 frames. ], batch size: 53, lr: 3.08e-03, grad_scale: 4.0 2023-05-01 10:36:00,581 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([1.7586, 2.6038, 2.6934, 4.7251, 3.4072, 4.1350, 1.7625, 2.8750], device='cuda:1'), covar=tensor([0.1386, 0.0944, 0.1241, 0.0161, 0.0344, 0.0367, 0.1698, 0.1031], device='cuda:1'), in_proj_covar=tensor([0.0168, 0.0175, 0.0196, 0.0195, 0.0207, 0.0219, 0.0204, 0.0194], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-05-01 10:36:08,364 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=217011.0, num_to_drop=1, layers_to_drop={2} 2023-05-01 10:36:29,800 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 22, batch 3900, loss[loss=0.2086, simple_loss=0.2881, pruned_loss=0.06456, over 12666.00 frames. ], tot_loss[loss=0.1713, simple_loss=0.2516, pruned_loss=0.0455, over 3284390.91 frames. ], batch size: 248, lr: 3.08e-03, grad_scale: 4.0 2023-05-01 10:37:22,071 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=217061.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 10:38:21,557 INFO [optim.py:368] (1/8) 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] (1/8) Epoch 22, batch 3950, loss[loss=0.1758, simple_loss=0.2494, pruned_loss=0.05109, over 16912.00 frames. ], tot_loss[loss=0.1712, simple_loss=0.2507, pruned_loss=0.04586, over 3287760.00 frames. ], batch size: 116, lr: 3.08e-03, grad_scale: 4.0 2023-05-01 10:39:35,757 INFO [train.py:904] (1/8) Epoch 22, batch 4000, loss[loss=0.1716, simple_loss=0.2576, pruned_loss=0.04281, over 16479.00 frames. ], tot_loss[loss=0.1719, simple_loss=0.2513, pruned_loss=0.04624, over 3274063.96 frames. ], batch size: 68, lr: 3.08e-03, grad_scale: 8.0 2023-05-01 10:39:45,292 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.2059, 4.5064, 4.6697, 4.6619, 4.7042, 4.3964, 4.1330, 4.2513], device='cuda:1'), covar=tensor([0.0629, 0.0729, 0.0613, 0.0704, 0.0669, 0.0654, 0.1269, 0.0686], device='cuda:1'), in_proj_covar=tensor([0.0412, 0.0456, 0.0442, 0.0410, 0.0490, 0.0464, 0.0547, 0.0372], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:1') 2023-05-01 10:40:37,440 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([5.2541, 5.2377, 5.0435, 4.3866, 5.1812, 1.9442, 4.9049, 4.6172], device='cuda:1'), covar=tensor([0.0057, 0.0053, 0.0164, 0.0333, 0.0061, 0.2938, 0.0103, 0.0237], device='cuda:1'), in_proj_covar=tensor([0.0172, 0.0161, 0.0205, 0.0184, 0.0183, 0.0213, 0.0195, 0.0179], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-01 10:40:43,096 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.18 vs. limit=2.0 2023-05-01 10:40:48,053 INFO [optim.py:368] (1/8) 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 10:40:49,972 INFO [train.py:904] (1/8) Epoch 22, batch 4050, loss[loss=0.1627, simple_loss=0.2435, pruned_loss=0.04096, over 16521.00 frames. ], tot_loss[loss=0.1716, simple_loss=0.2522, pruned_loss=0.04553, over 3273279.07 frames. ], batch size: 75, lr: 3.08e-03, grad_scale: 8.0 2023-05-01 10:41:55,378 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=217247.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 10:42:04,908 INFO [train.py:904] (1/8) Epoch 22, batch 4100, loss[loss=0.1927, simple_loss=0.2808, pruned_loss=0.05236, over 11982.00 frames. ], tot_loss[loss=0.1717, simple_loss=0.2536, pruned_loss=0.04491, over 3280799.81 frames. ], batch size: 246, lr: 3.08e-03, grad_scale: 8.0 2023-05-01 10:43:11,027 INFO [zipformer.py:625] (1/8) 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] (1/8) 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 10:43:23,202 INFO [train.py:904] (1/8) Epoch 22, batch 4150, loss[loss=0.1924, simple_loss=0.2851, pruned_loss=0.0499, over 16861.00 frames. ], tot_loss[loss=0.1771, simple_loss=0.26, pruned_loss=0.04707, over 3241165.71 frames. ], batch size: 109, lr: 3.07e-03, grad_scale: 8.0 2023-05-01 10:43:36,095 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=217311.0, num_to_drop=1, layers_to_drop={2} 2023-05-01 10:43:52,136 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=217321.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 10:44:11,932 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.63 vs. limit=2.0 2023-05-01 10:44:39,666 INFO [train.py:904] (1/8) Epoch 22, batch 4200, loss[loss=0.2002, simple_loss=0.2954, pruned_loss=0.05254, over 16910.00 frames. ], tot_loss[loss=0.1825, simple_loss=0.2672, pruned_loss=0.04886, over 3221008.65 frames. ], batch size: 42, lr: 3.07e-03, grad_scale: 8.0 2023-05-01 10:44:50,071 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=217359.0, num_to_drop=1, layers_to_drop={1} 2023-05-01 10:44:52,722 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=217361.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 10:45:12,326 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.4077, 4.6609, 4.5344, 4.5196, 4.2317, 4.1766, 4.1482, 4.7541], device='cuda:1'), covar=tensor([0.1058, 0.0783, 0.0822, 0.0753, 0.0756, 0.1415, 0.1131, 0.0787], device='cuda:1'), in_proj_covar=tensor([0.0684, 0.0839, 0.0691, 0.0638, 0.0531, 0.0541, 0.0702, 0.0657], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:1') 2023-05-01 10:45:32,814 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=217388.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 10:45:50,752 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.5162, 3.4780, 2.6848, 2.1186, 2.3449, 2.2386, 3.6033, 3.0495], device='cuda:1'), covar=tensor([0.3126, 0.0799, 0.2094, 0.3273, 0.3309, 0.2347, 0.0821, 0.1706], device='cuda:1'), in_proj_covar=tensor([0.0326, 0.0271, 0.0305, 0.0314, 0.0299, 0.0261, 0.0295, 0.0339], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:1') 2023-05-01 10:45:53,906 INFO [optim.py:368] (1/8) 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 10:45:55,231 INFO [train.py:904] (1/8) Epoch 22, batch 4250, loss[loss=0.1604, simple_loss=0.2571, pruned_loss=0.03189, over 16432.00 frames. ], tot_loss[loss=0.1841, simple_loss=0.2706, pruned_loss=0.04874, over 3193443.01 frames. ], batch size: 146, lr: 3.07e-03, grad_scale: 8.0 2023-05-01 10:46:04,663 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=217409.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 10:46:13,139 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.6046, 4.7453, 4.9452, 4.6250, 4.7254, 5.3120, 4.7768, 4.4909], device='cuda:1'), covar=tensor([0.1264, 0.1869, 0.1968, 0.2019, 0.2666, 0.0995, 0.1577, 0.2610], device='cuda:1'), in_proj_covar=tensor([0.0417, 0.0601, 0.0656, 0.0498, 0.0665, 0.0694, 0.0517, 0.0667], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-05-01 10:47:04,215 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=217449.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 10:47:09,152 INFO [train.py:904] (1/8) Epoch 22, batch 4300, loss[loss=0.1913, simple_loss=0.2843, pruned_loss=0.04919, over 12009.00 frames. ], tot_loss[loss=0.1836, simple_loss=0.2716, pruned_loss=0.04778, over 3197172.26 frames. ], batch size: 247, lr: 3.07e-03, grad_scale: 8.0 2023-05-01 10:48:07,838 INFO [zipformer.py:625] (1/8) 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] (1/8) Clipping_scale=2.0, grad-norm quartiles 1.458e+02 2.190e+02 2.513e+02 2.861e+02 5.953e+02, threshold=5.026e+02, percent-clipped=1.0 2023-05-01 10:48:24,301 INFO [train.py:904] (1/8) Epoch 22, batch 4350, loss[loss=0.185, simple_loss=0.2764, pruned_loss=0.0468, over 16845.00 frames. ], tot_loss[loss=0.1862, simple_loss=0.2751, pruned_loss=0.04867, over 3202725.84 frames. ], batch size: 42, lr: 3.07e-03, grad_scale: 8.0 2023-05-01 10:49:36,031 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.6237, 2.5330, 1.9016, 2.7081, 2.1310, 2.7683, 2.1175, 2.3369], device='cuda:1'), covar=tensor([0.0329, 0.0333, 0.1231, 0.0187, 0.0666, 0.0410, 0.1131, 0.0636], device='cuda:1'), in_proj_covar=tensor([0.0171, 0.0177, 0.0194, 0.0164, 0.0177, 0.0219, 0.0201, 0.0179], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-05-01 10:49:39,977 INFO [train.py:904] (1/8) Epoch 22, batch 4400, loss[loss=0.1809, simple_loss=0.2728, pruned_loss=0.04452, over 15402.00 frames. ], tot_loss[loss=0.1895, simple_loss=0.278, pruned_loss=0.05049, over 3190092.56 frames. ], batch size: 191, lr: 3.07e-03, grad_scale: 8.0 2023-05-01 10:49:41,122 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=217553.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 10:50:52,392 INFO [optim.py:368] (1/8) 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] (1/8) Epoch 22, batch 4450, loss[loss=0.1919, simple_loss=0.2828, pruned_loss=0.05057, over 16670.00 frames. ], tot_loss[loss=0.1927, simple_loss=0.2813, pruned_loss=0.05205, over 3182692.73 frames. ], batch size: 62, lr: 3.07e-03, grad_scale: 8.0 2023-05-01 10:51:20,386 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=217621.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 10:51:28,694 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.68 vs. limit=2.0 2023-05-01 10:52:08,355 INFO [train.py:904] (1/8) Epoch 22, batch 4500, loss[loss=0.1967, simple_loss=0.2873, pruned_loss=0.05309, over 16774.00 frames. ], tot_loss[loss=0.1938, simple_loss=0.2821, pruned_loss=0.05281, over 3184497.74 frames. ], batch size: 83, lr: 3.07e-03, grad_scale: 8.0 2023-05-01 10:52:24,016 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=217664.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 10:52:32,190 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=217669.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 10:53:18,212 INFO [optim.py:368] (1/8) 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 10:53:19,293 INFO [train.py:904] (1/8) Epoch 22, batch 4550, loss[loss=0.2092, simple_loss=0.2905, pruned_loss=0.06391, over 16423.00 frames. ], tot_loss[loss=0.1956, simple_loss=0.2835, pruned_loss=0.05388, over 3195813.14 frames. ], batch size: 68, lr: 3.07e-03, grad_scale: 8.0 2023-05-01 10:53:52,667 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=217725.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 10:54:16,268 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.7963, 3.8771, 2.3138, 4.7539, 3.0543, 4.6206, 2.4546, 3.1056], device='cuda:1'), covar=tensor([0.0289, 0.0352, 0.1798, 0.0112, 0.0853, 0.0375, 0.1683, 0.0823], device='cuda:1'), in_proj_covar=tensor([0.0172, 0.0178, 0.0195, 0.0164, 0.0178, 0.0221, 0.0203, 0.0180], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-05-01 10:54:19,808 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 22, batch 4600, loss[loss=0.1783, simple_loss=0.2713, pruned_loss=0.04264, over 16841.00 frames. ], tot_loss[loss=0.1961, simple_loss=0.284, pruned_loss=0.05407, over 3192979.64 frames. ], batch size: 96, lr: 3.07e-03, grad_scale: 8.0 2023-05-01 10:54:42,059 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.34 vs. limit=2.0 2023-05-01 10:55:24,421 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.31 vs. limit=2.0 2023-05-01 10:55:41,527 INFO [optim.py:368] (1/8) 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 10:55:42,806 INFO [train.py:904] (1/8) Epoch 22, batch 4650, loss[loss=0.176, simple_loss=0.2665, pruned_loss=0.04276, over 17243.00 frames. ], tot_loss[loss=0.1952, simple_loss=0.2828, pruned_loss=0.05383, over 3196547.97 frames. ], batch size: 45, lr: 3.07e-03, grad_scale: 8.0 2023-05-01 10:55:56,085 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=4.56 vs. limit=5.0 2023-05-01 10:56:25,453 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.16 vs. limit=2.0 2023-05-01 10:56:46,084 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=217848.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 10:56:52,992 INFO [train.py:904] (1/8) Epoch 22, batch 4700, loss[loss=0.1651, simple_loss=0.2528, pruned_loss=0.0387, over 16503.00 frames. ], tot_loss[loss=0.1926, simple_loss=0.28, pruned_loss=0.0526, over 3202141.55 frames. ], batch size: 62, lr: 3.07e-03, grad_scale: 8.0 2023-05-01 10:56:56,187 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.23 vs. limit=2.0 2023-05-01 10:57:59,171 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=217898.0, num_to_drop=1, layers_to_drop={0} 2023-05-01 10:58:04,665 INFO [optim.py:368] (1/8) 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] (1/8) Epoch 22, batch 4750, loss[loss=0.158, simple_loss=0.2443, pruned_loss=0.0359, over 16623.00 frames. ], tot_loss[loss=0.1881, simple_loss=0.2755, pruned_loss=0.05035, over 3216289.89 frames. ], batch size: 62, lr: 3.07e-03, grad_scale: 8.0 2023-05-01 10:58:58,958 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=2.82 vs. limit=5.0 2023-05-01 10:59:10,530 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.87 vs. limit=2.0 2023-05-01 10:59:17,330 INFO [train.py:904] (1/8) Epoch 22, batch 4800, loss[loss=0.1713, simple_loss=0.2667, pruned_loss=0.03791, over 16678.00 frames. ], tot_loss[loss=0.1848, simple_loss=0.2722, pruned_loss=0.04875, over 3207770.97 frames. ], batch size: 134, lr: 3.07e-03, grad_scale: 8.0 2023-05-01 10:59:27,609 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=217959.0, num_to_drop=1, layers_to_drop={2} 2023-05-01 10:59:47,031 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=2.54 vs. limit=5.0 2023-05-01 11:00:36,339 INFO [optim.py:368] (1/8) 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 11:00:36,355 INFO [train.py:904] (1/8) Epoch 22, batch 4850, loss[loss=0.1794, simple_loss=0.276, pruned_loss=0.04142, over 16703.00 frames. ], tot_loss[loss=0.1838, simple_loss=0.2724, pruned_loss=0.04757, over 3215972.07 frames. ], batch size: 134, lr: 3.07e-03, grad_scale: 8.0 2023-05-01 11:00:47,477 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.43 vs. limit=2.0 2023-05-01 11:01:01,606 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=218020.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 11:01:07,097 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.7248, 4.9328, 5.1518, 4.8916, 4.9713, 5.5386, 4.9897, 4.6852], device='cuda:1'), covar=tensor([0.0993, 0.1654, 0.1671, 0.1708, 0.2116, 0.0823, 0.1337, 0.2179], device='cuda:1'), in_proj_covar=tensor([0.0408, 0.0583, 0.0637, 0.0484, 0.0646, 0.0677, 0.0506, 0.0653], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-05-01 11:01:38,585 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=218044.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 11:01:51,560 INFO [train.py:904] (1/8) Epoch 22, batch 4900, loss[loss=0.1684, simple_loss=0.2565, pruned_loss=0.04013, over 16925.00 frames. ], tot_loss[loss=0.1825, simple_loss=0.2718, pruned_loss=0.04658, over 3202693.29 frames. ], batch size: 90, lr: 3.07e-03, grad_scale: 8.0 2023-05-01 11:02:49,276 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=218092.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 11:03:01,656 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=218100.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 11:03:05,447 INFO [optim.py:368] (1/8) 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] (1/8) Epoch 22, batch 4950, loss[loss=0.1734, simple_loss=0.2736, pruned_loss=0.03655, over 16637.00 frames. ], tot_loss[loss=0.1815, simple_loss=0.2714, pruned_loss=0.04583, over 3207980.76 frames. ], batch size: 89, lr: 3.07e-03, grad_scale: 8.0 2023-05-01 11:04:11,500 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 22, batch 5000, loss[loss=0.1898, simple_loss=0.2889, pruned_loss=0.04542, over 15440.00 frames. ], tot_loss[loss=0.183, simple_loss=0.2736, pruned_loss=0.04617, over 3213882.31 frames. ], batch size: 191, lr: 3.07e-03, grad_scale: 8.0 2023-05-01 11:04:30,072 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([1.9688, 3.8793, 3.8031, 2.3800, 3.4362, 3.9056, 3.4200, 2.1176], device='cuda:1'), covar=tensor([0.0647, 0.0046, 0.0051, 0.0435, 0.0105, 0.0083, 0.0112, 0.0470], device='cuda:1'), in_proj_covar=tensor([0.0134, 0.0082, 0.0084, 0.0132, 0.0098, 0.0109, 0.0094, 0.0127], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-05-01 11:04:30,102 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=218161.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 11:05:00,900 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=218182.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 11:05:11,568 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.6113, 4.7421, 4.9499, 4.7334, 4.8008, 5.3458, 4.8209, 4.4757], device='cuda:1'), covar=tensor([0.1222, 0.1906, 0.1917, 0.2001, 0.2606, 0.0924, 0.1560, 0.2578], device='cuda:1'), in_proj_covar=tensor([0.0409, 0.0582, 0.0638, 0.0485, 0.0648, 0.0677, 0.0508, 0.0655], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-05-01 11:05:17,026 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.8325, 4.8169, 4.7803, 3.9626, 4.7525, 1.6922, 4.4712, 4.5315], device='cuda:1'), covar=tensor([0.0113, 0.0098, 0.0166, 0.0551, 0.0130, 0.2973, 0.0164, 0.0242], device='cuda:1'), in_proj_covar=tensor([0.0167, 0.0156, 0.0199, 0.0179, 0.0177, 0.0208, 0.0189, 0.0173], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-01 11:05:21,350 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=218196.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 11:05:32,358 INFO [optim.py:368] (1/8) 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] (1/8) Epoch 22, batch 5050, loss[loss=0.1726, simple_loss=0.2586, pruned_loss=0.04333, over 16614.00 frames. ], tot_loss[loss=0.1829, simple_loss=0.2742, pruned_loss=0.04582, over 3225271.80 frames. ], batch size: 57, lr: 3.07e-03, grad_scale: 8.0 2023-05-01 11:05:42,383 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=218210.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 11:05:55,752 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.69 vs. limit=2.0 2023-05-01 11:06:29,680 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 22, batch 5100, loss[loss=0.1539, simple_loss=0.2424, pruned_loss=0.03271, over 17026.00 frames. ], tot_loss[loss=0.1815, simple_loss=0.2724, pruned_loss=0.04532, over 3232513.58 frames. ], batch size: 50, lr: 3.07e-03, grad_scale: 8.0 2023-05-01 11:06:45,777 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=218254.0, num_to_drop=1, layers_to_drop={2} 2023-05-01 11:07:10,578 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=218271.0, num_to_drop=1, layers_to_drop={3} 2023-05-01 11:07:47,133 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.40 vs. limit=2.0 2023-05-01 11:07:57,331 INFO [optim.py:368] (1/8) Clipping_scale=2.0, grad-norm quartiles 1.438e+02 1.925e+02 2.150e+02 2.531e+02 3.604e+02, threshold=4.300e+02, percent-clipped=0.0 2023-05-01 11:07:57,346 INFO [train.py:904] (1/8) Epoch 22, batch 5150, loss[loss=0.1596, simple_loss=0.2474, pruned_loss=0.03592, over 17274.00 frames. ], tot_loss[loss=0.181, simple_loss=0.2726, pruned_loss=0.04468, over 3238609.18 frames. ], batch size: 52, lr: 3.07e-03, grad_scale: 8.0 2023-05-01 11:08:23,111 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=218320.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 11:09:11,242 INFO [train.py:904] (1/8) Epoch 22, batch 5200, loss[loss=0.1828, simple_loss=0.2695, pruned_loss=0.04799, over 16732.00 frames. ], tot_loss[loss=0.1796, simple_loss=0.2708, pruned_loss=0.04419, over 3246863.02 frames. ], batch size: 124, lr: 3.07e-03, grad_scale: 8.0 2023-05-01 11:09:29,192 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.8142, 3.1626, 3.3022, 2.1365, 2.8645, 2.2199, 3.3568, 3.3968], device='cuda:1'), covar=tensor([0.0244, 0.0744, 0.0626, 0.1808, 0.0850, 0.0968, 0.0611, 0.0763], device='cuda:1'), in_proj_covar=tensor([0.0157, 0.0163, 0.0167, 0.0153, 0.0145, 0.0130, 0.0143, 0.0176], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:1') 2023-05-01 11:09:33,040 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=218368.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 11:09:36,297 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.5373, 2.5489, 2.5200, 4.4163, 2.4296, 2.9016, 2.6460, 2.8431], device='cuda:1'), covar=tensor([0.1292, 0.3325, 0.2770, 0.0426, 0.3739, 0.2379, 0.3320, 0.2674], device='cuda:1'), in_proj_covar=tensor([0.0401, 0.0447, 0.0366, 0.0326, 0.0433, 0.0514, 0.0419, 0.0522], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-01 11:10:03,908 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.21 vs. limit=2.0 2023-05-01 11:10:23,968 INFO [optim.py:368] (1/8) 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] (1/8) Epoch 22, batch 5250, loss[loss=0.1728, simple_loss=0.2674, pruned_loss=0.03909, over 16718.00 frames. ], tot_loss[loss=0.1778, simple_loss=0.2682, pruned_loss=0.04373, over 3237227.17 frames. ], batch size: 124, lr: 3.07e-03, grad_scale: 8.0 2023-05-01 11:10:38,924 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.22 vs. limit=2.0 2023-05-01 11:11:37,150 INFO [train.py:904] (1/8) Epoch 22, batch 5300, loss[loss=0.1543, simple_loss=0.2325, pruned_loss=0.03805, over 16656.00 frames. ], tot_loss[loss=0.1751, simple_loss=0.2647, pruned_loss=0.04278, over 3225767.21 frames. ], batch size: 62, lr: 3.07e-03, grad_scale: 8.0 2023-05-01 11:11:41,559 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=218456.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 11:11:47,305 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=2.73 vs. limit=5.0 2023-05-01 11:12:51,254 INFO [optim.py:368] (1/8) 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] (1/8) Epoch 22, batch 5350, loss[loss=0.1971, simple_loss=0.2915, pruned_loss=0.05132, over 16414.00 frames. ], tot_loss[loss=0.1736, simple_loss=0.2632, pruned_loss=0.04202, over 3227868.59 frames. ], batch size: 146, lr: 3.07e-03, grad_scale: 8.0 2023-05-01 11:13:00,630 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=218509.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 11:13:21,767 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.60 vs. limit=2.0 2023-05-01 11:13:43,231 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=218538.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 11:14:03,262 INFO [train.py:904] (1/8) Epoch 22, batch 5400, loss[loss=0.1923, simple_loss=0.2794, pruned_loss=0.05264, over 16602.00 frames. ], tot_loss[loss=0.1754, simple_loss=0.2656, pruned_loss=0.04257, over 3226351.65 frames. ], batch size: 62, lr: 3.07e-03, grad_scale: 8.0 2023-05-01 11:14:06,158 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=218554.0, num_to_drop=1, layers_to_drop={0} 2023-05-01 11:14:23,140 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=218566.0, num_to_drop=1, layers_to_drop={1} 2023-05-01 11:14:28,907 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=218570.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 11:14:34,756 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=218574.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 11:14:47,884 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.34 vs. limit=2.0 2023-05-01 11:15:18,766 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=218602.0, num_to_drop=1, layers_to_drop={0} 2023-05-01 11:15:19,489 INFO [optim.py:368] (1/8) 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 11:15:19,504 INFO [train.py:904] (1/8) Epoch 22, batch 5450, loss[loss=0.1989, simple_loss=0.2891, pruned_loss=0.05439, over 16777.00 frames. ], tot_loss[loss=0.1778, simple_loss=0.2684, pruned_loss=0.04367, over 3235681.84 frames. ], batch size: 83, lr: 3.07e-03, grad_scale: 8.0 2023-05-01 11:16:10,918 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=218635.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 11:16:37,998 INFO [train.py:904] (1/8) Epoch 22, batch 5500, loss[loss=0.2485, simple_loss=0.3216, pruned_loss=0.08774, over 11877.00 frames. ], tot_loss[loss=0.187, simple_loss=0.2763, pruned_loss=0.04879, over 3179168.01 frames. ], batch size: 248, lr: 3.07e-03, grad_scale: 8.0 2023-05-01 11:17:57,928 INFO [optim.py:368] (1/8) 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 11:17:57,944 INFO [train.py:904] (1/8) Epoch 22, batch 5550, loss[loss=0.2307, simple_loss=0.317, pruned_loss=0.07216, over 16417.00 frames. ], tot_loss[loss=0.1953, simple_loss=0.2833, pruned_loss=0.05365, over 3145957.60 frames. ], batch size: 146, lr: 3.06e-03, grad_scale: 8.0 2023-05-01 11:19:12,965 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.2388, 3.4044, 3.5393, 3.5054, 3.5325, 3.3764, 3.4013, 3.4508], device='cuda:1'), covar=tensor([0.0434, 0.0730, 0.0497, 0.0481, 0.0588, 0.0603, 0.0857, 0.0577], device='cuda:1'), in_proj_covar=tensor([0.0401, 0.0446, 0.0432, 0.0399, 0.0478, 0.0452, 0.0537, 0.0362], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:1') 2023-05-01 11:19:20,438 INFO [train.py:904] (1/8) Epoch 22, batch 5600, loss[loss=0.2781, simple_loss=0.3363, pruned_loss=0.11, over 11105.00 frames. ], tot_loss[loss=0.204, simple_loss=0.2896, pruned_loss=0.05922, over 3085673.65 frames. ], batch size: 247, lr: 3.06e-03, grad_scale: 8.0 2023-05-01 11:19:25,771 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=218756.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 11:20:41,871 INFO [optim.py:368] (1/8) 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 11:20:41,886 INFO [train.py:904] (1/8) Epoch 22, batch 5650, loss[loss=0.2827, simple_loss=0.3493, pruned_loss=0.108, over 10967.00 frames. ], tot_loss[loss=0.2107, simple_loss=0.2949, pruned_loss=0.06325, over 3057410.84 frames. ], batch size: 248, lr: 3.06e-03, grad_scale: 8.0 2023-05-01 11:20:44,085 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=218804.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 11:20:58,834 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.15 vs. limit=2.0 2023-05-01 11:21:35,456 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 22, batch 5700, loss[loss=0.2674, simple_loss=0.3303, pruned_loss=0.1022, over 11082.00 frames. ], tot_loss[loss=0.2117, simple_loss=0.2955, pruned_loss=0.06398, over 3067448.38 frames. ], batch size: 246, lr: 3.06e-03, grad_scale: 8.0 2023-05-01 11:22:16,171 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=218865.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 11:22:17,547 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=218866.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 11:22:17,786 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.8801, 2.1615, 2.4223, 3.1260, 2.2480, 2.3287, 2.3460, 2.2728], device='cuda:1'), covar=tensor([0.1343, 0.3060, 0.2291, 0.0667, 0.3604, 0.2260, 0.2915, 0.3110], device='cuda:1'), in_proj_covar=tensor([0.0402, 0.0447, 0.0367, 0.0326, 0.0435, 0.0516, 0.0419, 0.0522], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-01 11:22:32,454 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.11 vs. limit=2.0 2023-05-01 11:22:48,047 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=218886.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 11:23:13,876 INFO [optim.py:368] (1/8) 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] (1/8) Epoch 22, batch 5750, loss[loss=0.2, simple_loss=0.2906, pruned_loss=0.0547, over 16704.00 frames. ], tot_loss[loss=0.2148, simple_loss=0.2982, pruned_loss=0.06568, over 3030564.33 frames. ], batch size: 134, lr: 3.06e-03, grad_scale: 8.0 2023-05-01 11:23:24,487 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([1.6789, 2.4987, 2.3336, 3.6105, 2.6542, 3.7953, 1.3853, 2.8681], device='cuda:1'), covar=tensor([0.1408, 0.0854, 0.1369, 0.0240, 0.0260, 0.0439, 0.1818, 0.0817], device='cuda:1'), in_proj_covar=tensor([0.0166, 0.0173, 0.0194, 0.0190, 0.0204, 0.0214, 0.0201, 0.0191], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-05-01 11:23:29,755 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([1.7358, 2.6827, 2.4948, 4.2875, 3.0843, 3.9496, 1.5295, 3.0078], device='cuda:1'), covar=tensor([0.1438, 0.0817, 0.1312, 0.0189, 0.0308, 0.0470, 0.1750, 0.0843], device='cuda:1'), in_proj_covar=tensor([0.0166, 0.0173, 0.0194, 0.0190, 0.0204, 0.0214, 0.0201, 0.0191], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-05-01 11:23:32,116 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=218914.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 11:23:57,589 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=218930.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 11:24:34,389 INFO [train.py:904] (1/8) Epoch 22, batch 5800, loss[loss=0.1981, simple_loss=0.2875, pruned_loss=0.05434, over 16202.00 frames. ], tot_loss[loss=0.2132, simple_loss=0.2976, pruned_loss=0.06437, over 3035621.89 frames. ], batch size: 165, lr: 3.06e-03, grad_scale: 8.0 2023-05-01 11:24:54,978 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=218965.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 11:25:24,052 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=218984.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 11:25:52,955 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.6277, 3.8985, 2.8826, 2.2890, 2.6414, 2.4635, 4.2078, 3.4110], device='cuda:1'), covar=tensor([0.3102, 0.0624, 0.1851, 0.2740, 0.2566, 0.2049, 0.0444, 0.1315], device='cuda:1'), in_proj_covar=tensor([0.0326, 0.0269, 0.0305, 0.0314, 0.0297, 0.0259, 0.0296, 0.0337], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:1') 2023-05-01 11:25:53,553 INFO [optim.py:368] (1/8) 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] (1/8) Epoch 22, batch 5850, loss[loss=0.184, simple_loss=0.2817, pruned_loss=0.04318, over 16707.00 frames. ], tot_loss[loss=0.21, simple_loss=0.2952, pruned_loss=0.06246, over 3047611.68 frames. ], batch size: 76, lr: 3.06e-03, grad_scale: 8.0 2023-05-01 11:26:03,869 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.7680, 5.0422, 4.8099, 4.8347, 4.6093, 4.5795, 4.4938, 5.1339], device='cuda:1'), covar=tensor([0.1256, 0.0859, 0.1094, 0.0908, 0.0779, 0.1035, 0.1149, 0.0869], device='cuda:1'), in_proj_covar=tensor([0.0666, 0.0812, 0.0674, 0.0618, 0.0514, 0.0522, 0.0677, 0.0635], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:1') 2023-05-01 11:26:21,532 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=219021.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 11:26:22,752 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.5716, 4.8445, 4.6278, 4.6699, 4.4340, 4.3694, 4.3377, 4.9198], device='cuda:1'), covar=tensor([0.1193, 0.0829, 0.1105, 0.0898, 0.0779, 0.1272, 0.1158, 0.0851], device='cuda:1'), in_proj_covar=tensor([0.0665, 0.0811, 0.0674, 0.0618, 0.0514, 0.0522, 0.0676, 0.0634], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:1') 2023-05-01 11:26:29,992 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=219026.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 11:27:02,427 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=219045.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 11:27:14,925 INFO [train.py:904] (1/8) Epoch 22, batch 5900, loss[loss=0.2822, simple_loss=0.3339, pruned_loss=0.1152, over 11278.00 frames. ], tot_loss[loss=0.2085, simple_loss=0.2941, pruned_loss=0.06147, over 3075082.03 frames. ], batch size: 246, lr: 3.06e-03, grad_scale: 8.0 2023-05-01 11:27:49,385 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=219072.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 11:28:00,049 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.37 vs. limit=2.0 2023-05-01 11:28:04,724 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=219082.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 11:28:16,203 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.32 vs. limit=2.0 2023-05-01 11:28:35,884 INFO [optim.py:368] (1/8) 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] (1/8) Epoch 22, batch 5950, loss[loss=0.1857, simple_loss=0.2762, pruned_loss=0.04763, over 16692.00 frames. ], tot_loss[loss=0.2076, simple_loss=0.2951, pruned_loss=0.06005, over 3091132.56 frames. ], batch size: 134, lr: 3.06e-03, grad_scale: 8.0 2023-05-01 11:28:56,940 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.75 vs. limit=2.0 2023-05-01 11:29:24,461 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=219133.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 11:29:46,193 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.31 vs. limit=2.0 2023-05-01 11:29:57,664 INFO [train.py:904] (1/8) Epoch 22, batch 6000, loss[loss=0.1827, simple_loss=0.2772, pruned_loss=0.04404, over 16309.00 frames. ], tot_loss[loss=0.2068, simple_loss=0.2939, pruned_loss=0.05988, over 3095620.50 frames. ], batch size: 35, lr: 3.06e-03, grad_scale: 8.0 2023-05-01 11:29:57,664 INFO [train.py:929] (1/8) Computing validation loss 2023-05-01 11:30:07,616 INFO [train.py:938] (1/8) Epoch 22, validation: loss=0.1507, simple_loss=0.2632, pruned_loss=0.01907, over 944034.00 frames. 2023-05-01 11:30:07,617 INFO [train.py:939] (1/8) Maximum memory allocated so far is 18145MB 2023-05-01 11:30:27,713 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=219165.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 11:31:08,402 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=2.97 vs. limit=5.0 2023-05-01 11:31:28,923 INFO [optim.py:368] (1/8) Clipping_scale=2.0, grad-norm quartiles 1.979e+02 2.852e+02 3.411e+02 4.220e+02 7.320e+02, threshold=6.821e+02, percent-clipped=6.0 2023-05-01 11:31:28,945 INFO [train.py:904] (1/8) Epoch 22, batch 6050, loss[loss=0.2022, simple_loss=0.294, pruned_loss=0.0552, over 16668.00 frames. ], tot_loss[loss=0.2059, simple_loss=0.2926, pruned_loss=0.05958, over 3094582.28 frames. ], batch size: 134, lr: 3.06e-03, grad_scale: 8.0 2023-05-01 11:31:29,597 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=219203.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 11:31:44,912 INFO [zipformer.py:625] (1/8) 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] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=219230.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 11:32:46,223 INFO [train.py:904] (1/8) Epoch 22, batch 6100, loss[loss=0.1861, simple_loss=0.2826, pruned_loss=0.04479, over 16780.00 frames. ], tot_loss[loss=0.2043, simple_loss=0.2921, pruned_loss=0.0582, over 3114365.48 frames. ], batch size: 83, lr: 3.06e-03, grad_scale: 8.0 2023-05-01 11:33:05,467 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=219264.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 11:33:25,436 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.0143, 2.0991, 2.2159, 3.5733, 2.1046, 2.4179, 2.2167, 2.2852], device='cuda:1'), covar=tensor([0.1411, 0.3475, 0.2909, 0.0589, 0.4048, 0.2384, 0.3433, 0.3198], device='cuda:1'), in_proj_covar=tensor([0.0401, 0.0448, 0.0367, 0.0326, 0.0435, 0.0515, 0.0419, 0.0522], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-01 11:33:26,347 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=219278.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 11:33:32,097 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.1729, 2.4157, 2.0296, 2.2589, 2.8258, 2.4480, 2.7973, 2.9992], device='cuda:1'), covar=tensor([0.0148, 0.0410, 0.0531, 0.0439, 0.0267, 0.0376, 0.0227, 0.0258], device='cuda:1'), in_proj_covar=tensor([0.0208, 0.0230, 0.0223, 0.0223, 0.0232, 0.0231, 0.0232, 0.0226], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-01 11:34:04,107 INFO [train.py:904] (1/8) Epoch 22, batch 6150, loss[loss=0.1801, simple_loss=0.276, pruned_loss=0.0421, over 16760.00 frames. ], tot_loss[loss=0.2021, simple_loss=0.2894, pruned_loss=0.05739, over 3117397.43 frames. ], batch size: 83, lr: 3.06e-03, grad_scale: 4.0 2023-05-01 11:34:05,864 INFO [optim.py:368] (1/8) 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 11:34:33,692 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=219321.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 11:34:41,752 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([1.6908, 1.7825, 1.6071, 1.5157, 1.9454, 1.6166, 1.6014, 1.9198], device='cuda:1'), covar=tensor([0.0218, 0.0310, 0.0467, 0.0390, 0.0239, 0.0326, 0.0193, 0.0219], device='cuda:1'), in_proj_covar=tensor([0.0208, 0.0230, 0.0223, 0.0223, 0.0232, 0.0231, 0.0232, 0.0226], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-01 11:35:03,737 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=219340.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 11:35:06,817 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=219342.0, num_to_drop=1, layers_to_drop={1} 2023-05-01 11:35:22,807 INFO [train.py:904] (1/8) Epoch 22, batch 6200, loss[loss=0.2222, simple_loss=0.2964, pruned_loss=0.07402, over 16626.00 frames. ], tot_loss[loss=0.2011, simple_loss=0.2877, pruned_loss=0.05731, over 3100417.57 frames. ], batch size: 134, lr: 3.06e-03, grad_scale: 4.0 2023-05-01 11:36:02,464 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=219377.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 11:36:42,472 INFO [train.py:904] (1/8) Epoch 22, batch 6250, loss[loss=0.1895, simple_loss=0.2848, pruned_loss=0.04714, over 16804.00 frames. ], tot_loss[loss=0.2006, simple_loss=0.2872, pruned_loss=0.05703, over 3108873.08 frames. ], batch size: 83, lr: 3.06e-03, grad_scale: 4.0 2023-05-01 11:36:43,079 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=219403.0, num_to_drop=1, layers_to_drop={2} 2023-05-01 11:36:43,733 INFO [optim.py:368] (1/8) 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 11:36:48,727 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.5637, 3.8166, 2.8621, 2.2324, 2.4950, 2.4432, 4.0819, 3.3472], device='cuda:1'), covar=tensor([0.3115, 0.0655, 0.1846, 0.2990, 0.2805, 0.2079, 0.0435, 0.1330], device='cuda:1'), in_proj_covar=tensor([0.0327, 0.0269, 0.0305, 0.0314, 0.0298, 0.0259, 0.0296, 0.0337], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:1') 2023-05-01 11:36:53,019 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.74 vs. limit=2.0 2023-05-01 11:37:20,347 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=219428.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 11:37:31,807 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=4.11 vs. limit=5.0 2023-05-01 11:37:57,562 INFO [train.py:904] (1/8) Epoch 22, batch 6300, loss[loss=0.192, simple_loss=0.2899, pruned_loss=0.04706, over 16777.00 frames. ], tot_loss[loss=0.1997, simple_loss=0.2869, pruned_loss=0.05624, over 3121531.26 frames. ], batch size: 83, lr: 3.06e-03, grad_scale: 4.0 2023-05-01 11:37:58,168 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.9119, 3.7363, 3.7038, 4.0879, 4.1574, 3.8077, 4.0957, 4.1629], device='cuda:1'), covar=tensor([0.1648, 0.1413, 0.2241, 0.0948, 0.0886, 0.2248, 0.1255, 0.1081], device='cuda:1'), in_proj_covar=tensor([0.0629, 0.0780, 0.0903, 0.0789, 0.0598, 0.0627, 0.0650, 0.0751], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-01 11:38:47,139 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.9089, 2.1534, 2.4395, 3.1512, 2.2286, 2.4075, 2.3308, 2.2727], device='cuda:1'), covar=tensor([0.1383, 0.3382, 0.2235, 0.0708, 0.3839, 0.2120, 0.3147, 0.2933], device='cuda:1'), in_proj_covar=tensor([0.0401, 0.0448, 0.0367, 0.0326, 0.0436, 0.0516, 0.0420, 0.0522], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-01 11:38:55,964 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 22, batch 6350, loss[loss=0.1943, simple_loss=0.2873, pruned_loss=0.05059, over 16633.00 frames. ], tot_loss[loss=0.201, simple_loss=0.2876, pruned_loss=0.05717, over 3121448.80 frames. ], batch size: 76, lr: 3.06e-03, grad_scale: 4.0 2023-05-01 11:39:16,413 INFO [optim.py:368] (1/8) 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 11:40:28,623 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 22, batch 6400, loss[loss=0.1755, simple_loss=0.2642, pruned_loss=0.04338, over 16901.00 frames. ], tot_loss[loss=0.2019, simple_loss=0.2876, pruned_loss=0.05812, over 3112001.31 frames. ], batch size: 96, lr: 3.06e-03, grad_scale: 8.0 2023-05-01 11:40:34,161 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.4291, 3.5350, 3.2725, 2.9556, 3.0716, 3.4421, 3.2753, 3.1570], device='cuda:1'), covar=tensor([0.0721, 0.0775, 0.0350, 0.0340, 0.0616, 0.0576, 0.1659, 0.0628], device='cuda:1'), in_proj_covar=tensor([0.0292, 0.0430, 0.0342, 0.0339, 0.0348, 0.0395, 0.0236, 0.0408], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002], device='cuda:1') 2023-05-01 11:40:40,453 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=219559.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 11:41:15,283 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.1226, 2.4004, 2.3493, 2.7015, 1.8941, 3.1512, 1.8808, 2.7616], device='cuda:1'), covar=tensor([0.1128, 0.0592, 0.1077, 0.0168, 0.0119, 0.0452, 0.1511, 0.0707], device='cuda:1'), in_proj_covar=tensor([0.0167, 0.0174, 0.0195, 0.0190, 0.0207, 0.0216, 0.0203, 0.0193], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-05-01 11:41:45,411 INFO [train.py:904] (1/8) Epoch 22, batch 6450, loss[loss=0.1889, simple_loss=0.2815, pruned_loss=0.04813, over 15338.00 frames. ], tot_loss[loss=0.2014, simple_loss=0.2881, pruned_loss=0.05733, over 3108159.45 frames. ], batch size: 190, lr: 3.06e-03, grad_scale: 8.0 2023-05-01 11:41:47,188 INFO [optim.py:368] (1/8) 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 11:41:54,304 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.14 vs. limit=2.0 2023-05-01 11:42:13,356 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=219621.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 11:42:45,414 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 22, batch 6500, loss[loss=0.1979, simple_loss=0.2839, pruned_loss=0.05593, over 15288.00 frames. ], tot_loss[loss=0.1998, simple_loss=0.2857, pruned_loss=0.05692, over 3087512.27 frames. ], batch size: 190, lr: 3.06e-03, grad_scale: 8.0 2023-05-01 11:43:30,546 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=219669.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 11:43:42,829 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=219677.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 11:44:00,770 INFO [zipformer.py:625] (1/8) 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] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=219698.0, num_to_drop=1, layers_to_drop={2} 2023-05-01 11:44:26,639 INFO [train.py:904] (1/8) Epoch 22, batch 6550, loss[loss=0.2148, simple_loss=0.3103, pruned_loss=0.05963, over 16698.00 frames. ], tot_loss[loss=0.2021, simple_loss=0.2885, pruned_loss=0.05783, over 3088965.66 frames. ], batch size: 134, lr: 3.06e-03, grad_scale: 8.0 2023-05-01 11:44:28,427 INFO [optim.py:368] (1/8) 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 11:45:00,408 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=2.92 vs. limit=5.0 2023-05-01 11:45:03,111 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=219725.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 11:45:08,244 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=219728.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 11:45:46,189 INFO [train.py:904] (1/8) Epoch 22, batch 6600, loss[loss=0.2143, simple_loss=0.2892, pruned_loss=0.06969, over 11463.00 frames. ], tot_loss[loss=0.203, simple_loss=0.2902, pruned_loss=0.05794, over 3096380.25 frames. ], batch size: 247, lr: 3.06e-03, grad_scale: 8.0 2023-05-01 11:46:02,112 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=219763.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 11:46:23,295 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=219776.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 11:47:06,517 INFO [train.py:904] (1/8) Epoch 22, batch 6650, loss[loss=0.1912, simple_loss=0.2774, pruned_loss=0.05244, over 16744.00 frames. ], tot_loss[loss=0.2042, simple_loss=0.2907, pruned_loss=0.05886, over 3101682.29 frames. ], batch size: 124, lr: 3.06e-03, grad_scale: 8.0 2023-05-01 11:47:07,640 INFO [optim.py:368] (1/8) 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 11:47:21,131 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.85 vs. limit=2.0 2023-05-01 11:47:39,228 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=219824.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 11:48:10,762 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=219846.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 11:48:21,239 INFO [train.py:904] (1/8) Epoch 22, batch 6700, loss[loss=0.2011, simple_loss=0.2854, pruned_loss=0.05844, over 16621.00 frames. ], tot_loss[loss=0.2051, simple_loss=0.2905, pruned_loss=0.05981, over 3075471.89 frames. ], batch size: 68, lr: 3.06e-03, grad_scale: 8.0 2023-05-01 11:48:27,537 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=219857.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 11:48:30,106 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=219859.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 11:49:04,911 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.0042, 4.3971, 3.1589, 2.5460, 2.9691, 2.7510, 4.8054, 3.7479], device='cuda:1'), covar=tensor([0.2732, 0.0566, 0.1747, 0.2622, 0.2546, 0.1872, 0.0352, 0.1218], device='cuda:1'), in_proj_covar=tensor([0.0326, 0.0269, 0.0305, 0.0314, 0.0298, 0.0259, 0.0296, 0.0337], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:1') 2023-05-01 11:49:12,206 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.8792, 4.9216, 5.2623, 5.2402, 5.2985, 4.9722, 4.9084, 4.6333], device='cuda:1'), covar=tensor([0.0310, 0.0524, 0.0354, 0.0402, 0.0462, 0.0351, 0.0937, 0.0521], device='cuda:1'), in_proj_covar=tensor([0.0406, 0.0450, 0.0435, 0.0404, 0.0484, 0.0458, 0.0541, 0.0367], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:1') 2023-05-01 11:49:36,042 INFO [train.py:904] (1/8) Epoch 22, batch 6750, loss[loss=0.1802, simple_loss=0.2652, pruned_loss=0.04757, over 16609.00 frames. ], tot_loss[loss=0.2049, simple_loss=0.2898, pruned_loss=0.06003, over 3084021.38 frames. ], batch size: 62, lr: 3.06e-03, grad_scale: 8.0 2023-05-01 11:49:37,627 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=4.82 vs. limit=5.0 2023-05-01 11:49:37,870 INFO [optim.py:368] (1/8) 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 11:49:43,849 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=219907.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 11:50:00,715 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=219918.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 11:50:50,287 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([1.8672, 2.7058, 2.6131, 1.9781, 2.5272, 2.7120, 2.6430, 1.9956], device='cuda:1'), covar=tensor([0.0441, 0.0090, 0.0092, 0.0368, 0.0146, 0.0136, 0.0122, 0.0382], device='cuda:1'), in_proj_covar=tensor([0.0134, 0.0082, 0.0083, 0.0132, 0.0097, 0.0109, 0.0094, 0.0127], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-05-01 11:50:54,124 INFO [train.py:904] (1/8) Epoch 22, batch 6800, loss[loss=0.2294, simple_loss=0.3063, pruned_loss=0.07622, over 16196.00 frames. ], tot_loss[loss=0.2053, simple_loss=0.2902, pruned_loss=0.06021, over 3074250.48 frames. ], batch size: 165, lr: 3.06e-03, grad_scale: 8.0 2023-05-01 11:51:33,520 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.74 vs. limit=2.0 2023-05-01 11:52:05,044 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=219998.0, num_to_drop=1, layers_to_drop={0} 2023-05-01 11:52:15,002 INFO [train.py:904] (1/8) Epoch 22, batch 6850, loss[loss=0.2699, simple_loss=0.3294, pruned_loss=0.1052, over 11118.00 frames. ], tot_loss[loss=0.2062, simple_loss=0.2911, pruned_loss=0.06062, over 3071487.43 frames. ], batch size: 246, lr: 3.06e-03, grad_scale: 8.0 2023-05-01 11:52:16,798 INFO [optim.py:368] (1/8) 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 11:52:17,512 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=3.92 vs. limit=5.0 2023-05-01 11:52:56,871 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.21 vs. limit=2.0 2023-05-01 11:53:20,649 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=220046.0, num_to_drop=1, layers_to_drop={0} 2023-05-01 11:53:24,623 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.0762, 2.3776, 2.3224, 2.9225, 1.9704, 3.2050, 1.8563, 2.7016], device='cuda:1'), covar=tensor([0.1148, 0.0563, 0.1043, 0.0192, 0.0119, 0.0365, 0.1488, 0.0677], device='cuda:1'), in_proj_covar=tensor([0.0165, 0.0172, 0.0193, 0.0187, 0.0204, 0.0213, 0.0200, 0.0191], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-05-01 11:53:31,114 INFO [train.py:904] (1/8) Epoch 22, batch 6900, loss[loss=0.2224, simple_loss=0.3125, pruned_loss=0.06617, over 16912.00 frames. ], tot_loss[loss=0.207, simple_loss=0.2935, pruned_loss=0.06023, over 3063883.39 frames. ], batch size: 109, lr: 3.06e-03, grad_scale: 8.0 2023-05-01 11:53:59,160 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([1.9135, 2.7963, 2.6637, 4.8809, 3.6611, 4.2607, 1.6830, 3.0092], device='cuda:1'), covar=tensor([0.1332, 0.0796, 0.1215, 0.0129, 0.0298, 0.0390, 0.1651, 0.0872], device='cuda:1'), in_proj_covar=tensor([0.0165, 0.0172, 0.0193, 0.0188, 0.0204, 0.0213, 0.0200, 0.0191], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-05-01 11:54:14,063 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.22 vs. limit=2.0 2023-05-01 11:54:24,079 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=220087.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 11:54:41,874 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=220098.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 11:54:49,992 INFO [train.py:904] (1/8) Epoch 22, batch 6950, loss[loss=0.1919, simple_loss=0.2826, pruned_loss=0.05055, over 16637.00 frames. ], tot_loss[loss=0.2098, simple_loss=0.2954, pruned_loss=0.06208, over 3047546.06 frames. ], batch size: 57, lr: 3.06e-03, grad_scale: 8.0 2023-05-01 11:54:51,078 INFO [optim.py:368] (1/8) 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 11:55:14,107 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=220119.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 11:55:22,845 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.3782, 2.9268, 2.6763, 2.3279, 2.2708, 2.3134, 2.9540, 2.8574], device='cuda:1'), covar=tensor([0.2295, 0.0695, 0.1493, 0.2337, 0.2257, 0.2020, 0.0502, 0.1205], device='cuda:1'), in_proj_covar=tensor([0.0324, 0.0267, 0.0303, 0.0312, 0.0296, 0.0258, 0.0293, 0.0334], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:1') 2023-05-01 11:55:54,583 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=220146.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 11:55:57,626 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=220148.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 11:56:03,773 INFO [train.py:904] (1/8) Epoch 22, batch 7000, loss[loss=0.1731, simple_loss=0.2725, pruned_loss=0.03689, over 17016.00 frames. ], tot_loss[loss=0.208, simple_loss=0.295, pruned_loss=0.06054, over 3072462.02 frames. ], batch size: 50, lr: 3.05e-03, grad_scale: 8.0 2023-05-01 11:56:14,313 INFO [zipformer.py:625] (1/8) 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] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=220194.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 11:57:18,850 INFO [train.py:904] (1/8) Epoch 22, batch 7050, loss[loss=0.2083, simple_loss=0.2951, pruned_loss=0.06075, over 16748.00 frames. ], tot_loss[loss=0.2082, simple_loss=0.2955, pruned_loss=0.06045, over 3056958.24 frames. ], batch size: 124, lr: 3.05e-03, grad_scale: 4.0 2023-05-01 11:57:21,954 INFO [optim.py:368] (1/8) 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 11:57:34,674 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=220213.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 11:57:36,547 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.5287, 3.4661, 3.4485, 2.7489, 3.3307, 2.1278, 3.1374, 2.7672], device='cuda:1'), covar=tensor([0.0161, 0.0138, 0.0179, 0.0223, 0.0106, 0.2252, 0.0135, 0.0216], device='cuda:1'), in_proj_covar=tensor([0.0166, 0.0155, 0.0197, 0.0178, 0.0174, 0.0207, 0.0186, 0.0170], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-01 11:57:59,209 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=220229.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 11:58:36,513 INFO [train.py:904] (1/8) Epoch 22, batch 7100, loss[loss=0.2021, simple_loss=0.2895, pruned_loss=0.05736, over 16674.00 frames. ], tot_loss[loss=0.2083, simple_loss=0.2944, pruned_loss=0.06106, over 3025536.61 frames. ], batch size: 124, lr: 3.05e-03, grad_scale: 4.0 2023-05-01 11:58:45,439 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=220258.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 11:59:29,267 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.31 vs. limit=2.0 2023-05-01 11:59:35,732 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=220290.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 11:59:40,613 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([5.5593, 5.6034, 5.4229, 5.0148, 5.0753, 5.4892, 5.3729, 5.1637], device='cuda:1'), covar=tensor([0.0607, 0.0444, 0.0272, 0.0293, 0.0975, 0.0462, 0.0281, 0.0649], device='cuda:1'), in_proj_covar=tensor([0.0288, 0.0425, 0.0339, 0.0335, 0.0343, 0.0389, 0.0233, 0.0403], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-01 11:59:54,481 INFO [train.py:904] (1/8) Epoch 22, batch 7150, loss[loss=0.177, simple_loss=0.2735, pruned_loss=0.04022, over 16898.00 frames. ], tot_loss[loss=0.2072, simple_loss=0.2926, pruned_loss=0.0609, over 3032492.38 frames. ], batch size: 96, lr: 3.05e-03, grad_scale: 4.0 2023-05-01 11:59:58,136 INFO [optim.py:368] (1/8) 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 12:00:02,039 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.3658, 3.3513, 3.3647, 3.4570, 3.4906, 3.2863, 3.4622, 3.5442], device='cuda:1'), covar=tensor([0.1277, 0.0949, 0.1013, 0.0622, 0.0700, 0.2112, 0.1149, 0.0848], device='cuda:1'), in_proj_covar=tensor([0.0628, 0.0778, 0.0899, 0.0786, 0.0597, 0.0620, 0.0650, 0.0749], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-01 12:00:20,357 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=220319.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 12:01:04,574 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.8556, 4.8040, 4.5779, 3.9053, 4.7125, 1.7350, 4.4668, 4.3712], device='cuda:1'), covar=tensor([0.0095, 0.0094, 0.0209, 0.0440, 0.0107, 0.2914, 0.0137, 0.0253], device='cuda:1'), in_proj_covar=tensor([0.0166, 0.0156, 0.0198, 0.0179, 0.0175, 0.0208, 0.0187, 0.0170], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-01 12:01:07,949 INFO [train.py:904] (1/8) Epoch 22, batch 7200, loss[loss=0.1683, simple_loss=0.2548, pruned_loss=0.0409, over 16584.00 frames. ], tot_loss[loss=0.2038, simple_loss=0.2899, pruned_loss=0.05886, over 3033368.21 frames. ], batch size: 57, lr: 3.05e-03, grad_scale: 8.0 2023-05-01 12:02:28,580 INFO [train.py:904] (1/8) Epoch 22, batch 7250, loss[loss=0.2065, simple_loss=0.2878, pruned_loss=0.06261, over 15480.00 frames. ], tot_loss[loss=0.2017, simple_loss=0.2878, pruned_loss=0.05783, over 3030804.09 frames. ], batch size: 191, lr: 3.05e-03, grad_scale: 8.0 2023-05-01 12:02:30,892 INFO [optim.py:368] (1/8) 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 12:02:47,270 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.38 vs. limit=2.0 2023-05-01 12:02:53,037 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=220419.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 12:03:27,841 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.5020, 3.6148, 2.2099, 4.1682, 2.6595, 4.1060, 2.3018, 2.8374], device='cuda:1'), covar=tensor([0.0327, 0.0395, 0.1684, 0.0186, 0.0879, 0.0541, 0.1529, 0.0820], device='cuda:1'), in_proj_covar=tensor([0.0167, 0.0174, 0.0192, 0.0160, 0.0174, 0.0214, 0.0200, 0.0176], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-05-01 12:03:30,263 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 22, batch 7300, loss[loss=0.2007, simple_loss=0.2896, pruned_loss=0.05587, over 16654.00 frames. ], tot_loss[loss=0.2011, simple_loss=0.2873, pruned_loss=0.05741, over 3050611.00 frames. ], batch size: 134, lr: 3.05e-03, grad_scale: 8.0 2023-05-01 12:03:46,869 INFO [zipformer.py:625] (1/8) 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] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=220467.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 12:04:44,046 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.0338, 2.3535, 2.3286, 2.5186, 1.9558, 3.0851, 1.9177, 2.6628], device='cuda:1'), covar=tensor([0.1143, 0.0652, 0.1028, 0.0176, 0.0162, 0.0363, 0.1365, 0.0725], device='cuda:1'), in_proj_covar=tensor([0.0167, 0.0175, 0.0196, 0.0189, 0.0207, 0.0215, 0.0203, 0.0193], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-05-01 12:05:02,410 INFO [train.py:904] (1/8) Epoch 22, batch 7350, loss[loss=0.1993, simple_loss=0.2945, pruned_loss=0.05204, over 16818.00 frames. ], tot_loss[loss=0.2021, simple_loss=0.2882, pruned_loss=0.05802, over 3058383.65 frames. ], batch size: 83, lr: 3.05e-03, grad_scale: 8.0 2023-05-01 12:05:05,571 INFO [optim.py:368] (1/8) 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] (1/8) attn_weights_entropy = tensor([4.2582, 4.2300, 4.0817, 3.3007, 4.1713, 1.6869, 3.9194, 3.6851], device='cuda:1'), covar=tensor([0.0106, 0.0090, 0.0194, 0.0354, 0.0094, 0.2985, 0.0139, 0.0284], device='cuda:1'), in_proj_covar=tensor([0.0165, 0.0155, 0.0197, 0.0177, 0.0174, 0.0207, 0.0186, 0.0169], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-01 12:05:17,914 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 22, batch 7400, loss[loss=0.2022, simple_loss=0.2897, pruned_loss=0.05737, over 16211.00 frames. ], tot_loss[loss=0.2029, simple_loss=0.2894, pruned_loss=0.05823, over 3068399.89 frames. ], batch size: 35, lr: 3.05e-03, grad_scale: 8.0 2023-05-01 12:06:24,831 INFO [zipformer.py:625] (1/8) 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] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=220561.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 12:06:41,007 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.85 vs. limit=2.0 2023-05-01 12:06:58,934 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([1.6478, 2.5846, 2.2590, 3.8220, 2.5881, 3.7941, 1.5090, 2.6680], device='cuda:1'), covar=tensor([0.1478, 0.0891, 0.1401, 0.0174, 0.0229, 0.0452, 0.1807, 0.0936], device='cuda:1'), in_proj_covar=tensor([0.0168, 0.0176, 0.0197, 0.0191, 0.0208, 0.0217, 0.0204, 0.0194], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-05-01 12:07:02,589 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.8971, 2.7763, 2.7472, 2.1021, 2.6354, 2.1425, 2.7438, 2.9625], device='cuda:1'), covar=tensor([0.0308, 0.0846, 0.0586, 0.1887, 0.0887, 0.1003, 0.0618, 0.0775], device='cuda:1'), in_proj_covar=tensor([0.0156, 0.0165, 0.0168, 0.0154, 0.0146, 0.0131, 0.0144, 0.0177], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-05-01 12:07:13,078 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=220585.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 12:07:38,743 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.7958, 3.9691, 2.4893, 4.7712, 2.9808, 4.6161, 2.5213, 3.0587], device='cuda:1'), covar=tensor([0.0307, 0.0380, 0.1643, 0.0156, 0.0849, 0.0438, 0.1566, 0.0897], device='cuda:1'), in_proj_covar=tensor([0.0167, 0.0175, 0.0193, 0.0160, 0.0174, 0.0215, 0.0201, 0.0177], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-05-01 12:07:41,478 INFO [train.py:904] (1/8) Epoch 22, batch 7450, loss[loss=0.2056, simple_loss=0.302, pruned_loss=0.05459, over 15461.00 frames. ], tot_loss[loss=0.2044, simple_loss=0.2903, pruned_loss=0.0592, over 3064660.57 frames. ], batch size: 191, lr: 3.05e-03, grad_scale: 8.0 2023-05-01 12:07:43,936 INFO [optim.py:368] (1/8) 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 12:08:01,228 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=220614.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 12:08:04,706 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=220616.0, num_to_drop=1, layers_to_drop={2} 2023-05-01 12:09:00,773 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=220651.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 12:09:03,151 INFO [train.py:904] (1/8) Epoch 22, batch 7500, loss[loss=0.2364, simple_loss=0.3065, pruned_loss=0.08316, over 11472.00 frames. ], tot_loss[loss=0.2041, simple_loss=0.2905, pruned_loss=0.05882, over 3056593.50 frames. ], batch size: 248, lr: 3.05e-03, grad_scale: 8.0 2023-05-01 12:10:21,150 INFO [train.py:904] (1/8) Epoch 22, batch 7550, loss[loss=0.1817, simple_loss=0.2636, pruned_loss=0.04991, over 17063.00 frames. ], tot_loss[loss=0.2041, simple_loss=0.2897, pruned_loss=0.05927, over 3059566.03 frames. ], batch size: 55, lr: 3.05e-03, grad_scale: 8.0 2023-05-01 12:10:24,492 INFO [optim.py:368] (1/8) 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 12:10:35,487 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=220712.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 12:10:37,520 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=4.71 vs. limit=5.0 2023-05-01 12:11:10,609 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.2318, 1.5841, 1.9988, 2.2254, 2.3476, 2.5442, 1.7781, 2.4448], device='cuda:1'), covar=tensor([0.0244, 0.0526, 0.0321, 0.0360, 0.0315, 0.0187, 0.0510, 0.0162], device='cuda:1'), in_proj_covar=tensor([0.0184, 0.0189, 0.0176, 0.0180, 0.0193, 0.0150, 0.0193, 0.0148], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-01 12:11:23,180 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 22, batch 7600, loss[loss=0.2021, simple_loss=0.2817, pruned_loss=0.0612, over 16311.00 frames. ], tot_loss[loss=0.203, simple_loss=0.2886, pruned_loss=0.05866, over 3095718.80 frames. ], batch size: 165, lr: 3.05e-03, grad_scale: 8.0 2023-05-01 12:11:40,683 INFO [zipformer.py:625] (1/8) 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] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=220791.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 12:12:50,691 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.8263, 4.6703, 4.8628, 5.0371, 5.2378, 4.6024, 5.2120, 5.2084], device='cuda:1'), covar=tensor([0.1964, 0.1251, 0.1627, 0.0717, 0.0541, 0.1059, 0.0576, 0.0712], device='cuda:1'), in_proj_covar=tensor([0.0623, 0.0769, 0.0890, 0.0779, 0.0593, 0.0617, 0.0645, 0.0742], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-01 12:12:55,413 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=220802.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 12:12:56,366 INFO [train.py:904] (1/8) Epoch 22, batch 7650, loss[loss=0.1904, simple_loss=0.2794, pruned_loss=0.0507, over 16842.00 frames. ], tot_loss[loss=0.2046, simple_loss=0.2897, pruned_loss=0.05975, over 3079284.92 frames. ], batch size: 42, lr: 3.05e-03, grad_scale: 8.0 2023-05-01 12:12:59,168 INFO [optim.py:368] (1/8) 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 12:13:06,029 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.1038, 2.1977, 2.2369, 3.7786, 2.1452, 2.5141, 2.2903, 2.3588], device='cuda:1'), covar=tensor([0.1399, 0.3578, 0.3012, 0.0571, 0.4184, 0.2581, 0.3634, 0.3364], device='cuda:1'), in_proj_covar=tensor([0.0397, 0.0446, 0.0364, 0.0324, 0.0434, 0.0513, 0.0417, 0.0520], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-01 12:13:19,713 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.9379, 1.7776, 2.4272, 2.7616, 2.6937, 3.1384, 1.7950, 3.1275], device='cuda:1'), covar=tensor([0.0205, 0.0610, 0.0362, 0.0327, 0.0317, 0.0208, 0.0686, 0.0172], device='cuda:1'), in_proj_covar=tensor([0.0184, 0.0189, 0.0176, 0.0180, 0.0193, 0.0150, 0.0193, 0.0148], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-01 12:13:27,052 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=220822.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 12:14:11,871 INFO [train.py:904] (1/8) Epoch 22, batch 7700, loss[loss=0.1992, simple_loss=0.2837, pruned_loss=0.05733, over 16753.00 frames. ], tot_loss[loss=0.2049, simple_loss=0.2897, pruned_loss=0.06003, over 3079213.04 frames. ], batch size: 62, lr: 3.05e-03, grad_scale: 8.0 2023-05-01 12:14:18,164 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.0987, 1.5855, 1.9314, 2.0578, 2.2106, 2.3754, 1.7638, 2.2657], device='cuda:1'), covar=tensor([0.0249, 0.0490, 0.0290, 0.0363, 0.0321, 0.0214, 0.0525, 0.0186], device='cuda:1'), in_proj_covar=tensor([0.0185, 0.0190, 0.0177, 0.0181, 0.0194, 0.0151, 0.0194, 0.0149], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-01 12:14:47,158 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.82 vs. limit=2.0 2023-05-01 12:14:58,234 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=220883.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 12:15:01,407 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=220885.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 12:15:13,424 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=2.97 vs. limit=5.0 2023-05-01 12:15:29,148 INFO [train.py:904] (1/8) Epoch 22, batch 7750, loss[loss=0.1823, simple_loss=0.2785, pruned_loss=0.04304, over 16739.00 frames. ], tot_loss[loss=0.2048, simple_loss=0.2895, pruned_loss=0.06003, over 3066997.21 frames. ], batch size: 89, lr: 3.05e-03, grad_scale: 8.0 2023-05-01 12:15:30,818 INFO [zipformer.py:625] (1/8) 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] (1/8) 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] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=220911.0, num_to_drop=1, layers_to_drop={1} 2023-05-01 12:15:45,242 INFO [zipformer.py:625] (1/8) 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] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=220933.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 12:16:42,220 INFO [train.py:904] (1/8) Epoch 22, batch 7800, loss[loss=0.2027, simple_loss=0.2908, pruned_loss=0.05725, over 16831.00 frames. ], tot_loss[loss=0.2063, simple_loss=0.2907, pruned_loss=0.061, over 3064566.87 frames. ], batch size: 39, lr: 3.05e-03, grad_scale: 8.0 2023-05-01 12:16:56,567 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=220962.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 12:17:00,224 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 22, batch 7850, loss[loss=0.1758, simple_loss=0.2759, pruned_loss=0.03783, over 16809.00 frames. ], tot_loss[loss=0.2057, simple_loss=0.2907, pruned_loss=0.06036, over 3052725.66 frames. ], batch size: 102, lr: 3.05e-03, grad_scale: 8.0 2023-05-01 12:17:58,014 INFO [optim.py:368] (1/8) 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 12:18:02,914 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=221007.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 12:18:09,875 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.8037, 3.9118, 2.4806, 4.5442, 2.9416, 4.4649, 2.5344, 3.0815], device='cuda:1'), covar=tensor([0.0269, 0.0366, 0.1612, 0.0185, 0.0812, 0.0463, 0.1492, 0.0782], device='cuda:1'), in_proj_covar=tensor([0.0167, 0.0176, 0.0193, 0.0161, 0.0175, 0.0215, 0.0201, 0.0178], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-05-01 12:18:16,489 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.4946, 4.7970, 5.1017, 5.0709, 5.0582, 4.7439, 4.3961, 4.4713], device='cuda:1'), covar=tensor([0.0613, 0.0725, 0.0526, 0.0609, 0.0794, 0.0655, 0.1782, 0.0673], device='cuda:1'), in_proj_covar=tensor([0.0410, 0.0456, 0.0440, 0.0409, 0.0488, 0.0465, 0.0549, 0.0371], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:1') 2023-05-01 12:18:24,396 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.3950, 4.2600, 4.4605, 4.5768, 4.7775, 4.3167, 4.7171, 4.7658], device='cuda:1'), covar=tensor([0.1853, 0.1219, 0.1471, 0.0741, 0.0533, 0.1151, 0.0711, 0.0662], device='cuda:1'), in_proj_covar=tensor([0.0627, 0.0774, 0.0897, 0.0781, 0.0595, 0.0622, 0.0648, 0.0748], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-01 12:18:50,896 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=4.83 vs. limit=5.0 2023-05-01 12:18:53,624 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.5131, 3.6468, 2.2810, 4.1104, 2.7068, 4.0391, 2.2160, 2.7972], device='cuda:1'), covar=tensor([0.0292, 0.0411, 0.1669, 0.0203, 0.0877, 0.0598, 0.1681, 0.0879], device='cuda:1'), in_proj_covar=tensor([0.0167, 0.0175, 0.0193, 0.0161, 0.0176, 0.0215, 0.0201, 0.0178], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-05-01 12:19:09,446 INFO [train.py:904] (1/8) Epoch 22, batch 7900, loss[loss=0.1847, simple_loss=0.2808, pruned_loss=0.04431, over 16801.00 frames. ], tot_loss[loss=0.2049, simple_loss=0.2902, pruned_loss=0.05981, over 3058664.67 frames. ], batch size: 102, lr: 3.05e-03, grad_scale: 4.0 2023-05-01 12:19:16,622 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.22 vs. limit=2.0 2023-05-01 12:20:27,100 INFO [train.py:904] (1/8) Epoch 22, batch 7950, loss[loss=0.1885, simple_loss=0.2743, pruned_loss=0.05137, over 16747.00 frames. ], tot_loss[loss=0.2056, simple_loss=0.2908, pruned_loss=0.06014, over 3050643.13 frames. ], batch size: 76, lr: 3.05e-03, grad_scale: 4.0 2023-05-01 12:20:32,009 INFO [optim.py:368] (1/8) 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 12:21:30,209 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.8891, 2.1336, 2.3703, 3.1239, 2.1708, 2.3123, 2.3297, 2.2283], device='cuda:1'), covar=tensor([0.1304, 0.3193, 0.2441, 0.0710, 0.4193, 0.2225, 0.3118, 0.3339], device='cuda:1'), in_proj_covar=tensor([0.0398, 0.0446, 0.0363, 0.0325, 0.0436, 0.0514, 0.0418, 0.0519], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-01 12:21:41,819 INFO [train.py:904] (1/8) Epoch 22, batch 8000, loss[loss=0.2082, simple_loss=0.2952, pruned_loss=0.06064, over 17111.00 frames. ], tot_loss[loss=0.2073, simple_loss=0.2919, pruned_loss=0.06137, over 3034846.12 frames. ], batch size: 47, lr: 3.05e-03, grad_scale: 8.0 2023-05-01 12:22:18,530 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=221178.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 12:22:55,163 INFO [train.py:904] (1/8) Epoch 22, batch 8050, loss[loss=0.2503, simple_loss=0.3158, pruned_loss=0.09237, over 11944.00 frames. ], tot_loss[loss=0.2074, simple_loss=0.292, pruned_loss=0.06144, over 3031139.39 frames. ], batch size: 248, lr: 3.05e-03, grad_scale: 4.0 2023-05-01 12:23:01,254 INFO [optim.py:368] (1/8) 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 12:23:07,231 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=221211.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 12:23:26,659 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.8182, 2.7495, 2.7754, 2.1896, 2.6257, 2.1291, 2.6599, 2.9374], device='cuda:1'), covar=tensor([0.0317, 0.0794, 0.0555, 0.1787, 0.0890, 0.0977, 0.0673, 0.0765], device='cuda:1'), in_proj_covar=tensor([0.0154, 0.0163, 0.0165, 0.0152, 0.0144, 0.0129, 0.0142, 0.0174], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:1') 2023-05-01 12:24:08,786 INFO [train.py:904] (1/8) Epoch 22, batch 8100, loss[loss=0.2028, simple_loss=0.2983, pruned_loss=0.05359, over 16858.00 frames. ], tot_loss[loss=0.205, simple_loss=0.2905, pruned_loss=0.05978, over 3054949.10 frames. ], batch size: 90, lr: 3.05e-03, grad_scale: 4.0 2023-05-01 12:24:17,302 INFO [zipformer.py:625] (1/8) 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] (1/8) 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] (1/8) Epoch 22, batch 8150, loss[loss=0.1971, simple_loss=0.2793, pruned_loss=0.05742, over 16929.00 frames. ], tot_loss[loss=0.2028, simple_loss=0.2879, pruned_loss=0.05887, over 3061246.03 frames. ], batch size: 116, lr: 3.05e-03, grad_scale: 4.0 2023-05-01 12:25:31,011 INFO [optim.py:368] (1/8) 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 12:25:31,355 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=221307.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 12:26:30,163 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.2786, 1.5710, 1.9559, 2.1678, 2.3467, 2.5344, 1.7030, 2.4337], device='cuda:1'), covar=tensor([0.0259, 0.0521, 0.0357, 0.0418, 0.0322, 0.0204, 0.0579, 0.0147], device='cuda:1'), in_proj_covar=tensor([0.0185, 0.0190, 0.0176, 0.0181, 0.0194, 0.0151, 0.0193, 0.0149], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-01 12:26:41,082 INFO [train.py:904] (1/8) Epoch 22, batch 8200, loss[loss=0.1811, simple_loss=0.2723, pruned_loss=0.045, over 16829.00 frames. ], tot_loss[loss=0.2008, simple_loss=0.2855, pruned_loss=0.05805, over 3085880.79 frames. ], batch size: 90, lr: 3.05e-03, grad_scale: 4.0 2023-05-01 12:26:44,114 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 22, batch 8250, loss[loss=0.1737, simple_loss=0.2577, pruned_loss=0.04484, over 12076.00 frames. ], tot_loss[loss=0.197, simple_loss=0.2841, pruned_loss=0.05498, over 3071769.10 frames. ], batch size: 248, lr: 3.05e-03, grad_scale: 4.0 2023-05-01 12:28:05,605 INFO [optim.py:368] (1/8) Clipping_scale=2.0, grad-norm quartiles 1.564e+02 2.638e+02 3.060e+02 3.657e+02 7.056e+02, threshold=6.120e+02, percent-clipped=1.0 2023-05-01 12:28:31,775 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([5.4314, 5.8172, 5.5756, 5.6319, 5.2278, 5.2773, 5.1939, 5.9286], device='cuda:1'), covar=tensor([0.1267, 0.0790, 0.0982, 0.0784, 0.0933, 0.0643, 0.1181, 0.0823], device='cuda:1'), in_proj_covar=tensor([0.0666, 0.0809, 0.0672, 0.0614, 0.0509, 0.0524, 0.0678, 0.0633], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:1') 2023-05-01 12:29:17,813 INFO [train.py:904] (1/8) Epoch 22, batch 8300, loss[loss=0.1842, simple_loss=0.2789, pruned_loss=0.04478, over 15295.00 frames. ], tot_loss[loss=0.1929, simple_loss=0.2815, pruned_loss=0.05214, over 3070173.73 frames. ], batch size: 190, lr: 3.05e-03, grad_scale: 4.0 2023-05-01 12:29:44,895 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.6607, 4.8035, 4.5605, 4.1977, 4.0419, 4.7069, 4.5910, 4.3022], device='cuda:1'), covar=tensor([0.0687, 0.0647, 0.0401, 0.0414, 0.1299, 0.0566, 0.0431, 0.0828], device='cuda:1'), in_proj_covar=tensor([0.0287, 0.0424, 0.0337, 0.0332, 0.0342, 0.0388, 0.0232, 0.0402], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-01 12:29:57,550 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=221478.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 12:30:38,295 INFO [train.py:904] (1/8) Epoch 22, batch 8350, loss[loss=0.1872, simple_loss=0.2874, pruned_loss=0.04352, over 16420.00 frames. ], tot_loss[loss=0.1895, simple_loss=0.2797, pruned_loss=0.0497, over 3054855.70 frames. ], batch size: 146, lr: 3.05e-03, grad_scale: 4.0 2023-05-01 12:30:43,701 INFO [optim.py:368] (1/8) 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 12:30:44,190 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=221507.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 12:31:15,662 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=221526.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 12:31:56,139 INFO [train.py:904] (1/8) Epoch 22, batch 8400, loss[loss=0.1767, simple_loss=0.2703, pruned_loss=0.04155, over 16775.00 frames. ], tot_loss[loss=0.1871, simple_loss=0.2779, pruned_loss=0.04814, over 3052928.23 frames. ], batch size: 124, lr: 3.05e-03, grad_scale: 8.0 2023-05-01 12:32:05,129 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.36 vs. limit=2.0 2023-05-01 12:32:08,991 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=221560.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 12:32:21,044 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=221568.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 12:33:17,379 INFO [train.py:904] (1/8) Epoch 22, batch 8450, loss[loss=0.1764, simple_loss=0.2579, pruned_loss=0.04746, over 12371.00 frames. ], tot_loss[loss=0.1846, simple_loss=0.2759, pruned_loss=0.04665, over 3045377.81 frames. ], batch size: 248, lr: 3.04e-03, grad_scale: 8.0 2023-05-01 12:33:24,323 INFO [optim.py:368] (1/8) Clipping_scale=2.0, grad-norm quartiles 1.610e+02 2.296e+02 2.720e+02 3.439e+02 5.542e+02, threshold=5.440e+02, percent-clipped=2.0 2023-05-01 12:33:26,150 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=221608.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 12:34:26,704 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.3840, 3.2054, 3.4267, 1.8752, 3.5767, 3.6450, 2.9484, 2.8633], device='cuda:1'), covar=tensor([0.0793, 0.0283, 0.0227, 0.1217, 0.0088, 0.0164, 0.0428, 0.0460], device='cuda:1'), in_proj_covar=tensor([0.0145, 0.0105, 0.0095, 0.0135, 0.0078, 0.0121, 0.0126, 0.0127], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0004, 0.0003], device='cuda:1') 2023-05-01 12:34:38,832 INFO [train.py:904] (1/8) Epoch 22, batch 8500, loss[loss=0.1466, simple_loss=0.2306, pruned_loss=0.03129, over 12032.00 frames. ], tot_loss[loss=0.1806, simple_loss=0.2723, pruned_loss=0.04442, over 3048703.06 frames. ], batch size: 248, lr: 3.04e-03, grad_scale: 8.0 2023-05-01 12:36:02,530 INFO [train.py:904] (1/8) Epoch 22, batch 8550, loss[loss=0.17, simple_loss=0.2519, pruned_loss=0.04404, over 11870.00 frames. ], tot_loss[loss=0.1788, simple_loss=0.2703, pruned_loss=0.04371, over 3030871.31 frames. ], batch size: 246, lr: 3.04e-03, grad_scale: 8.0 2023-05-01 12:36:10,060 INFO [optim.py:368] (1/8) Clipping_scale=2.0, grad-norm quartiles 1.538e+02 2.073e+02 2.527e+02 2.950e+02 5.682e+02, threshold=5.053e+02, percent-clipped=1.0 2023-05-01 12:37:41,420 INFO [train.py:904] (1/8) Epoch 22, batch 8600, loss[loss=0.1758, simple_loss=0.2727, pruned_loss=0.03942, over 15409.00 frames. ], tot_loss[loss=0.1782, simple_loss=0.2709, pruned_loss=0.04275, over 3042469.60 frames. ], batch size: 191, lr: 3.04e-03, grad_scale: 8.0 2023-05-01 12:37:53,183 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=221758.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 12:38:37,153 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.3698, 2.0841, 2.7181, 3.2201, 3.0442, 3.6808, 2.0615, 3.7228], device='cuda:1'), covar=tensor([0.0153, 0.0621, 0.0340, 0.0260, 0.0266, 0.0150, 0.0726, 0.0117], device='cuda:1'), in_proj_covar=tensor([0.0184, 0.0190, 0.0176, 0.0180, 0.0192, 0.0150, 0.0192, 0.0147], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-01 12:38:59,197 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.4721, 1.9763, 1.7470, 1.6960, 2.2583, 1.9385, 1.9204, 2.3623], device='cuda:1'), covar=tensor([0.0199, 0.0443, 0.0570, 0.0528, 0.0310, 0.0421, 0.0206, 0.0312], device='cuda:1'), in_proj_covar=tensor([0.0202, 0.0227, 0.0221, 0.0220, 0.0228, 0.0226, 0.0226, 0.0221], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-01 12:39:21,614 INFO [train.py:904] (1/8) Epoch 22, batch 8650, loss[loss=0.1628, simple_loss=0.2533, pruned_loss=0.03613, over 11989.00 frames. ], tot_loss[loss=0.1754, simple_loss=0.2689, pruned_loss=0.04095, over 3053848.11 frames. ], batch size: 247, lr: 3.04e-03, grad_scale: 8.0 2023-05-01 12:39:30,557 INFO [optim.py:368] (1/8) 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 12:39:49,403 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.5355, 3.6501, 3.3985, 3.1432, 3.1424, 3.5525, 3.3238, 3.3116], device='cuda:1'), covar=tensor([0.0664, 0.0688, 0.0360, 0.0327, 0.0642, 0.0520, 0.1540, 0.0598], device='cuda:1'), in_proj_covar=tensor([0.0286, 0.0420, 0.0334, 0.0330, 0.0338, 0.0384, 0.0230, 0.0399], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-01 12:39:58,374 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 22, batch 8700, loss[loss=0.1575, simple_loss=0.2479, pruned_loss=0.03354, over 12604.00 frames. ], tot_loss[loss=0.1733, simple_loss=0.2664, pruned_loss=0.04016, over 3043593.68 frames. ], batch size: 248, lr: 3.04e-03, grad_scale: 8.0 2023-05-01 12:41:22,087 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.77 vs. limit=2.0 2023-05-01 12:41:27,670 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=221863.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 12:41:36,034 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.23 vs. limit=2.0 2023-05-01 12:42:16,275 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=221890.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 12:42:42,867 INFO [train.py:904] (1/8) Epoch 22, batch 8750, loss[loss=0.1856, simple_loss=0.2816, pruned_loss=0.0448, over 15219.00 frames. ], tot_loss[loss=0.1727, simple_loss=0.2661, pruned_loss=0.03965, over 3053756.94 frames. ], batch size: 190, lr: 3.04e-03, grad_scale: 8.0 2023-05-01 12:42:53,174 INFO [optim.py:368] (1/8) Clipping_scale=2.0, grad-norm quartiles 1.489e+02 2.058e+02 2.561e+02 3.126e+02 5.404e+02, threshold=5.122e+02, percent-clipped=2.0 2023-05-01 12:43:04,140 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.2138, 2.1573, 2.1024, 3.8478, 2.0693, 2.5081, 2.3149, 2.3054], device='cuda:1'), covar=tensor([0.1290, 0.3712, 0.3327, 0.0554, 0.4614, 0.2754, 0.3568, 0.3573], device='cuda:1'), in_proj_covar=tensor([0.0393, 0.0440, 0.0361, 0.0318, 0.0429, 0.0506, 0.0411, 0.0512], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-01 12:44:00,403 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.89 vs. limit=2.0 2023-05-01 12:44:31,209 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=221951.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 12:44:34,623 INFO [train.py:904] (1/8) Epoch 22, batch 8800, loss[loss=0.1812, simple_loss=0.2693, pruned_loss=0.0465, over 15315.00 frames. ], tot_loss[loss=0.1716, simple_loss=0.2654, pruned_loss=0.03893, over 3059400.14 frames. ], batch size: 190, lr: 3.04e-03, grad_scale: 8.0 2023-05-01 12:44:38,277 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.9984, 2.7785, 2.8607, 2.1356, 2.6028, 2.2205, 2.7230, 2.9771], device='cuda:1'), covar=tensor([0.0425, 0.1015, 0.0581, 0.1897, 0.0983, 0.0947, 0.0834, 0.0870], device='cuda:1'), in_proj_covar=tensor([0.0152, 0.0159, 0.0163, 0.0151, 0.0142, 0.0127, 0.0140, 0.0171], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:1') 2023-05-01 12:44:48,471 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([1.7325, 2.5432, 2.2004, 4.0833, 2.3054, 3.9218, 1.5651, 2.6574], device='cuda:1'), covar=tensor([0.1424, 0.0926, 0.1439, 0.0175, 0.0127, 0.0511, 0.1729, 0.0913], device='cuda:1'), in_proj_covar=tensor([0.0166, 0.0172, 0.0192, 0.0185, 0.0202, 0.0211, 0.0200, 0.0190], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-05-01 12:46:00,469 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.2144, 4.3735, 4.4919, 4.2586, 4.3590, 4.8494, 4.4142, 4.1409], device='cuda:1'), covar=tensor([0.1597, 0.1616, 0.1601, 0.2002, 0.2511, 0.0918, 0.1531, 0.2319], device='cuda:1'), in_proj_covar=tensor([0.0394, 0.0570, 0.0632, 0.0472, 0.0626, 0.0661, 0.0498, 0.0635], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-05-01 12:46:22,303 INFO [train.py:904] (1/8) Epoch 22, batch 8850, loss[loss=0.1499, simple_loss=0.2428, pruned_loss=0.02855, over 12313.00 frames. ], tot_loss[loss=0.1728, simple_loss=0.2687, pruned_loss=0.03843, over 3069442.38 frames. ], batch size: 247, lr: 3.04e-03, grad_scale: 8.0 2023-05-01 12:46:28,904 INFO [optim.py:368] (1/8) 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 12:46:35,322 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=222009.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 12:46:51,886 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.2143, 3.2589, 1.9725, 3.6429, 2.4350, 3.5835, 2.1689, 2.6382], device='cuda:1'), covar=tensor([0.0326, 0.0372, 0.1721, 0.0230, 0.0887, 0.0509, 0.1582, 0.0827], device='cuda:1'), in_proj_covar=tensor([0.0163, 0.0170, 0.0187, 0.0155, 0.0171, 0.0208, 0.0197, 0.0173], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0003, 0.0004, 0.0004, 0.0003], device='cuda:1') 2023-05-01 12:48:07,872 INFO [train.py:904] (1/8) Epoch 22, batch 8900, loss[loss=0.16, simple_loss=0.2535, pruned_loss=0.03325, over 12719.00 frames. ], tot_loss[loss=0.1718, simple_loss=0.2683, pruned_loss=0.03768, over 3060256.78 frames. ], batch size: 247, lr: 3.04e-03, grad_scale: 4.0 2023-05-01 12:48:41,318 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.6852, 3.8247, 2.3490, 4.3484, 2.9319, 4.2200, 2.4347, 3.0729], device='cuda:1'), covar=tensor([0.0281, 0.0348, 0.1583, 0.0202, 0.0785, 0.0448, 0.1588, 0.0705], device='cuda:1'), in_proj_covar=tensor([0.0163, 0.0171, 0.0188, 0.0155, 0.0171, 0.0208, 0.0198, 0.0174], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0003, 0.0004, 0.0004, 0.0003], device='cuda:1') 2023-05-01 12:48:42,837 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=222070.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 12:49:35,848 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.9828, 2.0587, 2.1617, 3.4547, 2.0871, 2.3435, 2.2011, 2.1887], device='cuda:1'), covar=tensor([0.1379, 0.3799, 0.3114, 0.0645, 0.4376, 0.2650, 0.3700, 0.3659], device='cuda:1'), in_proj_covar=tensor([0.0393, 0.0440, 0.0362, 0.0319, 0.0430, 0.0506, 0.0413, 0.0513], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-01 12:50:04,285 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.3769, 2.0115, 1.7738, 1.7630, 2.3099, 1.9910, 1.9449, 2.3722], device='cuda:1'), covar=tensor([0.0173, 0.0424, 0.0572, 0.0505, 0.0270, 0.0364, 0.0208, 0.0274], device='cuda:1'), in_proj_covar=tensor([0.0203, 0.0230, 0.0223, 0.0222, 0.0230, 0.0229, 0.0227, 0.0222], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-01 12:50:11,677 INFO [train.py:904] (1/8) Epoch 22, batch 8950, loss[loss=0.1605, simple_loss=0.2623, pruned_loss=0.02936, over 16330.00 frames. ], tot_loss[loss=0.1712, simple_loss=0.2674, pruned_loss=0.03751, over 3081614.84 frames. ], batch size: 146, lr: 3.04e-03, grad_scale: 4.0 2023-05-01 12:50:23,609 INFO [optim.py:368] (1/8) 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 12:50:37,675 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=222114.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 12:51:44,873 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.1711, 3.2123, 1.7394, 3.4856, 2.3770, 3.4360, 1.9564, 2.6508], device='cuda:1'), covar=tensor([0.0305, 0.0398, 0.1990, 0.0214, 0.0894, 0.0608, 0.1805, 0.0741], device='cuda:1'), in_proj_covar=tensor([0.0163, 0.0171, 0.0188, 0.0155, 0.0171, 0.0209, 0.0198, 0.0173], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0003, 0.0004, 0.0004, 0.0003], device='cuda:1') 2023-05-01 12:52:03,798 INFO [train.py:904] (1/8) Epoch 22, batch 9000, loss[loss=0.1466, simple_loss=0.243, pruned_loss=0.0251, over 16205.00 frames. ], tot_loss[loss=0.1681, simple_loss=0.2641, pruned_loss=0.03607, over 3094907.72 frames. ], batch size: 165, lr: 3.04e-03, grad_scale: 4.0 2023-05-01 12:52:03,798 INFO [train.py:929] (1/8) Computing validation loss 2023-05-01 12:52:14,708 INFO [train.py:938] (1/8) Epoch 22, validation: loss=0.1453, simple_loss=0.2492, pruned_loss=0.0207, over 944034.00 frames. 2023-05-01 12:52:14,709 INFO [train.py:939] (1/8) Maximum memory allocated so far is 18145MB 2023-05-01 12:52:36,448 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=222163.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 12:53:22,230 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=222185.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 12:53:58,628 INFO [train.py:904] (1/8) Epoch 22, batch 9050, loss[loss=0.1499, simple_loss=0.2477, pruned_loss=0.02602, over 17136.00 frames. ], tot_loss[loss=0.1682, simple_loss=0.264, pruned_loss=0.03618, over 3091954.60 frames. ], batch size: 47, lr: 3.04e-03, grad_scale: 4.0 2023-05-01 12:54:04,046 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=222205.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 12:54:09,040 INFO [optim.py:368] (1/8) 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 12:54:17,174 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=222211.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 12:54:18,921 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.0824, 3.0953, 1.9463, 3.3307, 2.3399, 3.3233, 2.1718, 2.6364], device='cuda:1'), covar=tensor([0.0285, 0.0384, 0.1560, 0.0228, 0.0808, 0.0540, 0.1486, 0.0690], device='cuda:1'), in_proj_covar=tensor([0.0163, 0.0170, 0.0187, 0.0155, 0.0170, 0.0208, 0.0197, 0.0173], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0003, 0.0004, 0.0004, 0.0003], device='cuda:1') 2023-05-01 12:54:30,545 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.96 vs. limit=2.0 2023-05-01 12:55:31,425 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=222246.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 12:55:31,568 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=222246.0, num_to_drop=1, layers_to_drop={0} 2023-05-01 12:55:44,664 INFO [train.py:904] (1/8) Epoch 22, batch 9100, loss[loss=0.1704, simple_loss=0.2689, pruned_loss=0.03598, over 16845.00 frames. ], tot_loss[loss=0.1695, simple_loss=0.2644, pruned_loss=0.03734, over 3089746.12 frames. ], batch size: 124, lr: 3.04e-03, grad_scale: 4.0 2023-05-01 12:56:11,338 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 22, batch 9150, loss[loss=0.1582, simple_loss=0.2558, pruned_loss=0.03035, over 16968.00 frames. ], tot_loss[loss=0.1691, simple_loss=0.2645, pruned_loss=0.03687, over 3074855.29 frames. ], batch size: 109, lr: 3.04e-03, grad_scale: 4.0 2023-05-01 12:57:53,761 INFO [optim.py:368] (1/8) 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 12:59:28,938 INFO [train.py:904] (1/8) Epoch 22, batch 9200, loss[loss=0.1462, simple_loss=0.2308, pruned_loss=0.03083, over 12558.00 frames. ], tot_loss[loss=0.1657, simple_loss=0.26, pruned_loss=0.03568, over 3087134.45 frames. ], batch size: 247, lr: 3.04e-03, grad_scale: 8.0 2023-05-01 12:59:52,287 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=222365.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 13:00:06,048 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=3.05 vs. limit=5.0 2023-05-01 13:00:13,798 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.4542, 3.3973, 3.4917, 3.5620, 3.5968, 3.3049, 3.5732, 3.6482], device='cuda:1'), covar=tensor([0.1274, 0.0980, 0.0958, 0.0638, 0.0669, 0.2548, 0.0896, 0.0829], device='cuda:1'), in_proj_covar=tensor([0.0607, 0.0752, 0.0868, 0.0760, 0.0579, 0.0604, 0.0630, 0.0728], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-01 13:00:36,351 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.59 vs. limit=2.0 2023-05-01 13:01:05,892 INFO [train.py:904] (1/8) Epoch 22, batch 9250, loss[loss=0.1584, simple_loss=0.2537, pruned_loss=0.03158, over 16226.00 frames. ], tot_loss[loss=0.1651, simple_loss=0.2593, pruned_loss=0.03548, over 3066988.82 frames. ], batch size: 165, lr: 3.04e-03, grad_scale: 8.0 2023-05-01 13:01:16,244 INFO [optim.py:368] (1/8) 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 13:01:28,803 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=222414.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 13:02:56,983 INFO [train.py:904] (1/8) Epoch 22, batch 9300, loss[loss=0.1542, simple_loss=0.2519, pruned_loss=0.02818, over 15686.00 frames. ], tot_loss[loss=0.1643, simple_loss=0.2582, pruned_loss=0.03522, over 3057639.14 frames. ], batch size: 194, lr: 3.04e-03, grad_scale: 8.0 2023-05-01 13:03:16,990 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=222462.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 13:04:40,773 INFO [train.py:904] (1/8) Epoch 22, batch 9350, loss[loss=0.1565, simple_loss=0.2416, pruned_loss=0.03572, over 12387.00 frames. ], tot_loss[loss=0.1642, simple_loss=0.2578, pruned_loss=0.03528, over 3053614.18 frames. ], batch size: 248, lr: 3.04e-03, grad_scale: 8.0 2023-05-01 13:04:49,912 INFO [optim.py:368] (1/8) 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 13:05:23,336 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=222523.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 13:05:57,376 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=222541.0, num_to_drop=1, layers_to_drop={2} 2023-05-01 13:06:08,605 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=222546.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 13:06:20,195 INFO [train.py:904] (1/8) Epoch 22, batch 9400, loss[loss=0.1344, simple_loss=0.2228, pruned_loss=0.02299, over 12461.00 frames. ], tot_loss[loss=0.1639, simple_loss=0.2576, pruned_loss=0.0351, over 3026550.73 frames. ], batch size: 247, lr: 3.04e-03, grad_scale: 8.0 2023-05-01 13:06:38,148 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=222561.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 13:06:55,383 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=222570.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 13:07:24,532 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=222584.0, num_to_drop=1, layers_to_drop={3} 2023-05-01 13:07:24,586 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.5796, 3.5658, 2.6961, 2.1463, 2.2172, 2.3073, 3.7694, 3.0855], device='cuda:1'), covar=tensor([0.2892, 0.0651, 0.1836, 0.3077, 0.3123, 0.2260, 0.0427, 0.1422], device='cuda:1'), in_proj_covar=tensor([0.0316, 0.0258, 0.0295, 0.0303, 0.0283, 0.0252, 0.0286, 0.0323], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:1') 2023-05-01 13:07:43,445 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=222594.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 13:07:56,646 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.5976, 4.7949, 4.9452, 4.7341, 4.8739, 5.3067, 4.8462, 4.5474], device='cuda:1'), covar=tensor([0.1142, 0.1627, 0.1743, 0.1892, 0.2118, 0.0807, 0.1269, 0.2269], device='cuda:1'), in_proj_covar=tensor([0.0389, 0.0563, 0.0624, 0.0466, 0.0622, 0.0652, 0.0493, 0.0628], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-05-01 13:08:00,593 INFO [train.py:904] (1/8) Epoch 22, batch 9450, loss[loss=0.1871, simple_loss=0.2737, pruned_loss=0.05024, over 12477.00 frames. ], tot_loss[loss=0.1656, simple_loss=0.2603, pruned_loss=0.03544, over 3046488.40 frames. ], batch size: 248, lr: 3.04e-03, grad_scale: 8.0 2023-05-01 13:08:08,412 INFO [optim.py:368] (1/8) 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 13:08:56,849 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=222631.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 13:09:04,063 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.1508, 2.3959, 2.1079, 2.2525, 2.7703, 2.4224, 2.5767, 2.9126], device='cuda:1'), covar=tensor([0.0159, 0.0477, 0.0568, 0.0540, 0.0300, 0.0453, 0.0248, 0.0277], device='cuda:1'), in_proj_covar=tensor([0.0201, 0.0228, 0.0221, 0.0220, 0.0229, 0.0228, 0.0224, 0.0220], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-01 13:09:40,759 INFO [train.py:904] (1/8) Epoch 22, batch 9500, loss[loss=0.15, simple_loss=0.2393, pruned_loss=0.03032, over 12689.00 frames. ], tot_loss[loss=0.1648, simple_loss=0.2593, pruned_loss=0.0351, over 3042186.99 frames. ], batch size: 249, lr: 3.04e-03, grad_scale: 8.0 2023-05-01 13:09:55,609 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.3947, 3.2849, 3.3332, 3.5268, 3.5428, 3.2590, 3.5501, 3.5777], device='cuda:1'), covar=tensor([0.1386, 0.1238, 0.1530, 0.0805, 0.0869, 0.3185, 0.1248, 0.1033], device='cuda:1'), in_proj_covar=tensor([0.0608, 0.0754, 0.0871, 0.0760, 0.0579, 0.0604, 0.0633, 0.0728], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-01 13:10:06,941 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=222665.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 13:10:59,209 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.1693, 3.4982, 3.4849, 2.4765, 3.2254, 3.5326, 3.3585, 2.1345], device='cuda:1'), covar=tensor([0.0533, 0.0051, 0.0055, 0.0356, 0.0096, 0.0095, 0.0080, 0.0457], device='cuda:1'), in_proj_covar=tensor([0.0132, 0.0080, 0.0081, 0.0130, 0.0095, 0.0106, 0.0091, 0.0125], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-05-01 13:11:22,397 INFO [train.py:904] (1/8) Epoch 22, batch 9550, loss[loss=0.1585, simple_loss=0.256, pruned_loss=0.03047, over 16711.00 frames. ], tot_loss[loss=0.1646, simple_loss=0.2591, pruned_loss=0.0351, over 3057829.05 frames. ], batch size: 76, lr: 3.04e-03, grad_scale: 8.0 2023-05-01 13:11:34,507 INFO [optim.py:368] (1/8) 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] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=222713.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 13:12:35,617 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.2480, 3.2905, 1.9370, 3.6058, 2.3815, 3.5551, 2.1488, 2.7716], device='cuda:1'), covar=tensor([0.0334, 0.0428, 0.1769, 0.0277, 0.0939, 0.0642, 0.1647, 0.0760], device='cuda:1'), in_proj_covar=tensor([0.0164, 0.0170, 0.0186, 0.0155, 0.0171, 0.0207, 0.0197, 0.0173], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0003, 0.0004, 0.0004, 0.0003], device='cuda:1') 2023-05-01 13:13:00,661 INFO [train.py:904] (1/8) Epoch 22, batch 9600, loss[loss=0.1604, simple_loss=0.2488, pruned_loss=0.03604, over 12388.00 frames. ], tot_loss[loss=0.1662, simple_loss=0.2606, pruned_loss=0.03589, over 3040105.80 frames. ], batch size: 247, lr: 3.04e-03, grad_scale: 8.0 2023-05-01 13:13:21,166 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.8907, 2.1402, 2.3332, 3.1917, 2.1651, 2.3498, 2.3062, 2.2078], device='cuda:1'), covar=tensor([0.1347, 0.3442, 0.2805, 0.0721, 0.4366, 0.2562, 0.3533, 0.3673], device='cuda:1'), in_proj_covar=tensor([0.0391, 0.0438, 0.0361, 0.0318, 0.0430, 0.0503, 0.0410, 0.0511], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-01 13:14:44,335 INFO [train.py:904] (1/8) Epoch 22, batch 9650, loss[loss=0.1536, simple_loss=0.2453, pruned_loss=0.03097, over 12274.00 frames. ], tot_loss[loss=0.1673, simple_loss=0.2618, pruned_loss=0.03637, over 3019549.95 frames. ], batch size: 248, lr: 3.04e-03, grad_scale: 4.0 2023-05-01 13:14:58,747 INFO [optim.py:368] (1/8) 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 13:16:03,740 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=222841.0, num_to_drop=1, layers_to_drop={1} 2023-05-01 13:16:27,395 INFO [train.py:904] (1/8) Epoch 22, batch 9700, loss[loss=0.1728, simple_loss=0.2557, pruned_loss=0.04492, over 12235.00 frames. ], tot_loss[loss=0.1666, simple_loss=0.2609, pruned_loss=0.03618, over 3019971.54 frames. ], batch size: 250, lr: 3.04e-03, grad_scale: 4.0 2023-05-01 13:16:43,306 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=222861.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 13:16:51,133 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.6962, 5.0001, 4.8389, 4.8228, 4.4967, 4.5014, 4.3997, 5.0618], device='cuda:1'), covar=tensor([0.1167, 0.0910, 0.0973, 0.0849, 0.0827, 0.1091, 0.1232, 0.0809], device='cuda:1'), in_proj_covar=tensor([0.0640, 0.0776, 0.0638, 0.0590, 0.0491, 0.0505, 0.0650, 0.0608], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:1') 2023-05-01 13:16:53,511 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([1.7903, 2.5587, 2.2441, 3.5578, 1.9270, 3.6166, 1.5664, 2.8110], device='cuda:1'), covar=tensor([0.1383, 0.0752, 0.1324, 0.0165, 0.0093, 0.0350, 0.1763, 0.0756], device='cuda:1'), in_proj_covar=tensor([0.0164, 0.0170, 0.0190, 0.0180, 0.0196, 0.0208, 0.0198, 0.0188], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0003], device='cuda:1') 2023-05-01 13:17:00,282 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.25 vs. limit=2.0 2023-05-01 13:17:22,890 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=222879.0, num_to_drop=1, layers_to_drop={3} 2023-05-01 13:17:42,990 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 22, batch 9750, loss[loss=0.1613, simple_loss=0.2617, pruned_loss=0.03044, over 16245.00 frames. ], tot_loss[loss=0.1665, simple_loss=0.2602, pruned_loss=0.03643, over 3029373.54 frames. ], batch size: 165, lr: 3.04e-03, grad_scale: 4.0 2023-05-01 13:18:18,157 INFO [optim.py:368] (1/8) 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] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=222909.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 13:18:52,118 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=222926.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 13:19:24,046 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=222940.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 13:19:45,653 INFO [train.py:904] (1/8) Epoch 22, batch 9800, loss[loss=0.1462, simple_loss=0.2496, pruned_loss=0.02146, over 16597.00 frames. ], tot_loss[loss=0.1658, simple_loss=0.2603, pruned_loss=0.03566, over 3038012.55 frames. ], batch size: 62, lr: 3.04e-03, grad_scale: 4.0 2023-05-01 13:20:00,895 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.7128, 4.7068, 5.0932, 5.0592, 5.0604, 4.7866, 4.6938, 4.6264], device='cuda:1'), covar=tensor([0.0291, 0.0569, 0.0363, 0.0363, 0.0507, 0.0412, 0.1066, 0.0410], device='cuda:1'), in_proj_covar=tensor([0.0394, 0.0437, 0.0425, 0.0393, 0.0468, 0.0446, 0.0526, 0.0358], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:1') 2023-05-01 13:20:15,146 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=222969.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 13:20:40,843 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=222984.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 13:21:03,232 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.4025, 3.4730, 3.6690, 3.6409, 3.6734, 3.4948, 3.5290, 3.5454], device='cuda:1'), covar=tensor([0.0389, 0.0781, 0.0491, 0.0484, 0.0515, 0.0598, 0.0812, 0.0484], device='cuda:1'), in_proj_covar=tensor([0.0393, 0.0436, 0.0424, 0.0393, 0.0467, 0.0445, 0.0526, 0.0357], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:1') 2023-05-01 13:21:23,014 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=223001.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 13:21:26,485 INFO [train.py:904] (1/8) Epoch 22, batch 9850, loss[loss=0.1651, simple_loss=0.2557, pruned_loss=0.03725, over 12629.00 frames. ], tot_loss[loss=0.1662, simple_loss=0.2615, pruned_loss=0.03541, over 3047878.88 frames. ], batch size: 248, lr: 3.04e-03, grad_scale: 4.0 2023-05-01 13:21:37,440 INFO [optim.py:368] (1/8) 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 13:22:19,872 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=223030.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 13:22:54,351 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.80 vs. limit=2.0 2023-05-01 13:23:01,421 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 22, batch 9900, loss[loss=0.1628, simple_loss=0.2482, pruned_loss=0.0387, over 12594.00 frames. ], tot_loss[loss=0.1659, simple_loss=0.2614, pruned_loss=0.03514, over 3045359.94 frames. ], batch size: 248, lr: 3.03e-03, grad_scale: 4.0 2023-05-01 13:23:43,521 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=223064.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 13:24:54,655 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.5175, 4.6695, 4.8323, 4.5807, 4.6668, 5.1935, 4.7063, 4.4258], device='cuda:1'), covar=tensor([0.1293, 0.1781, 0.2124, 0.1993, 0.2411, 0.0893, 0.1523, 0.2415], device='cuda:1'), in_proj_covar=tensor([0.0383, 0.0558, 0.0616, 0.0460, 0.0616, 0.0642, 0.0487, 0.0618], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-05-01 13:25:13,509 INFO [train.py:904] (1/8) Epoch 22, batch 9950, loss[loss=0.1604, simple_loss=0.2517, pruned_loss=0.0345, over 16712.00 frames. ], tot_loss[loss=0.1677, simple_loss=0.2641, pruned_loss=0.03558, over 3062174.70 frames. ], batch size: 62, lr: 3.03e-03, grad_scale: 4.0 2023-05-01 13:25:27,892 INFO [optim.py:368] (1/8) 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 13:26:07,061 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=223125.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 13:26:32,550 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.0162, 3.0078, 2.6458, 5.0396, 3.7430, 4.3885, 1.8766, 3.1478], device='cuda:1'), covar=tensor([0.1228, 0.0695, 0.1169, 0.0110, 0.0153, 0.0378, 0.1464, 0.0722], device='cuda:1'), in_proj_covar=tensor([0.0164, 0.0170, 0.0190, 0.0180, 0.0195, 0.0208, 0.0198, 0.0188], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0003], device='cuda:1') 2023-05-01 13:26:44,820 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.9273, 4.0574, 4.3730, 2.5791, 3.6372, 3.0255, 4.1930, 4.1946], device='cuda:1'), covar=tensor([0.0148, 0.0685, 0.0452, 0.1735, 0.0595, 0.0766, 0.0483, 0.0867], device='cuda:1'), in_proj_covar=tensor([0.0152, 0.0156, 0.0162, 0.0150, 0.0141, 0.0126, 0.0139, 0.0169], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:1') 2023-05-01 13:27:13,207 INFO [train.py:904] (1/8) Epoch 22, batch 10000, loss[loss=0.1803, simple_loss=0.2673, pruned_loss=0.04664, over 12959.00 frames. ], tot_loss[loss=0.1674, simple_loss=0.2633, pruned_loss=0.03575, over 3061311.33 frames. ], batch size: 248, lr: 3.03e-03, grad_scale: 8.0 2023-05-01 13:27:23,683 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.3823, 1.5646, 2.0338, 2.3338, 2.4041, 2.5774, 1.8145, 2.4360], device='cuda:1'), covar=tensor([0.0212, 0.0545, 0.0338, 0.0336, 0.0322, 0.0203, 0.0548, 0.0164], device='cuda:1'), in_proj_covar=tensor([0.0183, 0.0187, 0.0175, 0.0177, 0.0191, 0.0147, 0.0191, 0.0145], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-01 13:27:51,698 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=223173.0, num_to_drop=1, layers_to_drop={1} 2023-05-01 13:28:04,853 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=223179.0, num_to_drop=1, layers_to_drop={1} 2023-05-01 13:28:54,091 INFO [train.py:904] (1/8) Epoch 22, batch 10050, loss[loss=0.1754, simple_loss=0.2781, pruned_loss=0.03639, over 16179.00 frames. ], tot_loss[loss=0.1674, simple_loss=0.2634, pruned_loss=0.03564, over 3077426.13 frames. ], batch size: 165, lr: 3.03e-03, grad_scale: 8.0 2023-05-01 13:29:04,328 INFO [optim.py:368] (1/8) 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 13:29:28,246 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.50 vs. limit=2.0 2023-05-01 13:29:34,282 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.2913, 3.7583, 3.7673, 2.5687, 3.3714, 3.7654, 3.5387, 2.2800], device='cuda:1'), covar=tensor([0.0540, 0.0049, 0.0049, 0.0381, 0.0110, 0.0083, 0.0079, 0.0470], device='cuda:1'), in_proj_covar=tensor([0.0134, 0.0081, 0.0082, 0.0131, 0.0096, 0.0107, 0.0092, 0.0127], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-05-01 13:29:40,789 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=223226.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 13:29:42,094 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=223227.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 13:29:54,210 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=223234.0, num_to_drop=1, layers_to_drop={0} 2023-05-01 13:30:06,749 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.8575, 3.8656, 4.1521, 4.1236, 4.1230, 3.9361, 3.8929, 3.9659], device='cuda:1'), covar=tensor([0.0506, 0.1841, 0.0698, 0.0636, 0.0636, 0.0718, 0.1002, 0.0741], device='cuda:1'), in_proj_covar=tensor([0.0391, 0.0435, 0.0424, 0.0391, 0.0467, 0.0443, 0.0523, 0.0356], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:1') 2023-05-01 13:30:27,327 INFO [train.py:904] (1/8) Epoch 22, batch 10100, loss[loss=0.1594, simple_loss=0.2563, pruned_loss=0.03128, over 16346.00 frames. ], tot_loss[loss=0.1674, simple_loss=0.2634, pruned_loss=0.03572, over 3069344.45 frames. ], batch size: 146, lr: 3.03e-03, grad_scale: 8.0 2023-05-01 13:31:10,810 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=223274.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 13:31:32,869 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.73 vs. limit=2.0 2023-05-01 13:31:39,305 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=223296.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 13:31:42,349 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([1.9018, 2.7470, 2.6674, 1.9390, 2.5624, 2.7947, 2.5977, 1.7893], device='cuda:1'), covar=tensor([0.0457, 0.0093, 0.0097, 0.0395, 0.0149, 0.0117, 0.0139, 0.0584], device='cuda:1'), in_proj_covar=tensor([0.0134, 0.0081, 0.0082, 0.0131, 0.0096, 0.0106, 0.0092, 0.0128], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-05-01 13:31:44,081 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.1251, 2.5840, 2.6634, 1.9554, 2.8263, 2.8902, 2.4665, 2.5134], device='cuda:1'), covar=tensor([0.0626, 0.0236, 0.0226, 0.0959, 0.0117, 0.0200, 0.0466, 0.0420], device='cuda:1'), in_proj_covar=tensor([0.0143, 0.0103, 0.0092, 0.0135, 0.0077, 0.0119, 0.0124, 0.0126], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:1') 2023-05-01 13:32:13,429 INFO [train.py:904] (1/8) Epoch 23, batch 0, loss[loss=0.2251, simple_loss=0.2988, pruned_loss=0.07568, over 16753.00 frames. ], tot_loss[loss=0.2251, simple_loss=0.2988, pruned_loss=0.07568, over 16753.00 frames. ], batch size: 134, lr: 2.97e-03, grad_scale: 8.0 2023-05-01 13:32:13,430 INFO [train.py:929] (1/8) Computing validation loss 2023-05-01 13:32:20,846 INFO [train.py:938] (1/8) Epoch 23, validation: loss=0.1454, simple_loss=0.2487, pruned_loss=0.02111, over 944034.00 frames. 2023-05-01 13:32:20,847 INFO [train.py:939] (1/8) Maximum memory allocated so far is 18145MB 2023-05-01 13:32:28,407 INFO [optim.py:368] (1/8) 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 13:32:49,518 INFO [zipformer.py:625] (1/8) 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] (1/8) 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] (1/8) Epoch 23, batch 50, loss[loss=0.1848, simple_loss=0.2814, pruned_loss=0.04404, over 17067.00 frames. ], tot_loss[loss=0.1877, simple_loss=0.2721, pruned_loss=0.05163, over 736469.04 frames. ], batch size: 55, lr: 2.96e-03, grad_scale: 2.0 2023-05-01 13:33:49,483 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.5418, 2.1786, 1.6935, 2.0089, 2.5492, 2.3184, 2.5051, 2.6586], device='cuda:1'), covar=tensor([0.0272, 0.0465, 0.0706, 0.0555, 0.0307, 0.0390, 0.0274, 0.0333], device='cuda:1'), in_proj_covar=tensor([0.0204, 0.0232, 0.0224, 0.0223, 0.0232, 0.0230, 0.0227, 0.0223], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-01 13:34:24,350 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.6151, 4.4368, 4.6793, 4.8221, 5.0063, 4.5034, 4.9506, 4.9779], device='cuda:1'), covar=tensor([0.2028, 0.1410, 0.1663, 0.0834, 0.0727, 0.1043, 0.1220, 0.1078], device='cuda:1'), in_proj_covar=tensor([0.0613, 0.0754, 0.0874, 0.0763, 0.0580, 0.0606, 0.0635, 0.0732], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-01 13:34:32,081 INFO [train.py:904] (1/8) Epoch 23, batch 100, loss[loss=0.2179, simple_loss=0.2982, pruned_loss=0.06881, over 12132.00 frames. ], tot_loss[loss=0.1809, simple_loss=0.2659, pruned_loss=0.04799, over 1304154.72 frames. ], batch size: 246, lr: 2.96e-03, grad_scale: 2.0 2023-05-01 13:34:42,058 INFO [optim.py:368] (1/8) 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 13:34:44,260 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.3499, 3.5085, 3.9083, 2.2197, 3.1152, 2.4114, 3.6528, 3.6807], device='cuda:1'), covar=tensor([0.0282, 0.0959, 0.0518, 0.1963, 0.0848, 0.1005, 0.0646, 0.1092], device='cuda:1'), in_proj_covar=tensor([0.0153, 0.0158, 0.0164, 0.0151, 0.0143, 0.0128, 0.0140, 0.0170], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:1') 2023-05-01 13:34:55,449 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=223420.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 13:35:38,706 INFO [train.py:904] (1/8) Epoch 23, batch 150, loss[loss=0.1563, simple_loss=0.2517, pruned_loss=0.03049, over 17197.00 frames. ], tot_loss[loss=0.177, simple_loss=0.2635, pruned_loss=0.04528, over 1757024.18 frames. ], batch size: 46, lr: 2.96e-03, grad_scale: 2.0 2023-05-01 13:36:47,607 INFO [train.py:904] (1/8) Epoch 23, batch 200, loss[loss=0.1779, simple_loss=0.2735, pruned_loss=0.04114, over 17111.00 frames. ], tot_loss[loss=0.1774, simple_loss=0.2645, pruned_loss=0.04517, over 2094563.63 frames. ], batch size: 48, lr: 2.96e-03, grad_scale: 2.0 2023-05-01 13:36:57,903 INFO [optim.py:368] (1/8) 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 13:37:21,174 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=223529.0, num_to_drop=1, layers_to_drop={0} 2023-05-01 13:37:52,726 INFO [train.py:904] (1/8) Epoch 23, batch 250, loss[loss=0.1463, simple_loss=0.2392, pruned_loss=0.02673, over 17218.00 frames. ], tot_loss[loss=0.1748, simple_loss=0.2615, pruned_loss=0.04401, over 2367265.09 frames. ], batch size: 44, lr: 2.96e-03, grad_scale: 2.0 2023-05-01 13:38:29,062 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=223578.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 13:38:54,790 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=223596.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 13:39:04,044 INFO [train.py:904] (1/8) Epoch 23, batch 300, loss[loss=0.1665, simple_loss=0.2616, pruned_loss=0.03571, over 16728.00 frames. ], tot_loss[loss=0.1727, simple_loss=0.2589, pruned_loss=0.04328, over 2573058.79 frames. ], batch size: 57, lr: 2.96e-03, grad_scale: 2.0 2023-05-01 13:39:14,757 INFO [optim.py:368] (1/8) 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 13:39:20,811 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.91 vs. limit=2.0 2023-05-01 13:39:34,976 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=223625.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 13:39:49,530 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.86 vs. limit=2.0 2023-05-01 13:39:54,071 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=223639.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 13:39:55,303 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=223640.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 13:40:01,671 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=223644.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 13:40:13,646 INFO [train.py:904] (1/8) Epoch 23, batch 350, loss[loss=0.1801, simple_loss=0.2564, pruned_loss=0.05191, over 16575.00 frames. ], tot_loss[loss=0.1709, simple_loss=0.2573, pruned_loss=0.04222, over 2740298.01 frames. ], batch size: 75, lr: 2.96e-03, grad_scale: 2.0 2023-05-01 13:40:15,347 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.8844, 4.0516, 2.6260, 4.6555, 3.2165, 4.5554, 2.7401, 3.3387], device='cuda:1'), covar=tensor([0.0347, 0.0408, 0.1578, 0.0298, 0.0787, 0.0577, 0.1481, 0.0741], device='cuda:1'), in_proj_covar=tensor([0.0169, 0.0176, 0.0193, 0.0162, 0.0177, 0.0215, 0.0203, 0.0179], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-05-01 13:40:42,278 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=223673.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 13:41:03,083 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 23, batch 400, loss[loss=0.1629, simple_loss=0.2485, pruned_loss=0.0386, over 16908.00 frames. ], tot_loss[loss=0.1701, simple_loss=0.2556, pruned_loss=0.04232, over 2872684.49 frames. ], batch size: 96, lr: 2.96e-03, grad_scale: 4.0 2023-05-01 13:41:34,906 INFO [optim.py:368] (1/8) 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 13:41:46,591 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=223720.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 13:42:32,054 INFO [train.py:904] (1/8) Epoch 23, batch 450, loss[loss=0.1464, simple_loss=0.2272, pruned_loss=0.03279, over 16881.00 frames. ], tot_loss[loss=0.1688, simple_loss=0.2543, pruned_loss=0.0416, over 2980142.44 frames. ], batch size: 90, lr: 2.96e-03, grad_scale: 4.0 2023-05-01 13:42:43,846 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.80 vs. limit=2.0 2023-05-01 13:42:52,142 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=223768.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 13:43:40,984 INFO [train.py:904] (1/8) Epoch 23, batch 500, loss[loss=0.1349, simple_loss=0.2269, pruned_loss=0.02142, over 17188.00 frames. ], tot_loss[loss=0.1676, simple_loss=0.2533, pruned_loss=0.04098, over 3052846.21 frames. ], batch size: 46, lr: 2.96e-03, grad_scale: 2.0 2023-05-01 13:43:52,983 INFO [optim.py:368] (1/8) Clipping_scale=2.0, grad-norm quartiles 1.410e+02 2.054e+02 2.358e+02 2.805e+02 6.007e+02, threshold=4.715e+02, percent-clipped=4.0 2023-05-01 13:44:16,762 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=223829.0, num_to_drop=1, layers_to_drop={1} 2023-05-01 13:44:49,713 INFO [train.py:904] (1/8) Epoch 23, batch 550, loss[loss=0.173, simple_loss=0.2565, pruned_loss=0.04471, over 16990.00 frames. ], tot_loss[loss=0.1672, simple_loss=0.2529, pruned_loss=0.04073, over 3114606.61 frames. ], batch size: 41, lr: 2.96e-03, grad_scale: 2.0 2023-05-01 13:44:50,365 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.19 vs. limit=2.0 2023-05-01 13:45:15,333 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.0549, 2.1290, 2.3198, 3.6467, 2.1818, 2.4119, 2.2554, 2.3345], device='cuda:1'), covar=tensor([0.1602, 0.3873, 0.3198, 0.0759, 0.4012, 0.2698, 0.4051, 0.3141], device='cuda:1'), in_proj_covar=tensor([0.0403, 0.0452, 0.0371, 0.0328, 0.0440, 0.0518, 0.0423, 0.0527], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-01 13:45:24,265 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=223877.0, num_to_drop=1, layers_to_drop={1} 2023-05-01 13:45:59,107 INFO [train.py:904] (1/8) Epoch 23, batch 600, loss[loss=0.1639, simple_loss=0.2373, pruned_loss=0.04525, over 16640.00 frames. ], tot_loss[loss=0.1669, simple_loss=0.2526, pruned_loss=0.04059, over 3155917.96 frames. ], batch size: 134, lr: 2.96e-03, grad_scale: 2.0 2023-05-01 13:46:10,991 INFO [optim.py:368] (1/8) 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 13:46:11,393 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([5.6589, 6.0389, 5.8239, 5.8074, 5.3959, 5.4524, 5.4432, 6.1778], device='cuda:1'), covar=tensor([0.1390, 0.0972, 0.1080, 0.0847, 0.0999, 0.0725, 0.1183, 0.0873], device='cuda:1'), in_proj_covar=tensor([0.0677, 0.0826, 0.0678, 0.0625, 0.0519, 0.0531, 0.0694, 0.0643], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:1') 2023-05-01 13:46:41,947 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 23, batch 650, loss[loss=0.1932, simple_loss=0.2821, pruned_loss=0.05219, over 17042.00 frames. ], tot_loss[loss=0.1653, simple_loss=0.2506, pruned_loss=0.03997, over 3189582.88 frames. ], batch size: 53, lr: 2.96e-03, grad_scale: 2.0 2023-05-01 13:48:20,586 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=4.51 vs. limit=5.0 2023-05-01 13:48:22,013 INFO [train.py:904] (1/8) Epoch 23, batch 700, loss[loss=0.1743, simple_loss=0.2636, pruned_loss=0.04257, over 16668.00 frames. ], tot_loss[loss=0.1647, simple_loss=0.2499, pruned_loss=0.03973, over 3213363.17 frames. ], batch size: 89, lr: 2.96e-03, grad_scale: 2.0 2023-05-01 13:48:35,491 INFO [optim.py:368] (1/8) 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 13:48:40,950 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.23 vs. limit=2.0 2023-05-01 13:49:33,342 INFO [train.py:904] (1/8) Epoch 23, batch 750, loss[loss=0.1579, simple_loss=0.2373, pruned_loss=0.03927, over 16870.00 frames. ], tot_loss[loss=0.1649, simple_loss=0.2502, pruned_loss=0.03982, over 3227104.68 frames. ], batch size: 90, lr: 2.96e-03, grad_scale: 2.0 2023-05-01 13:50:13,517 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.24 vs. limit=2.0 2023-05-01 13:50:41,982 INFO [train.py:904] (1/8) Epoch 23, batch 800, loss[loss=0.1646, simple_loss=0.2412, pruned_loss=0.044, over 16346.00 frames. ], tot_loss[loss=0.165, simple_loss=0.2502, pruned_loss=0.03995, over 3251778.43 frames. ], batch size: 165, lr: 2.96e-03, grad_scale: 4.0 2023-05-01 13:50:54,818 INFO [optim.py:368] (1/8) 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 13:51:34,927 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([5.2747, 5.2418, 4.9928, 4.4906, 5.0145, 1.9575, 4.7601, 4.7886], device='cuda:1'), covar=tensor([0.0097, 0.0086, 0.0223, 0.0395, 0.0116, 0.2896, 0.0159, 0.0255], device='cuda:1'), in_proj_covar=tensor([0.0169, 0.0160, 0.0202, 0.0177, 0.0179, 0.0213, 0.0191, 0.0174], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-01 13:51:49,873 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([1.7238, 2.2643, 2.3897, 3.5842, 2.9271, 3.7487, 1.4821, 2.8147], device='cuda:1'), covar=tensor([0.1349, 0.0834, 0.1190, 0.0204, 0.0168, 0.0394, 0.1658, 0.0826], device='cuda:1'), in_proj_covar=tensor([0.0168, 0.0174, 0.0193, 0.0188, 0.0202, 0.0214, 0.0203, 0.0192], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-05-01 13:51:51,762 INFO [train.py:904] (1/8) Epoch 23, batch 850, loss[loss=0.1445, simple_loss=0.2229, pruned_loss=0.03307, over 16495.00 frames. ], tot_loss[loss=0.1646, simple_loss=0.2497, pruned_loss=0.03975, over 3272015.37 frames. ], batch size: 146, lr: 2.96e-03, grad_scale: 4.0 2023-05-01 13:52:00,990 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.7766, 3.9597, 2.9447, 2.2661, 2.4990, 2.4360, 4.0438, 3.3750], device='cuda:1'), covar=tensor([0.2788, 0.0576, 0.1765, 0.3149, 0.2814, 0.2142, 0.0512, 0.1551], device='cuda:1'), in_proj_covar=tensor([0.0329, 0.0269, 0.0307, 0.0315, 0.0296, 0.0262, 0.0298, 0.0340], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:1') 2023-05-01 13:52:06,592 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.26 vs. limit=2.0 2023-05-01 13:52:42,669 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.83 vs. limit=2.0 2023-05-01 13:53:00,739 INFO [train.py:904] (1/8) Epoch 23, batch 900, loss[loss=0.1521, simple_loss=0.2328, pruned_loss=0.03569, over 12466.00 frames. ], tot_loss[loss=0.1632, simple_loss=0.2485, pruned_loss=0.03895, over 3277283.35 frames. ], batch size: 246, lr: 2.96e-03, grad_scale: 4.0 2023-05-01 13:53:07,514 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.17 vs. limit=2.0 2023-05-01 13:53:14,897 INFO [optim.py:368] (1/8) 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 13:53:15,203 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=224212.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 13:53:46,096 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 23, batch 950, loss[loss=0.1654, simple_loss=0.2591, pruned_loss=0.03582, over 17161.00 frames. ], tot_loss[loss=0.1631, simple_loss=0.2489, pruned_loss=0.0387, over 3286981.31 frames. ], batch size: 46, lr: 2.96e-03, grad_scale: 4.0 2023-05-01 13:54:33,545 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=4.98 vs. limit=5.0 2023-05-01 13:54:38,608 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=224273.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 13:54:48,371 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.2398, 1.6242, 2.0206, 2.1016, 2.2872, 2.3349, 1.8732, 2.2768], device='cuda:1'), covar=tensor([0.0265, 0.0494, 0.0288, 0.0361, 0.0336, 0.0291, 0.0482, 0.0201], device='cuda:1'), in_proj_covar=tensor([0.0191, 0.0193, 0.0180, 0.0185, 0.0197, 0.0154, 0.0196, 0.0152], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-01 13:54:50,479 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=224282.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 13:55:20,379 INFO [train.py:904] (1/8) Epoch 23, batch 1000, loss[loss=0.158, simple_loss=0.2544, pruned_loss=0.03076, over 17102.00 frames. ], tot_loss[loss=0.1624, simple_loss=0.2474, pruned_loss=0.03866, over 3300166.65 frames. ], batch size: 49, lr: 2.96e-03, grad_scale: 4.0 2023-05-01 13:55:33,531 INFO [optim.py:368] (1/8) 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 13:56:31,386 INFO [train.py:904] (1/8) Epoch 23, batch 1050, loss[loss=0.1679, simple_loss=0.2445, pruned_loss=0.04565, over 16679.00 frames. ], tot_loss[loss=0.1623, simple_loss=0.2474, pruned_loss=0.0386, over 3302815.44 frames. ], batch size: 83, lr: 2.96e-03, grad_scale: 4.0 2023-05-01 13:57:42,154 INFO [train.py:904] (1/8) Epoch 23, batch 1100, loss[loss=0.1654, simple_loss=0.2464, pruned_loss=0.04225, over 15374.00 frames. ], tot_loss[loss=0.1623, simple_loss=0.248, pruned_loss=0.03831, over 3309808.93 frames. ], batch size: 190, lr: 2.96e-03, grad_scale: 4.0 2023-05-01 13:57:54,067 INFO [optim.py:368] (1/8) 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 13:58:26,149 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.9573, 4.7662, 4.9443, 5.1747, 5.3531, 4.8061, 5.3134, 5.3338], device='cuda:1'), covar=tensor([0.2164, 0.1423, 0.2116, 0.1015, 0.0785, 0.0989, 0.0790, 0.0744], device='cuda:1'), in_proj_covar=tensor([0.0661, 0.0813, 0.0941, 0.0825, 0.0624, 0.0651, 0.0679, 0.0787], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-01 13:58:51,564 INFO [train.py:904] (1/8) Epoch 23, batch 1150, loss[loss=0.1405, simple_loss=0.2216, pruned_loss=0.02966, over 16781.00 frames. ], tot_loss[loss=0.1616, simple_loss=0.2475, pruned_loss=0.03785, over 3317721.77 frames. ], batch size: 39, lr: 2.96e-03, grad_scale: 4.0 2023-05-01 13:59:20,448 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.8327, 4.0615, 3.1089, 2.4257, 2.6467, 2.6770, 4.2203, 3.4674], device='cuda:1'), covar=tensor([0.2898, 0.0624, 0.1796, 0.3089, 0.2791, 0.2089, 0.0501, 0.1569], device='cuda:1'), in_proj_covar=tensor([0.0330, 0.0270, 0.0308, 0.0316, 0.0298, 0.0264, 0.0299, 0.0342], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:1') 2023-05-01 13:59:35,960 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=4.90 vs. limit=5.0 2023-05-01 14:00:00,970 INFO [train.py:904] (1/8) Epoch 23, batch 1200, loss[loss=0.1384, simple_loss=0.2218, pruned_loss=0.02757, over 16716.00 frames. ], tot_loss[loss=0.1609, simple_loss=0.2465, pruned_loss=0.03765, over 3327743.33 frames. ], batch size: 39, lr: 2.96e-03, grad_scale: 8.0 2023-05-01 14:00:14,522 INFO [optim.py:368] (1/8) 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 14:01:10,499 INFO [train.py:904] (1/8) Epoch 23, batch 1250, loss[loss=0.1866, simple_loss=0.2544, pruned_loss=0.05942, over 16822.00 frames. ], tot_loss[loss=0.1617, simple_loss=0.2467, pruned_loss=0.03832, over 3322490.63 frames. ], batch size: 102, lr: 2.96e-03, grad_scale: 8.0 2023-05-01 14:01:13,788 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([5.2065, 4.9095, 4.9327, 5.3887, 5.5795, 4.8875, 5.4491, 5.5107], device='cuda:1'), covar=tensor([0.1819, 0.1338, 0.2431, 0.0889, 0.0696, 0.1096, 0.0994, 0.0889], device='cuda:1'), in_proj_covar=tensor([0.0660, 0.0812, 0.0941, 0.0824, 0.0623, 0.0650, 0.0677, 0.0787], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-01 14:01:22,909 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=224562.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 14:01:31,198 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=224568.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 14:01:51,592 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=5.02 vs. limit=5.0 2023-05-01 14:02:20,413 INFO [train.py:904] (1/8) Epoch 23, batch 1300, loss[loss=0.1566, simple_loss=0.2443, pruned_loss=0.03443, over 16527.00 frames. ], tot_loss[loss=0.1617, simple_loss=0.2468, pruned_loss=0.03829, over 3319553.54 frames. ], batch size: 75, lr: 2.96e-03, grad_scale: 8.0 2023-05-01 14:02:33,658 INFO [optim.py:368] (1/8) Clipping_scale=2.0, grad-norm quartiles 1.337e+02 2.195e+02 2.523e+02 3.004e+02 8.579e+02, threshold=5.045e+02, percent-clipped=3.0 2023-05-01 14:02:49,045 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=224623.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 14:03:03,944 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.9552, 5.0938, 4.8474, 4.4522, 4.0720, 5.0478, 4.9474, 4.5409], device='cuda:1'), covar=tensor([0.1002, 0.0746, 0.0583, 0.0521, 0.2321, 0.0592, 0.0457, 0.0973], device='cuda:1'), in_proj_covar=tensor([0.0303, 0.0446, 0.0355, 0.0352, 0.0361, 0.0412, 0.0241, 0.0423], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002], device='cuda:1') 2023-05-01 14:03:29,364 INFO [train.py:904] (1/8) Epoch 23, batch 1350, loss[loss=0.1556, simple_loss=0.2463, pruned_loss=0.03248, over 17228.00 frames. ], tot_loss[loss=0.1618, simple_loss=0.2472, pruned_loss=0.03821, over 3313994.48 frames. ], batch size: 45, lr: 2.96e-03, grad_scale: 8.0 2023-05-01 14:04:40,082 INFO [train.py:904] (1/8) Epoch 23, batch 1400, loss[loss=0.1523, simple_loss=0.2282, pruned_loss=0.03823, over 16735.00 frames. ], tot_loss[loss=0.1614, simple_loss=0.2466, pruned_loss=0.03813, over 3324216.65 frames. ], batch size: 83, lr: 2.96e-03, grad_scale: 8.0 2023-05-01 14:04:52,698 INFO [optim.py:368] (1/8) 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 14:05:13,952 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=224726.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 14:05:33,090 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=224740.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 14:05:46,771 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.8365, 4.0981, 4.2512, 2.8288, 3.6684, 4.1739, 3.8131, 2.5990], device='cuda:1'), covar=tensor([0.0448, 0.0124, 0.0053, 0.0407, 0.0131, 0.0100, 0.0098, 0.0447], device='cuda:1'), in_proj_covar=tensor([0.0138, 0.0086, 0.0086, 0.0136, 0.0100, 0.0111, 0.0097, 0.0132], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0003, 0.0004], device='cuda:1') 2023-05-01 14:05:50,560 INFO [train.py:904] (1/8) Epoch 23, batch 1450, loss[loss=0.1886, simple_loss=0.2535, pruned_loss=0.06191, over 16723.00 frames. ], tot_loss[loss=0.161, simple_loss=0.246, pruned_loss=0.03804, over 3323668.55 frames. ], batch size: 124, lr: 2.96e-03, grad_scale: 8.0 2023-05-01 14:06:01,575 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.8026, 3.8394, 2.5816, 4.5077, 3.0918, 4.4666, 2.6100, 3.1549], device='cuda:1'), covar=tensor([0.0350, 0.0475, 0.1522, 0.0272, 0.0812, 0.0536, 0.1526, 0.0816], device='cuda:1'), in_proj_covar=tensor([0.0174, 0.0180, 0.0196, 0.0169, 0.0179, 0.0220, 0.0206, 0.0182], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-05-01 14:06:39,643 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=224787.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 14:06:48,458 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.93 vs. limit=2.0 2023-05-01 14:06:59,972 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=224801.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 14:07:01,745 INFO [train.py:904] (1/8) Epoch 23, batch 1500, loss[loss=0.165, simple_loss=0.2473, pruned_loss=0.04137, over 16505.00 frames. ], tot_loss[loss=0.1614, simple_loss=0.246, pruned_loss=0.03836, over 3306970.42 frames. ], batch size: 75, lr: 2.96e-03, grad_scale: 8.0 2023-05-01 14:07:16,528 INFO [optim.py:368] (1/8) 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 14:08:14,249 INFO [train.py:904] (1/8) Epoch 23, batch 1550, loss[loss=0.1812, simple_loss=0.2609, pruned_loss=0.05072, over 16265.00 frames. ], tot_loss[loss=0.1626, simple_loss=0.2473, pruned_loss=0.03896, over 3320442.00 frames. ], batch size: 165, lr: 2.96e-03, grad_scale: 4.0 2023-05-01 14:08:34,696 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=224868.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 14:09:23,917 INFO [train.py:904] (1/8) Epoch 23, batch 1600, loss[loss=0.1623, simple_loss=0.2595, pruned_loss=0.03251, over 17111.00 frames. ], tot_loss[loss=0.165, simple_loss=0.2494, pruned_loss=0.04028, over 3312839.73 frames. ], batch size: 48, lr: 2.95e-03, grad_scale: 8.0 2023-05-01 14:09:36,820 INFO [optim.py:368] (1/8) 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 14:09:41,254 INFO [zipformer.py:625] (1/8) 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] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=224918.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 14:10:32,920 INFO [train.py:904] (1/8) Epoch 23, batch 1650, loss[loss=0.1581, simple_loss=0.2563, pruned_loss=0.03, over 17250.00 frames. ], tot_loss[loss=0.1665, simple_loss=0.2511, pruned_loss=0.04096, over 3307737.04 frames. ], batch size: 52, lr: 2.95e-03, grad_scale: 8.0 2023-05-01 14:10:33,897 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.74 vs. limit=2.0 2023-05-01 14:10:37,437 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=4.84 vs. limit=5.0 2023-05-01 14:11:03,831 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.9813, 5.5303, 5.6507, 5.2868, 5.4293, 6.0324, 5.4547, 5.1218], device='cuda:1'), covar=tensor([0.1138, 0.1935, 0.2449, 0.2176, 0.2606, 0.1001, 0.1592, 0.2478], device='cuda:1'), in_proj_covar=tensor([0.0418, 0.0614, 0.0677, 0.0503, 0.0678, 0.0703, 0.0531, 0.0675], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-05-01 14:11:41,679 INFO [train.py:904] (1/8) Epoch 23, batch 1700, loss[loss=0.2016, simple_loss=0.2834, pruned_loss=0.05995, over 15461.00 frames. ], tot_loss[loss=0.1681, simple_loss=0.2538, pruned_loss=0.04122, over 3317239.01 frames. ], batch size: 190, lr: 2.95e-03, grad_scale: 8.0 2023-05-01 14:11:56,160 INFO [optim.py:368] (1/8) 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 14:12:46,748 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([5.6592, 6.0574, 5.7658, 5.8662, 5.4190, 5.4746, 5.5312, 6.2127], device='cuda:1'), covar=tensor([0.1625, 0.1062, 0.1477, 0.0959, 0.1029, 0.0724, 0.1374, 0.0975], device='cuda:1'), in_proj_covar=tensor([0.0699, 0.0852, 0.0698, 0.0645, 0.0537, 0.0545, 0.0715, 0.0665], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:1') 2023-05-01 14:12:52,530 INFO [train.py:904] (1/8) Epoch 23, batch 1750, loss[loss=0.1664, simple_loss=0.2646, pruned_loss=0.03408, over 17111.00 frames. ], tot_loss[loss=0.1683, simple_loss=0.2545, pruned_loss=0.04108, over 3318634.02 frames. ], batch size: 47, lr: 2.95e-03, grad_scale: 8.0 2023-05-01 14:13:32,942 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=225082.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 14:13:52,571 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 23, batch 1800, loss[loss=0.1648, simple_loss=0.2629, pruned_loss=0.03333, over 17121.00 frames. ], tot_loss[loss=0.1688, simple_loss=0.2556, pruned_loss=0.04102, over 3315828.72 frames. ], batch size: 49, lr: 2.95e-03, grad_scale: 8.0 2023-05-01 14:14:02,316 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=225103.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 14:14:15,807 INFO [optim.py:368] (1/8) 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 14:14:57,186 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=225142.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 14:15:07,447 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.98 vs. limit=2.0 2023-05-01 14:15:12,126 INFO [train.py:904] (1/8) Epoch 23, batch 1850, loss[loss=0.1784, simple_loss=0.2637, pruned_loss=0.0465, over 16246.00 frames. ], tot_loss[loss=0.1699, simple_loss=0.2567, pruned_loss=0.04153, over 3323488.24 frames. ], batch size: 165, lr: 2.95e-03, grad_scale: 8.0 2023-05-01 14:15:27,761 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=225164.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 14:15:40,518 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([5.5285, 5.9320, 5.6612, 5.7207, 5.2588, 5.4227, 5.2895, 6.0528], device='cuda:1'), covar=tensor([0.1649, 0.1026, 0.1250, 0.0979, 0.1169, 0.0648, 0.1433, 0.1048], device='cuda:1'), in_proj_covar=tensor([0.0697, 0.0848, 0.0697, 0.0645, 0.0535, 0.0543, 0.0713, 0.0662], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:1') 2023-05-01 14:15:56,962 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([5.4118, 5.4006, 5.1426, 4.5962, 5.2031, 2.0688, 4.9573, 5.1119], device='cuda:1'), covar=tensor([0.0084, 0.0078, 0.0213, 0.0412, 0.0108, 0.2844, 0.0147, 0.0203], device='cuda:1'), in_proj_covar=tensor([0.0175, 0.0165, 0.0207, 0.0184, 0.0185, 0.0218, 0.0197, 0.0180], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-01 14:16:22,244 INFO [train.py:904] (1/8) Epoch 23, batch 1900, loss[loss=0.187, simple_loss=0.2723, pruned_loss=0.05085, over 16452.00 frames. ], tot_loss[loss=0.1688, simple_loss=0.2559, pruned_loss=0.04085, over 3329054.60 frames. ], batch size: 68, lr: 2.95e-03, grad_scale: 8.0 2023-05-01 14:16:22,756 INFO [zipformer.py:625] (1/8) 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] (1/8) 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 14:16:43,965 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 23, batch 1950, loss[loss=0.1562, simple_loss=0.2516, pruned_loss=0.03046, over 17175.00 frames. ], tot_loss[loss=0.1679, simple_loss=0.2556, pruned_loss=0.04013, over 3324072.69 frames. ], batch size: 46, lr: 2.95e-03, grad_scale: 8.0 2023-05-01 14:17:50,055 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=225266.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 14:18:23,652 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.2597, 2.0370, 2.2067, 3.9964, 2.0753, 2.3154, 2.1931, 2.2043], device='cuda:1'), covar=tensor([0.1635, 0.4864, 0.3629, 0.0738, 0.5716, 0.3564, 0.4350, 0.4702], device='cuda:1'), in_proj_covar=tensor([0.0407, 0.0456, 0.0374, 0.0333, 0.0441, 0.0526, 0.0428, 0.0533], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-01 14:18:42,396 INFO [train.py:904] (1/8) Epoch 23, batch 2000, loss[loss=0.1821, simple_loss=0.2547, pruned_loss=0.05474, over 16771.00 frames. ], tot_loss[loss=0.1676, simple_loss=0.2553, pruned_loss=0.03997, over 3320556.02 frames. ], batch size: 134, lr: 2.95e-03, grad_scale: 8.0 2023-05-01 14:18:56,051 INFO [optim.py:368] (1/8) 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 14:19:47,222 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.9908, 3.0759, 2.8361, 3.0033, 3.3233, 3.1217, 3.6174, 3.4735], device='cuda:1'), covar=tensor([0.0136, 0.0409, 0.0469, 0.0401, 0.0299, 0.0375, 0.0297, 0.0289], device='cuda:1'), in_proj_covar=tensor([0.0221, 0.0242, 0.0233, 0.0232, 0.0243, 0.0242, 0.0244, 0.0238], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-01 14:19:52,847 INFO [train.py:904] (1/8) Epoch 23, batch 2050, loss[loss=0.1732, simple_loss=0.2636, pruned_loss=0.04147, over 16777.00 frames. ], tot_loss[loss=0.1681, simple_loss=0.2554, pruned_loss=0.04035, over 3310833.62 frames. ], batch size: 57, lr: 2.95e-03, grad_scale: 8.0 2023-05-01 14:19:57,164 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.8209, 3.9785, 3.0105, 2.3416, 2.6345, 2.4282, 4.0890, 3.4672], device='cuda:1'), covar=tensor([0.2633, 0.0546, 0.1667, 0.2917, 0.2539, 0.2139, 0.0487, 0.1425], device='cuda:1'), in_proj_covar=tensor([0.0329, 0.0271, 0.0309, 0.0317, 0.0300, 0.0265, 0.0300, 0.0344], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:1') 2023-05-01 14:20:06,188 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.81 vs. limit=2.0 2023-05-01 14:20:35,645 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=225382.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 14:20:41,066 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.3692, 2.3695, 2.3746, 4.3242, 2.2223, 2.7716, 2.4084, 2.4985], device='cuda:1'), covar=tensor([0.1376, 0.3643, 0.3206, 0.0537, 0.4215, 0.2594, 0.3639, 0.3683], device='cuda:1'), in_proj_covar=tensor([0.0407, 0.0456, 0.0374, 0.0333, 0.0441, 0.0526, 0.0428, 0.0533], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-01 14:20:54,901 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=225396.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 14:21:04,288 INFO [train.py:904] (1/8) Epoch 23, batch 2100, loss[loss=0.1518, simple_loss=0.2402, pruned_loss=0.03169, over 16738.00 frames. ], tot_loss[loss=0.1691, simple_loss=0.2561, pruned_loss=0.04105, over 3323185.95 frames. ], batch size: 39, lr: 2.95e-03, grad_scale: 8.0 2023-05-01 14:21:09,752 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([5.3869, 5.7360, 5.5682, 5.6308, 5.2609, 5.2175, 5.2336, 5.8952], device='cuda:1'), covar=tensor([0.1530, 0.1036, 0.1098, 0.0814, 0.0903, 0.0711, 0.1043, 0.0940], device='cuda:1'), in_proj_covar=tensor([0.0700, 0.0852, 0.0699, 0.0648, 0.0537, 0.0545, 0.0715, 0.0666], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:1') 2023-05-01 14:21:18,934 INFO [optim.py:368] (1/8) 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 14:21:44,082 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=225430.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 14:21:45,506 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.0224, 3.0636, 2.7660, 2.8613, 3.3340, 3.0262, 3.5634, 3.4657], device='cuda:1'), covar=tensor([0.0123, 0.0375, 0.0459, 0.0418, 0.0282, 0.0360, 0.0282, 0.0256], device='cuda:1'), in_proj_covar=tensor([0.0222, 0.0243, 0.0233, 0.0233, 0.0243, 0.0243, 0.0245, 0.0239], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-01 14:21:56,546 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=225439.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 14:22:02,520 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.1753, 2.6110, 2.1493, 2.4674, 2.9441, 2.7146, 3.1118, 3.0802], device='cuda:1'), covar=tensor([0.0223, 0.0419, 0.0583, 0.0478, 0.0283, 0.0378, 0.0226, 0.0280], device='cuda:1'), in_proj_covar=tensor([0.0222, 0.0243, 0.0233, 0.0233, 0.0244, 0.0243, 0.0245, 0.0239], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-01 14:22:03,498 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=225444.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 14:22:13,590 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=4.58 vs. limit=5.0 2023-05-01 14:22:15,762 INFO [train.py:904] (1/8) Epoch 23, batch 2150, loss[loss=0.1809, simple_loss=0.2577, pruned_loss=0.05201, over 16780.00 frames. ], tot_loss[loss=0.17, simple_loss=0.2568, pruned_loss=0.04159, over 3324924.33 frames. ], batch size: 83, lr: 2.95e-03, grad_scale: 8.0 2023-05-01 14:22:18,707 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=3.29 vs. limit=5.0 2023-05-01 14:22:25,206 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=225459.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 14:22:29,418 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=225462.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 14:23:02,144 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.86 vs. limit=2.0 2023-05-01 14:23:04,210 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.2523, 3.2093, 1.9827, 3.4619, 2.4819, 3.4543, 2.1692, 2.6933], device='cuda:1'), covar=tensor([0.0286, 0.0447, 0.1577, 0.0320, 0.0810, 0.0725, 0.1469, 0.0694], device='cuda:1'), in_proj_covar=tensor([0.0173, 0.0180, 0.0196, 0.0170, 0.0180, 0.0222, 0.0206, 0.0182], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-05-01 14:23:18,681 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=225498.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 14:23:22,060 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=225500.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 14:23:26,036 INFO [train.py:904] (1/8) Epoch 23, batch 2200, loss[loss=0.1632, simple_loss=0.2464, pruned_loss=0.04002, over 16778.00 frames. ], tot_loss[loss=0.1704, simple_loss=0.2573, pruned_loss=0.0418, over 3320463.16 frames. ], batch size: 102, lr: 2.95e-03, grad_scale: 8.0 2023-05-01 14:23:29,386 INFO [zipformer.py:625] (1/8) 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] (1/8) 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 14:23:53,514 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=225522.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 14:23:54,696 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=225523.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 14:24:36,368 INFO [train.py:904] (1/8) Epoch 23, batch 2250, loss[loss=0.1779, simple_loss=0.2746, pruned_loss=0.04063, over 17020.00 frames. ], tot_loss[loss=0.1709, simple_loss=0.2576, pruned_loss=0.04208, over 3322342.01 frames. ], batch size: 53, lr: 2.95e-03, grad_scale: 8.0 2023-05-01 14:24:45,178 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=225559.0, num_to_drop=1, layers_to_drop={1} 2023-05-01 14:24:54,889 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=225566.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 14:24:59,366 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=3.40 vs. limit=5.0 2023-05-01 14:25:18,248 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=225583.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 14:25:46,974 INFO [train.py:904] (1/8) Epoch 23, batch 2300, loss[loss=0.1537, simple_loss=0.2463, pruned_loss=0.03062, over 17141.00 frames. ], tot_loss[loss=0.1716, simple_loss=0.2582, pruned_loss=0.04248, over 3314433.56 frames. ], batch size: 48, lr: 2.95e-03, grad_scale: 8.0 2023-05-01 14:26:01,554 INFO [optim.py:368] (1/8) 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] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=225620.0, num_to_drop=1, layers_to_drop={1} 2023-05-01 14:26:32,062 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.9886, 2.1166, 2.1971, 3.5481, 2.1099, 2.3703, 2.2120, 2.2448], device='cuda:1'), covar=tensor([0.1576, 0.4041, 0.3278, 0.0792, 0.4198, 0.2853, 0.3919, 0.3524], device='cuda:1'), in_proj_covar=tensor([0.0409, 0.0459, 0.0376, 0.0334, 0.0443, 0.0529, 0.0431, 0.0536], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-01 14:26:58,883 INFO [train.py:904] (1/8) Epoch 23, batch 2350, loss[loss=0.1519, simple_loss=0.2329, pruned_loss=0.03546, over 16829.00 frames. ], tot_loss[loss=0.1715, simple_loss=0.2582, pruned_loss=0.04242, over 3314176.94 frames. ], batch size: 102, lr: 2.95e-03, grad_scale: 8.0 2023-05-01 14:27:58,017 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.8294, 5.0714, 5.2122, 5.0010, 4.9317, 5.5998, 5.1111, 4.8185], device='cuda:1'), covar=tensor([0.1377, 0.2014, 0.2308, 0.2150, 0.3050, 0.1021, 0.1572, 0.2429], device='cuda:1'), in_proj_covar=tensor([0.0420, 0.0613, 0.0678, 0.0505, 0.0680, 0.0700, 0.0531, 0.0676], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-05-01 14:28:10,319 INFO [train.py:904] (1/8) Epoch 23, batch 2400, loss[loss=0.1744, simple_loss=0.2557, pruned_loss=0.04658, over 16736.00 frames. ], tot_loss[loss=0.1723, simple_loss=0.2589, pruned_loss=0.04284, over 3315300.20 frames. ], batch size: 83, lr: 2.95e-03, grad_scale: 8.0 2023-05-01 14:28:23,215 INFO [optim.py:368] (1/8) 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 14:29:17,843 INFO [train.py:904] (1/8) Epoch 23, batch 2450, loss[loss=0.1688, simple_loss=0.2664, pruned_loss=0.03565, over 17122.00 frames. ], tot_loss[loss=0.1728, simple_loss=0.2604, pruned_loss=0.04263, over 3318942.36 frames. ], batch size: 49, lr: 2.95e-03, grad_scale: 8.0 2023-05-01 14:29:26,378 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=225759.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 14:30:17,161 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=225795.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 14:30:22,196 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=225798.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 14:30:29,832 INFO [train.py:904] (1/8) Epoch 23, batch 2500, loss[loss=0.172, simple_loss=0.251, pruned_loss=0.04652, over 16920.00 frames. ], tot_loss[loss=0.1728, simple_loss=0.2605, pruned_loss=0.04256, over 3307686.68 frames. ], batch size: 90, lr: 2.95e-03, grad_scale: 8.0 2023-05-01 14:30:36,107 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=225807.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 14:30:43,416 INFO [optim.py:368] (1/8) 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 14:30:50,741 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=225818.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 14:31:21,447 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.2027, 4.1648, 4.1288, 3.5543, 4.1448, 1.6113, 3.9438, 3.6803], device='cuda:1'), covar=tensor([0.0145, 0.0129, 0.0198, 0.0294, 0.0115, 0.3128, 0.0156, 0.0267], device='cuda:1'), in_proj_covar=tensor([0.0174, 0.0165, 0.0208, 0.0184, 0.0186, 0.0218, 0.0198, 0.0180], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-01 14:31:28,974 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 23, batch 2550, loss[loss=0.1813, simple_loss=0.2549, pruned_loss=0.05382, over 16852.00 frames. ], tot_loss[loss=0.1728, simple_loss=0.2604, pruned_loss=0.04256, over 3314114.51 frames. ], batch size: 116, lr: 2.95e-03, grad_scale: 8.0 2023-05-01 14:31:49,609 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=225861.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 14:31:50,905 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=225862.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 14:32:14,112 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=225878.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 14:32:38,066 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 23, batch 2600, loss[loss=0.1959, simple_loss=0.2806, pruned_loss=0.05556, over 12251.00 frames. ], tot_loss[loss=0.1723, simple_loss=0.2597, pruned_loss=0.04239, over 3305873.93 frames. ], batch size: 246, lr: 2.95e-03, grad_scale: 8.0 2023-05-01 14:33:03,045 INFO [optim.py:368] (1/8) 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 14:33:06,085 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=225915.0, num_to_drop=1, layers_to_drop={2} 2023-05-01 14:33:16,544 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 23, batch 2650, loss[loss=0.1724, simple_loss=0.277, pruned_loss=0.03386, over 16655.00 frames. ], tot_loss[loss=0.1723, simple_loss=0.2606, pruned_loss=0.042, over 3299218.74 frames. ], batch size: 62, lr: 2.95e-03, grad_scale: 8.0 2023-05-01 14:34:04,304 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=225956.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 14:34:32,964 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.5674, 4.5957, 4.9089, 4.8989, 4.9464, 4.6155, 4.6194, 4.4491], device='cuda:1'), covar=tensor([0.0370, 0.0760, 0.0451, 0.0484, 0.0560, 0.0487, 0.0887, 0.0631], device='cuda:1'), in_proj_covar=tensor([0.0425, 0.0474, 0.0461, 0.0426, 0.0507, 0.0483, 0.0566, 0.0385], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:1') 2023-05-01 14:35:12,270 INFO [train.py:904] (1/8) Epoch 23, batch 2700, loss[loss=0.1602, simple_loss=0.2581, pruned_loss=0.03113, over 17225.00 frames. ], tot_loss[loss=0.171, simple_loss=0.2599, pruned_loss=0.04102, over 3313544.78 frames. ], batch size: 52, lr: 2.95e-03, grad_scale: 8.0 2023-05-01 14:35:25,728 INFO [optim.py:368] (1/8) 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 14:36:03,413 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.1228, 4.1630, 4.4885, 4.4635, 4.4964, 4.2003, 4.2235, 4.1305], device='cuda:1'), covar=tensor([0.0432, 0.0659, 0.0418, 0.0457, 0.0599, 0.0477, 0.0859, 0.0682], device='cuda:1'), in_proj_covar=tensor([0.0427, 0.0476, 0.0463, 0.0428, 0.0508, 0.0485, 0.0569, 0.0386], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:1') 2023-05-01 14:36:06,351 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=226041.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 14:36:23,070 INFO [train.py:904] (1/8) Epoch 23, batch 2750, loss[loss=0.1749, simple_loss=0.2764, pruned_loss=0.03673, over 16702.00 frames. ], tot_loss[loss=0.1709, simple_loss=0.2598, pruned_loss=0.041, over 3311526.46 frames. ], batch size: 62, lr: 2.95e-03, grad_scale: 8.0 2023-05-01 14:37:21,835 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=226095.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 14:37:28,016 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.5871, 2.3467, 1.9575, 2.1464, 2.6639, 2.4709, 2.6192, 2.7808], device='cuda:1'), covar=tensor([0.0237, 0.0419, 0.0545, 0.0491, 0.0255, 0.0365, 0.0202, 0.0287], device='cuda:1'), in_proj_covar=tensor([0.0223, 0.0243, 0.0231, 0.0233, 0.0244, 0.0242, 0.0244, 0.0239], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-01 14:37:31,347 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 23, batch 2800, loss[loss=0.1632, simple_loss=0.2503, pruned_loss=0.03802, over 16272.00 frames. ], tot_loss[loss=0.1709, simple_loss=0.2598, pruned_loss=0.04104, over 3314530.23 frames. ], batch size: 165, lr: 2.95e-03, grad_scale: 8.0 2023-05-01 14:37:47,346 INFO [optim.py:368] (1/8) 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 14:37:54,717 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=226118.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 14:38:25,674 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.6484, 2.3863, 2.4219, 4.4217, 2.3794, 2.8156, 2.4560, 2.5970], device='cuda:1'), covar=tensor([0.1279, 0.3876, 0.3212, 0.0501, 0.4359, 0.2661, 0.3711, 0.3695], device='cuda:1'), in_proj_covar=tensor([0.0407, 0.0456, 0.0373, 0.0334, 0.0440, 0.0525, 0.0427, 0.0533], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-01 14:38:29,330 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=226143.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 14:38:42,392 INFO [train.py:904] (1/8) Epoch 23, batch 2850, loss[loss=0.1473, simple_loss=0.2375, pruned_loss=0.02854, over 17020.00 frames. ], tot_loss[loss=0.17, simple_loss=0.2586, pruned_loss=0.04064, over 3318437.38 frames. ], batch size: 41, lr: 2.95e-03, grad_scale: 8.0 2023-05-01 14:38:54,257 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=226161.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 14:39:00,760 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=226166.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 14:39:18,180 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=226178.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 14:39:23,525 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.18 vs. limit=2.0 2023-05-01 14:39:39,781 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([1.9176, 2.9578, 2.6231, 4.3265, 3.6168, 4.1733, 1.7153, 3.1382], device='cuda:1'), covar=tensor([0.1336, 0.0625, 0.1108, 0.0153, 0.0155, 0.0376, 0.1523, 0.0717], device='cuda:1'), in_proj_covar=tensor([0.0167, 0.0174, 0.0194, 0.0193, 0.0204, 0.0217, 0.0203, 0.0192], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-05-01 14:39:46,460 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.6486, 3.6007, 2.8630, 2.2464, 2.3107, 2.2835, 3.7391, 3.1705], device='cuda:1'), covar=tensor([0.2846, 0.0623, 0.1614, 0.3110, 0.3186, 0.2238, 0.0505, 0.1741], device='cuda:1'), in_proj_covar=tensor([0.0330, 0.0272, 0.0309, 0.0318, 0.0301, 0.0265, 0.0301, 0.0347], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:1') 2023-05-01 14:39:51,835 INFO [train.py:904] (1/8) Epoch 23, batch 2900, loss[loss=0.1564, simple_loss=0.2421, pruned_loss=0.03528, over 15899.00 frames. ], tot_loss[loss=0.1703, simple_loss=0.2581, pruned_loss=0.04131, over 3317397.19 frames. ], batch size: 35, lr: 2.95e-03, grad_scale: 8.0 2023-05-01 14:40:00,214 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=226209.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 14:40:00,386 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.6995, 3.7223, 3.8851, 2.7499, 3.4998, 3.9786, 3.6308, 2.3970], device='cuda:1'), covar=tensor([0.0495, 0.0232, 0.0066, 0.0403, 0.0139, 0.0102, 0.0118, 0.0518], device='cuda:1'), in_proj_covar=tensor([0.0138, 0.0086, 0.0087, 0.0136, 0.0100, 0.0112, 0.0098, 0.0132], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0003, 0.0004], device='cuda:1') 2023-05-01 14:40:05,815 INFO [optim.py:368] (1/8) 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 14:40:09,262 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=226215.0, num_to_drop=1, layers_to_drop={0} 2023-05-01 14:40:13,513 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=226218.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 14:40:24,106 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=226226.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 14:40:33,425 INFO [zipformer.py:625] (1/8) 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] (1/8) 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] (1/8) Epoch 23, batch 2950, loss[loss=0.1401, simple_loss=0.2294, pruned_loss=0.02536, over 17203.00 frames. ], tot_loss[loss=0.1699, simple_loss=0.2563, pruned_loss=0.0417, over 3318342.79 frames. ], batch size: 44, lr: 2.95e-03, grad_scale: 8.0 2023-05-01 14:41:14,091 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=226263.0, num_to_drop=1, layers_to_drop={1} 2023-05-01 14:41:56,435 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 23, batch 3000, loss[loss=0.1652, simple_loss=0.2631, pruned_loss=0.03364, over 17060.00 frames. ], tot_loss[loss=0.1702, simple_loss=0.2562, pruned_loss=0.04214, over 3316306.95 frames. ], batch size: 50, lr: 2.95e-03, grad_scale: 8.0 2023-05-01 14:42:08,778 INFO [train.py:929] (1/8) Computing validation loss 2023-05-01 14:42:17,866 INFO [train.py:938] (1/8) Epoch 23, validation: loss=0.1344, simple_loss=0.2397, pruned_loss=0.01456, over 944034.00 frames. 2023-05-01 14:42:17,866 INFO [train.py:939] (1/8) Maximum memory allocated so far is 18145MB 2023-05-01 14:42:30,994 INFO [optim.py:368] (1/8) Clipping_scale=2.0, grad-norm quartiles 1.698e+02 2.313e+02 2.743e+02 3.282e+02 6.136e+02, threshold=5.486e+02, percent-clipped=4.0 2023-05-01 14:42:48,575 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([5.2512, 5.7172, 5.8850, 5.5695, 5.5822, 6.1621, 5.7433, 5.4876], device='cuda:1'), covar=tensor([0.0875, 0.1945, 0.2457, 0.2136, 0.2858, 0.0988, 0.1523, 0.2359], device='cuda:1'), in_proj_covar=tensor([0.0423, 0.0618, 0.0681, 0.0508, 0.0684, 0.0705, 0.0533, 0.0679], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-05-01 14:43:05,635 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.38 vs. limit=2.0 2023-05-01 14:43:26,896 INFO [train.py:904] (1/8) Epoch 23, batch 3050, loss[loss=0.154, simple_loss=0.2469, pruned_loss=0.03059, over 17125.00 frames. ], tot_loss[loss=0.1697, simple_loss=0.2561, pruned_loss=0.04165, over 3322189.14 frames. ], batch size: 48, lr: 2.95e-03, grad_scale: 8.0 2023-05-01 14:44:29,374 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=226397.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 14:44:37,979 INFO [train.py:904] (1/8) Epoch 23, batch 3100, loss[loss=0.1592, simple_loss=0.2462, pruned_loss=0.03609, over 16833.00 frames. ], tot_loss[loss=0.1696, simple_loss=0.2558, pruned_loss=0.04168, over 3322820.42 frames. ], batch size: 42, lr: 2.94e-03, grad_scale: 8.0 2023-05-01 14:44:51,600 INFO [optim.py:368] (1/8) 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 14:45:47,148 INFO [train.py:904] (1/8) Epoch 23, batch 3150, loss[loss=0.1597, simple_loss=0.2422, pruned_loss=0.0386, over 16752.00 frames. ], tot_loss[loss=0.169, simple_loss=0.2552, pruned_loss=0.04137, over 3330716.07 frames. ], batch size: 134, lr: 2.94e-03, grad_scale: 8.0 2023-05-01 14:46:54,851 INFO [train.py:904] (1/8) Epoch 23, batch 3200, loss[loss=0.1702, simple_loss=0.2701, pruned_loss=0.03518, over 17109.00 frames. ], tot_loss[loss=0.1689, simple_loss=0.2549, pruned_loss=0.0414, over 3331362.31 frames. ], batch size: 47, lr: 2.94e-03, grad_scale: 8.0 2023-05-01 14:47:07,327 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.0144, 1.9672, 2.5665, 2.8661, 2.8093, 3.4017, 2.5400, 3.3279], device='cuda:1'), covar=tensor([0.0271, 0.0556, 0.0338, 0.0350, 0.0346, 0.0186, 0.0426, 0.0165], device='cuda:1'), in_proj_covar=tensor([0.0198, 0.0198, 0.0185, 0.0190, 0.0203, 0.0160, 0.0201, 0.0157], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-01 14:47:09,850 INFO [optim.py:368] (1/8) 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 14:47:15,792 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=226518.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 14:48:01,430 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=226551.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 14:48:04,026 INFO [train.py:904] (1/8) Epoch 23, batch 3250, loss[loss=0.1625, simple_loss=0.26, pruned_loss=0.03245, over 17133.00 frames. ], tot_loss[loss=0.1685, simple_loss=0.2547, pruned_loss=0.04112, over 3332220.85 frames. ], batch size: 47, lr: 2.94e-03, grad_scale: 8.0 2023-05-01 14:48:05,752 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.3354, 3.6843, 4.0955, 2.1241, 3.2733, 2.6435, 3.8386, 3.9230], device='cuda:1'), covar=tensor([0.0309, 0.0861, 0.0451, 0.2084, 0.0782, 0.0948, 0.0616, 0.0908], device='cuda:1'), in_proj_covar=tensor([0.0160, 0.0169, 0.0170, 0.0156, 0.0148, 0.0132, 0.0146, 0.0181], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0005], device='cuda:1') 2023-05-01 14:48:05,973 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.28 vs. limit=2.0 2023-05-01 14:48:22,284 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=226566.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 14:48:54,453 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=226589.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 14:49:00,080 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.8085, 1.8844, 2.3592, 2.6312, 2.6728, 2.6910, 2.0078, 2.8850], device='cuda:1'), covar=tensor([0.0189, 0.0505, 0.0376, 0.0298, 0.0331, 0.0307, 0.0573, 0.0189], device='cuda:1'), in_proj_covar=tensor([0.0198, 0.0198, 0.0185, 0.0191, 0.0204, 0.0161, 0.0202, 0.0157], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-01 14:49:08,849 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=226599.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 14:49:14,395 INFO [train.py:904] (1/8) Epoch 23, batch 3300, loss[loss=0.2058, simple_loss=0.2832, pruned_loss=0.06425, over 15414.00 frames. ], tot_loss[loss=0.1678, simple_loss=0.2546, pruned_loss=0.04049, over 3339545.59 frames. ], batch size: 190, lr: 2.94e-03, grad_scale: 8.0 2023-05-01 14:49:27,925 INFO [optim.py:368] (1/8) 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 14:50:22,520 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.6804, 3.7438, 2.5999, 2.2761, 2.3816, 2.1655, 3.7061, 3.1176], device='cuda:1'), covar=tensor([0.2860, 0.0599, 0.2010, 0.3005, 0.2747, 0.2527, 0.0701, 0.1632], device='cuda:1'), in_proj_covar=tensor([0.0331, 0.0273, 0.0310, 0.0319, 0.0303, 0.0266, 0.0302, 0.0347], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:1') 2023-05-01 14:50:23,120 INFO [train.py:904] (1/8) Epoch 23, batch 3350, loss[loss=0.1956, simple_loss=0.2729, pruned_loss=0.05917, over 16784.00 frames. ], tot_loss[loss=0.1678, simple_loss=0.2551, pruned_loss=0.04032, over 3340734.74 frames. ], batch size: 124, lr: 2.94e-03, grad_scale: 8.0 2023-05-01 14:51:25,638 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=226697.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 14:51:34,603 INFO [train.py:904] (1/8) Epoch 23, batch 3400, loss[loss=0.177, simple_loss=0.2524, pruned_loss=0.05076, over 16854.00 frames. ], tot_loss[loss=0.1692, simple_loss=0.2561, pruned_loss=0.04119, over 3334537.30 frames. ], batch size: 116, lr: 2.94e-03, grad_scale: 8.0 2023-05-01 14:51:47,767 INFO [optim.py:368] (1/8) 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] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=226745.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 14:52:43,442 INFO [train.py:904] (1/8) Epoch 23, batch 3450, loss[loss=0.1631, simple_loss=0.2587, pruned_loss=0.03379, over 17153.00 frames. ], tot_loss[loss=0.1678, simple_loss=0.2544, pruned_loss=0.04064, over 3337215.84 frames. ], batch size: 49, lr: 2.94e-03, grad_scale: 8.0 2023-05-01 14:53:01,643 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.2805, 4.0502, 4.5183, 2.3173, 4.7061, 4.7724, 3.5000, 3.7415], device='cuda:1'), covar=tensor([0.0639, 0.0234, 0.0233, 0.1166, 0.0081, 0.0175, 0.0408, 0.0397], device='cuda:1'), in_proj_covar=tensor([0.0149, 0.0110, 0.0100, 0.0141, 0.0083, 0.0129, 0.0130, 0.0132], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-05-01 14:53:01,681 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.8920, 2.8612, 2.4752, 2.7846, 3.1469, 2.9248, 3.5402, 3.3829], device='cuda:1'), covar=tensor([0.0149, 0.0458, 0.0591, 0.0474, 0.0314, 0.0445, 0.0241, 0.0316], device='cuda:1'), in_proj_covar=tensor([0.0227, 0.0246, 0.0234, 0.0236, 0.0246, 0.0245, 0.0247, 0.0242], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-01 14:53:38,893 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.25 vs. limit=2.0 2023-05-01 14:53:52,946 INFO [train.py:904] (1/8) Epoch 23, batch 3500, loss[loss=0.1925, simple_loss=0.2748, pruned_loss=0.05513, over 12184.00 frames. ], tot_loss[loss=0.1667, simple_loss=0.2532, pruned_loss=0.0401, over 3330434.33 frames. ], batch size: 248, lr: 2.94e-03, grad_scale: 8.0 2023-05-01 14:54:07,205 INFO [optim.py:368] (1/8) 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 14:54:34,970 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.8784, 4.3016, 2.9734, 2.3554, 2.7771, 2.5871, 4.7218, 3.6272], device='cuda:1'), covar=tensor([0.3089, 0.0657, 0.2006, 0.3184, 0.2942, 0.2215, 0.0375, 0.1399], device='cuda:1'), in_proj_covar=tensor([0.0331, 0.0273, 0.0310, 0.0319, 0.0303, 0.0266, 0.0302, 0.0347], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:1') 2023-05-01 14:55:00,965 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.1447, 2.5930, 2.0751, 2.4555, 2.9318, 2.7096, 3.0262, 3.0871], device='cuda:1'), covar=tensor([0.0201, 0.0458, 0.0584, 0.0442, 0.0284, 0.0361, 0.0272, 0.0300], device='cuda:1'), in_proj_covar=tensor([0.0226, 0.0247, 0.0234, 0.0236, 0.0246, 0.0245, 0.0247, 0.0242], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-01 14:55:03,886 INFO [train.py:904] (1/8) Epoch 23, batch 3550, loss[loss=0.1558, simple_loss=0.2507, pruned_loss=0.03045, over 17017.00 frames. ], tot_loss[loss=0.1656, simple_loss=0.2522, pruned_loss=0.03951, over 3329454.78 frames. ], batch size: 50, lr: 2.94e-03, grad_scale: 16.0 2023-05-01 14:55:52,844 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=226889.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 14:56:12,760 INFO [train.py:904] (1/8) Epoch 23, batch 3600, loss[loss=0.162, simple_loss=0.2579, pruned_loss=0.03304, over 17010.00 frames. ], tot_loss[loss=0.165, simple_loss=0.2513, pruned_loss=0.03929, over 3329677.66 frames. ], batch size: 50, lr: 2.94e-03, grad_scale: 8.0 2023-05-01 14:56:28,214 INFO [optim.py:368] (1/8) 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 14:56:58,539 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.20 vs. limit=2.0 2023-05-01 14:57:02,339 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 23, batch 3650, loss[loss=0.1574, simple_loss=0.2297, pruned_loss=0.04256, over 16780.00 frames. ], tot_loss[loss=0.165, simple_loss=0.2506, pruned_loss=0.03976, over 3319199.32 frames. ], batch size: 124, lr: 2.94e-03, grad_scale: 8.0 2023-05-01 14:58:19,546 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=226989.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 14:58:40,559 INFO [train.py:904] (1/8) Epoch 23, batch 3700, loss[loss=0.1447, simple_loss=0.2316, pruned_loss=0.02888, over 15433.00 frames. ], tot_loss[loss=0.1663, simple_loss=0.2498, pruned_loss=0.0414, over 3271813.67 frames. ], batch size: 191, lr: 2.94e-03, grad_scale: 8.0 2023-05-01 14:58:56,577 INFO [optim.py:368] (1/8) 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 14:59:50,711 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=227050.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 14:59:53,687 INFO [train.py:904] (1/8) Epoch 23, batch 3750, loss[loss=0.1858, simple_loss=0.2621, pruned_loss=0.05476, over 16731.00 frames. ], tot_loss[loss=0.1677, simple_loss=0.2498, pruned_loss=0.04278, over 3260462.15 frames. ], batch size: 134, lr: 2.94e-03, grad_scale: 8.0 2023-05-01 15:01:07,855 INFO [train.py:904] (1/8) Epoch 23, batch 3800, loss[loss=0.1791, simple_loss=0.262, pruned_loss=0.0481, over 16502.00 frames. ], tot_loss[loss=0.1702, simple_loss=0.2515, pruned_loss=0.04451, over 3267668.15 frames. ], batch size: 68, lr: 2.94e-03, grad_scale: 8.0 2023-05-01 15:01:19,301 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.5017, 3.5224, 2.7672, 2.1980, 2.2608, 2.2679, 3.5500, 3.0479], device='cuda:1'), covar=tensor([0.2928, 0.0671, 0.1791, 0.3124, 0.2857, 0.2238, 0.0543, 0.1622], device='cuda:1'), in_proj_covar=tensor([0.0331, 0.0273, 0.0310, 0.0319, 0.0303, 0.0266, 0.0302, 0.0346], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:1') 2023-05-01 15:01:25,256 INFO [optim.py:368] (1/8) 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 15:01:37,519 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.7796, 3.0149, 3.1460, 2.0822, 2.7424, 2.2808, 3.4060, 3.3146], device='cuda:1'), covar=tensor([0.0296, 0.0891, 0.0691, 0.2019, 0.0968, 0.1059, 0.0531, 0.0842], device='cuda:1'), in_proj_covar=tensor([0.0161, 0.0170, 0.0170, 0.0156, 0.0147, 0.0132, 0.0145, 0.0181], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0005], device='cuda:1') 2023-05-01 15:02:20,284 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.5738, 2.7409, 2.0815, 2.4519, 3.0513, 2.7429, 3.1406, 3.3211], device='cuda:1'), covar=tensor([0.0098, 0.0389, 0.0625, 0.0490, 0.0267, 0.0382, 0.0213, 0.0225], device='cuda:1'), in_proj_covar=tensor([0.0223, 0.0243, 0.0231, 0.0233, 0.0242, 0.0242, 0.0244, 0.0240], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-01 15:02:21,617 INFO [train.py:904] (1/8) Epoch 23, batch 3850, loss[loss=0.1907, simple_loss=0.2697, pruned_loss=0.05584, over 16659.00 frames. ], tot_loss[loss=0.171, simple_loss=0.2517, pruned_loss=0.04514, over 3269006.90 frames. ], batch size: 57, lr: 2.94e-03, grad_scale: 8.0 2023-05-01 15:03:34,968 INFO [train.py:904] (1/8) Epoch 23, batch 3900, loss[loss=0.189, simple_loss=0.2669, pruned_loss=0.05555, over 12322.00 frames. ], tot_loss[loss=0.1712, simple_loss=0.2512, pruned_loss=0.04564, over 3273455.50 frames. ], batch size: 247, lr: 2.94e-03, grad_scale: 8.0 2023-05-01 15:03:51,651 INFO [optim.py:368] (1/8) Clipping_scale=2.0, grad-norm quartiles 1.638e+02 2.230e+02 2.530e+02 3.049e+02 6.179e+02, threshold=5.060e+02, percent-clipped=1.0 2023-05-01 15:04:47,722 INFO [train.py:904] (1/8) Epoch 23, batch 3950, loss[loss=0.1716, simple_loss=0.2404, pruned_loss=0.05139, over 16536.00 frames. ], tot_loss[loss=0.171, simple_loss=0.2507, pruned_loss=0.04568, over 3274584.15 frames. ], batch size: 75, lr: 2.94e-03, grad_scale: 8.0 2023-05-01 15:04:55,290 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.4984, 3.7738, 3.9555, 2.6357, 3.6370, 4.0678, 3.6517, 2.2364], device='cuda:1'), covar=tensor([0.0515, 0.0146, 0.0056, 0.0409, 0.0107, 0.0097, 0.0100, 0.0511], device='cuda:1'), in_proj_covar=tensor([0.0136, 0.0084, 0.0087, 0.0134, 0.0099, 0.0111, 0.0097, 0.0130], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0003, 0.0004], device='cuda:1') 2023-05-01 15:05:43,438 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([5.8430, 6.1469, 5.9378, 6.0856, 5.6725, 5.3947, 5.6355, 6.2953], device='cuda:1'), covar=tensor([0.1215, 0.0750, 0.0968, 0.0730, 0.0744, 0.0560, 0.1005, 0.0697], device='cuda:1'), in_proj_covar=tensor([0.0699, 0.0848, 0.0701, 0.0647, 0.0536, 0.0541, 0.0713, 0.0665], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:1') 2023-05-01 15:05:55,579 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.40 vs. limit=2.0 2023-05-01 15:06:00,628 INFO [train.py:904] (1/8) Epoch 23, batch 4000, loss[loss=0.174, simple_loss=0.2482, pruned_loss=0.04989, over 16724.00 frames. ], tot_loss[loss=0.1719, simple_loss=0.2514, pruned_loss=0.04617, over 3280545.13 frames. ], batch size: 124, lr: 2.94e-03, grad_scale: 8.0 2023-05-01 15:06:17,146 INFO [optim.py:368] (1/8) 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 15:07:01,808 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=227345.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 15:07:13,386 INFO [train.py:904] (1/8) Epoch 23, batch 4050, loss[loss=0.1623, simple_loss=0.2444, pruned_loss=0.04003, over 16445.00 frames. ], tot_loss[loss=0.1711, simple_loss=0.2517, pruned_loss=0.0452, over 3292575.90 frames. ], batch size: 35, lr: 2.94e-03, grad_scale: 8.0 2023-05-01 15:08:24,236 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([5.2152, 5.1963, 4.9923, 4.0060, 5.1728, 1.6484, 4.9084, 4.6536], device='cuda:1'), covar=tensor([0.0090, 0.0096, 0.0207, 0.0513, 0.0098, 0.3611, 0.0134, 0.0303], device='cuda:1'), in_proj_covar=tensor([0.0174, 0.0166, 0.0207, 0.0184, 0.0185, 0.0214, 0.0196, 0.0180], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-01 15:08:27,054 INFO [train.py:904] (1/8) Epoch 23, batch 4100, loss[loss=0.1858, simple_loss=0.2781, pruned_loss=0.04678, over 16541.00 frames. ], tot_loss[loss=0.1717, simple_loss=0.2537, pruned_loss=0.04491, over 3289830.17 frames. ], batch size: 68, lr: 2.94e-03, grad_scale: 8.0 2023-05-01 15:08:37,448 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.4050, 3.0019, 2.6868, 2.3152, 2.3027, 2.3483, 3.0013, 2.8961], device='cuda:1'), covar=tensor([0.2513, 0.0677, 0.1537, 0.2439, 0.2207, 0.2001, 0.0475, 0.1138], device='cuda:1'), in_proj_covar=tensor([0.0331, 0.0274, 0.0311, 0.0320, 0.0305, 0.0267, 0.0302, 0.0348], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:1') 2023-05-01 15:08:39,123 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=227411.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 15:08:42,768 INFO [optim.py:368] (1/8) 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 15:09:01,335 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.31 vs. limit=2.0 2023-05-01 15:09:19,572 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.29 vs. limit=2.0 2023-05-01 15:09:25,361 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.37 vs. limit=2.0 2023-05-01 15:09:45,383 INFO [train.py:904] (1/8) Epoch 23, batch 4150, loss[loss=0.2211, simple_loss=0.2976, pruned_loss=0.07228, over 11801.00 frames. ], tot_loss[loss=0.1776, simple_loss=0.2606, pruned_loss=0.04723, over 3248577.32 frames. ], batch size: 247, lr: 2.94e-03, grad_scale: 8.0 2023-05-01 15:10:15,648 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=227472.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 15:10:49,717 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.7303, 4.9249, 5.0843, 4.7940, 4.7876, 5.4322, 4.9302, 4.5929], device='cuda:1'), covar=tensor([0.1101, 0.1781, 0.1761, 0.1945, 0.2870, 0.0951, 0.1596, 0.2576], device='cuda:1'), in_proj_covar=tensor([0.0420, 0.0611, 0.0672, 0.0504, 0.0671, 0.0698, 0.0528, 0.0675], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-05-01 15:10:56,027 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.9057, 2.1330, 2.2089, 3.4011, 2.0820, 2.3778, 2.2419, 2.2672], device='cuda:1'), covar=tensor([0.1476, 0.3479, 0.2838, 0.0643, 0.4093, 0.2480, 0.3502, 0.3223], device='cuda:1'), in_proj_covar=tensor([0.0407, 0.0458, 0.0373, 0.0333, 0.0439, 0.0528, 0.0428, 0.0535], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-01 15:11:03,932 INFO [train.py:904] (1/8) Epoch 23, batch 4200, loss[loss=0.2011, simple_loss=0.2941, pruned_loss=0.05405, over 16726.00 frames. ], tot_loss[loss=0.1828, simple_loss=0.2675, pruned_loss=0.04898, over 3209023.59 frames. ], batch size: 57, lr: 2.94e-03, grad_scale: 8.0 2023-05-01 15:11:20,462 INFO [optim.py:368] (1/8) 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] (1/8) Epoch 23, batch 4250, loss[loss=0.1677, simple_loss=0.2667, pruned_loss=0.03435, over 16449.00 frames. ], tot_loss[loss=0.185, simple_loss=0.2713, pruned_loss=0.04928, over 3187316.34 frames. ], batch size: 68, lr: 2.94e-03, grad_scale: 8.0 2023-05-01 15:12:40,716 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=4.36 vs. limit=5.0 2023-05-01 15:13:36,401 INFO [train.py:904] (1/8) Epoch 23, batch 4300, loss[loss=0.1912, simple_loss=0.2844, pruned_loss=0.04898, over 16438.00 frames. ], tot_loss[loss=0.1847, simple_loss=0.2726, pruned_loss=0.04843, over 3194912.21 frames. ], batch size: 75, lr: 2.94e-03, grad_scale: 4.0 2023-05-01 15:13:55,081 INFO [optim.py:368] (1/8) 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 15:14:42,058 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=227645.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 15:14:51,651 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([1.9803, 3.5564, 3.4757, 2.1502, 3.2999, 3.5600, 3.2200, 2.0307], device='cuda:1'), covar=tensor([0.0572, 0.0044, 0.0059, 0.0468, 0.0092, 0.0088, 0.0115, 0.0435], device='cuda:1'), in_proj_covar=tensor([0.0136, 0.0084, 0.0087, 0.0134, 0.0099, 0.0111, 0.0097, 0.0130], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:1') 2023-05-01 15:14:53,553 INFO [train.py:904] (1/8) Epoch 23, batch 4350, loss[loss=0.1943, simple_loss=0.2803, pruned_loss=0.05413, over 11679.00 frames. ], tot_loss[loss=0.1867, simple_loss=0.275, pruned_loss=0.04922, over 3168337.19 frames. ], batch size: 246, lr: 2.94e-03, grad_scale: 4.0 2023-05-01 15:15:17,291 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.12 vs. limit=2.0 2023-05-01 15:15:55,219 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=227693.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 15:15:55,359 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=227693.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 15:16:09,820 INFO [train.py:904] (1/8) Epoch 23, batch 4400, loss[loss=0.1922, simple_loss=0.279, pruned_loss=0.05269, over 16889.00 frames. ], tot_loss[loss=0.1892, simple_loss=0.2772, pruned_loss=0.05057, over 3156496.19 frames. ], batch size: 116, lr: 2.94e-03, grad_scale: 8.0 2023-05-01 15:16:27,159 INFO [optim.py:368] (1/8) 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 15:17:22,038 INFO [train.py:904] (1/8) Epoch 23, batch 4450, loss[loss=0.2026, simple_loss=0.2893, pruned_loss=0.05793, over 16854.00 frames. ], tot_loss[loss=0.1923, simple_loss=0.2808, pruned_loss=0.05188, over 3175313.87 frames. ], batch size: 42, lr: 2.94e-03, grad_scale: 8.0 2023-05-01 15:17:23,592 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=227754.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 15:17:41,686 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=227767.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 15:18:19,165 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=2.93 vs. limit=5.0 2023-05-01 15:18:35,095 INFO [train.py:904] (1/8) Epoch 23, batch 4500, loss[loss=0.1814, simple_loss=0.2735, pruned_loss=0.04467, over 16761.00 frames. ], tot_loss[loss=0.1932, simple_loss=0.2815, pruned_loss=0.05249, over 3182412.44 frames. ], batch size: 76, lr: 2.94e-03, grad_scale: 8.0 2023-05-01 15:18:52,359 INFO [optim.py:368] (1/8) 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 15:19:11,660 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=227827.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 15:19:48,151 INFO [train.py:904] (1/8) Epoch 23, batch 4550, loss[loss=0.1895, simple_loss=0.2819, pruned_loss=0.04852, over 17264.00 frames. ], tot_loss[loss=0.1947, simple_loss=0.2827, pruned_loss=0.05338, over 3195144.56 frames. ], batch size: 52, lr: 2.94e-03, grad_scale: 8.0 2023-05-01 15:20:39,328 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=227888.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 15:21:00,782 INFO [train.py:904] (1/8) Epoch 23, batch 4600, loss[loss=0.1796, simple_loss=0.2694, pruned_loss=0.04494, over 16658.00 frames. ], tot_loss[loss=0.1951, simple_loss=0.2835, pruned_loss=0.05337, over 3212977.14 frames. ], batch size: 57, lr: 2.94e-03, grad_scale: 8.0 2023-05-01 15:21:18,784 INFO [optim.py:368] (1/8) 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 15:21:53,607 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=4.30 vs. limit=5.0 2023-05-01 15:22:14,194 INFO [train.py:904] (1/8) Epoch 23, batch 4650, loss[loss=0.2009, simple_loss=0.2842, pruned_loss=0.05883, over 16403.00 frames. ], tot_loss[loss=0.1949, simple_loss=0.2826, pruned_loss=0.05359, over 3202991.40 frames. ], batch size: 146, lr: 2.93e-03, grad_scale: 8.0 2023-05-01 15:23:28,874 INFO [train.py:904] (1/8) Epoch 23, batch 4700, loss[loss=0.1753, simple_loss=0.2594, pruned_loss=0.04561, over 16427.00 frames. ], tot_loss[loss=0.1925, simple_loss=0.2798, pruned_loss=0.05258, over 3192315.74 frames. ], batch size: 35, lr: 2.93e-03, grad_scale: 8.0 2023-05-01 15:23:34,968 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=228007.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 15:23:45,414 INFO [optim.py:368] (1/8) 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 15:24:08,038 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.0251, 3.7770, 3.7564, 2.1874, 3.3852, 3.7765, 3.3953, 1.9217], device='cuda:1'), covar=tensor([0.0645, 0.0048, 0.0051, 0.0504, 0.0094, 0.0097, 0.0121, 0.0537], device='cuda:1'), in_proj_covar=tensor([0.0137, 0.0085, 0.0087, 0.0135, 0.0100, 0.0112, 0.0097, 0.0131], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0003, 0.0004], device='cuda:1') 2023-05-01 15:24:34,809 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 23, batch 4750, loss[loss=0.165, simple_loss=0.2516, pruned_loss=0.03925, over 16644.00 frames. ], tot_loss[loss=0.1878, simple_loss=0.2755, pruned_loss=0.05011, over 3212779.58 frames. ], batch size: 62, lr: 2.93e-03, grad_scale: 8.0 2023-05-01 15:25:02,136 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=228067.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 15:25:03,396 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=228068.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 15:25:19,461 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=228080.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 15:25:54,062 INFO [train.py:904] (1/8) Epoch 23, batch 4800, loss[loss=0.1614, simple_loss=0.2586, pruned_loss=0.03207, over 16802.00 frames. ], tot_loss[loss=0.1836, simple_loss=0.2718, pruned_loss=0.04764, over 3227687.02 frames. ], batch size: 83, lr: 2.93e-03, grad_scale: 8.0 2023-05-01 15:26:10,775 INFO [optim.py:368] (1/8) 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] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=228115.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 15:26:37,280 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.46 vs. limit=2.0 2023-05-01 15:26:50,591 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=228141.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 15:27:07,821 INFO [train.py:904] (1/8) Epoch 23, batch 4850, loss[loss=0.1737, simple_loss=0.2759, pruned_loss=0.03578, over 16840.00 frames. ], tot_loss[loss=0.1829, simple_loss=0.272, pruned_loss=0.04695, over 3212291.63 frames. ], batch size: 102, lr: 2.93e-03, grad_scale: 8.0 2023-05-01 15:27:30,820 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.2719, 4.2815, 4.4584, 4.1992, 4.3332, 4.8236, 4.3636, 4.0421], device='cuda:1'), covar=tensor([0.1590, 0.1772, 0.1857, 0.2025, 0.2446, 0.1031, 0.1564, 0.2507], device='cuda:1'), in_proj_covar=tensor([0.0412, 0.0597, 0.0655, 0.0491, 0.0654, 0.0686, 0.0513, 0.0660], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-05-01 15:27:42,663 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([1.7851, 3.2363, 3.2302, 1.9622, 2.9758, 3.2274, 3.0311, 1.9966], device='cuda:1'), covar=tensor([0.0681, 0.0062, 0.0068, 0.0512, 0.0105, 0.0116, 0.0115, 0.0488], device='cuda:1'), in_proj_covar=tensor([0.0137, 0.0085, 0.0087, 0.0135, 0.0099, 0.0112, 0.0096, 0.0131], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0003, 0.0004], device='cuda:1') 2023-05-01 15:27:54,516 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=228183.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 15:28:23,601 INFO [train.py:904] (1/8) Epoch 23, batch 4900, loss[loss=0.184, simple_loss=0.2768, pruned_loss=0.04557, over 16387.00 frames. ], tot_loss[loss=0.1821, simple_loss=0.2719, pruned_loss=0.0462, over 3190678.09 frames. ], batch size: 165, lr: 2.93e-03, grad_scale: 8.0 2023-05-01 15:28:42,121 INFO [optim.py:368] (1/8) 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 15:28:54,668 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.63 vs. limit=2.0 2023-05-01 15:29:10,635 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.5960, 4.4524, 4.6667, 4.8178, 4.9794, 4.4899, 4.9513, 5.0165], device='cuda:1'), covar=tensor([0.1781, 0.1209, 0.1474, 0.0688, 0.0523, 0.1055, 0.0622, 0.0550], device='cuda:1'), in_proj_covar=tensor([0.0645, 0.0800, 0.0923, 0.0807, 0.0614, 0.0642, 0.0666, 0.0770], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-01 15:29:22,775 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.1965, 4.2762, 4.1193, 3.8159, 3.8392, 4.2099, 3.8816, 3.9561], device='cuda:1'), covar=tensor([0.0605, 0.0464, 0.0306, 0.0290, 0.0826, 0.0447, 0.0795, 0.0591], device='cuda:1'), in_proj_covar=tensor([0.0297, 0.0438, 0.0348, 0.0346, 0.0351, 0.0403, 0.0237, 0.0413], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002], device='cuda:1') 2023-05-01 15:29:36,711 INFO [train.py:904] (1/8) Epoch 23, batch 4950, loss[loss=0.2151, simple_loss=0.3016, pruned_loss=0.06426, over 16756.00 frames. ], tot_loss[loss=0.1815, simple_loss=0.2716, pruned_loss=0.04567, over 3183990.10 frames. ], batch size: 124, lr: 2.93e-03, grad_scale: 8.0 2023-05-01 15:29:50,809 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.37 vs. limit=2.0 2023-05-01 15:30:18,704 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.22 vs. limit=2.0 2023-05-01 15:30:51,758 INFO [train.py:904] (1/8) Epoch 23, batch 5000, loss[loss=0.1914, simple_loss=0.2859, pruned_loss=0.04839, over 16702.00 frames. ], tot_loss[loss=0.1826, simple_loss=0.2733, pruned_loss=0.04594, over 3186235.12 frames. ], batch size: 134, lr: 2.93e-03, grad_scale: 8.0 2023-05-01 15:30:58,379 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.4321, 3.4800, 2.1400, 3.9643, 2.6379, 3.9241, 2.2595, 2.8208], device='cuda:1'), covar=tensor([0.0295, 0.0348, 0.1629, 0.0129, 0.0851, 0.0480, 0.1487, 0.0798], device='cuda:1'), in_proj_covar=tensor([0.0170, 0.0175, 0.0192, 0.0164, 0.0175, 0.0215, 0.0199, 0.0178], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-05-01 15:31:09,998 INFO [optim.py:368] (1/8) 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 15:31:30,369 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=228329.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 15:31:36,869 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=228333.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 15:31:54,872 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.7907, 2.6559, 2.1251, 2.6014, 3.0635, 2.7623, 3.2450, 3.3011], device='cuda:1'), covar=tensor([0.0089, 0.0456, 0.0620, 0.0417, 0.0274, 0.0406, 0.0242, 0.0252], device='cuda:1'), in_proj_covar=tensor([0.0217, 0.0237, 0.0227, 0.0230, 0.0238, 0.0237, 0.0238, 0.0235], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-01 15:31:58,783 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=228349.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 15:32:04,601 INFO [train.py:904] (1/8) Epoch 23, batch 5050, loss[loss=0.1649, simple_loss=0.2522, pruned_loss=0.03875, over 16323.00 frames. ], tot_loss[loss=0.1828, simple_loss=0.2739, pruned_loss=0.04585, over 3188431.70 frames. ], batch size: 35, lr: 2.93e-03, grad_scale: 8.0 2023-05-01 15:32:12,029 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.36 vs. limit=2.0 2023-05-01 15:32:19,783 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=228363.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 15:32:30,609 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.9425, 2.7707, 2.8110, 2.0640, 2.6306, 2.1450, 2.7074, 2.9180], device='cuda:1'), covar=tensor([0.0324, 0.0756, 0.0568, 0.1823, 0.0858, 0.0903, 0.0712, 0.0706], device='cuda:1'), in_proj_covar=tensor([0.0158, 0.0167, 0.0169, 0.0154, 0.0146, 0.0130, 0.0144, 0.0178], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:1') 2023-05-01 15:32:40,897 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.23 vs. limit=2.0 2023-05-01 15:32:58,974 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=228390.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 15:33:04,525 INFO [zipformer.py:625] (1/8) 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] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=228397.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 15:33:17,238 INFO [train.py:904] (1/8) Epoch 23, batch 5100, loss[loss=0.1985, simple_loss=0.2857, pruned_loss=0.05559, over 15336.00 frames. ], tot_loss[loss=0.1809, simple_loss=0.2718, pruned_loss=0.045, over 3204591.36 frames. ], batch size: 192, lr: 2.93e-03, grad_scale: 8.0 2023-05-01 15:33:34,757 INFO [optim.py:368] (1/8) 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 15:33:55,500 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=228429.0, num_to_drop=1, layers_to_drop={0} 2023-05-01 15:34:05,834 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=228436.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 15:34:30,926 INFO [train.py:904] (1/8) Epoch 23, batch 5150, loss[loss=0.1707, simple_loss=0.2579, pruned_loss=0.04172, over 16674.00 frames. ], tot_loss[loss=0.18, simple_loss=0.2717, pruned_loss=0.04413, over 3198425.99 frames. ], batch size: 57, lr: 2.93e-03, grad_scale: 8.0 2023-05-01 15:34:54,549 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.2785, 2.5540, 1.9894, 2.3729, 2.9315, 2.5982, 2.9160, 3.1150], device='cuda:1'), covar=tensor([0.0140, 0.0483, 0.0636, 0.0462, 0.0268, 0.0400, 0.0218, 0.0276], device='cuda:1'), in_proj_covar=tensor([0.0217, 0.0238, 0.0228, 0.0231, 0.0239, 0.0238, 0.0239, 0.0236], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-01 15:34:57,759 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([1.8253, 3.1860, 3.1871, 2.0321, 2.9610, 3.1891, 2.9875, 1.8145], device='cuda:1'), covar=tensor([0.0674, 0.0068, 0.0072, 0.0505, 0.0114, 0.0120, 0.0138, 0.0610], device='cuda:1'), in_proj_covar=tensor([0.0137, 0.0085, 0.0087, 0.0135, 0.0100, 0.0112, 0.0096, 0.0131], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0003, 0.0004], device='cuda:1') 2023-05-01 15:35:15,457 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=228483.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 15:35:26,038 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=228490.0, num_to_drop=1, layers_to_drop={1} 2023-05-01 15:35:44,365 INFO [train.py:904] (1/8) Epoch 23, batch 5200, loss[loss=0.1621, simple_loss=0.258, pruned_loss=0.03312, over 16657.00 frames. ], tot_loss[loss=0.1791, simple_loss=0.2706, pruned_loss=0.04375, over 3196985.36 frames. ], batch size: 134, lr: 2.93e-03, grad_scale: 8.0 2023-05-01 15:35:47,936 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.1124, 2.9173, 3.1681, 1.6696, 3.2902, 3.3111, 2.7462, 2.4936], device='cuda:1'), covar=tensor([0.0864, 0.0272, 0.0163, 0.1307, 0.0094, 0.0174, 0.0402, 0.0549], device='cuda:1'), in_proj_covar=tensor([0.0149, 0.0109, 0.0098, 0.0139, 0.0082, 0.0126, 0.0128, 0.0131], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0004, 0.0004], device='cuda:1') 2023-05-01 15:36:01,315 INFO [optim.py:368] (1/8) 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] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=228531.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 15:36:57,633 INFO [train.py:904] (1/8) Epoch 23, batch 5250, loss[loss=0.1808, simple_loss=0.2636, pruned_loss=0.04905, over 17215.00 frames. ], tot_loss[loss=0.1773, simple_loss=0.2681, pruned_loss=0.04331, over 3208184.70 frames. ], batch size: 45, lr: 2.93e-03, grad_scale: 8.0 2023-05-01 15:37:59,496 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.7348, 4.1187, 3.1268, 2.5204, 2.8122, 2.7623, 4.4059, 3.5722], device='cuda:1'), covar=tensor([0.3058, 0.0635, 0.1810, 0.2751, 0.2715, 0.1893, 0.0489, 0.1388], device='cuda:1'), in_proj_covar=tensor([0.0330, 0.0271, 0.0308, 0.0316, 0.0300, 0.0263, 0.0300, 0.0342], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:1') 2023-05-01 15:38:10,576 INFO [train.py:904] (1/8) Epoch 23, batch 5300, loss[loss=0.1724, simple_loss=0.2585, pruned_loss=0.04312, over 16525.00 frames. ], tot_loss[loss=0.174, simple_loss=0.2641, pruned_loss=0.04192, over 3217915.96 frames. ], batch size: 75, lr: 2.93e-03, grad_scale: 8.0 2023-05-01 15:38:28,435 INFO [optim.py:368] (1/8) 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 15:38:28,860 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.5764, 3.5222, 3.5356, 2.8899, 3.4290, 1.9682, 3.2767, 2.9251], device='cuda:1'), covar=tensor([0.0155, 0.0155, 0.0174, 0.0315, 0.0121, 0.2470, 0.0159, 0.0264], device='cuda:1'), in_proj_covar=tensor([0.0166, 0.0160, 0.0199, 0.0177, 0.0178, 0.0206, 0.0188, 0.0172], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-01 15:39:23,394 INFO [train.py:904] (1/8) Epoch 23, batch 5350, loss[loss=0.1688, simple_loss=0.2558, pruned_loss=0.04088, over 16628.00 frames. ], tot_loss[loss=0.1733, simple_loss=0.2631, pruned_loss=0.04174, over 3214700.47 frames. ], batch size: 62, lr: 2.93e-03, grad_scale: 8.0 2023-05-01 15:39:38,458 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=228663.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 15:39:48,373 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.8322, 5.0540, 5.2266, 5.0286, 5.0309, 5.5959, 5.0210, 4.7477], device='cuda:1'), covar=tensor([0.1071, 0.1772, 0.1831, 0.1799, 0.2632, 0.0843, 0.1539, 0.2596], device='cuda:1'), in_proj_covar=tensor([0.0410, 0.0590, 0.0647, 0.0485, 0.0649, 0.0678, 0.0508, 0.0655], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-05-01 15:40:10,181 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=228685.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 15:40:16,736 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=228689.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 15:40:36,799 INFO [train.py:904] (1/8) Epoch 23, batch 5400, loss[loss=0.1869, simple_loss=0.285, pruned_loss=0.04444, over 16923.00 frames. ], tot_loss[loss=0.1752, simple_loss=0.2656, pruned_loss=0.04236, over 3205471.62 frames. ], batch size: 109, lr: 2.93e-03, grad_scale: 8.0 2023-05-01 15:40:48,792 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=228711.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 15:40:54,348 INFO [optim.py:368] (1/8) 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 15:41:26,156 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.5523, 4.4539, 4.3469, 2.8600, 3.7368, 4.3678, 3.7849, 2.4183], device='cuda:1'), covar=tensor([0.0559, 0.0032, 0.0039, 0.0394, 0.0104, 0.0074, 0.0101, 0.0464], device='cuda:1'), in_proj_covar=tensor([0.0136, 0.0084, 0.0086, 0.0134, 0.0099, 0.0110, 0.0096, 0.0130], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:1') 2023-05-01 15:41:27,291 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=228736.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 15:41:28,723 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.5608, 2.2679, 1.7847, 2.0778, 2.5745, 2.2925, 2.3592, 2.7491], device='cuda:1'), covar=tensor([0.0177, 0.0413, 0.0587, 0.0480, 0.0267, 0.0397, 0.0244, 0.0289], device='cuda:1'), in_proj_covar=tensor([0.0217, 0.0239, 0.0229, 0.0231, 0.0239, 0.0239, 0.0238, 0.0237], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-01 15:41:54,047 INFO [train.py:904] (1/8) Epoch 23, batch 5450, loss[loss=0.209, simple_loss=0.301, pruned_loss=0.05847, over 16149.00 frames. ], tot_loss[loss=0.1784, simple_loss=0.2689, pruned_loss=0.04397, over 3194459.80 frames. ], batch size: 165, lr: 2.93e-03, grad_scale: 8.0 2023-05-01 15:41:57,261 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.75 vs. limit=2.0 2023-05-01 15:42:42,358 INFO [zipformer.py:625] (1/8) 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] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=228785.0, num_to_drop=1, layers_to_drop={3} 2023-05-01 15:43:12,590 INFO [train.py:904] (1/8) Epoch 23, batch 5500, loss[loss=0.2044, simple_loss=0.2966, pruned_loss=0.05606, over 16298.00 frames. ], tot_loss[loss=0.1863, simple_loss=0.276, pruned_loss=0.04833, over 3153564.98 frames. ], batch size: 146, lr: 2.93e-03, grad_scale: 8.0 2023-05-01 15:43:32,426 INFO [optim.py:368] (1/8) 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 15:44:13,428 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([5.2258, 5.1759, 5.0388, 4.3344, 5.1165, 1.9579, 4.8648, 4.7235], device='cuda:1'), covar=tensor([0.0091, 0.0078, 0.0182, 0.0394, 0.0094, 0.2585, 0.0129, 0.0227], device='cuda:1'), in_proj_covar=tensor([0.0167, 0.0160, 0.0200, 0.0178, 0.0177, 0.0206, 0.0189, 0.0172], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-01 15:44:31,570 INFO [train.py:904] (1/8) Epoch 23, batch 5550, loss[loss=0.2078, simple_loss=0.2938, pruned_loss=0.06092, over 16225.00 frames. ], tot_loss[loss=0.1946, simple_loss=0.2826, pruned_loss=0.05325, over 3122222.88 frames. ], batch size: 165, lr: 2.93e-03, grad_scale: 8.0 2023-05-01 15:45:24,021 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.8369, 4.8495, 4.7293, 4.3654, 4.3681, 4.7577, 4.6225, 4.5225], device='cuda:1'), covar=tensor([0.0610, 0.0522, 0.0288, 0.0327, 0.0942, 0.0493, 0.0420, 0.0634], device='cuda:1'), in_proj_covar=tensor([0.0297, 0.0440, 0.0347, 0.0345, 0.0353, 0.0404, 0.0236, 0.0413], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002], device='cuda:1') 2023-05-01 15:45:45,318 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([1.8322, 2.9458, 2.6200, 4.7236, 3.4763, 4.1527, 1.8628, 3.1019], device='cuda:1'), covar=tensor([0.1328, 0.0750, 0.1228, 0.0176, 0.0327, 0.0436, 0.1557, 0.0811], device='cuda:1'), in_proj_covar=tensor([0.0168, 0.0175, 0.0196, 0.0193, 0.0206, 0.0215, 0.0204, 0.0194], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-05-01 15:45:52,971 INFO [train.py:904] (1/8) Epoch 23, batch 5600, loss[loss=0.2096, simple_loss=0.2923, pruned_loss=0.06343, over 16597.00 frames. ], tot_loss[loss=0.1996, simple_loss=0.2864, pruned_loss=0.05643, over 3104140.92 frames. ], batch size: 134, lr: 2.93e-03, grad_scale: 8.0 2023-05-01 15:45:58,190 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.8296, 3.8800, 2.9818, 2.4312, 2.7431, 2.5572, 4.2059, 3.5807], device='cuda:1'), covar=tensor([0.2669, 0.0630, 0.1720, 0.2590, 0.2487, 0.1960, 0.0459, 0.1133], device='cuda:1'), in_proj_covar=tensor([0.0329, 0.0271, 0.0307, 0.0317, 0.0299, 0.0263, 0.0299, 0.0341], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:1') 2023-05-01 15:46:10,855 INFO [optim.py:368] (1/8) 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 15:46:45,080 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=228935.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 15:46:53,950 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.44 vs. limit=2.0 2023-05-01 15:47:13,520 INFO [train.py:904] (1/8) Epoch 23, batch 5650, loss[loss=0.194, simple_loss=0.2813, pruned_loss=0.05336, over 16525.00 frames. ], tot_loss[loss=0.2055, simple_loss=0.291, pruned_loss=0.06001, over 3077890.84 frames. ], batch size: 68, lr: 2.93e-03, grad_scale: 8.0 2023-05-01 15:47:48,088 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.0271, 2.1294, 2.2377, 3.5375, 2.0871, 2.4848, 2.2533, 2.2904], device='cuda:1'), covar=tensor([0.1457, 0.3403, 0.2831, 0.0649, 0.4220, 0.2451, 0.3421, 0.3328], device='cuda:1'), in_proj_covar=tensor([0.0403, 0.0450, 0.0368, 0.0327, 0.0433, 0.0518, 0.0421, 0.0526], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-01 15:48:04,689 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=228985.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 15:48:10,635 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=228989.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 15:48:21,835 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=228996.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 15:48:32,721 INFO [train.py:904] (1/8) Epoch 23, batch 5700, loss[loss=0.2717, simple_loss=0.3399, pruned_loss=0.1017, over 11308.00 frames. ], tot_loss[loss=0.2075, simple_loss=0.2924, pruned_loss=0.06128, over 3082935.44 frames. ], batch size: 248, lr: 2.93e-03, grad_scale: 8.0 2023-05-01 15:48:50,700 INFO [optim.py:368] (1/8) Clipping_scale=2.0, grad-norm quartiles 2.137e+02 3.035e+02 3.772e+02 4.443e+02 6.175e+02, threshold=7.544e+02, percent-clipped=0.0 2023-05-01 15:49:18,826 INFO [zipformer.py:625] (1/8) 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] (1/8) 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] (1/8) Epoch 23, batch 5750, loss[loss=0.2173, simple_loss=0.302, pruned_loss=0.06625, over 16843.00 frames. ], tot_loss[loss=0.2107, simple_loss=0.2955, pruned_loss=0.0629, over 3070743.91 frames. ], batch size: 42, lr: 2.93e-03, grad_scale: 8.0 2023-05-01 15:50:19,289 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.9008, 2.1443, 2.4408, 3.1366, 2.2521, 2.3976, 2.3623, 2.2778], device='cuda:1'), covar=tensor([0.1410, 0.3261, 0.2310, 0.0729, 0.3687, 0.2200, 0.2979, 0.2870], device='cuda:1'), in_proj_covar=tensor([0.0401, 0.0449, 0.0367, 0.0326, 0.0432, 0.0517, 0.0420, 0.0524], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-01 15:50:23,763 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.42 vs. limit=2.0 2023-05-01 15:50:44,210 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=229085.0, num_to_drop=1, layers_to_drop={1} 2023-05-01 15:51:13,973 INFO [train.py:904] (1/8) Epoch 23, batch 5800, loss[loss=0.1822, simple_loss=0.2787, pruned_loss=0.04286, over 16722.00 frames. ], tot_loss[loss=0.2093, simple_loss=0.2951, pruned_loss=0.06176, over 3077057.65 frames. ], batch size: 124, lr: 2.93e-03, grad_scale: 8.0 2023-05-01 15:51:32,213 INFO [optim.py:368] (1/8) Clipping_scale=2.0, grad-norm quartiles 1.995e+02 2.918e+02 3.323e+02 4.052e+02 8.500e+02, threshold=6.645e+02, percent-clipped=1.0 2023-05-01 15:51:42,021 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.6451, 3.4568, 3.8881, 1.8908, 4.0301, 4.0565, 3.0330, 3.0516], device='cuda:1'), covar=tensor([0.0763, 0.0259, 0.0196, 0.1325, 0.0076, 0.0144, 0.0428, 0.0474], device='cuda:1'), in_proj_covar=tensor([0.0149, 0.0109, 0.0099, 0.0139, 0.0082, 0.0127, 0.0128, 0.0131], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0004, 0.0004], device='cuda:1') 2023-05-01 15:52:01,460 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=229133.0, num_to_drop=1, layers_to_drop={0} 2023-05-01 15:52:31,388 INFO [train.py:904] (1/8) Epoch 23, batch 5850, loss[loss=0.1873, simple_loss=0.2763, pruned_loss=0.04918, over 16749.00 frames. ], tot_loss[loss=0.2067, simple_loss=0.2929, pruned_loss=0.06028, over 3081622.14 frames. ], batch size: 83, lr: 2.93e-03, grad_scale: 8.0 2023-05-01 15:52:44,229 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.21 vs. limit=2.0 2023-05-01 15:53:50,472 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.0159, 3.3131, 3.1019, 5.3932, 4.1247, 4.6136, 2.1913, 3.7142], device='cuda:1'), covar=tensor([0.1280, 0.0679, 0.1053, 0.0134, 0.0311, 0.0308, 0.1406, 0.0602], device='cuda:1'), in_proj_covar=tensor([0.0168, 0.0175, 0.0196, 0.0193, 0.0206, 0.0216, 0.0204, 0.0194], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-05-01 15:53:53,137 INFO [train.py:904] (1/8) Epoch 23, batch 5900, loss[loss=0.2111, simple_loss=0.2838, pruned_loss=0.06916, over 11659.00 frames. ], tot_loss[loss=0.2067, simple_loss=0.2923, pruned_loss=0.06049, over 3060809.76 frames. ], batch size: 248, lr: 2.93e-03, grad_scale: 8.0 2023-05-01 15:54:16,066 INFO [optim.py:368] (1/8) 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] (1/8) Epoch 23, batch 5950, loss[loss=0.2249, simple_loss=0.3108, pruned_loss=0.06947, over 17007.00 frames. ], tot_loss[loss=0.2056, simple_loss=0.2925, pruned_loss=0.05932, over 3057379.60 frames. ], batch size: 50, lr: 2.93e-03, grad_scale: 8.0 2023-05-01 15:55:56,342 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.8973, 2.1552, 2.4539, 3.1356, 2.2344, 2.3661, 2.3527, 2.2930], device='cuda:1'), covar=tensor([0.1365, 0.3138, 0.2326, 0.0728, 0.3791, 0.2323, 0.2927, 0.3159], device='cuda:1'), in_proj_covar=tensor([0.0401, 0.0449, 0.0367, 0.0325, 0.0433, 0.0516, 0.0420, 0.0525], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-01 15:56:18,207 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=229291.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 15:56:33,829 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=229300.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 15:56:37,757 INFO [train.py:904] (1/8) Epoch 23, batch 6000, loss[loss=0.2067, simple_loss=0.2937, pruned_loss=0.05991, over 16986.00 frames. ], tot_loss[loss=0.2049, simple_loss=0.2916, pruned_loss=0.05912, over 3059016.21 frames. ], batch size: 55, lr: 2.93e-03, grad_scale: 8.0 2023-05-01 15:56:37,757 INFO [train.py:929] (1/8) Computing validation loss 2023-05-01 15:56:49,491 INFO [train.py:938] (1/8) Epoch 23, validation: loss=0.1497, simple_loss=0.2623, pruned_loss=0.01859, over 944034.00 frames. 2023-05-01 15:56:49,492 INFO [train.py:939] (1/8) Maximum memory allocated so far is 18145MB 2023-05-01 15:57:01,392 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([1.7015, 2.6104, 2.4805, 3.7916, 2.8107, 3.8988, 1.5385, 2.8424], device='cuda:1'), covar=tensor([0.1376, 0.0796, 0.1225, 0.0184, 0.0219, 0.0391, 0.1720, 0.0810], device='cuda:1'), in_proj_covar=tensor([0.0169, 0.0176, 0.0197, 0.0193, 0.0207, 0.0216, 0.0204, 0.0194], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-05-01 15:57:07,745 INFO [optim.py:368] (1/8) 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 15:58:06,099 INFO [train.py:904] (1/8) Epoch 23, batch 6050, loss[loss=0.1929, simple_loss=0.2686, pruned_loss=0.05859, over 11626.00 frames. ], tot_loss[loss=0.2033, simple_loss=0.2899, pruned_loss=0.05828, over 3072999.39 frames. ], batch size: 246, lr: 2.93e-03, grad_scale: 8.0 2023-05-01 15:58:20,410 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=229361.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 15:59:05,150 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.51 vs. limit=2.0 2023-05-01 15:59:19,833 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.19 vs. limit=2.0 2023-05-01 15:59:21,990 INFO [train.py:904] (1/8) Epoch 23, batch 6100, loss[loss=0.2208, simple_loss=0.2905, pruned_loss=0.07554, over 11128.00 frames. ], tot_loss[loss=0.2033, simple_loss=0.2903, pruned_loss=0.05818, over 3051909.32 frames. ], batch size: 247, lr: 2.93e-03, grad_scale: 8.0 2023-05-01 15:59:40,507 INFO [optim.py:368] (1/8) 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 16:00:22,339 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 23, batch 6150, loss[loss=0.1982, simple_loss=0.2863, pruned_loss=0.05506, over 17038.00 frames. ], tot_loss[loss=0.2015, simple_loss=0.2885, pruned_loss=0.0573, over 3073722.73 frames. ], batch size: 41, lr: 2.93e-03, grad_scale: 8.0 2023-05-01 16:01:15,182 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 23, batch 6200, loss[loss=0.2054, simple_loss=0.2911, pruned_loss=0.05984, over 16405.00 frames. ], tot_loss[loss=0.2002, simple_loss=0.2867, pruned_loss=0.05686, over 3083522.08 frames. ], batch size: 146, lr: 2.92e-03, grad_scale: 4.0 2023-05-01 16:01:55,799 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=229504.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 16:02:14,504 INFO [optim.py:368] (1/8) 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 16:02:49,428 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 23, batch 6250, loss[loss=0.2237, simple_loss=0.2974, pruned_loss=0.07507, over 11621.00 frames. ], tot_loss[loss=0.1987, simple_loss=0.2857, pruned_loss=0.05588, over 3106927.11 frames. ], batch size: 247, lr: 2.92e-03, grad_scale: 4.0 2023-05-01 16:03:17,498 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.1129, 4.1743, 2.5335, 4.8336, 3.2069, 4.7257, 2.6406, 3.3209], device='cuda:1'), covar=tensor([0.0222, 0.0334, 0.1712, 0.0275, 0.0777, 0.0605, 0.1552, 0.0746], device='cuda:1'), in_proj_covar=tensor([0.0171, 0.0176, 0.0194, 0.0165, 0.0177, 0.0217, 0.0202, 0.0180], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-05-01 16:03:22,964 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.6020, 3.7998, 2.6772, 2.2038, 2.5320, 2.2984, 3.9039, 3.3563], device='cuda:1'), covar=tensor([0.3235, 0.0623, 0.2085, 0.3072, 0.2790, 0.2292, 0.0588, 0.1351], device='cuda:1'), in_proj_covar=tensor([0.0331, 0.0271, 0.0308, 0.0317, 0.0299, 0.0263, 0.0300, 0.0342], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:1') 2023-05-01 16:03:48,901 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=229577.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 16:04:09,693 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=229591.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 16:04:28,569 INFO [train.py:904] (1/8) Epoch 23, batch 6300, loss[loss=0.2037, simple_loss=0.2866, pruned_loss=0.06039, over 15326.00 frames. ], tot_loss[loss=0.198, simple_loss=0.2855, pruned_loss=0.05526, over 3106863.67 frames. ], batch size: 190, lr: 2.92e-03, grad_scale: 4.0 2023-05-01 16:04:29,020 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.5430, 3.5009, 3.5030, 2.7437, 3.3640, 2.0377, 3.1319, 2.7740], device='cuda:1'), covar=tensor([0.0175, 0.0156, 0.0195, 0.0250, 0.0111, 0.2448, 0.0147, 0.0277], device='cuda:1'), in_proj_covar=tensor([0.0168, 0.0160, 0.0201, 0.0179, 0.0178, 0.0208, 0.0189, 0.0171], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-01 16:04:50,639 INFO [optim.py:368] (1/8) 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 16:05:27,294 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=229638.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 16:05:28,546 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=229639.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 16:05:41,504 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.4899, 3.4987, 2.8010, 2.1855, 2.3037, 2.2665, 3.6317, 3.1789], device='cuda:1'), covar=tensor([0.3130, 0.0671, 0.1787, 0.2965, 0.2716, 0.2249, 0.0561, 0.1422], device='cuda:1'), in_proj_covar=tensor([0.0330, 0.0270, 0.0306, 0.0316, 0.0298, 0.0262, 0.0299, 0.0341], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:1') 2023-05-01 16:05:48,007 INFO [train.py:904] (1/8) Epoch 23, batch 6350, loss[loss=0.2953, simple_loss=0.3443, pruned_loss=0.1231, over 11653.00 frames. ], tot_loss[loss=0.2012, simple_loss=0.2875, pruned_loss=0.05747, over 3085438.03 frames. ], batch size: 247, lr: 2.92e-03, grad_scale: 4.0 2023-05-01 16:05:48,994 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.25 vs. limit=2.0 2023-05-01 16:05:49,846 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.6888, 4.5325, 4.7091, 4.8971, 5.0698, 4.5441, 5.0678, 5.0746], device='cuda:1'), covar=tensor([0.1953, 0.1317, 0.1840, 0.0790, 0.0592, 0.0964, 0.0554, 0.0664], device='cuda:1'), in_proj_covar=tensor([0.0635, 0.0787, 0.0906, 0.0792, 0.0604, 0.0631, 0.0653, 0.0758], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-01 16:05:53,056 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=229656.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 16:05:57,416 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([1.9171, 2.7454, 2.5831, 1.9680, 2.5859, 2.6899, 2.5452, 1.9580], device='cuda:1'), covar=tensor([0.0451, 0.0104, 0.0092, 0.0375, 0.0133, 0.0135, 0.0127, 0.0413], device='cuda:1'), in_proj_covar=tensor([0.0135, 0.0084, 0.0085, 0.0134, 0.0098, 0.0111, 0.0095, 0.0129], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:1') 2023-05-01 16:05:57,433 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=229659.0, num_to_drop=1, layers_to_drop={1} 2023-05-01 16:06:52,307 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.1662, 2.0821, 2.7636, 3.0894, 2.9534, 3.6107, 2.2298, 3.5059], device='cuda:1'), covar=tensor([0.0215, 0.0547, 0.0323, 0.0326, 0.0318, 0.0148, 0.0569, 0.0133], device='cuda:1'), in_proj_covar=tensor([0.0192, 0.0194, 0.0180, 0.0185, 0.0200, 0.0156, 0.0199, 0.0155], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-01 16:07:04,876 INFO [train.py:904] (1/8) Epoch 23, batch 6400, loss[loss=0.1729, simple_loss=0.2649, pruned_loss=0.04044, over 16795.00 frames. ], tot_loss[loss=0.2017, simple_loss=0.2873, pruned_loss=0.05805, over 3091076.20 frames. ], batch size: 83, lr: 2.92e-03, grad_scale: 8.0 2023-05-01 16:07:11,481 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([5.0786, 5.6400, 5.7955, 5.4731, 5.5916, 6.0717, 5.5621, 5.3491], device='cuda:1'), covar=tensor([0.1010, 0.1723, 0.2296, 0.1960, 0.2508, 0.1064, 0.1807, 0.2468], device='cuda:1'), in_proj_covar=tensor([0.0414, 0.0598, 0.0659, 0.0492, 0.0655, 0.0690, 0.0515, 0.0662], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-05-01 16:07:14,766 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.6112, 4.6392, 4.4831, 4.1437, 4.1850, 4.5735, 4.3082, 4.2787], device='cuda:1'), covar=tensor([0.0590, 0.0553, 0.0307, 0.0342, 0.0898, 0.0463, 0.0597, 0.0638], device='cuda:1'), in_proj_covar=tensor([0.0293, 0.0437, 0.0343, 0.0339, 0.0346, 0.0398, 0.0233, 0.0409], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002], device='cuda:1') 2023-05-01 16:07:24,870 INFO [optim.py:368] (1/8) 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 16:07:32,397 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=229720.0, num_to_drop=1, layers_to_drop={2} 2023-05-01 16:07:48,424 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([5.0914, 5.0922, 4.9542, 4.1200, 4.9716, 1.7298, 4.7245, 4.6219], device='cuda:1'), covar=tensor([0.0143, 0.0142, 0.0211, 0.0513, 0.0132, 0.2955, 0.0245, 0.0262], device='cuda:1'), in_proj_covar=tensor([0.0169, 0.0160, 0.0201, 0.0179, 0.0178, 0.0208, 0.0190, 0.0172], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-01 16:08:16,898 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.1784, 3.6572, 3.6636, 2.1973, 3.0476, 2.5527, 3.6129, 3.8582], device='cuda:1'), covar=tensor([0.0332, 0.0747, 0.0652, 0.2158, 0.0956, 0.1015, 0.0666, 0.0920], device='cuda:1'), in_proj_covar=tensor([0.0157, 0.0166, 0.0168, 0.0154, 0.0146, 0.0131, 0.0143, 0.0177], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-05-01 16:08:19,149 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.9935, 4.0853, 4.2919, 4.2665, 4.2970, 4.0736, 4.0677, 4.0262], device='cuda:1'), covar=tensor([0.0364, 0.0591, 0.0474, 0.0485, 0.0452, 0.0435, 0.0896, 0.0533], device='cuda:1'), in_proj_covar=tensor([0.0414, 0.0460, 0.0447, 0.0416, 0.0492, 0.0468, 0.0554, 0.0374], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:1') 2023-05-01 16:08:21,138 INFO [train.py:904] (1/8) Epoch 23, batch 6450, loss[loss=0.1962, simple_loss=0.2694, pruned_loss=0.06155, over 11778.00 frames. ], tot_loss[loss=0.2007, simple_loss=0.2871, pruned_loss=0.05717, over 3094367.90 frames. ], batch size: 250, lr: 2.92e-03, grad_scale: 8.0 2023-05-01 16:08:31,332 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.85 vs. limit=2.0 2023-05-01 16:09:34,335 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=229799.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 16:09:40,107 INFO [train.py:904] (1/8) Epoch 23, batch 6500, loss[loss=0.2333, simple_loss=0.2982, pruned_loss=0.08424, over 11470.00 frames. ], tot_loss[loss=0.1989, simple_loss=0.285, pruned_loss=0.05642, over 3105673.34 frames. ], batch size: 247, lr: 2.92e-03, grad_scale: 8.0 2023-05-01 16:09:59,297 INFO [optim.py:368] (1/8) 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] (1/8) 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] (1/8) Epoch 23, batch 6550, loss[loss=0.1863, simple_loss=0.2916, pruned_loss=0.04047, over 16281.00 frames. ], tot_loss[loss=0.2019, simple_loss=0.2882, pruned_loss=0.05774, over 3092078.40 frames. ], batch size: 165, lr: 2.92e-03, grad_scale: 8.0 2023-05-01 16:12:02,453 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 23, batch 6600, loss[loss=0.2018, simple_loss=0.2924, pruned_loss=0.05557, over 16935.00 frames. ], tot_loss[loss=0.2034, simple_loss=0.2905, pruned_loss=0.05815, over 3088095.19 frames. ], batch size: 109, lr: 2.92e-03, grad_scale: 8.0 2023-05-01 16:12:35,465 INFO [optim.py:368] (1/8) 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 16:12:43,346 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.2566, 2.2799, 2.2628, 3.9285, 2.1993, 2.7231, 2.3829, 2.4040], device='cuda:1'), covar=tensor([0.1383, 0.3529, 0.2977, 0.0554, 0.4185, 0.2379, 0.3509, 0.3334], device='cuda:1'), in_proj_covar=tensor([0.0405, 0.0452, 0.0369, 0.0327, 0.0437, 0.0520, 0.0424, 0.0529], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-01 16:13:01,778 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=229933.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 16:13:16,251 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.8825, 5.1827, 5.3356, 5.1438, 5.1580, 5.7148, 5.1678, 4.9537], device='cuda:1'), covar=tensor([0.1104, 0.1837, 0.2162, 0.1921, 0.2463, 0.0956, 0.1683, 0.2351], device='cuda:1'), in_proj_covar=tensor([0.0417, 0.0603, 0.0664, 0.0496, 0.0660, 0.0692, 0.0519, 0.0668], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-05-01 16:13:33,862 INFO [train.py:904] (1/8) Epoch 23, batch 6650, loss[loss=0.174, simple_loss=0.2677, pruned_loss=0.04017, over 16765.00 frames. ], tot_loss[loss=0.2027, simple_loss=0.2897, pruned_loss=0.05784, over 3117255.43 frames. ], batch size: 83, lr: 2.92e-03, grad_scale: 8.0 2023-05-01 16:13:37,581 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=229955.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 16:13:38,748 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=229956.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 16:14:14,213 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.3634, 3.0054, 2.5401, 2.2473, 2.2423, 2.1193, 2.9601, 2.7966], device='cuda:1'), covar=tensor([0.2703, 0.0846, 0.1882, 0.2602, 0.2673, 0.2574, 0.0667, 0.1420], device='cuda:1'), in_proj_covar=tensor([0.0329, 0.0269, 0.0305, 0.0315, 0.0297, 0.0262, 0.0299, 0.0340], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:1') 2023-05-01 16:14:26,029 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=229987.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 16:14:52,372 INFO [train.py:904] (1/8) Epoch 23, batch 6700, loss[loss=0.2384, simple_loss=0.3073, pruned_loss=0.08482, over 11393.00 frames. ], tot_loss[loss=0.2026, simple_loss=0.2888, pruned_loss=0.0582, over 3094630.74 frames. ], batch size: 247, lr: 2.92e-03, grad_scale: 8.0 2023-05-01 16:14:54,791 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=230004.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 16:15:11,260 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=230015.0, num_to_drop=1, layers_to_drop={3} 2023-05-01 16:15:11,780 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.33 vs. limit=2.0 2023-05-01 16:15:12,081 INFO [optim.py:368] (1/8) 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 16:15:23,303 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.9440, 2.7181, 2.8864, 2.1320, 2.7002, 2.1464, 2.8013, 2.8806], device='cuda:1'), covar=tensor([0.0293, 0.0786, 0.0481, 0.1813, 0.0803, 0.0916, 0.0597, 0.0803], device='cuda:1'), in_proj_covar=tensor([0.0157, 0.0167, 0.0169, 0.0154, 0.0147, 0.0131, 0.0143, 0.0178], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-05-01 16:16:02,766 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=230048.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 16:16:09,702 INFO [train.py:904] (1/8) Epoch 23, batch 6750, loss[loss=0.1802, simple_loss=0.2647, pruned_loss=0.04789, over 17012.00 frames. ], tot_loss[loss=0.2019, simple_loss=0.2878, pruned_loss=0.05798, over 3108319.49 frames. ], batch size: 55, lr: 2.92e-03, grad_scale: 8.0 2023-05-01 16:16:22,524 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.52 vs. limit=2.0 2023-05-01 16:16:39,340 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.3608, 4.3555, 4.2111, 3.4578, 4.2808, 1.7026, 4.0649, 3.8646], device='cuda:1'), covar=tensor([0.0099, 0.0091, 0.0197, 0.0323, 0.0089, 0.2897, 0.0123, 0.0275], device='cuda:1'), in_proj_covar=tensor([0.0167, 0.0160, 0.0200, 0.0178, 0.0176, 0.0207, 0.0189, 0.0171], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-01 16:17:00,519 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.3610, 3.3239, 3.3962, 3.4847, 3.5115, 3.2908, 3.4930, 3.5718], device='cuda:1'), covar=tensor([0.1179, 0.0917, 0.0975, 0.0627, 0.0646, 0.2198, 0.1019, 0.0764], device='cuda:1'), in_proj_covar=tensor([0.0628, 0.0781, 0.0898, 0.0789, 0.0599, 0.0624, 0.0649, 0.0754], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-01 16:17:19,754 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 23, batch 6800, loss[loss=0.2596, simple_loss=0.3199, pruned_loss=0.09967, over 11535.00 frames. ], tot_loss[loss=0.2036, simple_loss=0.2888, pruned_loss=0.05917, over 3076825.45 frames. ], batch size: 246, lr: 2.92e-03, grad_scale: 8.0 2023-05-01 16:17:43,640 INFO [optim.py:368] (1/8) 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 16:18:11,074 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=230133.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 16:18:31,927 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=230147.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 16:18:36,450 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=230150.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 16:18:40,521 INFO [train.py:904] (1/8) Epoch 23, batch 6850, loss[loss=0.2576, simple_loss=0.3274, pruned_loss=0.0939, over 11483.00 frames. ], tot_loss[loss=0.2046, simple_loss=0.2906, pruned_loss=0.05933, over 3092392.36 frames. ], batch size: 248, lr: 2.92e-03, grad_scale: 8.0 2023-05-01 16:18:51,523 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.7698, 4.9556, 5.0946, 4.8611, 4.8368, 5.5115, 4.9471, 4.6127], device='cuda:1'), covar=tensor([0.1026, 0.1925, 0.2596, 0.2113, 0.2611, 0.0884, 0.1663, 0.2538], device='cuda:1'), in_proj_covar=tensor([0.0416, 0.0598, 0.0662, 0.0495, 0.0658, 0.0689, 0.0517, 0.0664], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-05-01 16:19:23,353 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=230181.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 16:19:56,852 INFO [train.py:904] (1/8) Epoch 23, batch 6900, loss[loss=0.1981, simple_loss=0.2893, pruned_loss=0.05344, over 17042.00 frames. ], tot_loss[loss=0.2041, simple_loss=0.2921, pruned_loss=0.0581, over 3108201.29 frames. ], batch size: 53, lr: 2.92e-03, grad_scale: 4.0 2023-05-01 16:20:10,763 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=230211.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 16:20:13,233 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.6222, 4.1680, 4.1923, 2.7986, 3.7582, 4.2448, 3.8407, 2.3118], device='cuda:1'), covar=tensor([0.0510, 0.0058, 0.0049, 0.0392, 0.0099, 0.0114, 0.0092, 0.0479], device='cuda:1'), in_proj_covar=tensor([0.0137, 0.0086, 0.0087, 0.0136, 0.0100, 0.0112, 0.0097, 0.0131], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0003, 0.0004], device='cuda:1') 2023-05-01 16:20:20,076 INFO [optim.py:368] (1/8) 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 16:20:45,536 INFO [zipformer.py:625] (1/8) 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] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=230241.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 16:21:13,081 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=230250.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 16:21:17,992 INFO [train.py:904] (1/8) Epoch 23, batch 6950, loss[loss=0.2047, simple_loss=0.2957, pruned_loss=0.05685, over 16814.00 frames. ], tot_loss[loss=0.2061, simple_loss=0.2935, pruned_loss=0.05929, over 3106048.15 frames. ], batch size: 116, lr: 2.92e-03, grad_scale: 4.0 2023-05-01 16:21:29,091 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.4835, 3.4748, 3.4826, 2.7370, 3.3124, 2.0941, 3.1250, 2.8592], device='cuda:1'), covar=tensor([0.0174, 0.0148, 0.0192, 0.0231, 0.0111, 0.2393, 0.0148, 0.0270], device='cuda:1'), in_proj_covar=tensor([0.0167, 0.0160, 0.0201, 0.0178, 0.0176, 0.0207, 0.0189, 0.0171], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-01 16:22:01,067 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=230281.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 16:22:34,144 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=230302.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 16:22:34,887 INFO [train.py:904] (1/8) Epoch 23, batch 7000, loss[loss=0.2356, simple_loss=0.3, pruned_loss=0.08559, over 11438.00 frames. ], tot_loss[loss=0.2061, simple_loss=0.2936, pruned_loss=0.0593, over 3091461.92 frames. ], batch size: 248, lr: 2.92e-03, grad_scale: 4.0 2023-05-01 16:22:52,289 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([1.7643, 2.9253, 2.7397, 4.8178, 3.4991, 4.2662, 1.8228, 3.1216], device='cuda:1'), covar=tensor([0.1442, 0.0749, 0.1182, 0.0153, 0.0271, 0.0401, 0.1627, 0.0802], device='cuda:1'), in_proj_covar=tensor([0.0170, 0.0177, 0.0198, 0.0193, 0.0208, 0.0217, 0.0205, 0.0195], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-05-01 16:22:53,571 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=230315.0, num_to_drop=1, layers_to_drop={1} 2023-05-01 16:22:56,093 INFO [optim.py:368] (1/8) 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 16:23:17,266 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.24 vs. limit=2.0 2023-05-01 16:23:34,442 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.94 vs. limit=2.0 2023-05-01 16:23:35,274 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 23, batch 7050, loss[loss=0.2161, simple_loss=0.2865, pruned_loss=0.07289, over 11345.00 frames. ], tot_loss[loss=0.2064, simple_loss=0.2938, pruned_loss=0.05948, over 3067901.45 frames. ], batch size: 250, lr: 2.92e-03, grad_scale: 4.0 2023-05-01 16:24:05,843 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=230363.0, num_to_drop=1, layers_to_drop={0} 2023-05-01 16:24:12,806 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.35 vs. limit=2.0 2023-05-01 16:25:02,695 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.9642, 3.8657, 4.0412, 4.1594, 4.2452, 3.8160, 4.1996, 4.2915], device='cuda:1'), covar=tensor([0.1738, 0.1279, 0.1381, 0.0703, 0.0621, 0.1706, 0.0873, 0.0726], device='cuda:1'), in_proj_covar=tensor([0.0624, 0.0778, 0.0893, 0.0783, 0.0596, 0.0621, 0.0647, 0.0752], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-01 16:25:07,480 INFO [train.py:904] (1/8) Epoch 23, batch 7100, loss[loss=0.1926, simple_loss=0.284, pruned_loss=0.05057, over 17029.00 frames. ], tot_loss[loss=0.2044, simple_loss=0.2917, pruned_loss=0.05856, over 3083289.14 frames. ], batch size: 50, lr: 2.92e-03, grad_scale: 4.0 2023-05-01 16:25:19,821 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.2243, 3.6953, 3.6784, 2.4222, 3.3499, 3.7046, 3.3991, 2.1394], device='cuda:1'), covar=tensor([0.0560, 0.0055, 0.0060, 0.0437, 0.0117, 0.0127, 0.0111, 0.0470], device='cuda:1'), in_proj_covar=tensor([0.0137, 0.0086, 0.0087, 0.0136, 0.0100, 0.0113, 0.0097, 0.0131], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0003, 0.0004], device='cuda:1') 2023-05-01 16:25:30,796 INFO [optim.py:368] (1/8) 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 16:26:01,631 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.71 vs. limit=2.0 2023-05-01 16:26:24,960 INFO [train.py:904] (1/8) Epoch 23, batch 7150, loss[loss=0.2214, simple_loss=0.3075, pruned_loss=0.06766, over 16782.00 frames. ], tot_loss[loss=0.2039, simple_loss=0.2902, pruned_loss=0.05879, over 3070741.76 frames. ], batch size: 124, lr: 2.92e-03, grad_scale: 4.0 2023-05-01 16:26:48,723 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=230468.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 16:27:17,322 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.5110, 4.1236, 4.0814, 2.7931, 3.6049, 4.1281, 3.6861, 2.2999], device='cuda:1'), covar=tensor([0.0546, 0.0047, 0.0051, 0.0408, 0.0114, 0.0111, 0.0102, 0.0473], device='cuda:1'), in_proj_covar=tensor([0.0137, 0.0086, 0.0087, 0.0136, 0.0100, 0.0112, 0.0097, 0.0131], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0003, 0.0004], device='cuda:1') 2023-05-01 16:27:17,662 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.28 vs. limit=2.0 2023-05-01 16:27:42,212 INFO [train.py:904] (1/8) Epoch 23, batch 7200, loss[loss=0.1861, simple_loss=0.2817, pruned_loss=0.04523, over 16423.00 frames. ], tot_loss[loss=0.2023, simple_loss=0.2889, pruned_loss=0.05788, over 3042489.12 frames. ], batch size: 146, lr: 2.92e-03, grad_scale: 8.0 2023-05-01 16:27:47,310 INFO [zipformer.py:625] (1/8) 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] (1/8) 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 16:28:16,805 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=230525.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 16:28:22,554 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=230529.0, num_to_drop=1, layers_to_drop={2} 2023-05-01 16:28:29,590 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=230533.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 16:28:58,913 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=230550.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 16:29:02,531 INFO [train.py:904] (1/8) Epoch 23, batch 7250, loss[loss=0.1773, simple_loss=0.2591, pruned_loss=0.04769, over 17015.00 frames. ], tot_loss[loss=0.2001, simple_loss=0.2864, pruned_loss=0.0569, over 3044512.39 frames. ], batch size: 55, lr: 2.92e-03, grad_scale: 8.0 2023-05-01 16:29:51,471 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=230586.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 16:29:58,449 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.8346, 4.8314, 5.2188, 5.1946, 5.2265, 4.9175, 4.8752, 4.6577], device='cuda:1'), covar=tensor([0.0367, 0.0567, 0.0416, 0.0403, 0.0541, 0.0385, 0.1016, 0.0534], device='cuda:1'), in_proj_covar=tensor([0.0416, 0.0462, 0.0448, 0.0416, 0.0493, 0.0470, 0.0557, 0.0375], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:1') 2023-05-01 16:30:03,496 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=230594.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 16:30:08,116 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=230597.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 16:30:10,071 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=230598.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 16:30:18,070 INFO [train.py:904] (1/8) Epoch 23, batch 7300, loss[loss=0.208, simple_loss=0.2759, pruned_loss=0.07008, over 11124.00 frames. ], tot_loss[loss=0.1997, simple_loss=0.2861, pruned_loss=0.05669, over 3063688.07 frames. ], batch size: 246, lr: 2.92e-03, grad_scale: 8.0 2023-05-01 16:30:21,183 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.83 vs. limit=2.0 2023-05-01 16:30:39,631 INFO [optim.py:368] (1/8) 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 16:30:57,553 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.8013, 3.9786, 3.0610, 2.4183, 2.8173, 2.6497, 4.5424, 3.5076], device='cuda:1'), covar=tensor([0.2813, 0.0637, 0.1661, 0.2436, 0.2458, 0.1901, 0.0370, 0.1237], device='cuda:1'), in_proj_covar=tensor([0.0331, 0.0271, 0.0308, 0.0318, 0.0300, 0.0263, 0.0300, 0.0341], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:1') 2023-05-01 16:30:59,101 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.24 vs. limit=2.0 2023-05-01 16:31:19,773 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 23, batch 7350, loss[loss=0.2018, simple_loss=0.2873, pruned_loss=0.05816, over 16460.00 frames. ], tot_loss[loss=0.2011, simple_loss=0.2872, pruned_loss=0.0575, over 3060603.08 frames. ], batch size: 68, lr: 2.92e-03, grad_scale: 8.0 2023-05-01 16:31:34,901 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.6155, 1.7993, 2.2863, 2.5462, 2.5250, 2.9593, 2.0245, 2.8354], device='cuda:1'), covar=tensor([0.0234, 0.0543, 0.0305, 0.0343, 0.0356, 0.0189, 0.0512, 0.0168], device='cuda:1'), in_proj_covar=tensor([0.0190, 0.0193, 0.0179, 0.0185, 0.0199, 0.0155, 0.0197, 0.0154], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-01 16:32:33,454 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 23, batch 7400, loss[loss=0.2366, simple_loss=0.3063, pruned_loss=0.08348, over 11651.00 frames. ], tot_loss[loss=0.2032, simple_loss=0.2887, pruned_loss=0.05881, over 3029112.98 frames. ], batch size: 248, lr: 2.92e-03, grad_scale: 8.0 2023-05-01 16:33:16,062 INFO [optim.py:368] (1/8) 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 16:34:13,430 INFO [train.py:904] (1/8) Epoch 23, batch 7450, loss[loss=0.1921, simple_loss=0.2866, pruned_loss=0.04878, over 16395.00 frames. ], tot_loss[loss=0.2046, simple_loss=0.2898, pruned_loss=0.05965, over 3041599.15 frames. ], batch size: 146, lr: 2.92e-03, grad_scale: 8.0 2023-05-01 16:34:59,233 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.5129, 4.0169, 4.0962, 2.6967, 3.6731, 4.1658, 3.6751, 2.3441], device='cuda:1'), covar=tensor([0.0524, 0.0066, 0.0057, 0.0412, 0.0109, 0.0110, 0.0095, 0.0465], device='cuda:1'), in_proj_covar=tensor([0.0136, 0.0086, 0.0086, 0.0135, 0.0099, 0.0111, 0.0096, 0.0130], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0003, 0.0004], device='cuda:1') 2023-05-01 16:35:18,610 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=230792.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 16:35:35,339 INFO [train.py:904] (1/8) Epoch 23, batch 7500, loss[loss=0.2057, simple_loss=0.2883, pruned_loss=0.06155, over 16903.00 frames. ], tot_loss[loss=0.2046, simple_loss=0.2904, pruned_loss=0.05942, over 3035426.94 frames. ], batch size: 109, lr: 2.92e-03, grad_scale: 8.0 2023-05-01 16:35:39,122 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.4804, 2.2094, 1.8181, 1.9921, 2.5494, 2.1664, 2.2430, 2.6494], device='cuda:1'), covar=tensor([0.0228, 0.0477, 0.0563, 0.0488, 0.0256, 0.0408, 0.0217, 0.0277], device='cuda:1'), in_proj_covar=tensor([0.0211, 0.0233, 0.0224, 0.0226, 0.0235, 0.0232, 0.0232, 0.0230], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-01 16:35:40,974 INFO [zipformer.py:625] (1/8) 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] (1/8) 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 16:36:09,660 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=230824.0, num_to_drop=1, layers_to_drop={1} 2023-05-01 16:36:52,031 INFO [train.py:904] (1/8) Epoch 23, batch 7550, loss[loss=0.1965, simple_loss=0.2783, pruned_loss=0.05735, over 17048.00 frames. ], tot_loss[loss=0.2045, simple_loss=0.2896, pruned_loss=0.05968, over 3034305.54 frames. ], batch size: 50, lr: 2.92e-03, grad_scale: 8.0 2023-05-01 16:36:52,519 INFO [zipformer.py:625] (1/8) 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] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=230854.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 16:37:01,916 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.0243, 3.0270, 1.9288, 3.2523, 2.3871, 3.3173, 2.0898, 2.5020], device='cuda:1'), covar=tensor([0.0344, 0.0429, 0.1579, 0.0245, 0.0816, 0.0600, 0.1540, 0.0805], device='cuda:1'), in_proj_covar=tensor([0.0171, 0.0176, 0.0194, 0.0164, 0.0176, 0.0217, 0.0203, 0.0180], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-05-01 16:37:33,400 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=230881.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 16:37:45,189 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=230889.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 16:37:47,070 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=230890.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 16:37:58,367 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 23, batch 7600, loss[loss=0.2308, simple_loss=0.2957, pruned_loss=0.08298, over 11417.00 frames. ], tot_loss[loss=0.2044, simple_loss=0.289, pruned_loss=0.0599, over 3034630.27 frames. ], batch size: 247, lr: 2.92e-03, grad_scale: 8.0 2023-05-01 16:38:27,912 INFO [optim.py:368] (1/8) 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] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=230945.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 16:39:19,836 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=230951.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 16:39:21,941 INFO [train.py:904] (1/8) Epoch 23, batch 7650, loss[loss=0.2333, simple_loss=0.3089, pruned_loss=0.07882, over 11478.00 frames. ], tot_loss[loss=0.2044, simple_loss=0.2891, pruned_loss=0.05984, over 3061250.70 frames. ], batch size: 246, lr: 2.92e-03, grad_scale: 8.0 2023-05-01 16:39:26,315 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.13 vs. limit=2.0 2023-05-01 16:39:53,032 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.32 vs. limit=2.0 2023-05-01 16:40:12,983 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.8812, 2.1324, 2.4176, 3.1503, 2.2112, 2.3206, 2.3409, 2.2609], device='cuda:1'), covar=tensor([0.1360, 0.3298, 0.2472, 0.0706, 0.3980, 0.2407, 0.3182, 0.3462], device='cuda:1'), in_proj_covar=tensor([0.0399, 0.0449, 0.0367, 0.0324, 0.0433, 0.0515, 0.0420, 0.0524], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-01 16:40:35,934 INFO [train.py:904] (1/8) Epoch 23, batch 7700, loss[loss=0.1882, simple_loss=0.2788, pruned_loss=0.04876, over 16902.00 frames. ], tot_loss[loss=0.2039, simple_loss=0.2888, pruned_loss=0.05952, over 3068979.97 frames. ], batch size: 90, lr: 2.92e-03, grad_scale: 8.0 2023-05-01 16:40:43,305 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([1.8796, 2.9626, 2.6246, 4.8243, 3.6894, 4.2681, 1.6712, 3.2505], device='cuda:1'), covar=tensor([0.1359, 0.0751, 0.1304, 0.0173, 0.0329, 0.0409, 0.1718, 0.0808], device='cuda:1'), in_proj_covar=tensor([0.0171, 0.0178, 0.0199, 0.0194, 0.0209, 0.0218, 0.0206, 0.0197], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-05-01 16:40:57,683 INFO [optim.py:368] (1/8) Clipping_scale=2.0, grad-norm quartiles 2.025e+02 2.849e+02 3.493e+02 4.384e+02 8.642e+02, threshold=6.986e+02, percent-clipped=7.0 2023-05-01 16:41:53,923 INFO [train.py:904] (1/8) Epoch 23, batch 7750, loss[loss=0.1995, simple_loss=0.286, pruned_loss=0.05651, over 16997.00 frames. ], tot_loss[loss=0.2031, simple_loss=0.2883, pruned_loss=0.05894, over 3068473.19 frames. ], batch size: 41, lr: 2.92e-03, grad_scale: 8.0 2023-05-01 16:43:09,649 INFO [train.py:904] (1/8) Epoch 23, batch 7800, loss[loss=0.1855, simple_loss=0.2738, pruned_loss=0.04859, over 16364.00 frames. ], tot_loss[loss=0.2031, simple_loss=0.2883, pruned_loss=0.05892, over 3073871.05 frames. ], batch size: 146, lr: 2.91e-03, grad_scale: 8.0 2023-05-01 16:43:30,321 INFO [optim.py:368] (1/8) Clipping_scale=2.0, grad-norm quartiles 2.091e+02 2.907e+02 3.422e+02 4.020e+02 9.124e+02, threshold=6.845e+02, percent-clipped=1.0 2023-05-01 16:43:41,679 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=231124.0, num_to_drop=1, layers_to_drop={1} 2023-05-01 16:43:57,918 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.5248, 4.8404, 4.6201, 4.6250, 4.3602, 4.3342, 4.3110, 4.8982], device='cuda:1'), covar=tensor([0.1250, 0.0877, 0.1002, 0.1010, 0.0878, 0.1336, 0.1254, 0.0973], device='cuda:1'), in_proj_covar=tensor([0.0675, 0.0817, 0.0680, 0.0629, 0.0520, 0.0532, 0.0691, 0.0647], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:1') 2023-05-01 16:44:17,272 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=231148.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 16:44:24,904 INFO [train.py:904] (1/8) Epoch 23, batch 7850, loss[loss=0.2054, simple_loss=0.2987, pruned_loss=0.056, over 16800.00 frames. ], tot_loss[loss=0.2032, simple_loss=0.2892, pruned_loss=0.05859, over 3086922.20 frames. ], batch size: 83, lr: 2.91e-03, grad_scale: 8.0 2023-05-01 16:44:52,236 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=231172.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 16:44:52,345 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.8350, 4.8974, 5.2954, 5.2310, 5.2835, 4.9498, 4.8834, 4.6005], device='cuda:1'), covar=tensor([0.0375, 0.0457, 0.0337, 0.0413, 0.0476, 0.0361, 0.1025, 0.0528], device='cuda:1'), in_proj_covar=tensor([0.0414, 0.0459, 0.0445, 0.0414, 0.0490, 0.0468, 0.0553, 0.0374], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:1') 2023-05-01 16:45:05,901 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=231181.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 16:45:16,869 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=231189.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 16:45:37,200 INFO [train.py:904] (1/8) Epoch 23, batch 7900, loss[loss=0.2381, simple_loss=0.3162, pruned_loss=0.07995, over 16705.00 frames. ], tot_loss[loss=0.2026, simple_loss=0.2887, pruned_loss=0.0583, over 3101366.87 frames. ], batch size: 134, lr: 2.91e-03, grad_scale: 8.0 2023-05-01 16:45:57,446 INFO [optim.py:368] (1/8) 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 16:45:58,621 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=4.95 vs. limit=5.0 2023-05-01 16:46:16,871 INFO [zipformer.py:625] (1/8) 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] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=231237.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 16:46:44,785 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=231246.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 16:46:52,153 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=231251.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 16:46:54,844 INFO [train.py:904] (1/8) Epoch 23, batch 7950, loss[loss=0.2182, simple_loss=0.3002, pruned_loss=0.06806, over 16658.00 frames. ], tot_loss[loss=0.2035, simple_loss=0.2893, pruned_loss=0.05883, over 3092953.45 frames. ], batch size: 62, lr: 2.91e-03, grad_scale: 8.0 2023-05-01 16:48:11,517 INFO [train.py:904] (1/8) Epoch 23, batch 8000, loss[loss=0.2202, simple_loss=0.3007, pruned_loss=0.06988, over 16710.00 frames. ], tot_loss[loss=0.2052, simple_loss=0.2905, pruned_loss=0.05993, over 3080819.25 frames. ], batch size: 134, lr: 2.91e-03, grad_scale: 8.0 2023-05-01 16:48:15,228 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([1.7856, 1.8301, 1.6370, 1.4959, 1.9660, 1.6208, 1.6803, 1.9079], device='cuda:1'), covar=tensor([0.0180, 0.0261, 0.0387, 0.0346, 0.0214, 0.0255, 0.0185, 0.0197], device='cuda:1'), in_proj_covar=tensor([0.0211, 0.0233, 0.0224, 0.0227, 0.0234, 0.0231, 0.0232, 0.0231], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-01 16:48:26,430 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=231312.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 16:48:33,450 INFO [optim.py:368] (1/8) 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 16:49:18,003 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.2615, 2.9715, 3.3317, 1.7900, 3.3960, 3.4980, 2.7370, 2.6010], device='cuda:1'), covar=tensor([0.0793, 0.0292, 0.0187, 0.1146, 0.0095, 0.0181, 0.0470, 0.0457], device='cuda:1'), in_proj_covar=tensor([0.0147, 0.0108, 0.0098, 0.0138, 0.0081, 0.0125, 0.0127, 0.0129], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0004, 0.0003], device='cuda:1') 2023-05-01 16:49:18,250 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.12 vs. limit=2.0 2023-05-01 16:49:26,909 INFO [train.py:904] (1/8) Epoch 23, batch 8050, loss[loss=0.2365, simple_loss=0.3069, pruned_loss=0.08304, over 11527.00 frames. ], tot_loss[loss=0.2048, simple_loss=0.2907, pruned_loss=0.05948, over 3084590.15 frames. ], batch size: 246, lr: 2.91e-03, grad_scale: 8.0 2023-05-01 16:49:28,229 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.12 vs. limit=2.0 2023-05-01 16:50:42,758 INFO [train.py:904] (1/8) Epoch 23, batch 8100, loss[loss=0.2327, simple_loss=0.2972, pruned_loss=0.08414, over 11468.00 frames. ], tot_loss[loss=0.2039, simple_loss=0.29, pruned_loss=0.05894, over 3081765.59 frames. ], batch size: 246, lr: 2.91e-03, grad_scale: 8.0 2023-05-01 16:51:04,319 INFO [optim.py:368] (1/8) 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 16:51:23,013 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.83 vs. limit=2.0 2023-05-01 16:51:51,035 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=231448.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 16:51:59,308 INFO [train.py:904] (1/8) Epoch 23, batch 8150, loss[loss=0.1696, simple_loss=0.2537, pruned_loss=0.04281, over 16438.00 frames. ], tot_loss[loss=0.2015, simple_loss=0.2876, pruned_loss=0.05777, over 3090350.22 frames. ], batch size: 68, lr: 2.91e-03, grad_scale: 8.0 2023-05-01 16:52:17,486 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.17 vs. limit=2.0 2023-05-01 16:52:43,858 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.1926, 3.7588, 3.7791, 2.3755, 3.4205, 3.7542, 3.3975, 2.0623], device='cuda:1'), covar=tensor([0.0583, 0.0056, 0.0053, 0.0449, 0.0110, 0.0113, 0.0113, 0.0492], device='cuda:1'), in_proj_covar=tensor([0.0136, 0.0085, 0.0086, 0.0134, 0.0098, 0.0111, 0.0096, 0.0130], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:1') 2023-05-01 16:53:05,170 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=231496.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 16:53:14,921 INFO [train.py:904] (1/8) Epoch 23, batch 8200, loss[loss=0.1719, simple_loss=0.2627, pruned_loss=0.0406, over 16707.00 frames. ], tot_loss[loss=0.2, simple_loss=0.2854, pruned_loss=0.05732, over 3101782.39 frames. ], batch size: 134, lr: 2.91e-03, grad_scale: 8.0 2023-05-01 16:53:38,110 INFO [optim.py:368] (1/8) 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 16:54:26,251 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 23, batch 8250, loss[loss=0.1683, simple_loss=0.2571, pruned_loss=0.03979, over 12150.00 frames. ], tot_loss[loss=0.197, simple_loss=0.2847, pruned_loss=0.05463, over 3105648.00 frames. ], batch size: 246, lr: 2.91e-03, grad_scale: 8.0 2023-05-01 16:55:13,528 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.17 vs. limit=2.0 2023-05-01 16:55:42,909 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=231594.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 16:55:43,185 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.6510, 3.7734, 2.9379, 2.2069, 2.3252, 2.3996, 3.9702, 3.2643], device='cuda:1'), covar=tensor([0.2831, 0.0524, 0.1629, 0.2988, 0.2901, 0.2271, 0.0387, 0.1297], device='cuda:1'), in_proj_covar=tensor([0.0330, 0.0269, 0.0307, 0.0316, 0.0299, 0.0263, 0.0297, 0.0339], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:1') 2023-05-01 16:55:57,538 INFO [train.py:904] (1/8) Epoch 23, batch 8300, loss[loss=0.1761, simple_loss=0.274, pruned_loss=0.03909, over 16255.00 frames. ], tot_loss[loss=0.1933, simple_loss=0.2824, pruned_loss=0.05208, over 3098248.48 frames. ], batch size: 165, lr: 2.91e-03, grad_scale: 4.0 2023-05-01 16:56:04,247 INFO [zipformer.py:625] (1/8) 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] (1/8) 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 16:57:11,268 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=231647.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 16:57:19,276 INFO [train.py:904] (1/8) Epoch 23, batch 8350, loss[loss=0.1938, simple_loss=0.2736, pruned_loss=0.05703, over 11956.00 frames. ], tot_loss[loss=0.1903, simple_loss=0.2807, pruned_loss=0.04999, over 3059482.01 frames. ], batch size: 248, lr: 2.91e-03, grad_scale: 4.0 2023-05-01 16:57:34,329 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.9210, 4.9411, 5.3432, 5.3008, 5.3088, 5.0288, 4.8921, 4.8094], device='cuda:1'), covar=tensor([0.0357, 0.0685, 0.0395, 0.0404, 0.0466, 0.0436, 0.1043, 0.0450], device='cuda:1'), in_proj_covar=tensor([0.0410, 0.0456, 0.0442, 0.0410, 0.0487, 0.0463, 0.0548, 0.0370], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:1') 2023-05-01 16:58:41,947 INFO [train.py:904] (1/8) Epoch 23, batch 8400, loss[loss=0.1728, simple_loss=0.2691, pruned_loss=0.03823, over 16787.00 frames. ], tot_loss[loss=0.1863, simple_loss=0.2777, pruned_loss=0.0475, over 3057855.49 frames. ], batch size: 124, lr: 2.91e-03, grad_scale: 8.0 2023-05-01 16:58:50,853 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=231708.0, num_to_drop=1, layers_to_drop={3} 2023-05-01 16:59:06,589 INFO [optim.py:368] (1/8) 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 16:59:39,036 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=3.06 vs. limit=5.0 2023-05-01 17:00:04,853 INFO [train.py:904] (1/8) Epoch 23, batch 8450, loss[loss=0.1641, simple_loss=0.2615, pruned_loss=0.03341, over 16231.00 frames. ], tot_loss[loss=0.1842, simple_loss=0.2759, pruned_loss=0.04626, over 3033863.55 frames. ], batch size: 165, lr: 2.91e-03, grad_scale: 8.0 2023-05-01 17:00:40,468 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([1.7660, 1.3615, 1.6398, 1.7083, 1.8848, 1.8737, 1.6807, 1.7560], device='cuda:1'), covar=tensor([0.0291, 0.0424, 0.0230, 0.0294, 0.0275, 0.0196, 0.0444, 0.0157], device='cuda:1'), in_proj_covar=tensor([0.0186, 0.0190, 0.0176, 0.0181, 0.0194, 0.0152, 0.0194, 0.0151], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-01 17:00:41,661 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.6606, 4.6135, 4.4270, 3.8262, 4.4977, 1.7891, 4.2756, 4.2304], device='cuda:1'), covar=tensor([0.0105, 0.0120, 0.0203, 0.0368, 0.0115, 0.2814, 0.0151, 0.0247], device='cuda:1'), in_proj_covar=tensor([0.0167, 0.0159, 0.0200, 0.0177, 0.0176, 0.0208, 0.0188, 0.0169], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-01 17:01:00,683 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.9877, 3.8537, 4.0786, 4.1697, 4.2721, 3.8419, 4.2335, 4.3096], device='cuda:1'), covar=tensor([0.1748, 0.1208, 0.1329, 0.0747, 0.0662, 0.1608, 0.0767, 0.0791], device='cuda:1'), in_proj_covar=tensor([0.0623, 0.0775, 0.0888, 0.0780, 0.0596, 0.0622, 0.0649, 0.0757], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-01 17:01:24,902 INFO [train.py:904] (1/8) Epoch 23, batch 8500, loss[loss=0.1489, simple_loss=0.2368, pruned_loss=0.03055, over 11970.00 frames. ], tot_loss[loss=0.1802, simple_loss=0.2723, pruned_loss=0.04408, over 3041185.06 frames. ], batch size: 246, lr: 2.91e-03, grad_scale: 8.0 2023-05-01 17:01:42,096 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.0615, 2.3863, 2.0351, 2.1767, 2.7115, 2.4026, 2.6003, 2.8731], device='cuda:1'), covar=tensor([0.0188, 0.0446, 0.0521, 0.0510, 0.0288, 0.0410, 0.0245, 0.0285], device='cuda:1'), in_proj_covar=tensor([0.0207, 0.0229, 0.0221, 0.0223, 0.0231, 0.0229, 0.0228, 0.0226], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-01 17:01:48,516 INFO [optim.py:368] (1/8) 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 17:02:48,224 INFO [train.py:904] (1/8) Epoch 23, batch 8550, loss[loss=0.1886, simple_loss=0.2858, pruned_loss=0.04563, over 16643.00 frames. ], tot_loss[loss=0.1788, simple_loss=0.2709, pruned_loss=0.04341, over 3035948.79 frames. ], batch size: 134, lr: 2.91e-03, grad_scale: 8.0 2023-05-01 17:03:20,164 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.0711, 3.4797, 3.7778, 2.0627, 3.1214, 2.4247, 3.5156, 3.6959], device='cuda:1'), covar=tensor([0.0265, 0.0784, 0.0474, 0.2071, 0.0742, 0.0977, 0.0602, 0.0860], device='cuda:1'), in_proj_covar=tensor([0.0152, 0.0161, 0.0164, 0.0151, 0.0142, 0.0127, 0.0140, 0.0172], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:1') 2023-05-01 17:03:31,346 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.6218, 2.0627, 1.7418, 1.8311, 2.3471, 1.9924, 1.9825, 2.4540], device='cuda:1'), covar=tensor([0.0220, 0.0486, 0.0570, 0.0528, 0.0320, 0.0468, 0.0224, 0.0319], device='cuda:1'), in_proj_covar=tensor([0.0207, 0.0230, 0.0221, 0.0223, 0.0231, 0.0229, 0.0229, 0.0226], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-01 17:04:28,271 INFO [train.py:904] (1/8) Epoch 23, batch 8600, loss[loss=0.1637, simple_loss=0.2514, pruned_loss=0.03796, over 12702.00 frames. ], tot_loss[loss=0.1778, simple_loss=0.2711, pruned_loss=0.04224, over 3055773.61 frames. ], batch size: 248, lr: 2.91e-03, grad_scale: 8.0 2023-05-01 17:04:36,593 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=231907.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 17:05:00,378 INFO [optim.py:368] (1/8) Clipping_scale=2.0, grad-norm quartiles 1.562e+02 2.309e+02 2.625e+02 3.302e+02 5.565e+02, threshold=5.250e+02, percent-clipped=1.0 2023-05-01 17:06:03,590 INFO [train.py:904] (1/8) Epoch 23, batch 8650, loss[loss=0.159, simple_loss=0.2598, pruned_loss=0.02912, over 16559.00 frames. ], tot_loss[loss=0.1751, simple_loss=0.2691, pruned_loss=0.04061, over 3060915.36 frames. ], batch size: 75, lr: 2.91e-03, grad_scale: 8.0 2023-05-01 17:06:10,094 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=231955.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 17:06:24,880 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=4.15 vs. limit=5.0 2023-05-01 17:07:53,209 INFO [train.py:904] (1/8) Epoch 23, batch 8700, loss[loss=0.1732, simple_loss=0.2709, pruned_loss=0.03774, over 15388.00 frames. ], tot_loss[loss=0.1727, simple_loss=0.2665, pruned_loss=0.03947, over 3049938.01 frames. ], batch size: 190, lr: 2.91e-03, grad_scale: 4.0 2023-05-01 17:07:54,259 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=232003.0, num_to_drop=1, layers_to_drop={0} 2023-05-01 17:08:23,061 INFO [optim.py:368] (1/8) 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] (1/8) Epoch 23, batch 8750, loss[loss=0.1982, simple_loss=0.2937, pruned_loss=0.05137, over 16939.00 frames. ], tot_loss[loss=0.172, simple_loss=0.2663, pruned_loss=0.03883, over 3056205.84 frames. ], batch size: 116, lr: 2.91e-03, grad_scale: 4.0 2023-05-01 17:10:27,891 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.4585, 3.4963, 3.7104, 3.6744, 3.7124, 3.5131, 3.5612, 3.5937], device='cuda:1'), covar=tensor([0.0373, 0.0840, 0.0482, 0.0472, 0.0433, 0.0537, 0.0740, 0.0457], device='cuda:1'), in_proj_covar=tensor([0.0404, 0.0449, 0.0437, 0.0405, 0.0481, 0.0456, 0.0539, 0.0367], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:1') 2023-05-01 17:10:56,275 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.52 vs. limit=2.0 2023-05-01 17:11:15,140 INFO [train.py:904] (1/8) Epoch 23, batch 8800, loss[loss=0.1738, simple_loss=0.27, pruned_loss=0.03883, over 17015.00 frames. ], tot_loss[loss=0.1703, simple_loss=0.2647, pruned_loss=0.03793, over 3048696.79 frames. ], batch size: 109, lr: 2.91e-03, grad_scale: 8.0 2023-05-01 17:11:33,043 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.8295, 4.7979, 4.5839, 4.0478, 4.6773, 1.8139, 4.4639, 4.4087], device='cuda:1'), covar=tensor([0.0080, 0.0082, 0.0193, 0.0323, 0.0105, 0.2676, 0.0123, 0.0224], device='cuda:1'), in_proj_covar=tensor([0.0166, 0.0158, 0.0198, 0.0175, 0.0174, 0.0206, 0.0186, 0.0168], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-01 17:11:46,651 INFO [optim.py:368] (1/8) 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 17:12:19,577 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.29 vs. limit=2.0 2023-05-01 17:12:57,921 INFO [train.py:904] (1/8) Epoch 23, batch 8850, loss[loss=0.1922, simple_loss=0.2937, pruned_loss=0.04529, over 16460.00 frames. ], tot_loss[loss=0.1714, simple_loss=0.2673, pruned_loss=0.0377, over 3041599.87 frames. ], batch size: 147, lr: 2.91e-03, grad_scale: 8.0 2023-05-01 17:12:58,769 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.4617, 3.3917, 3.5149, 3.5577, 3.6037, 3.3392, 3.5610, 3.6675], device='cuda:1'), covar=tensor([0.1182, 0.0918, 0.0946, 0.0614, 0.0604, 0.2277, 0.0852, 0.0676], device='cuda:1'), in_proj_covar=tensor([0.0615, 0.0763, 0.0877, 0.0772, 0.0587, 0.0615, 0.0641, 0.0746], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-01 17:14:39,540 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.27 vs. limit=2.0 2023-05-01 17:14:42,083 INFO [train.py:904] (1/8) Epoch 23, batch 8900, loss[loss=0.19, simple_loss=0.2853, pruned_loss=0.04732, over 16228.00 frames. ], tot_loss[loss=0.1705, simple_loss=0.2672, pruned_loss=0.03692, over 3046245.08 frames. ], batch size: 165, lr: 2.91e-03, grad_scale: 8.0 2023-05-01 17:15:12,042 INFO [optim.py:368] (1/8) 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 17:16:33,783 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=232247.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 17:16:45,324 INFO [train.py:904] (1/8) Epoch 23, batch 8950, loss[loss=0.153, simple_loss=0.2494, pruned_loss=0.02833, over 16258.00 frames. ], tot_loss[loss=0.1701, simple_loss=0.2663, pruned_loss=0.03695, over 3045202.83 frames. ], batch size: 165, lr: 2.91e-03, grad_scale: 8.0 2023-05-01 17:17:32,852 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.6611, 2.6064, 1.8706, 2.7774, 2.1192, 2.7934, 2.1291, 2.3851], device='cuda:1'), covar=tensor([0.0320, 0.0366, 0.1364, 0.0234, 0.0684, 0.0511, 0.1317, 0.0684], device='cuda:1'), in_proj_covar=tensor([0.0166, 0.0171, 0.0189, 0.0158, 0.0172, 0.0209, 0.0198, 0.0176], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0003, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-05-01 17:18:31,849 INFO [train.py:904] (1/8) Epoch 23, batch 9000, loss[loss=0.1667, simple_loss=0.2563, pruned_loss=0.03852, over 17208.00 frames. ], tot_loss[loss=0.1673, simple_loss=0.263, pruned_loss=0.03573, over 3062199.64 frames. ], batch size: 44, lr: 2.91e-03, grad_scale: 8.0 2023-05-01 17:18:31,849 INFO [train.py:929] (1/8) Computing validation loss 2023-05-01 17:18:42,672 INFO [train.py:938] (1/8) Epoch 23, validation: loss=0.1452, simple_loss=0.249, pruned_loss=0.02066, over 944034.00 frames. 2023-05-01 17:18:42,672 INFO [train.py:939] (1/8) Maximum memory allocated so far is 18145MB 2023-05-01 17:18:44,321 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=232303.0, num_to_drop=1, layers_to_drop={2} 2023-05-01 17:18:54,211 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=232308.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 17:19:18,067 INFO [optim.py:368] (1/8) 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 17:20:22,846 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.0432, 4.0141, 3.9634, 3.2309, 3.9811, 1.7724, 3.7839, 3.6555], device='cuda:1'), covar=tensor([0.0156, 0.0145, 0.0198, 0.0341, 0.0137, 0.2900, 0.0202, 0.0321], device='cuda:1'), in_proj_covar=tensor([0.0165, 0.0157, 0.0197, 0.0174, 0.0174, 0.0206, 0.0186, 0.0168], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-01 17:20:24,510 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=232351.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 17:20:27,725 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=4.35 vs. limit=5.0 2023-05-01 17:20:28,118 INFO [train.py:904] (1/8) Epoch 23, batch 9050, loss[loss=0.1878, simple_loss=0.2707, pruned_loss=0.05244, over 12848.00 frames. ], tot_loss[loss=0.1688, simple_loss=0.2647, pruned_loss=0.03649, over 3066345.73 frames. ], batch size: 248, lr: 2.91e-03, grad_scale: 8.0 2023-05-01 17:21:23,412 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=232381.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 17:21:29,529 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.9526, 2.1649, 2.3646, 3.1950, 2.1668, 2.3618, 2.3325, 2.2536], device='cuda:1'), covar=tensor([0.1276, 0.3584, 0.2812, 0.0684, 0.4455, 0.2549, 0.3612, 0.3625], device='cuda:1'), in_proj_covar=tensor([0.0396, 0.0445, 0.0366, 0.0320, 0.0430, 0.0509, 0.0416, 0.0517], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-01 17:22:12,828 INFO [train.py:904] (1/8) Epoch 23, batch 9100, loss[loss=0.1938, simple_loss=0.2927, pruned_loss=0.04748, over 16969.00 frames. ], tot_loss[loss=0.1691, simple_loss=0.2645, pruned_loss=0.03688, over 3068631.00 frames. ], batch size: 109, lr: 2.91e-03, grad_scale: 4.0 2023-05-01 17:22:46,300 INFO [optim.py:368] (1/8) 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 17:23:46,459 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=232442.0, num_to_drop=1, layers_to_drop={2} 2023-05-01 17:23:49,264 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([1.9462, 3.2416, 3.2792, 2.1282, 3.0815, 3.3451, 3.1639, 1.6432], device='cuda:1'), covar=tensor([0.0644, 0.0096, 0.0096, 0.0514, 0.0154, 0.0139, 0.0137, 0.0759], device='cuda:1'), in_proj_covar=tensor([0.0134, 0.0084, 0.0084, 0.0133, 0.0098, 0.0109, 0.0094, 0.0129], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-05-01 17:24:09,371 INFO [train.py:904] (1/8) Epoch 23, batch 9150, loss[loss=0.1531, simple_loss=0.2486, pruned_loss=0.02877, over 16302.00 frames. ], tot_loss[loss=0.1689, simple_loss=0.2647, pruned_loss=0.03662, over 3059914.67 frames. ], batch size: 146, lr: 2.91e-03, grad_scale: 4.0 2023-05-01 17:25:00,528 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=232476.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 17:25:07,515 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.6807, 3.8057, 2.3737, 4.3938, 2.8046, 4.2995, 2.5477, 3.0708], device='cuda:1'), covar=tensor([0.0305, 0.0353, 0.1567, 0.0179, 0.0858, 0.0376, 0.1453, 0.0771], device='cuda:1'), in_proj_covar=tensor([0.0166, 0.0171, 0.0189, 0.0158, 0.0172, 0.0209, 0.0199, 0.0176], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0003, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-05-01 17:25:11,462 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.51 vs. limit=2.0 2023-05-01 17:25:52,924 INFO [train.py:904] (1/8) Epoch 23, batch 9200, loss[loss=0.1664, simple_loss=0.255, pruned_loss=0.03888, over 16613.00 frames. ], tot_loss[loss=0.1659, simple_loss=0.2607, pruned_loss=0.0356, over 3081627.53 frames. ], batch size: 57, lr: 2.91e-03, grad_scale: 8.0 2023-05-01 17:26:21,384 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=232518.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 17:26:24,111 INFO [optim.py:368] (1/8) 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 17:26:57,072 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=232537.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 17:27:27,732 INFO [train.py:904] (1/8) Epoch 23, batch 9250, loss[loss=0.1542, simple_loss=0.2559, pruned_loss=0.0262, over 15434.00 frames. ], tot_loss[loss=0.1658, simple_loss=0.2605, pruned_loss=0.0356, over 3080840.58 frames. ], batch size: 190, lr: 2.91e-03, grad_scale: 8.0 2023-05-01 17:28:23,247 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=232579.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 17:29:06,404 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.70 vs. limit=2.0 2023-05-01 17:29:17,466 INFO [train.py:904] (1/8) Epoch 23, batch 9300, loss[loss=0.166, simple_loss=0.2589, pruned_loss=0.03661, over 16610.00 frames. ], tot_loss[loss=0.1649, simple_loss=0.2592, pruned_loss=0.03527, over 3076489.01 frames. ], batch size: 148, lr: 2.91e-03, grad_scale: 8.0 2023-05-01 17:29:18,375 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=232603.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 17:29:58,635 INFO [optim.py:368] (1/8) 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 17:30:08,235 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.34 vs. limit=2.0 2023-05-01 17:30:23,864 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.0852, 3.9329, 4.1652, 4.2542, 4.3789, 3.9744, 4.3581, 4.4225], device='cuda:1'), covar=tensor([0.1770, 0.1326, 0.1484, 0.0777, 0.0611, 0.1343, 0.0762, 0.0717], device='cuda:1'), in_proj_covar=tensor([0.0612, 0.0758, 0.0869, 0.0769, 0.0584, 0.0611, 0.0636, 0.0743], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-01 17:31:04,797 INFO [train.py:904] (1/8) Epoch 23, batch 9350, loss[loss=0.1512, simple_loss=0.2361, pruned_loss=0.03318, over 12357.00 frames. ], tot_loss[loss=0.1644, simple_loss=0.2587, pruned_loss=0.03511, over 3072277.98 frames. ], batch size: 248, lr: 2.91e-03, grad_scale: 8.0 2023-05-01 17:31:59,213 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.9962, 2.6770, 2.9227, 2.1261, 2.7375, 2.1981, 2.7541, 2.9028], device='cuda:1'), covar=tensor([0.0276, 0.0930, 0.0485, 0.1810, 0.0783, 0.0976, 0.0583, 0.0832], device='cuda:1'), in_proj_covar=tensor([0.0151, 0.0158, 0.0163, 0.0150, 0.0142, 0.0126, 0.0139, 0.0170], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:1') 2023-05-01 17:32:47,939 INFO [train.py:904] (1/8) Epoch 23, batch 9400, loss[loss=0.1733, simple_loss=0.28, pruned_loss=0.03331, over 16700.00 frames. ], tot_loss[loss=0.1643, simple_loss=0.2586, pruned_loss=0.03497, over 3069155.29 frames. ], batch size: 76, lr: 2.90e-03, grad_scale: 8.0 2023-05-01 17:33:21,842 INFO [optim.py:368] (1/8) 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] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=232737.0, num_to_drop=1, layers_to_drop={3} 2023-05-01 17:34:29,649 INFO [train.py:904] (1/8) Epoch 23, batch 9450, loss[loss=0.1525, simple_loss=0.2514, pruned_loss=0.02676, over 16439.00 frames. ], tot_loss[loss=0.1647, simple_loss=0.2595, pruned_loss=0.03495, over 3059633.18 frames. ], batch size: 146, lr: 2.90e-03, grad_scale: 8.0 2023-05-01 17:36:00,016 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.4543, 3.0723, 2.7685, 2.2424, 2.1647, 2.2958, 3.0827, 2.8164], device='cuda:1'), covar=tensor([0.2819, 0.0640, 0.1541, 0.2873, 0.2925, 0.2363, 0.0439, 0.1436], device='cuda:1'), in_proj_covar=tensor([0.0322, 0.0262, 0.0299, 0.0309, 0.0286, 0.0257, 0.0289, 0.0329], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:1') 2023-05-01 17:36:04,900 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.3422, 4.2773, 4.1432, 3.3714, 4.2521, 1.6998, 3.9779, 3.8341], device='cuda:1'), covar=tensor([0.0102, 0.0118, 0.0201, 0.0365, 0.0119, 0.3096, 0.0154, 0.0336], device='cuda:1'), in_proj_covar=tensor([0.0165, 0.0157, 0.0196, 0.0172, 0.0173, 0.0205, 0.0185, 0.0167], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-01 17:36:09,231 INFO [train.py:904] (1/8) Epoch 23, batch 9500, loss[loss=0.1673, simple_loss=0.262, pruned_loss=0.03628, over 15241.00 frames. ], tot_loss[loss=0.164, simple_loss=0.2589, pruned_loss=0.03453, over 3079908.07 frames. ], batch size: 190, lr: 2.90e-03, grad_scale: 8.0 2023-05-01 17:36:44,544 INFO [optim.py:368] (1/8) 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 17:37:07,735 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=232832.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 17:37:54,041 INFO [train.py:904] (1/8) Epoch 23, batch 9550, loss[loss=0.161, simple_loss=0.2544, pruned_loss=0.03382, over 12609.00 frames. ], tot_loss[loss=0.1636, simple_loss=0.2585, pruned_loss=0.03438, over 3095944.79 frames. ], batch size: 248, lr: 2.90e-03, grad_scale: 8.0 2023-05-01 17:38:37,447 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=232874.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 17:38:37,652 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.6069, 3.0471, 3.1791, 1.9656, 2.7801, 2.1932, 3.0853, 3.2549], device='cuda:1'), covar=tensor([0.0358, 0.0844, 0.0628, 0.2244, 0.0905, 0.1052, 0.0770, 0.0972], device='cuda:1'), in_proj_covar=tensor([0.0151, 0.0157, 0.0163, 0.0150, 0.0141, 0.0126, 0.0139, 0.0169], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:1') 2023-05-01 17:39:33,897 INFO [train.py:904] (1/8) Epoch 23, batch 9600, loss[loss=0.1679, simple_loss=0.2532, pruned_loss=0.0413, over 12153.00 frames. ], tot_loss[loss=0.1653, simple_loss=0.2598, pruned_loss=0.03537, over 3074830.23 frames. ], batch size: 247, lr: 2.90e-03, grad_scale: 8.0 2023-05-01 17:39:34,576 INFO [zipformer.py:625] (1/8) 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] (1/8) 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 17:41:17,233 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=232951.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 17:41:20,710 INFO [train.py:904] (1/8) Epoch 23, batch 9650, loss[loss=0.1571, simple_loss=0.2571, pruned_loss=0.02854, over 16713.00 frames. ], tot_loss[loss=0.167, simple_loss=0.2625, pruned_loss=0.03576, over 3088208.13 frames. ], batch size: 76, lr: 2.90e-03, grad_scale: 8.0 2023-05-01 17:41:22,195 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([5.1880, 5.2520, 5.6674, 5.6397, 5.6522, 5.3371, 5.2582, 5.1253], device='cuda:1'), covar=tensor([0.0356, 0.0736, 0.0367, 0.0399, 0.0520, 0.0391, 0.0942, 0.0403], device='cuda:1'), in_proj_covar=tensor([0.0399, 0.0440, 0.0432, 0.0396, 0.0473, 0.0448, 0.0528, 0.0361], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:1') 2023-05-01 17:42:41,578 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=232990.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 17:43:00,274 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.30 vs. limit=2.0 2023-05-01 17:43:09,181 INFO [train.py:904] (1/8) Epoch 23, batch 9700, loss[loss=0.1589, simple_loss=0.2626, pruned_loss=0.02764, over 16593.00 frames. ], tot_loss[loss=0.1665, simple_loss=0.2617, pruned_loss=0.03569, over 3080557.25 frames. ], batch size: 89, lr: 2.90e-03, grad_scale: 8.0 2023-05-01 17:43:25,043 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([1.9015, 1.4276, 1.7473, 1.7620, 1.9120, 1.9545, 1.6751, 1.9355], device='cuda:1'), covar=tensor([0.0271, 0.0459, 0.0249, 0.0332, 0.0329, 0.0251, 0.0455, 0.0154], device='cuda:1'), in_proj_covar=tensor([0.0184, 0.0188, 0.0174, 0.0177, 0.0193, 0.0150, 0.0192, 0.0147], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-01 17:43:40,104 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=233018.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 17:43:43,779 INFO [optim.py:368] (1/8) 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 17:44:22,048 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=233037.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 17:44:50,779 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=233051.0, num_to_drop=1, layers_to_drop={1} 2023-05-01 17:44:53,546 INFO [train.py:904] (1/8) Epoch 23, batch 9750, loss[loss=0.1629, simple_loss=0.2474, pruned_loss=0.03921, over 12423.00 frames. ], tot_loss[loss=0.166, simple_loss=0.2606, pruned_loss=0.03576, over 3079786.08 frames. ], batch size: 250, lr: 2.90e-03, grad_scale: 8.0 2023-05-01 17:45:31,732 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.71 vs. limit=2.0 2023-05-01 17:45:43,933 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=233079.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 17:45:59,986 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 23, batch 9800, loss[loss=0.1627, simple_loss=0.2668, pruned_loss=0.02935, over 16672.00 frames. ], tot_loss[loss=0.165, simple_loss=0.26, pruned_loss=0.03499, over 3067886.22 frames. ], batch size: 134, lr: 2.90e-03, grad_scale: 8.0 2023-05-01 17:47:03,693 INFO [optim.py:368] (1/8) 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 17:47:07,702 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.5534, 3.5228, 3.4768, 2.8394, 3.4292, 1.9661, 3.2049, 2.9064], device='cuda:1'), covar=tensor([0.0125, 0.0122, 0.0175, 0.0202, 0.0091, 0.2528, 0.0122, 0.0272], device='cuda:1'), in_proj_covar=tensor([0.0165, 0.0157, 0.0196, 0.0171, 0.0173, 0.0206, 0.0185, 0.0166], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-01 17:47:26,800 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=233132.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 17:48:16,805 INFO [train.py:904] (1/8) Epoch 23, batch 9850, loss[loss=0.1699, simple_loss=0.265, pruned_loss=0.03745, over 16907.00 frames. ], tot_loss[loss=0.1651, simple_loss=0.2608, pruned_loss=0.03472, over 3047766.95 frames. ], batch size: 116, lr: 2.90e-03, grad_scale: 8.0 2023-05-01 17:48:20,837 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.2678, 4.3379, 4.4589, 4.3411, 4.3440, 4.8298, 4.3824, 4.0507], device='cuda:1'), covar=tensor([0.1695, 0.1992, 0.2197, 0.2025, 0.2740, 0.1069, 0.1614, 0.2698], device='cuda:1'), in_proj_covar=tensor([0.0388, 0.0570, 0.0631, 0.0471, 0.0628, 0.0657, 0.0491, 0.0627], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-05-01 17:49:01,131 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=233174.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 17:49:12,419 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.50 vs. limit=2.0 2023-05-01 17:49:13,852 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 23, batch 9900, loss[loss=0.1496, simple_loss=0.2407, pruned_loss=0.02926, over 12413.00 frames. ], tot_loss[loss=0.1655, simple_loss=0.2616, pruned_loss=0.03474, over 3057105.39 frames. ], batch size: 248, lr: 2.90e-03, grad_scale: 8.0 2023-05-01 17:50:46,347 INFO [optim.py:368] (1/8) 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] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=233222.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 17:51:06,664 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.4989, 3.5523, 2.0523, 4.1054, 2.7252, 3.9399, 2.1628, 2.7769], device='cuda:1'), covar=tensor([0.0317, 0.0409, 0.1792, 0.0206, 0.0813, 0.0585, 0.1732, 0.0823], device='cuda:1'), in_proj_covar=tensor([0.0163, 0.0168, 0.0186, 0.0155, 0.0169, 0.0205, 0.0197, 0.0173], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0003, 0.0004, 0.0004, 0.0003], device='cuda:1') 2023-05-01 17:51:37,812 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.8119, 2.7398, 2.4412, 2.4437, 3.1385, 2.7968, 3.2866, 3.3094], device='cuda:1'), covar=tensor([0.0129, 0.0504, 0.0526, 0.0521, 0.0283, 0.0471, 0.0260, 0.0252], device='cuda:1'), in_proj_covar=tensor([0.0205, 0.0229, 0.0221, 0.0223, 0.0229, 0.0230, 0.0223, 0.0223], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-01 17:52:06,036 INFO [train.py:904] (1/8) Epoch 23, batch 9950, loss[loss=0.1662, simple_loss=0.2666, pruned_loss=0.03288, over 16674.00 frames. ], tot_loss[loss=0.1664, simple_loss=0.2632, pruned_loss=0.03486, over 3061359.18 frames. ], batch size: 134, lr: 2.90e-03, grad_scale: 8.0 2023-05-01 17:52:20,240 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=233259.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 17:54:07,453 INFO [train.py:904] (1/8) Epoch 23, batch 10000, loss[loss=0.1732, simple_loss=0.2618, pruned_loss=0.04231, over 12840.00 frames. ], tot_loss[loss=0.1656, simple_loss=0.2619, pruned_loss=0.03466, over 3054406.64 frames. ], batch size: 250, lr: 2.90e-03, grad_scale: 8.0 2023-05-01 17:54:33,474 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.83 vs. limit=2.0 2023-05-01 17:54:40,344 INFO [optim.py:368] (1/8) 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 17:54:41,013 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=233320.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 17:55:36,793 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=233346.0, num_to_drop=1, layers_to_drop={3} 2023-05-01 17:55:47,979 INFO [train.py:904] (1/8) Epoch 23, batch 10050, loss[loss=0.1779, simple_loss=0.2718, pruned_loss=0.04194, over 12321.00 frames. ], tot_loss[loss=0.1664, simple_loss=0.2626, pruned_loss=0.03507, over 3058705.02 frames. ], batch size: 247, lr: 2.90e-03, grad_scale: 8.0 2023-05-01 17:56:12,201 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.9130, 4.2132, 4.0456, 4.1053, 3.7002, 3.8596, 3.8427, 4.2207], device='cuda:1'), covar=tensor([0.1171, 0.0975, 0.0947, 0.0799, 0.0843, 0.1660, 0.1019, 0.0955], device='cuda:1'), in_proj_covar=tensor([0.0655, 0.0794, 0.0656, 0.0608, 0.0504, 0.0515, 0.0670, 0.0628], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:1') 2023-05-01 17:56:22,778 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.6180, 3.9564, 3.9934, 2.8541, 3.5008, 3.9827, 3.6863, 2.2476], device='cuda:1'), covar=tensor([0.0494, 0.0047, 0.0047, 0.0349, 0.0114, 0.0086, 0.0079, 0.0523], device='cuda:1'), in_proj_covar=tensor([0.0133, 0.0083, 0.0083, 0.0132, 0.0097, 0.0107, 0.0093, 0.0128], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-05-01 17:56:30,650 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=233374.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 17:56:37,746 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.7973, 2.1921, 1.8973, 1.9209, 2.4789, 2.1776, 2.2380, 2.6101], device='cuda:1'), covar=tensor([0.0210, 0.0497, 0.0599, 0.0536, 0.0332, 0.0425, 0.0250, 0.0316], device='cuda:1'), in_proj_covar=tensor([0.0204, 0.0229, 0.0221, 0.0223, 0.0229, 0.0230, 0.0224, 0.0223], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-01 17:57:20,830 INFO [train.py:904] (1/8) Epoch 23, batch 10100, loss[loss=0.159, simple_loss=0.2531, pruned_loss=0.03249, over 16173.00 frames. ], tot_loss[loss=0.1665, simple_loss=0.2626, pruned_loss=0.03516, over 3053609.53 frames. ], batch size: 165, lr: 2.90e-03, grad_scale: 8.0 2023-05-01 17:57:53,980 INFO [optim.py:368] (1/8) 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] (1/8) Epoch 24, batch 0, loss[loss=0.1594, simple_loss=0.2427, pruned_loss=0.03803, over 17247.00 frames. ], tot_loss[loss=0.1594, simple_loss=0.2427, pruned_loss=0.03803, over 17247.00 frames. ], batch size: 45, lr: 2.84e-03, grad_scale: 8.0 2023-05-01 17:59:06,335 INFO [train.py:929] (1/8) Computing validation loss 2023-05-01 17:59:14,240 INFO [train.py:938] (1/8) Epoch 24, validation: loss=0.145, simple_loss=0.2483, pruned_loss=0.02085, over 944034.00 frames. 2023-05-01 17:59:14,241 INFO [train.py:939] (1/8) Maximum memory allocated so far is 18145MB 2023-05-01 17:59:59,958 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.8531, 2.0198, 2.3732, 2.7520, 2.6541, 3.1870, 2.0144, 3.3066], device='cuda:1'), covar=tensor([0.0303, 0.0601, 0.0463, 0.0398, 0.0447, 0.0264, 0.0629, 0.0201], device='cuda:1'), in_proj_covar=tensor([0.0186, 0.0190, 0.0176, 0.0179, 0.0195, 0.0151, 0.0194, 0.0149], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-01 18:00:06,412 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.94 vs. limit=2.0 2023-05-01 18:00:23,694 INFO [train.py:904] (1/8) Epoch 24, batch 50, loss[loss=0.1779, simple_loss=0.2709, pruned_loss=0.04239, over 17052.00 frames. ], tot_loss[loss=0.1848, simple_loss=0.2702, pruned_loss=0.04968, over 751967.25 frames. ], batch size: 53, lr: 2.84e-03, grad_scale: 1.0 2023-05-01 18:00:44,686 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.7210, 2.5290, 2.3317, 2.5190, 2.9763, 2.6883, 3.2314, 3.1573], device='cuda:1'), covar=tensor([0.0169, 0.0578, 0.0589, 0.0520, 0.0327, 0.0472, 0.0284, 0.0338], device='cuda:1'), in_proj_covar=tensor([0.0208, 0.0232, 0.0223, 0.0224, 0.0231, 0.0232, 0.0226, 0.0225], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-01 18:00:52,610 INFO [optim.py:368] (1/8) 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] (1/8) Epoch 24, batch 100, loss[loss=0.1555, simple_loss=0.2403, pruned_loss=0.03533, over 16815.00 frames. ], tot_loss[loss=0.179, simple_loss=0.2651, pruned_loss=0.04642, over 1319543.16 frames. ], batch size: 39, lr: 2.84e-03, grad_scale: 1.0 2023-05-01 18:01:53,744 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.6829, 3.7821, 2.3114, 4.2260, 2.9667, 4.1324, 2.4832, 3.1242], device='cuda:1'), covar=tensor([0.0325, 0.0410, 0.1660, 0.0369, 0.0859, 0.0641, 0.1534, 0.0768], device='cuda:1'), in_proj_covar=tensor([0.0166, 0.0171, 0.0189, 0.0159, 0.0172, 0.0210, 0.0199, 0.0176], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0003, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-05-01 18:02:40,657 INFO [train.py:904] (1/8) Epoch 24, batch 150, loss[loss=0.1552, simple_loss=0.2355, pruned_loss=0.03749, over 16838.00 frames. ], tot_loss[loss=0.1758, simple_loss=0.2624, pruned_loss=0.04458, over 1768478.82 frames. ], batch size: 42, lr: 2.84e-03, grad_scale: 1.0 2023-05-01 18:02:56,019 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=233615.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 18:03:08,347 INFO [optim.py:368] (1/8) 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 18:03:30,575 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=233639.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 18:03:39,352 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=233646.0, num_to_drop=1, layers_to_drop={1} 2023-05-01 18:03:48,629 INFO [train.py:904] (1/8) Epoch 24, batch 200, loss[loss=0.1527, simple_loss=0.2456, pruned_loss=0.02994, over 16812.00 frames. ], tot_loss[loss=0.1758, simple_loss=0.2629, pruned_loss=0.04437, over 2110373.16 frames. ], batch size: 42, lr: 2.84e-03, grad_scale: 1.0 2023-05-01 18:04:17,982 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=233674.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 18:04:34,590 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.6598, 2.5051, 1.9959, 2.2786, 2.8475, 2.6378, 3.2365, 3.1698], device='cuda:1'), covar=tensor([0.0175, 0.0675, 0.0772, 0.0684, 0.0426, 0.0547, 0.0374, 0.0356], device='cuda:1'), in_proj_covar=tensor([0.0215, 0.0238, 0.0228, 0.0230, 0.0237, 0.0237, 0.0234, 0.0232], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-01 18:04:45,849 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=233694.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 18:04:54,281 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=233700.0, num_to_drop=1, layers_to_drop={0} 2023-05-01 18:04:58,754 INFO [train.py:904] (1/8) Epoch 24, batch 250, loss[loss=0.1784, simple_loss=0.2604, pruned_loss=0.04821, over 16655.00 frames. ], tot_loss[loss=0.1754, simple_loss=0.2613, pruned_loss=0.04481, over 2383092.67 frames. ], batch size: 134, lr: 2.84e-03, grad_scale: 1.0 2023-05-01 18:05:07,844 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=233710.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 18:05:25,968 INFO [zipformer.py:625] (1/8) 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] (1/8) 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 18:05:48,916 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.18 vs. limit=2.0 2023-05-01 18:06:08,092 INFO [train.py:904] (1/8) Epoch 24, batch 300, loss[loss=0.1674, simple_loss=0.2391, pruned_loss=0.04781, over 16902.00 frames. ], tot_loss[loss=0.1719, simple_loss=0.2582, pruned_loss=0.04285, over 2601827.46 frames. ], batch size: 116, lr: 2.84e-03, grad_scale: 1.0 2023-05-01 18:06:33,472 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=233771.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 18:07:16,410 INFO [train.py:904] (1/8) Epoch 24, batch 350, loss[loss=0.1564, simple_loss=0.2414, pruned_loss=0.03568, over 16477.00 frames. ], tot_loss[loss=0.1695, simple_loss=0.256, pruned_loss=0.04147, over 2762887.75 frames. ], batch size: 68, lr: 2.84e-03, grad_scale: 1.0 2023-05-01 18:07:43,747 INFO [optim.py:368] (1/8) 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 18:08:25,442 INFO [train.py:904] (1/8) Epoch 24, batch 400, loss[loss=0.1868, simple_loss=0.2719, pruned_loss=0.05084, over 16244.00 frames. ], tot_loss[loss=0.1689, simple_loss=0.2545, pruned_loss=0.04168, over 2891553.27 frames. ], batch size: 165, lr: 2.84e-03, grad_scale: 2.0 2023-05-01 18:09:32,975 INFO [train.py:904] (1/8) Epoch 24, batch 450, loss[loss=0.1753, simple_loss=0.2701, pruned_loss=0.04022, over 16760.00 frames. ], tot_loss[loss=0.1677, simple_loss=0.2529, pruned_loss=0.04128, over 2983774.62 frames. ], batch size: 57, lr: 2.84e-03, grad_scale: 2.0 2023-05-01 18:09:44,108 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([1.8133, 2.7279, 2.5248, 4.8253, 3.5263, 4.1155, 1.6917, 2.9218], device='cuda:1'), covar=tensor([0.1427, 0.0920, 0.1394, 0.0214, 0.0244, 0.0492, 0.1703, 0.0918], device='cuda:1'), in_proj_covar=tensor([0.0167, 0.0173, 0.0194, 0.0189, 0.0199, 0.0212, 0.0203, 0.0192], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-05-01 18:09:47,294 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.9226, 2.0747, 2.4808, 2.9249, 2.7009, 3.3496, 2.4142, 3.3610], device='cuda:1'), covar=tensor([0.0314, 0.0577, 0.0392, 0.0357, 0.0401, 0.0229, 0.0483, 0.0192], device='cuda:1'), in_proj_covar=tensor([0.0191, 0.0194, 0.0180, 0.0184, 0.0200, 0.0156, 0.0198, 0.0153], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-01 18:09:50,117 INFO [zipformer.py:625] (1/8) 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] (1/8) 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] (1/8) Epoch 24, batch 500, loss[loss=0.16, simple_loss=0.2434, pruned_loss=0.03835, over 16288.00 frames. ], tot_loss[loss=0.1661, simple_loss=0.2515, pruned_loss=0.04032, over 3063640.94 frames. ], batch size: 165, lr: 2.83e-03, grad_scale: 2.0 2023-05-01 18:10:53,522 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=233963.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 18:11:37,663 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.4124, 4.6748, 4.5447, 4.4807, 4.2616, 4.2203, 4.2453, 4.7423], device='cuda:1'), covar=tensor([0.1265, 0.0942, 0.0940, 0.0854, 0.0769, 0.1433, 0.1095, 0.0928], device='cuda:1'), in_proj_covar=tensor([0.0686, 0.0831, 0.0685, 0.0638, 0.0528, 0.0536, 0.0703, 0.0655], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:1') 2023-05-01 18:11:37,667 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=233995.0, num_to_drop=1, layers_to_drop={3} 2023-05-01 18:11:50,840 INFO [train.py:904] (1/8) Epoch 24, batch 550, loss[loss=0.157, simple_loss=0.2415, pruned_loss=0.03629, over 16676.00 frames. ], tot_loss[loss=0.1657, simple_loss=0.2517, pruned_loss=0.03985, over 3123819.66 frames. ], batch size: 89, lr: 2.83e-03, grad_scale: 2.0 2023-05-01 18:12:17,097 INFO [optim.py:368] (1/8) 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 18:12:20,393 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=234025.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 18:12:57,924 INFO [train.py:904] (1/8) Epoch 24, batch 600, loss[loss=0.1633, simple_loss=0.251, pruned_loss=0.03783, over 16401.00 frames. ], tot_loss[loss=0.165, simple_loss=0.2508, pruned_loss=0.03964, over 3164226.36 frames. ], batch size: 68, lr: 2.83e-03, grad_scale: 2.0 2023-05-01 18:13:17,435 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=234066.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 18:13:44,751 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=234086.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 18:14:05,354 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.30 vs. limit=2.0 2023-05-01 18:14:08,032 INFO [train.py:904] (1/8) Epoch 24, batch 650, loss[loss=0.1373, simple_loss=0.2273, pruned_loss=0.02367, over 16842.00 frames. ], tot_loss[loss=0.1644, simple_loss=0.25, pruned_loss=0.03944, over 3195207.41 frames. ], batch size: 42, lr: 2.83e-03, grad_scale: 2.0 2023-05-01 18:14:10,929 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.51 vs. limit=2.0 2023-05-01 18:14:36,190 INFO [optim.py:368] (1/8) Clipping_scale=2.0, grad-norm quartiles 1.365e+02 2.015e+02 2.288e+02 2.913e+02 5.815e+02, threshold=4.575e+02, percent-clipped=2.0 2023-05-01 18:15:16,420 INFO [train.py:904] (1/8) Epoch 24, batch 700, loss[loss=0.1502, simple_loss=0.2433, pruned_loss=0.0285, over 17093.00 frames. ], tot_loss[loss=0.1639, simple_loss=0.2494, pruned_loss=0.03921, over 3223824.59 frames. ], batch size: 47, lr: 2.83e-03, grad_scale: 2.0 2023-05-01 18:15:24,296 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([5.4484, 5.8935, 5.6244, 5.6892, 5.2047, 5.2865, 5.3020, 6.0459], device='cuda:1'), covar=tensor([0.1756, 0.1134, 0.1194, 0.0855, 0.1077, 0.0772, 0.1299, 0.1077], device='cuda:1'), in_proj_covar=tensor([0.0685, 0.0833, 0.0686, 0.0638, 0.0529, 0.0536, 0.0705, 0.0655], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:1') 2023-05-01 18:16:23,752 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 24, batch 750, loss[loss=0.1725, simple_loss=0.2671, pruned_loss=0.03896, over 16725.00 frames. ], tot_loss[loss=0.1641, simple_loss=0.2498, pruned_loss=0.03921, over 3251750.87 frames. ], batch size: 62, lr: 2.83e-03, grad_scale: 2.0 2023-05-01 18:16:44,265 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.43 vs. limit=2.0 2023-05-01 18:16:45,101 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.2113, 4.5655, 4.5887, 3.3823, 3.7744, 4.5268, 4.0569, 2.8450], device='cuda:1'), covar=tensor([0.0429, 0.0067, 0.0045, 0.0342, 0.0142, 0.0100, 0.0092, 0.0440], device='cuda:1'), in_proj_covar=tensor([0.0136, 0.0086, 0.0087, 0.0135, 0.0099, 0.0111, 0.0096, 0.0131], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0003, 0.0004], device='cuda:1') 2023-05-01 18:16:52,424 INFO [optim.py:368] (1/8) 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 18:17:04,947 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=234231.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 18:17:33,977 INFO [train.py:904] (1/8) Epoch 24, batch 800, loss[loss=0.1724, simple_loss=0.2662, pruned_loss=0.03924, over 17063.00 frames. ], tot_loss[loss=0.163, simple_loss=0.2485, pruned_loss=0.03872, over 3272831.61 frames. ], batch size: 53, lr: 2.83e-03, grad_scale: 4.0 2023-05-01 18:17:48,280 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=234263.0, num_to_drop=1, layers_to_drop={1} 2023-05-01 18:17:59,561 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.48 vs. limit=2.0 2023-05-01 18:18:29,151 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=234292.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 18:18:32,351 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=234295.0, num_to_drop=1, layers_to_drop={2} 2023-05-01 18:18:43,643 INFO [train.py:904] (1/8) Epoch 24, batch 850, loss[loss=0.1821, simple_loss=0.2794, pruned_loss=0.04237, over 17064.00 frames. ], tot_loss[loss=0.1628, simple_loss=0.2484, pruned_loss=0.03858, over 3287414.24 frames. ], batch size: 55, lr: 2.83e-03, grad_scale: 4.0 2023-05-01 18:19:11,881 INFO [optim.py:368] (1/8) Clipping_scale=2.0, grad-norm quartiles 1.464e+02 2.010e+02 2.479e+02 2.918e+02 4.237e+02, threshold=4.958e+02, percent-clipped=0.0 2023-05-01 18:19:40,588 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=234343.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 18:19:48,835 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.5071, 4.4321, 4.4202, 4.1306, 4.1555, 4.4462, 4.2368, 4.2315], device='cuda:1'), covar=tensor([0.0692, 0.0882, 0.0342, 0.0322, 0.0848, 0.0490, 0.0489, 0.0644], device='cuda:1'), in_proj_covar=tensor([0.0300, 0.0444, 0.0351, 0.0350, 0.0353, 0.0407, 0.0238, 0.0420], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002], device='cuda:1') 2023-05-01 18:19:52,506 INFO [train.py:904] (1/8) Epoch 24, batch 900, loss[loss=0.1909, simple_loss=0.2673, pruned_loss=0.05719, over 11993.00 frames. ], tot_loss[loss=0.1617, simple_loss=0.2472, pruned_loss=0.03806, over 3292331.10 frames. ], batch size: 246, lr: 2.83e-03, grad_scale: 4.0 2023-05-01 18:20:11,468 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=234366.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 18:20:33,658 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=234381.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 18:21:02,821 INFO [train.py:904] (1/8) Epoch 24, batch 950, loss[loss=0.1614, simple_loss=0.2379, pruned_loss=0.04242, over 16460.00 frames. ], tot_loss[loss=0.1629, simple_loss=0.2483, pruned_loss=0.03875, over 3297207.98 frames. ], batch size: 146, lr: 2.83e-03, grad_scale: 4.0 2023-05-01 18:21:17,243 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=234414.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 18:21:30,309 INFO [optim.py:368] (1/8) Clipping_scale=2.0, grad-norm quartiles 1.542e+02 2.144e+02 2.493e+02 3.109e+02 1.071e+03, threshold=4.986e+02, percent-clipped=5.0 2023-05-01 18:21:37,390 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.9213, 2.1294, 2.4645, 2.8498, 2.6838, 3.3173, 2.3288, 3.3524], device='cuda:1'), covar=tensor([0.0289, 0.0518, 0.0381, 0.0355, 0.0390, 0.0219, 0.0500, 0.0178], device='cuda:1'), in_proj_covar=tensor([0.0192, 0.0195, 0.0181, 0.0186, 0.0200, 0.0158, 0.0198, 0.0155], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-01 18:21:38,421 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.8644, 4.9807, 5.3453, 5.3456, 5.3511, 4.9897, 4.9451, 4.7480], device='cuda:1'), covar=tensor([0.0385, 0.0542, 0.0383, 0.0390, 0.0451, 0.0432, 0.0954, 0.0529], device='cuda:1'), in_proj_covar=tensor([0.0427, 0.0473, 0.0462, 0.0424, 0.0503, 0.0483, 0.0565, 0.0386], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:1') 2023-05-01 18:22:10,607 INFO [train.py:904] (1/8) Epoch 24, batch 1000, loss[loss=0.1611, simple_loss=0.2319, pruned_loss=0.04508, over 16859.00 frames. ], tot_loss[loss=0.1612, simple_loss=0.2467, pruned_loss=0.03787, over 3292901.56 frames. ], batch size: 116, lr: 2.83e-03, grad_scale: 4.0 2023-05-01 18:23:22,529 INFO [train.py:904] (1/8) Epoch 24, batch 1050, loss[loss=0.134, simple_loss=0.221, pruned_loss=0.0235, over 16768.00 frames. ], tot_loss[loss=0.1609, simple_loss=0.2463, pruned_loss=0.03773, over 3292798.82 frames. ], batch size: 39, lr: 2.83e-03, grad_scale: 4.0 2023-05-01 18:23:50,904 INFO [optim.py:368] (1/8) 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 18:24:23,696 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.6877, 3.7320, 2.9207, 2.2370, 2.4167, 2.3585, 3.8595, 3.2302], device='cuda:1'), covar=tensor([0.2730, 0.0611, 0.1701, 0.3114, 0.2767, 0.2243, 0.0496, 0.1606], device='cuda:1'), in_proj_covar=tensor([0.0330, 0.0270, 0.0307, 0.0317, 0.0297, 0.0265, 0.0297, 0.0341], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:1') 2023-05-01 18:24:30,871 INFO [train.py:904] (1/8) Epoch 24, batch 1100, loss[loss=0.1704, simple_loss=0.2382, pruned_loss=0.05136, over 16403.00 frames. ], tot_loss[loss=0.1604, simple_loss=0.2456, pruned_loss=0.03759, over 3301605.96 frames. ], batch size: 146, lr: 2.83e-03, grad_scale: 4.0 2023-05-01 18:24:37,559 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=234558.0, num_to_drop=1, layers_to_drop={1} 2023-05-01 18:24:59,148 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([5.2702, 5.2440, 5.1110, 4.5188, 4.6812, 5.1183, 5.1173, 4.7423], device='cuda:1'), covar=tensor([0.0592, 0.0491, 0.0340, 0.0394, 0.1232, 0.0497, 0.0418, 0.0815], device='cuda:1'), in_proj_covar=tensor([0.0304, 0.0449, 0.0355, 0.0354, 0.0358, 0.0413, 0.0241, 0.0426], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002], device='cuda:1') 2023-05-01 18:25:17,444 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=234587.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 18:25:38,659 INFO [train.py:904] (1/8) Epoch 24, batch 1150, loss[loss=0.1614, simple_loss=0.2573, pruned_loss=0.03276, over 17127.00 frames. ], tot_loss[loss=0.1607, simple_loss=0.2463, pruned_loss=0.03754, over 3294331.30 frames. ], batch size: 48, lr: 2.83e-03, grad_scale: 4.0 2023-05-01 18:26:06,112 INFO [optim.py:368] (1/8) 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 18:26:25,976 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.3381, 4.3550, 4.6848, 4.6827, 4.7488, 4.4251, 4.4689, 4.3239], device='cuda:1'), covar=tensor([0.0388, 0.0697, 0.0409, 0.0393, 0.0467, 0.0480, 0.0773, 0.0581], device='cuda:1'), in_proj_covar=tensor([0.0425, 0.0471, 0.0461, 0.0421, 0.0501, 0.0480, 0.0563, 0.0383], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:1') 2023-05-01 18:26:46,784 INFO [train.py:904] (1/8) Epoch 24, batch 1200, loss[loss=0.155, simple_loss=0.2309, pruned_loss=0.03952, over 16118.00 frames. ], tot_loss[loss=0.1602, simple_loss=0.2455, pruned_loss=0.03741, over 3308432.55 frames. ], batch size: 164, lr: 2.83e-03, grad_scale: 8.0 2023-05-01 18:27:09,101 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=234669.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 18:27:25,054 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=234681.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 18:27:53,735 INFO [train.py:904] (1/8) Epoch 24, batch 1250, loss[loss=0.1761, simple_loss=0.2487, pruned_loss=0.05174, over 16791.00 frames. ], tot_loss[loss=0.1614, simple_loss=0.2458, pruned_loss=0.03853, over 3308629.73 frames. ], batch size: 124, lr: 2.83e-03, grad_scale: 8.0 2023-05-01 18:28:14,394 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.1468, 2.2307, 2.3711, 3.7640, 2.2393, 2.5312, 2.3119, 2.3505], device='cuda:1'), covar=tensor([0.1621, 0.3657, 0.3065, 0.0724, 0.4080, 0.2572, 0.3993, 0.3270], device='cuda:1'), in_proj_covar=tensor([0.0409, 0.0459, 0.0377, 0.0331, 0.0441, 0.0525, 0.0430, 0.0536], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-01 18:28:21,122 INFO [optim.py:368] (1/8) 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 18:28:25,276 INFO [zipformer.py:625] (1/8) 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] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=234729.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 18:28:31,734 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=234730.0, num_to_drop=1, layers_to_drop={1} 2023-05-01 18:28:35,885 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.6123, 3.6126, 2.6521, 2.2170, 2.3373, 2.1880, 3.6799, 3.0407], device='cuda:1'), covar=tensor([0.2978, 0.0659, 0.2099, 0.3046, 0.2857, 0.2534, 0.0629, 0.1754], device='cuda:1'), in_proj_covar=tensor([0.0331, 0.0271, 0.0309, 0.0318, 0.0298, 0.0266, 0.0298, 0.0342], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:1') 2023-05-01 18:29:02,399 INFO [train.py:904] (1/8) Epoch 24, batch 1300, loss[loss=0.1493, simple_loss=0.2436, pruned_loss=0.02756, over 17203.00 frames. ], tot_loss[loss=0.1615, simple_loss=0.2459, pruned_loss=0.03851, over 3302278.13 frames. ], batch size: 46, lr: 2.83e-03, grad_scale: 8.0 2023-05-01 18:29:06,505 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=3.69 vs. limit=5.0 2023-05-01 18:29:41,665 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=234782.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 18:29:47,294 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=234787.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 18:30:10,000 INFO [train.py:904] (1/8) Epoch 24, batch 1350, loss[loss=0.1791, simple_loss=0.2563, pruned_loss=0.05093, over 16894.00 frames. ], tot_loss[loss=0.1617, simple_loss=0.2465, pruned_loss=0.03844, over 3313588.43 frames. ], batch size: 109, lr: 2.83e-03, grad_scale: 8.0 2023-05-01 18:30:10,289 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([5.0532, 5.5738, 5.7277, 5.4911, 5.5496, 6.1134, 5.5838, 5.2568], device='cuda:1'), covar=tensor([0.0963, 0.1731, 0.2221, 0.1998, 0.2452, 0.0896, 0.1622, 0.2445], device='cuda:1'), in_proj_covar=tensor([0.0419, 0.0613, 0.0677, 0.0508, 0.0675, 0.0707, 0.0528, 0.0677], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-05-01 18:30:38,934 INFO [optim.py:368] (1/8) 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 18:31:07,103 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=234843.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 18:31:19,625 INFO [train.py:904] (1/8) Epoch 24, batch 1400, loss[loss=0.1706, simple_loss=0.2667, pruned_loss=0.03725, over 17114.00 frames. ], tot_loss[loss=0.1612, simple_loss=0.246, pruned_loss=0.0382, over 3308821.30 frames. ], batch size: 49, lr: 2.83e-03, grad_scale: 8.0 2023-05-01 18:31:27,985 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=234858.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 18:32:00,220 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=234882.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 18:32:07,288 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 24, batch 1450, loss[loss=0.1718, simple_loss=0.2407, pruned_loss=0.05139, over 16909.00 frames. ], tot_loss[loss=0.1612, simple_loss=0.2459, pruned_loss=0.0383, over 3318719.22 frames. ], batch size: 90, lr: 2.83e-03, grad_scale: 4.0 2023-05-01 18:32:34,693 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=234906.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 18:32:58,961 INFO [optim.py:368] (1/8) 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] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=234935.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 18:33:26,764 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=234943.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 18:33:38,963 INFO [train.py:904] (1/8) Epoch 24, batch 1500, loss[loss=0.1798, simple_loss=0.249, pruned_loss=0.05527, over 16468.00 frames. ], tot_loss[loss=0.161, simple_loss=0.2455, pruned_loss=0.03821, over 3319358.94 frames. ], batch size: 75, lr: 2.83e-03, grad_scale: 4.0 2023-05-01 18:33:45,642 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=234958.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 18:33:49,671 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=234960.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 18:34:49,724 INFO [train.py:904] (1/8) Epoch 24, batch 1550, loss[loss=0.1933, simple_loss=0.2692, pruned_loss=0.05874, over 16470.00 frames. ], tot_loss[loss=0.1621, simple_loss=0.2468, pruned_loss=0.03876, over 3327417.82 frames. ], batch size: 75, lr: 2.83e-03, grad_scale: 4.0 2023-05-01 18:35:12,930 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=235019.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 18:35:15,734 INFO [zipformer.py:625] (1/8) 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] (1/8) 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] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=235025.0, num_to_drop=1, layers_to_drop={0} 2023-05-01 18:35:34,013 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.89 vs. limit=2.0 2023-05-01 18:35:58,105 INFO [train.py:904] (1/8) Epoch 24, batch 1600, loss[loss=0.1751, simple_loss=0.2684, pruned_loss=0.04087, over 17057.00 frames. ], tot_loss[loss=0.1637, simple_loss=0.2485, pruned_loss=0.03945, over 3325760.73 frames. ], batch size: 55, lr: 2.83e-03, grad_scale: 8.0 2023-05-01 18:36:37,257 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.14 vs. limit=2.0 2023-05-01 18:36:37,994 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=235082.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 18:37:06,937 INFO [train.py:904] (1/8) Epoch 24, batch 1650, loss[loss=0.1417, simple_loss=0.2298, pruned_loss=0.02678, over 17235.00 frames. ], tot_loss[loss=0.1648, simple_loss=0.2497, pruned_loss=0.03997, over 3315087.58 frames. ], batch size: 45, lr: 2.83e-03, grad_scale: 8.0 2023-05-01 18:37:25,545 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.8376, 2.1254, 2.4671, 2.8400, 2.7636, 3.2812, 2.2409, 3.3266], device='cuda:1'), covar=tensor([0.0327, 0.0523, 0.0392, 0.0392, 0.0384, 0.0239, 0.0538, 0.0172], device='cuda:1'), in_proj_covar=tensor([0.0195, 0.0197, 0.0183, 0.0188, 0.0203, 0.0160, 0.0200, 0.0158], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-01 18:37:32,031 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=4.89 vs. limit=5.0 2023-05-01 18:37:35,278 INFO [optim.py:368] (1/8) 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 18:37:39,407 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.80 vs. limit=2.0 2023-05-01 18:37:43,734 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.0666, 4.1034, 4.4261, 4.3925, 4.4260, 4.1504, 4.1627, 4.0986], device='cuda:1'), covar=tensor([0.0400, 0.0685, 0.0388, 0.0417, 0.0529, 0.0487, 0.0833, 0.0647], device='cuda:1'), in_proj_covar=tensor([0.0425, 0.0473, 0.0461, 0.0424, 0.0503, 0.0481, 0.0566, 0.0385], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:1') 2023-05-01 18:37:54,826 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=235138.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 18:38:16,053 INFO [train.py:904] (1/8) Epoch 24, batch 1700, loss[loss=0.1751, simple_loss=0.2676, pruned_loss=0.0413, over 17124.00 frames. ], tot_loss[loss=0.1653, simple_loss=0.251, pruned_loss=0.03984, over 3326647.88 frames. ], batch size: 48, lr: 2.83e-03, grad_scale: 8.0 2023-05-01 18:38:43,697 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.18 vs. limit=2.0 2023-05-01 18:38:48,892 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.5883, 2.3684, 1.9057, 2.1493, 2.7173, 2.4511, 2.5871, 2.7773], device='cuda:1'), covar=tensor([0.0277, 0.0432, 0.0555, 0.0466, 0.0228, 0.0333, 0.0236, 0.0278], device='cuda:1'), in_proj_covar=tensor([0.0226, 0.0244, 0.0233, 0.0234, 0.0244, 0.0243, 0.0243, 0.0240], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-01 18:38:50,987 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.9476, 2.5320, 1.9241, 2.3503, 2.9449, 2.7036, 2.8696, 3.0210], device='cuda:1'), covar=tensor([0.0283, 0.0458, 0.0664, 0.0501, 0.0262, 0.0368, 0.0293, 0.0302], device='cuda:1'), in_proj_covar=tensor([0.0226, 0.0244, 0.0233, 0.0234, 0.0244, 0.0243, 0.0243, 0.0240], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-01 18:39:18,112 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.18 vs. limit=2.0 2023-05-01 18:39:24,668 INFO [train.py:904] (1/8) Epoch 24, batch 1750, loss[loss=0.1855, simple_loss=0.2753, pruned_loss=0.04788, over 17076.00 frames. ], tot_loss[loss=0.1659, simple_loss=0.2517, pruned_loss=0.04001, over 3333993.86 frames. ], batch size: 53, lr: 2.83e-03, grad_scale: 8.0 2023-05-01 18:39:52,420 INFO [optim.py:368] (1/8) 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 18:40:12,948 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=235238.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 18:40:33,174 INFO [train.py:904] (1/8) Epoch 24, batch 1800, loss[loss=0.1564, simple_loss=0.2556, pruned_loss=0.02865, over 17229.00 frames. ], tot_loss[loss=0.1665, simple_loss=0.2532, pruned_loss=0.03995, over 3323005.06 frames. ], batch size: 45, lr: 2.83e-03, grad_scale: 8.0 2023-05-01 18:40:36,398 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=235255.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 18:41:40,804 INFO [train.py:904] (1/8) Epoch 24, batch 1850, loss[loss=0.1746, simple_loss=0.2506, pruned_loss=0.0493, over 16923.00 frames. ], tot_loss[loss=0.1666, simple_loss=0.2536, pruned_loss=0.03983, over 3320085.04 frames. ], batch size: 109, lr: 2.83e-03, grad_scale: 4.0 2023-05-01 18:41:56,547 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=235314.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 18:41:59,455 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=235316.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 18:41:59,606 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=235316.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 18:42:11,306 INFO [optim.py:368] (1/8) 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 18:42:11,682 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=235325.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 18:42:49,548 INFO [train.py:904] (1/8) Epoch 24, batch 1900, loss[loss=0.1808, simple_loss=0.2554, pruned_loss=0.05306, over 16295.00 frames. ], tot_loss[loss=0.166, simple_loss=0.2532, pruned_loss=0.03939, over 3324662.40 frames. ], batch size: 165, lr: 2.83e-03, grad_scale: 4.0 2023-05-01 18:43:01,064 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=235360.0, num_to_drop=1, layers_to_drop={0} 2023-05-01 18:43:17,587 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=235373.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 18:43:25,960 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=4.60 vs. limit=5.0 2023-05-01 18:43:31,656 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=235382.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 18:44:00,955 INFO [train.py:904] (1/8) Epoch 24, batch 1950, loss[loss=0.1501, simple_loss=0.242, pruned_loss=0.02904, over 17231.00 frames. ], tot_loss[loss=0.1658, simple_loss=0.2533, pruned_loss=0.0391, over 3322826.54 frames. ], batch size: 45, lr: 2.83e-03, grad_scale: 4.0 2023-05-01 18:44:26,840 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=235421.0, num_to_drop=1, layers_to_drop={2} 2023-05-01 18:44:30,633 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.17 vs. limit=2.0 2023-05-01 18:44:31,114 INFO [optim.py:368] (1/8) Clipping_scale=2.0, grad-norm quartiles 1.458e+02 2.100e+02 2.476e+02 2.993e+02 5.203e+02, threshold=4.952e+02, percent-clipped=1.0 2023-05-01 18:44:37,916 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=235430.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 18:44:49,015 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 24, batch 2000, loss[loss=0.1986, simple_loss=0.287, pruned_loss=0.05511, over 11654.00 frames. ], tot_loss[loss=0.1655, simple_loss=0.253, pruned_loss=0.03895, over 3321394.82 frames. ], batch size: 246, lr: 2.83e-03, grad_scale: 8.0 2023-05-01 18:45:53,863 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=235486.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 18:46:15,190 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.5544, 3.4870, 3.6773, 2.6402, 3.3533, 3.7962, 3.5080, 2.2858], device='cuda:1'), covar=tensor([0.0485, 0.0197, 0.0067, 0.0396, 0.0129, 0.0104, 0.0114, 0.0483], device='cuda:1'), in_proj_covar=tensor([0.0137, 0.0087, 0.0087, 0.0135, 0.0100, 0.0112, 0.0097, 0.0131], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0003, 0.0004], device='cuda:1') 2023-05-01 18:46:16,400 INFO [train.py:904] (1/8) Epoch 24, batch 2050, loss[loss=0.1874, simple_loss=0.257, pruned_loss=0.05885, over 16866.00 frames. ], tot_loss[loss=0.1664, simple_loss=0.254, pruned_loss=0.03939, over 3327002.65 frames. ], batch size: 109, lr: 2.83e-03, grad_scale: 8.0 2023-05-01 18:46:46,284 INFO [optim.py:368] (1/8) 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 18:47:03,575 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=235538.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 18:47:13,568 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.93 vs. limit=2.0 2023-05-01 18:47:24,546 INFO [train.py:904] (1/8) Epoch 24, batch 2100, loss[loss=0.1486, simple_loss=0.234, pruned_loss=0.03157, over 16995.00 frames. ], tot_loss[loss=0.1686, simple_loss=0.2553, pruned_loss=0.04089, over 3319873.98 frames. ], batch size: 41, lr: 2.83e-03, grad_scale: 8.0 2023-05-01 18:48:10,073 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=235586.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 18:48:31,092 INFO [train.py:904] (1/8) Epoch 24, batch 2150, loss[loss=0.1435, simple_loss=0.2263, pruned_loss=0.0304, over 16025.00 frames. ], tot_loss[loss=0.1689, simple_loss=0.256, pruned_loss=0.04089, over 3313658.48 frames. ], batch size: 35, lr: 2.82e-03, grad_scale: 8.0 2023-05-01 18:48:43,103 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=235611.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 18:48:46,868 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=235614.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 18:48:49,945 INFO [zipformer.py:625] (1/8) 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] (1/8) 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 18:49:41,310 INFO [train.py:904] (1/8) Epoch 24, batch 2200, loss[loss=0.1598, simple_loss=0.2499, pruned_loss=0.03486, over 16831.00 frames. ], tot_loss[loss=0.1698, simple_loss=0.2568, pruned_loss=0.04137, over 3299603.80 frames. ], batch size: 42, lr: 2.82e-03, grad_scale: 8.0 2023-05-01 18:49:42,229 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.23 vs. limit=2.0 2023-05-01 18:49:53,762 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=235662.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 18:49:57,650 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=235664.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 18:50:18,219 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=235679.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 18:50:50,800 INFO [train.py:904] (1/8) Epoch 24, batch 2250, loss[loss=0.1653, simple_loss=0.2638, pruned_loss=0.03342, over 17049.00 frames. ], tot_loss[loss=0.1713, simple_loss=0.258, pruned_loss=0.04233, over 3295725.77 frames. ], batch size: 50, lr: 2.82e-03, grad_scale: 8.0 2023-05-01 18:51:08,547 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=235716.0, num_to_drop=1, layers_to_drop={3} 2023-05-01 18:51:20,155 INFO [optim.py:368] (1/8) 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 18:51:41,268 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=235740.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 18:51:43,883 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.69 vs. limit=2.0 2023-05-01 18:51:57,980 INFO [train.py:904] (1/8) Epoch 24, batch 2300, loss[loss=0.1726, simple_loss=0.2519, pruned_loss=0.04664, over 16781.00 frames. ], tot_loss[loss=0.1712, simple_loss=0.2581, pruned_loss=0.04212, over 3300871.99 frames. ], batch size: 124, lr: 2.82e-03, grad_scale: 8.0 2023-05-01 18:52:50,338 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 24, batch 2350, loss[loss=0.1559, simple_loss=0.2404, pruned_loss=0.03568, over 17237.00 frames. ], tot_loss[loss=0.1709, simple_loss=0.2582, pruned_loss=0.04185, over 3312521.99 frames. ], batch size: 45, lr: 2.82e-03, grad_scale: 8.0 2023-05-01 18:53:37,790 INFO [optim.py:368] (1/8) 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 18:53:50,031 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.1077, 3.8426, 4.2648, 2.3406, 4.5486, 4.5834, 3.2794, 3.5124], device='cuda:1'), covar=tensor([0.0678, 0.0267, 0.0232, 0.1108, 0.0075, 0.0181, 0.0457, 0.0396], device='cuda:1'), in_proj_covar=tensor([0.0146, 0.0108, 0.0098, 0.0138, 0.0081, 0.0127, 0.0127, 0.0129], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0004, 0.0003], device='cuda:1') 2023-05-01 18:54:03,707 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.90 vs. limit=2.0 2023-05-01 18:54:04,497 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.3934, 2.4468, 2.3834, 4.3248, 2.3113, 2.7711, 2.4318, 2.5544], device='cuda:1'), covar=tensor([0.1346, 0.3659, 0.3196, 0.0549, 0.4224, 0.2594, 0.3731, 0.3747], device='cuda:1'), in_proj_covar=tensor([0.0411, 0.0462, 0.0378, 0.0334, 0.0442, 0.0529, 0.0432, 0.0539], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-01 18:54:14,680 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=235852.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 18:54:15,306 INFO [train.py:904] (1/8) Epoch 24, batch 2400, loss[loss=0.2274, simple_loss=0.2951, pruned_loss=0.07981, over 16665.00 frames. ], tot_loss[loss=0.1708, simple_loss=0.2582, pruned_loss=0.04175, over 3319256.70 frames. ], batch size: 124, lr: 2.82e-03, grad_scale: 8.0 2023-05-01 18:55:22,142 INFO [train.py:904] (1/8) Epoch 24, batch 2450, loss[loss=0.177, simple_loss=0.2723, pruned_loss=0.04089, over 17040.00 frames. ], tot_loss[loss=0.1704, simple_loss=0.2585, pruned_loss=0.04118, over 3325150.53 frames. ], batch size: 53, lr: 2.82e-03, grad_scale: 8.0 2023-05-01 18:55:33,241 INFO [zipformer.py:625] (1/8) 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] (1/8) 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 18:56:28,797 INFO [train.py:904] (1/8) Epoch 24, batch 2500, loss[loss=0.156, simple_loss=0.2498, pruned_loss=0.03113, over 17115.00 frames. ], tot_loss[loss=0.17, simple_loss=0.2582, pruned_loss=0.04092, over 3328878.64 frames. ], batch size: 49, lr: 2.82e-03, grad_scale: 8.0 2023-05-01 18:56:37,232 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=235959.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 18:57:24,311 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.5360, 3.5996, 3.3130, 2.9547, 3.1864, 3.4774, 3.2995, 3.3341], device='cuda:1'), covar=tensor([0.0624, 0.0679, 0.0305, 0.0287, 0.0502, 0.0459, 0.1376, 0.0507], device='cuda:1'), in_proj_covar=tensor([0.0311, 0.0462, 0.0363, 0.0362, 0.0367, 0.0421, 0.0246, 0.0438], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-05-01 18:57:24,906 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.88 vs. limit=2.0 2023-05-01 18:57:41,344 INFO [train.py:904] (1/8) Epoch 24, batch 2550, loss[loss=0.1485, simple_loss=0.2446, pruned_loss=0.02622, over 17225.00 frames. ], tot_loss[loss=0.1701, simple_loss=0.2582, pruned_loss=0.04094, over 3314135.95 frames. ], batch size: 45, lr: 2.82e-03, grad_scale: 8.0 2023-05-01 18:57:58,514 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=236016.0, num_to_drop=1, layers_to_drop={2} 2023-05-01 18:58:01,805 INFO [zipformer.py:625] (1/8) 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] (1/8) 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 18:58:24,320 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=236035.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 18:58:36,215 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([5.0251, 5.3279, 5.1048, 5.1068, 4.8131, 4.7463, 4.7495, 5.4373], device='cuda:1'), covar=tensor([0.1253, 0.0864, 0.0986, 0.0843, 0.0911, 0.1109, 0.1211, 0.0879], device='cuda:1'), in_proj_covar=tensor([0.0714, 0.0869, 0.0714, 0.0668, 0.0552, 0.0556, 0.0731, 0.0679], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-01 18:58:49,127 INFO [train.py:904] (1/8) Epoch 24, batch 2600, loss[loss=0.1554, simple_loss=0.2538, pruned_loss=0.02854, over 17103.00 frames. ], tot_loss[loss=0.1698, simple_loss=0.2584, pruned_loss=0.04056, over 3314114.08 frames. ], batch size: 49, lr: 2.82e-03, grad_scale: 8.0 2023-05-01 18:59:01,921 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.1965, 3.3697, 3.4173, 2.4006, 3.1559, 3.5523, 3.2502, 1.9998], device='cuda:1'), covar=tensor([0.0553, 0.0115, 0.0071, 0.0398, 0.0128, 0.0104, 0.0114, 0.0507], device='cuda:1'), in_proj_covar=tensor([0.0136, 0.0087, 0.0087, 0.0135, 0.0100, 0.0112, 0.0097, 0.0131], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0003, 0.0004], device='cuda:1') 2023-05-01 18:59:03,926 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=236064.0, num_to_drop=1, layers_to_drop={0} 2023-05-01 18:59:25,494 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=236079.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 18:59:58,180 INFO [train.py:904] (1/8) Epoch 24, batch 2650, loss[loss=0.1643, simple_loss=0.2532, pruned_loss=0.0377, over 16420.00 frames. ], tot_loss[loss=0.1705, simple_loss=0.2597, pruned_loss=0.04067, over 3315094.36 frames. ], batch size: 68, lr: 2.82e-03, grad_scale: 8.0 2023-05-01 19:00:27,606 INFO [optim.py:368] (1/8) 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 19:00:58,522 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=236147.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 19:01:05,910 INFO [train.py:904] (1/8) Epoch 24, batch 2700, loss[loss=0.1937, simple_loss=0.2769, pruned_loss=0.05523, over 16892.00 frames. ], tot_loss[loss=0.1697, simple_loss=0.2593, pruned_loss=0.04, over 3311735.00 frames. ], batch size: 109, lr: 2.82e-03, grad_scale: 8.0 2023-05-01 19:01:44,429 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.7279, 3.9019, 2.6219, 4.5557, 3.1616, 4.5228, 2.7580, 3.3465], device='cuda:1'), covar=tensor([0.0383, 0.0450, 0.1516, 0.0296, 0.0806, 0.0551, 0.1376, 0.0703], device='cuda:1'), in_proj_covar=tensor([0.0176, 0.0182, 0.0198, 0.0174, 0.0180, 0.0224, 0.0207, 0.0185], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-05-01 19:02:15,469 INFO [train.py:904] (1/8) Epoch 24, batch 2750, loss[loss=0.1628, simple_loss=0.2493, pruned_loss=0.03816, over 16864.00 frames. ], tot_loss[loss=0.169, simple_loss=0.2586, pruned_loss=0.03973, over 3314508.74 frames. ], batch size: 102, lr: 2.82e-03, grad_scale: 8.0 2023-05-01 19:02:44,668 INFO [optim.py:368] (1/8) 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 19:03:13,304 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=4.91 vs. limit=5.0 2023-05-01 19:03:22,962 INFO [train.py:904] (1/8) Epoch 24, batch 2800, loss[loss=0.1662, simple_loss=0.2486, pruned_loss=0.04192, over 16727.00 frames. ], tot_loss[loss=0.1679, simple_loss=0.2577, pruned_loss=0.03903, over 3323489.17 frames. ], batch size: 89, lr: 2.82e-03, grad_scale: 8.0 2023-05-01 19:03:37,568 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.5812, 4.8340, 4.6619, 4.6535, 4.4230, 4.3564, 4.3580, 4.9110], device='cuda:1'), covar=tensor([0.1158, 0.0886, 0.0986, 0.0818, 0.0803, 0.1337, 0.1082, 0.0937], device='cuda:1'), in_proj_covar=tensor([0.0716, 0.0873, 0.0717, 0.0670, 0.0554, 0.0558, 0.0733, 0.0682], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-01 19:04:32,914 INFO [train.py:904] (1/8) Epoch 24, batch 2850, loss[loss=0.2, simple_loss=0.2892, pruned_loss=0.05542, over 16714.00 frames. ], tot_loss[loss=0.1679, simple_loss=0.2571, pruned_loss=0.03936, over 3311179.69 frames. ], batch size: 57, lr: 2.82e-03, grad_scale: 8.0 2023-05-01 19:05:04,031 INFO [optim.py:368] (1/8) 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 19:05:17,527 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=236335.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 19:05:43,130 INFO [train.py:904] (1/8) Epoch 24, batch 2900, loss[loss=0.1708, simple_loss=0.2501, pruned_loss=0.04573, over 16400.00 frames. ], tot_loss[loss=0.1684, simple_loss=0.2567, pruned_loss=0.04005, over 3316472.66 frames. ], batch size: 146, lr: 2.82e-03, grad_scale: 8.0 2023-05-01 19:05:49,338 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.30 vs. limit=2.0 2023-05-01 19:06:12,649 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=236374.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 19:06:16,693 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([5.2168, 5.1170, 5.5613, 5.5157, 5.6589, 5.2642, 5.1774, 5.0389], device='cuda:1'), covar=tensor([0.0403, 0.0781, 0.0509, 0.0628, 0.0562, 0.0527, 0.1145, 0.0524], device='cuda:1'), in_proj_covar=tensor([0.0430, 0.0478, 0.0465, 0.0429, 0.0509, 0.0485, 0.0570, 0.0389], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:1') 2023-05-01 19:06:25,482 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=236383.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 19:06:48,669 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.0827, 2.2342, 2.7241, 3.0810, 2.9396, 3.5920, 2.5383, 3.6134], device='cuda:1'), covar=tensor([0.0293, 0.0537, 0.0344, 0.0348, 0.0347, 0.0217, 0.0491, 0.0186], device='cuda:1'), in_proj_covar=tensor([0.0197, 0.0197, 0.0185, 0.0190, 0.0205, 0.0163, 0.0201, 0.0159], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-01 19:06:53,586 INFO [train.py:904] (1/8) Epoch 24, batch 2950, loss[loss=0.1788, simple_loss=0.2613, pruned_loss=0.04813, over 16740.00 frames. ], tot_loss[loss=0.1692, simple_loss=0.2561, pruned_loss=0.04112, over 3307187.28 frames. ], batch size: 124, lr: 2.82e-03, grad_scale: 8.0 2023-05-01 19:07:24,081 INFO [optim.py:368] (1/8) 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 19:07:50,673 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.9934, 3.8563, 4.2857, 2.3128, 4.5254, 4.5554, 3.4196, 3.5262], device='cuda:1'), covar=tensor([0.0751, 0.0274, 0.0222, 0.1190, 0.0072, 0.0188, 0.0392, 0.0406], device='cuda:1'), in_proj_covar=tensor([0.0147, 0.0108, 0.0099, 0.0138, 0.0082, 0.0128, 0.0128, 0.0130], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0004, 0.0004], device='cuda:1') 2023-05-01 19:07:54,808 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([5.5280, 5.4854, 5.3601, 4.8526, 4.9689, 5.4031, 5.3517, 4.9702], device='cuda:1'), covar=tensor([0.0539, 0.0488, 0.0294, 0.0349, 0.1111, 0.0468, 0.0269, 0.0800], device='cuda:1'), in_proj_covar=tensor([0.0316, 0.0468, 0.0368, 0.0367, 0.0372, 0.0427, 0.0251, 0.0444], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-05-01 19:07:54,812 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=236447.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 19:08:02,711 INFO [train.py:904] (1/8) Epoch 24, batch 3000, loss[loss=0.1489, simple_loss=0.2353, pruned_loss=0.03124, over 16796.00 frames. ], tot_loss[loss=0.1692, simple_loss=0.2562, pruned_loss=0.04111, over 3318716.82 frames. ], batch size: 39, lr: 2.82e-03, grad_scale: 8.0 2023-05-01 19:08:02,711 INFO [train.py:929] (1/8) Computing validation loss 2023-05-01 19:08:12,054 INFO [train.py:938] (1/8) Epoch 24, validation: loss=0.1342, simple_loss=0.2393, pruned_loss=0.0145, over 944034.00 frames. 2023-05-01 19:08:12,054 INFO [train.py:939] (1/8) Maximum memory allocated so far is 18145MB 2023-05-01 19:09:12,138 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=236495.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 19:09:22,991 INFO [train.py:904] (1/8) Epoch 24, batch 3050, loss[loss=0.1736, simple_loss=0.2547, pruned_loss=0.04625, over 16686.00 frames. ], tot_loss[loss=0.1689, simple_loss=0.2557, pruned_loss=0.04109, over 3308765.79 frames. ], batch size: 89, lr: 2.82e-03, grad_scale: 8.0 2023-05-01 19:09:53,424 INFO [optim.py:368] (1/8) 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 19:10:32,488 INFO [train.py:904] (1/8) Epoch 24, batch 3100, loss[loss=0.1643, simple_loss=0.245, pruned_loss=0.04175, over 16308.00 frames. ], tot_loss[loss=0.1692, simple_loss=0.256, pruned_loss=0.04116, over 3314846.30 frames. ], batch size: 36, lr: 2.82e-03, grad_scale: 8.0 2023-05-01 19:11:43,886 INFO [train.py:904] (1/8) Epoch 24, batch 3150, loss[loss=0.164, simple_loss=0.2408, pruned_loss=0.04362, over 16714.00 frames. ], tot_loss[loss=0.1688, simple_loss=0.2552, pruned_loss=0.0412, over 3309257.17 frames. ], batch size: 89, lr: 2.82e-03, grad_scale: 8.0 2023-05-01 19:11:46,055 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.25 vs. limit=2.0 2023-05-01 19:12:13,744 INFO [optim.py:368] (1/8) 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 19:12:52,325 INFO [train.py:904] (1/8) Epoch 24, batch 3200, loss[loss=0.1745, simple_loss=0.2589, pruned_loss=0.04507, over 12569.00 frames. ], tot_loss[loss=0.1678, simple_loss=0.2538, pruned_loss=0.04093, over 3311836.41 frames. ], batch size: 246, lr: 2.82e-03, grad_scale: 8.0 2023-05-01 19:13:14,999 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([5.2104, 5.5376, 5.2947, 5.3034, 5.0374, 5.0068, 4.9603, 5.6440], device='cuda:1'), covar=tensor([0.1186, 0.0866, 0.0995, 0.0906, 0.0834, 0.0880, 0.1122, 0.0839], device='cuda:1'), in_proj_covar=tensor([0.0711, 0.0868, 0.0712, 0.0666, 0.0552, 0.0554, 0.0727, 0.0677], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:1') 2023-05-01 19:13:15,149 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([1.8264, 2.9788, 2.7897, 4.5432, 3.8271, 4.2030, 1.5690, 3.4170], device='cuda:1'), covar=tensor([0.1359, 0.0663, 0.1086, 0.0191, 0.0242, 0.0329, 0.1652, 0.0657], device='cuda:1'), in_proj_covar=tensor([0.0169, 0.0177, 0.0197, 0.0197, 0.0207, 0.0218, 0.0205, 0.0195], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-05-01 19:13:21,953 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 24, batch 3250, loss[loss=0.1502, simple_loss=0.2337, pruned_loss=0.03335, over 15834.00 frames. ], tot_loss[loss=0.168, simple_loss=0.2542, pruned_loss=0.04086, over 3310397.03 frames. ], batch size: 35, lr: 2.82e-03, grad_scale: 8.0 2023-05-01 19:14:26,977 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=236722.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 19:14:31,667 INFO [optim.py:368] (1/8) 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 19:15:07,170 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.47 vs. limit=2.0 2023-05-01 19:15:11,546 INFO [train.py:904] (1/8) Epoch 24, batch 3300, loss[loss=0.1529, simple_loss=0.2497, pruned_loss=0.02803, over 17097.00 frames. ], tot_loss[loss=0.1692, simple_loss=0.2557, pruned_loss=0.04133, over 3316156.23 frames. ], batch size: 49, lr: 2.82e-03, grad_scale: 8.0 2023-05-01 19:16:21,167 INFO [train.py:904] (1/8) Epoch 24, batch 3350, loss[loss=0.1315, simple_loss=0.2173, pruned_loss=0.02288, over 16837.00 frames. ], tot_loss[loss=0.169, simple_loss=0.2557, pruned_loss=0.04114, over 3319721.08 frames. ], batch size: 39, lr: 2.82e-03, grad_scale: 8.0 2023-05-01 19:16:25,278 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.6791, 2.3462, 1.8342, 2.0138, 2.6651, 2.4666, 2.6837, 2.8307], device='cuda:1'), covar=tensor([0.0274, 0.0508, 0.0659, 0.0615, 0.0288, 0.0424, 0.0257, 0.0348], device='cuda:1'), in_proj_covar=tensor([0.0229, 0.0245, 0.0234, 0.0236, 0.0245, 0.0246, 0.0247, 0.0243], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-01 19:16:51,747 INFO [optim.py:368] (1/8) 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 19:17:26,801 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=3.06 vs. limit=5.0 2023-05-01 19:17:33,207 INFO [train.py:904] (1/8) Epoch 24, batch 3400, loss[loss=0.1547, simple_loss=0.2437, pruned_loss=0.03283, over 17116.00 frames. ], tot_loss[loss=0.1691, simple_loss=0.2563, pruned_loss=0.04102, over 3324565.06 frames. ], batch size: 48, lr: 2.82e-03, grad_scale: 8.0 2023-05-01 19:18:31,042 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=236892.0, num_to_drop=1, layers_to_drop={1} 2023-05-01 19:18:36,742 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.21 vs. limit=2.0 2023-05-01 19:18:42,274 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([5.2307, 5.7923, 5.8964, 5.5931, 5.6967, 6.2517, 5.7967, 5.4740], device='cuda:1'), covar=tensor([0.0897, 0.1873, 0.2425, 0.2169, 0.2779, 0.1006, 0.1479, 0.2462], device='cuda:1'), in_proj_covar=tensor([0.0426, 0.0626, 0.0692, 0.0518, 0.0689, 0.0719, 0.0540, 0.0685], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-05-01 19:18:45,596 INFO [train.py:904] (1/8) Epoch 24, batch 3450, loss[loss=0.1706, simple_loss=0.2538, pruned_loss=0.04366, over 16826.00 frames. ], tot_loss[loss=0.1677, simple_loss=0.2546, pruned_loss=0.04046, over 3323889.42 frames. ], batch size: 102, lr: 2.82e-03, grad_scale: 8.0 2023-05-01 19:19:04,688 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([5.3932, 5.7620, 5.4921, 5.5507, 5.1518, 5.0931, 5.1445, 5.8873], device='cuda:1'), covar=tensor([0.1459, 0.1054, 0.1104, 0.0928, 0.0977, 0.0888, 0.1272, 0.0936], device='cuda:1'), in_proj_covar=tensor([0.0718, 0.0873, 0.0720, 0.0673, 0.0557, 0.0559, 0.0733, 0.0684], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-01 19:19:15,854 INFO [optim.py:368] (1/8) 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 19:19:51,412 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.47 vs. limit=2.0 2023-05-01 19:19:55,267 INFO [train.py:904] (1/8) Epoch 24, batch 3500, loss[loss=0.1918, simple_loss=0.2794, pruned_loss=0.05208, over 16690.00 frames. ], tot_loss[loss=0.1674, simple_loss=0.2539, pruned_loss=0.04039, over 3322443.84 frames. ], batch size: 124, lr: 2.82e-03, grad_scale: 8.0 2023-05-01 19:19:56,869 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=236953.0, num_to_drop=1, layers_to_drop={1} 2023-05-01 19:21:06,713 INFO [train.py:904] (1/8) Epoch 24, batch 3550, loss[loss=0.1646, simple_loss=0.251, pruned_loss=0.03904, over 16505.00 frames. ], tot_loss[loss=0.1661, simple_loss=0.2525, pruned_loss=0.03983, over 3319774.98 frames. ], batch size: 68, lr: 2.82e-03, grad_scale: 8.0 2023-05-01 19:21:35,731 INFO [optim.py:368] (1/8) 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 19:21:47,364 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.2649, 2.6647, 2.0165, 2.3347, 2.9595, 2.7052, 3.0741, 3.1267], device='cuda:1'), covar=tensor([0.0194, 0.0399, 0.0587, 0.0461, 0.0272, 0.0359, 0.0262, 0.0265], device='cuda:1'), in_proj_covar=tensor([0.0228, 0.0243, 0.0233, 0.0235, 0.0244, 0.0244, 0.0246, 0.0242], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-01 19:22:15,045 INFO [train.py:904] (1/8) Epoch 24, batch 3600, loss[loss=0.1585, simple_loss=0.2496, pruned_loss=0.03374, over 17079.00 frames. ], tot_loss[loss=0.1661, simple_loss=0.2522, pruned_loss=0.04, over 3317561.26 frames. ], batch size: 55, lr: 2.82e-03, grad_scale: 8.0 2023-05-01 19:23:26,122 INFO [train.py:904] (1/8) Epoch 24, batch 3650, loss[loss=0.1486, simple_loss=0.2324, pruned_loss=0.03238, over 16794.00 frames. ], tot_loss[loss=0.1655, simple_loss=0.2507, pruned_loss=0.0402, over 3318089.13 frames. ], batch size: 39, lr: 2.82e-03, grad_scale: 8.0 2023-05-01 19:23:58,725 INFO [optim.py:368] (1/8) 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 19:24:25,215 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.72 vs. limit=2.0 2023-05-01 19:24:39,858 INFO [train.py:904] (1/8) Epoch 24, batch 3700, loss[loss=0.1604, simple_loss=0.2419, pruned_loss=0.03943, over 15434.00 frames. ], tot_loss[loss=0.1661, simple_loss=0.2495, pruned_loss=0.04134, over 3287880.46 frames. ], batch size: 190, lr: 2.82e-03, grad_scale: 4.0 2023-05-01 19:25:53,323 INFO [train.py:904] (1/8) Epoch 24, batch 3750, loss[loss=0.1784, simple_loss=0.2572, pruned_loss=0.04981, over 16479.00 frames. ], tot_loss[loss=0.1687, simple_loss=0.2511, pruned_loss=0.04309, over 3260228.33 frames. ], batch size: 146, lr: 2.82e-03, grad_scale: 4.0 2023-05-01 19:26:10,803 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.8229, 3.7911, 3.9056, 3.7107, 3.8515, 4.2851, 3.9420, 3.5767], device='cuda:1'), covar=tensor([0.2273, 0.2379, 0.2400, 0.2458, 0.2707, 0.1776, 0.1564, 0.2492], device='cuda:1'), in_proj_covar=tensor([0.0423, 0.0624, 0.0686, 0.0516, 0.0684, 0.0716, 0.0539, 0.0679], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-05-01 19:26:25,688 INFO [optim.py:368] (1/8) 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 19:26:57,492 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=237248.0, num_to_drop=1, layers_to_drop={1} 2023-05-01 19:27:03,606 INFO [train.py:904] (1/8) Epoch 24, batch 3800, loss[loss=0.194, simple_loss=0.291, pruned_loss=0.04848, over 16586.00 frames. ], tot_loss[loss=0.1708, simple_loss=0.2528, pruned_loss=0.04442, over 3245247.64 frames. ], batch size: 62, lr: 2.82e-03, grad_scale: 4.0 2023-05-01 19:27:40,015 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.9810, 2.0833, 2.3355, 3.5833, 2.1225, 2.3608, 2.1977, 2.2449], device='cuda:1'), covar=tensor([0.1614, 0.4018, 0.3145, 0.0787, 0.4178, 0.2822, 0.4042, 0.3322], device='cuda:1'), in_proj_covar=tensor([0.0416, 0.0465, 0.0381, 0.0337, 0.0444, 0.0534, 0.0435, 0.0544], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-01 19:28:05,968 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.31 vs. limit=2.0 2023-05-01 19:28:20,060 INFO [train.py:904] (1/8) Epoch 24, batch 3850, loss[loss=0.1627, simple_loss=0.2441, pruned_loss=0.04068, over 16463.00 frames. ], tot_loss[loss=0.1709, simple_loss=0.2523, pruned_loss=0.04475, over 3259420.70 frames. ], batch size: 68, lr: 2.81e-03, grad_scale: 4.0 2023-05-01 19:28:52,884 INFO [optim.py:368] (1/8) 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 19:29:02,078 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.9381, 4.9177, 4.7817, 3.9272, 4.9321, 1.9000, 4.6356, 4.3718], device='cuda:1'), covar=tensor([0.0152, 0.0139, 0.0245, 0.0546, 0.0128, 0.3181, 0.0182, 0.0332], device='cuda:1'), in_proj_covar=tensor([0.0175, 0.0167, 0.0210, 0.0186, 0.0185, 0.0214, 0.0197, 0.0180], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-01 19:29:18,610 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.8607, 3.5521, 4.0552, 2.1344, 4.1057, 4.1171, 3.3773, 3.0522], device='cuda:1'), covar=tensor([0.0707, 0.0271, 0.0153, 0.1197, 0.0086, 0.0189, 0.0334, 0.0448], device='cuda:1'), in_proj_covar=tensor([0.0148, 0.0110, 0.0100, 0.0140, 0.0083, 0.0129, 0.0129, 0.0131], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-05-01 19:29:30,846 INFO [train.py:904] (1/8) Epoch 24, batch 3900, loss[loss=0.1747, simple_loss=0.2531, pruned_loss=0.04815, over 16510.00 frames. ], tot_loss[loss=0.1707, simple_loss=0.2514, pruned_loss=0.04504, over 3269102.65 frames. ], batch size: 68, lr: 2.81e-03, grad_scale: 4.0 2023-05-01 19:30:02,066 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.28 vs. limit=2.0 2023-05-01 19:30:42,748 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.6521, 1.9073, 2.2767, 2.4963, 2.5779, 2.5524, 1.9496, 2.7577], device='cuda:1'), covar=tensor([0.0193, 0.0525, 0.0335, 0.0346, 0.0337, 0.0360, 0.0547, 0.0190], device='cuda:1'), in_proj_covar=tensor([0.0197, 0.0196, 0.0186, 0.0190, 0.0205, 0.0164, 0.0201, 0.0160], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-01 19:30:43,384 INFO [train.py:904] (1/8) Epoch 24, batch 3950, loss[loss=0.1787, simple_loss=0.2542, pruned_loss=0.05155, over 16726.00 frames. ], tot_loss[loss=0.1709, simple_loss=0.2507, pruned_loss=0.04549, over 3281026.76 frames. ], batch size: 134, lr: 2.81e-03, grad_scale: 4.0 2023-05-01 19:31:17,988 INFO [optim.py:368] (1/8) 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 19:31:50,783 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.7202, 3.7770, 2.4247, 4.0791, 2.9964, 4.0832, 2.4835, 3.0713], device='cuda:1'), covar=tensor([0.0279, 0.0381, 0.1557, 0.0328, 0.0696, 0.0626, 0.1438, 0.0655], device='cuda:1'), in_proj_covar=tensor([0.0174, 0.0181, 0.0197, 0.0174, 0.0179, 0.0223, 0.0205, 0.0184], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-05-01 19:31:57,069 INFO [train.py:904] (1/8) Epoch 24, batch 4000, loss[loss=0.1597, simple_loss=0.238, pruned_loss=0.04073, over 16863.00 frames. ], tot_loss[loss=0.1714, simple_loss=0.251, pruned_loss=0.0459, over 3282208.03 frames. ], batch size: 116, lr: 2.81e-03, grad_scale: 8.0 2023-05-01 19:32:02,749 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.53 vs. limit=2.0 2023-05-01 19:32:53,172 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 24, batch 4050, loss[loss=0.1519, simple_loss=0.2401, pruned_loss=0.03187, over 16481.00 frames. ], tot_loss[loss=0.1711, simple_loss=0.2517, pruned_loss=0.04521, over 3280201.70 frames. ], batch size: 75, lr: 2.81e-03, grad_scale: 8.0 2023-05-01 19:33:43,734 INFO [optim.py:368] (1/8) Clipping_scale=2.0, grad-norm quartiles 1.357e+02 1.828e+02 2.122e+02 2.508e+02 6.067e+02, threshold=4.243e+02, percent-clipped=1.0 2023-05-01 19:34:17,701 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=237548.0, num_to_drop=1, layers_to_drop={1} 2023-05-01 19:34:23,390 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=237552.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 19:34:24,475 INFO [train.py:904] (1/8) Epoch 24, batch 4100, loss[loss=0.1734, simple_loss=0.2578, pruned_loss=0.04443, over 17206.00 frames. ], tot_loss[loss=0.1712, simple_loss=0.2534, pruned_loss=0.04455, over 3272017.98 frames. ], batch size: 46, lr: 2.81e-03, grad_scale: 8.0 2023-05-01 19:34:38,873 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=237563.0, num_to_drop=1, layers_to_drop={0} 2023-05-01 19:35:07,370 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.5535, 3.5745, 2.6745, 2.3346, 2.4391, 2.4279, 3.8606, 3.2295], device='cuda:1'), covar=tensor([0.3042, 0.0719, 0.1884, 0.2437, 0.2611, 0.2034, 0.0465, 0.1325], device='cuda:1'), in_proj_covar=tensor([0.0332, 0.0274, 0.0311, 0.0322, 0.0306, 0.0269, 0.0301, 0.0346], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:1') 2023-05-01 19:35:26,109 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.2788, 4.3538, 4.5025, 4.2743, 4.3645, 4.8548, 4.4310, 4.0803], device='cuda:1'), covar=tensor([0.1584, 0.1768, 0.1817, 0.1917, 0.2439, 0.0989, 0.1506, 0.2461], device='cuda:1'), in_proj_covar=tensor([0.0422, 0.0621, 0.0679, 0.0513, 0.0679, 0.0711, 0.0534, 0.0676], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-05-01 19:35:30,061 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=237596.0, num_to_drop=1, layers_to_drop={1} 2023-05-01 19:35:40,164 INFO [train.py:904] (1/8) Epoch 24, batch 4150, loss[loss=0.1929, simple_loss=0.2833, pruned_loss=0.05119, over 17031.00 frames. ], tot_loss[loss=0.1769, simple_loss=0.2604, pruned_loss=0.04667, over 3241009.34 frames. ], batch size: 53, lr: 2.81e-03, grad_scale: 8.0 2023-05-01 19:36:14,075 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=237624.0, num_to_drop=1, layers_to_drop={0} 2023-05-01 19:36:16,367 INFO [optim.py:368] (1/8) 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 19:36:56,392 INFO [train.py:904] (1/8) Epoch 24, batch 4200, loss[loss=0.2062, simple_loss=0.2936, pruned_loss=0.05945, over 17036.00 frames. ], tot_loss[loss=0.1816, simple_loss=0.2671, pruned_loss=0.04804, over 3226622.96 frames. ], batch size: 53, lr: 2.81e-03, grad_scale: 8.0 2023-05-01 19:36:59,492 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.8665, 2.0781, 2.4453, 3.0887, 2.2375, 2.2861, 2.3051, 2.2230], device='cuda:1'), covar=tensor([0.1430, 0.3655, 0.2329, 0.0714, 0.3722, 0.2427, 0.3266, 0.3418], device='cuda:1'), in_proj_covar=tensor([0.0411, 0.0462, 0.0377, 0.0333, 0.0440, 0.0531, 0.0432, 0.0540], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-01 19:38:10,756 INFO [train.py:904] (1/8) Epoch 24, batch 4250, loss[loss=0.1866, simple_loss=0.2815, pruned_loss=0.04584, over 16641.00 frames. ], tot_loss[loss=0.1829, simple_loss=0.2703, pruned_loss=0.04776, over 3212910.72 frames. ], batch size: 62, lr: 2.81e-03, grad_scale: 8.0 2023-05-01 19:38:16,760 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=4.21 vs. limit=5.0 2023-05-01 19:38:45,324 INFO [optim.py:368] (1/8) 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 19:38:48,873 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.7887, 3.9767, 2.9333, 2.4863, 2.7479, 2.7094, 4.2092, 3.4783], device='cuda:1'), covar=tensor([0.2796, 0.0649, 0.1857, 0.2706, 0.2664, 0.1993, 0.0490, 0.1292], device='cuda:1'), in_proj_covar=tensor([0.0330, 0.0272, 0.0309, 0.0320, 0.0304, 0.0267, 0.0299, 0.0342], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:1') 2023-05-01 19:39:16,351 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.65 vs. limit=2.0 2023-05-01 19:39:26,151 INFO [train.py:904] (1/8) Epoch 24, batch 4300, loss[loss=0.1969, simple_loss=0.2847, pruned_loss=0.05456, over 11919.00 frames. ], tot_loss[loss=0.1829, simple_loss=0.2717, pruned_loss=0.04703, over 3219891.21 frames. ], batch size: 246, lr: 2.81e-03, grad_scale: 8.0 2023-05-01 19:40:01,145 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.49 vs. limit=2.0 2023-05-01 19:40:23,403 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=237790.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 19:40:41,734 INFO [train.py:904] (1/8) Epoch 24, batch 4350, loss[loss=0.2039, simple_loss=0.2944, pruned_loss=0.05671, over 16737.00 frames. ], tot_loss[loss=0.1855, simple_loss=0.2747, pruned_loss=0.04817, over 3200497.93 frames. ], batch size: 124, lr: 2.81e-03, grad_scale: 8.0 2023-05-01 19:41:14,400 INFO [optim.py:368] (1/8) 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 19:41:21,576 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.49 vs. limit=2.0 2023-05-01 19:41:35,355 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([5.1473, 5.4050, 5.1638, 5.1616, 4.9473, 4.8331, 4.7615, 5.5212], device='cuda:1'), covar=tensor([0.1019, 0.0716, 0.1000, 0.0876, 0.0678, 0.0811, 0.1046, 0.0745], device='cuda:1'), in_proj_covar=tensor([0.0695, 0.0841, 0.0697, 0.0653, 0.0537, 0.0542, 0.0710, 0.0660], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:1') 2023-05-01 19:41:45,867 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=237847.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 19:41:52,210 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=237851.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 19:41:54,770 INFO [train.py:904] (1/8) Epoch 24, batch 4400, loss[loss=0.1689, simple_loss=0.2696, pruned_loss=0.0341, over 16904.00 frames. ], tot_loss[loss=0.1883, simple_loss=0.2774, pruned_loss=0.04959, over 3208349.40 frames. ], batch size: 96, lr: 2.81e-03, grad_scale: 8.0 2023-05-01 19:42:22,879 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.8502, 4.6958, 4.9239, 5.0523, 5.1650, 4.6443, 5.2118, 5.2154], device='cuda:1'), covar=tensor([0.1476, 0.1162, 0.1188, 0.0528, 0.0420, 0.0872, 0.0456, 0.0455], device='cuda:1'), in_proj_covar=tensor([0.0663, 0.0823, 0.0943, 0.0826, 0.0632, 0.0655, 0.0683, 0.0796], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-01 19:43:05,638 INFO [train.py:904] (1/8) Epoch 24, batch 4450, loss[loss=0.1791, simple_loss=0.2722, pruned_loss=0.04298, over 16707.00 frames. ], tot_loss[loss=0.1911, simple_loss=0.2806, pruned_loss=0.05076, over 3213325.29 frames. ], batch size: 89, lr: 2.81e-03, grad_scale: 8.0 2023-05-01 19:43:18,849 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=237912.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 19:43:28,553 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=237919.0, num_to_drop=1, layers_to_drop={1} 2023-05-01 19:43:38,900 INFO [optim.py:368] (1/8) 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 19:43:44,178 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.6071, 4.7520, 4.9201, 4.7064, 4.8303, 5.3486, 4.8088, 4.4865], device='cuda:1'), covar=tensor([0.1148, 0.1734, 0.1833, 0.1988, 0.2355, 0.0873, 0.1386, 0.2295], device='cuda:1'), in_proj_covar=tensor([0.0415, 0.0609, 0.0668, 0.0504, 0.0665, 0.0697, 0.0525, 0.0663], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-05-01 19:44:16,768 INFO [train.py:904] (1/8) Epoch 24, batch 4500, loss[loss=0.2224, simple_loss=0.3022, pruned_loss=0.0713, over 17200.00 frames. ], tot_loss[loss=0.1924, simple_loss=0.2813, pruned_loss=0.05172, over 3230186.41 frames. ], batch size: 45, lr: 2.81e-03, grad_scale: 8.0 2023-05-01 19:44:25,604 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.1220, 4.1032, 3.9901, 3.2061, 3.9709, 1.8493, 3.7727, 3.3546], device='cuda:1'), covar=tensor([0.0090, 0.0071, 0.0176, 0.0270, 0.0076, 0.3088, 0.0108, 0.0294], device='cuda:1'), in_proj_covar=tensor([0.0173, 0.0166, 0.0208, 0.0184, 0.0183, 0.0213, 0.0196, 0.0178], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-01 19:44:46,731 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=237973.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 19:45:32,218 INFO [train.py:904] (1/8) Epoch 24, batch 4550, loss[loss=0.2103, simple_loss=0.2965, pruned_loss=0.06204, over 16772.00 frames. ], tot_loss[loss=0.1939, simple_loss=0.2824, pruned_loss=0.05268, over 3235550.72 frames. ], batch size: 124, lr: 2.81e-03, grad_scale: 8.0 2023-05-01 19:45:54,310 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([1.8940, 2.8989, 2.7351, 4.6826, 3.6169, 4.1256, 1.7170, 3.1700], device='cuda:1'), covar=tensor([0.1265, 0.0751, 0.1143, 0.0111, 0.0300, 0.0351, 0.1629, 0.0740], device='cuda:1'), in_proj_covar=tensor([0.0169, 0.0178, 0.0197, 0.0195, 0.0207, 0.0217, 0.0205, 0.0196], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-05-01 19:46:04,569 INFO [optim.py:368] (1/8) 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 19:46:44,385 INFO [train.py:904] (1/8) Epoch 24, batch 4600, loss[loss=0.2031, simple_loss=0.2926, pruned_loss=0.05675, over 16485.00 frames. ], tot_loss[loss=0.1943, simple_loss=0.2832, pruned_loss=0.05274, over 3239727.21 frames. ], batch size: 68, lr: 2.81e-03, grad_scale: 8.0 2023-05-01 19:47:00,849 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=238064.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 19:47:15,140 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=238074.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 19:47:56,654 INFO [train.py:904] (1/8) Epoch 24, batch 4650, loss[loss=0.1984, simple_loss=0.2841, pruned_loss=0.05633, over 15356.00 frames. ], tot_loss[loss=0.1935, simple_loss=0.2822, pruned_loss=0.05242, over 3240439.19 frames. ], batch size: 191, lr: 2.81e-03, grad_scale: 8.0 2023-05-01 19:48:29,603 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=238125.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 19:48:30,377 INFO [optim.py:368] (1/8) 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 19:48:30,778 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=238126.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 19:48:43,685 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=238135.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 19:48:59,346 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=238146.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 19:49:00,519 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 24, batch 4700, loss[loss=0.1723, simple_loss=0.2622, pruned_loss=0.04121, over 16814.00 frames. ], tot_loss[loss=0.1914, simple_loss=0.2798, pruned_loss=0.05151, over 3229844.59 frames. ], batch size: 116, lr: 2.81e-03, grad_scale: 8.0 2023-05-01 19:49:42,075 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.3241, 2.6440, 2.9127, 1.8703, 2.6041, 1.9311, 2.9332, 2.9298], device='cuda:1'), covar=tensor([0.0325, 0.1048, 0.0705, 0.2291, 0.1066, 0.1184, 0.0771, 0.1100], device='cuda:1'), in_proj_covar=tensor([0.0158, 0.0168, 0.0169, 0.0155, 0.0147, 0.0131, 0.0144, 0.0180], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-05-01 19:49:58,158 INFO [zipformer.py:625] (1/8) 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] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=238195.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 19:50:20,671 INFO [train.py:904] (1/8) Epoch 24, batch 4750, loss[loss=0.1477, simple_loss=0.2436, pruned_loss=0.02588, over 16824.00 frames. ], tot_loss[loss=0.1868, simple_loss=0.2752, pruned_loss=0.04926, over 3223552.04 frames. ], batch size: 102, lr: 2.81e-03, grad_scale: 8.0 2023-05-01 19:50:43,679 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=238219.0, num_to_drop=1, layers_to_drop={2} 2023-05-01 19:50:53,741 INFO [optim.py:368] (1/8) 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 19:50:55,568 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([1.9758, 3.4170, 3.3885, 2.0818, 3.0517, 3.4036, 3.1209, 1.9107], device='cuda:1'), covar=tensor([0.0649, 0.0061, 0.0065, 0.0499, 0.0126, 0.0114, 0.0129, 0.0537], device='cuda:1'), in_proj_covar=tensor([0.0136, 0.0086, 0.0087, 0.0134, 0.0099, 0.0111, 0.0096, 0.0129], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:1') 2023-05-01 19:51:31,486 INFO [train.py:904] (1/8) Epoch 24, batch 4800, loss[loss=0.1751, simple_loss=0.2679, pruned_loss=0.04112, over 16667.00 frames. ], tot_loss[loss=0.1834, simple_loss=0.272, pruned_loss=0.04745, over 3212938.99 frames. ], batch size: 76, lr: 2.81e-03, grad_scale: 8.0 2023-05-01 19:51:52,764 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=238267.0, num_to_drop=1, layers_to_drop={1} 2023-05-01 19:51:54,087 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=238268.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 19:52:41,836 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.4708, 3.3872, 2.6106, 2.1343, 2.1863, 2.3079, 3.4634, 2.9678], device='cuda:1'), covar=tensor([0.3102, 0.0695, 0.1960, 0.3272, 0.2786, 0.2247, 0.0594, 0.1461], device='cuda:1'), in_proj_covar=tensor([0.0330, 0.0272, 0.0308, 0.0319, 0.0301, 0.0266, 0.0299, 0.0342], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:1') 2023-05-01 19:52:46,484 INFO [train.py:904] (1/8) Epoch 24, batch 4850, loss[loss=0.1907, simple_loss=0.2863, pruned_loss=0.04754, over 16257.00 frames. ], tot_loss[loss=0.183, simple_loss=0.2726, pruned_loss=0.04674, over 3200226.87 frames. ], batch size: 165, lr: 2.81e-03, grad_scale: 8.0 2023-05-01 19:53:22,733 INFO [optim.py:368] (1/8) 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] (1/8) Epoch 24, batch 4900, loss[loss=0.1632, simple_loss=0.2644, pruned_loss=0.03105, over 16754.00 frames. ], tot_loss[loss=0.1813, simple_loss=0.2718, pruned_loss=0.04541, over 3191062.43 frames. ], batch size: 83, lr: 2.81e-03, grad_scale: 8.0 2023-05-01 19:54:13,312 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.58 vs. limit=2.0 2023-05-01 19:54:42,082 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.1997, 2.3458, 2.3688, 4.0160, 2.2560, 2.6970, 2.4135, 2.4727], device='cuda:1'), covar=tensor([0.1416, 0.3466, 0.2950, 0.0521, 0.3893, 0.2448, 0.3747, 0.3104], device='cuda:1'), in_proj_covar=tensor([0.0411, 0.0461, 0.0377, 0.0331, 0.0440, 0.0529, 0.0431, 0.0538], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-01 19:55:17,228 INFO [train.py:904] (1/8) Epoch 24, batch 4950, loss[loss=0.1882, simple_loss=0.2855, pruned_loss=0.04547, over 17133.00 frames. ], tot_loss[loss=0.1809, simple_loss=0.2716, pruned_loss=0.04508, over 3184789.46 frames. ], batch size: 47, lr: 2.81e-03, grad_scale: 8.0 2023-05-01 19:55:27,488 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.0290, 2.3384, 1.8580, 2.1355, 2.6988, 2.3455, 2.5769, 2.8733], device='cuda:1'), covar=tensor([0.0180, 0.0531, 0.0711, 0.0557, 0.0306, 0.0484, 0.0256, 0.0332], device='cuda:1'), in_proj_covar=tensor([0.0222, 0.0239, 0.0230, 0.0231, 0.0241, 0.0240, 0.0241, 0.0238], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-01 19:55:41,590 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=238420.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 19:55:41,750 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=238420.0, num_to_drop=1, layers_to_drop={1} 2023-05-01 19:55:49,724 INFO [optim.py:368] (1/8) 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 19:55:56,821 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=238430.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 19:56:20,205 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=238446.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 19:56:29,597 INFO [train.py:904] (1/8) Epoch 24, batch 5000, loss[loss=0.1788, simple_loss=0.2668, pruned_loss=0.04537, over 17069.00 frames. ], tot_loss[loss=0.1812, simple_loss=0.2727, pruned_loss=0.04483, over 3197497.93 frames. ], batch size: 53, lr: 2.81e-03, grad_scale: 8.0 2023-05-01 19:57:11,139 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=238481.0, num_to_drop=1, layers_to_drop={3} 2023-05-01 19:57:11,995 INFO [zipformer.py:625] (1/8) 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] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=238494.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 19:57:42,730 INFO [train.py:904] (1/8) Epoch 24, batch 5050, loss[loss=0.1739, simple_loss=0.2686, pruned_loss=0.03965, over 16878.00 frames. ], tot_loss[loss=0.1815, simple_loss=0.2736, pruned_loss=0.04471, over 3202140.58 frames. ], batch size: 96, lr: 2.81e-03, grad_scale: 8.0 2023-05-01 19:58:15,454 INFO [optim.py:368] (1/8) 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 19:58:24,081 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([5.0024, 5.0733, 5.3329, 5.3088, 5.3551, 5.0432, 4.9832, 4.7567], device='cuda:1'), covar=tensor([0.0283, 0.0503, 0.0389, 0.0369, 0.0515, 0.0325, 0.0972, 0.0450], device='cuda:1'), in_proj_covar=tensor([0.0411, 0.0459, 0.0449, 0.0411, 0.0492, 0.0467, 0.0549, 0.0373], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:1') 2023-05-01 19:58:29,306 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([1.9940, 3.4232, 3.3863, 2.0762, 3.0110, 3.3858, 3.1051, 1.8985], device='cuda:1'), covar=tensor([0.0627, 0.0061, 0.0061, 0.0492, 0.0126, 0.0115, 0.0130, 0.0521], device='cuda:1'), in_proj_covar=tensor([0.0137, 0.0086, 0.0087, 0.0134, 0.0100, 0.0111, 0.0097, 0.0130], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:1') 2023-05-01 19:58:39,898 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.8727, 2.7293, 2.3337, 2.7218, 3.1989, 2.8303, 3.3920, 3.4653], device='cuda:1'), covar=tensor([0.0085, 0.0510, 0.0584, 0.0442, 0.0261, 0.0437, 0.0242, 0.0258], device='cuda:1'), in_proj_covar=tensor([0.0222, 0.0240, 0.0231, 0.0232, 0.0241, 0.0240, 0.0242, 0.0238], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-01 19:58:51,279 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.3646, 2.4395, 2.4155, 4.1312, 2.3096, 2.8442, 2.4562, 2.6486], device='cuda:1'), covar=tensor([0.1382, 0.3539, 0.2990, 0.0541, 0.3903, 0.2423, 0.3506, 0.2934], device='cuda:1'), in_proj_covar=tensor([0.0412, 0.0463, 0.0378, 0.0333, 0.0442, 0.0530, 0.0433, 0.0540], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-01 19:58:53,423 INFO [train.py:904] (1/8) Epoch 24, batch 5100, loss[loss=0.2108, simple_loss=0.295, pruned_loss=0.06326, over 12327.00 frames. ], tot_loss[loss=0.1814, simple_loss=0.2731, pruned_loss=0.04484, over 3202673.42 frames. ], batch size: 246, lr: 2.81e-03, grad_scale: 8.0 2023-05-01 19:59:04,892 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.0154, 3.0213, 1.8995, 3.2694, 2.3502, 3.2972, 2.2029, 2.5217], device='cuda:1'), covar=tensor([0.0318, 0.0435, 0.1669, 0.0193, 0.0862, 0.0517, 0.1406, 0.0779], device='cuda:1'), in_proj_covar=tensor([0.0172, 0.0178, 0.0193, 0.0166, 0.0176, 0.0217, 0.0202, 0.0181], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-05-01 19:59:16,695 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=238568.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 19:59:18,366 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.37 vs. limit=2.0 2023-05-01 20:00:07,249 INFO [train.py:904] (1/8) Epoch 24, batch 5150, loss[loss=0.1765, simple_loss=0.2678, pruned_loss=0.04257, over 12024.00 frames. ], tot_loss[loss=0.1804, simple_loss=0.2725, pruned_loss=0.04416, over 3195237.06 frames. ], batch size: 247, lr: 2.81e-03, grad_scale: 8.0 2023-05-01 20:00:26,122 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=238616.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 20:00:38,784 INFO [optim.py:368] (1/8) 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 20:00:57,305 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.83 vs. limit=2.0 2023-05-01 20:01:06,631 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([1.6706, 2.7213, 2.4281, 4.2760, 2.8554, 3.9292, 1.6143, 2.9032], device='cuda:1'), covar=tensor([0.1374, 0.0757, 0.1278, 0.0147, 0.0173, 0.0390, 0.1700, 0.0789], device='cuda:1'), in_proj_covar=tensor([0.0168, 0.0177, 0.0195, 0.0193, 0.0205, 0.0215, 0.0204, 0.0194], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-05-01 20:01:17,675 INFO [train.py:904] (1/8) Epoch 24, batch 5200, loss[loss=0.1849, simple_loss=0.2762, pruned_loss=0.0468, over 16411.00 frames. ], tot_loss[loss=0.1791, simple_loss=0.2707, pruned_loss=0.04372, over 3193864.93 frames. ], batch size: 146, lr: 2.81e-03, grad_scale: 8.0 2023-05-01 20:01:22,756 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.3813, 2.8431, 3.0030, 1.9968, 2.7168, 2.0421, 3.0617, 3.0928], device='cuda:1'), covar=tensor([0.0268, 0.0815, 0.0650, 0.1930, 0.0892, 0.1086, 0.0602, 0.0799], device='cuda:1'), in_proj_covar=tensor([0.0157, 0.0166, 0.0168, 0.0154, 0.0146, 0.0130, 0.0144, 0.0178], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:1') 2023-05-01 20:01:39,334 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.6085, 3.7005, 2.2506, 4.2024, 2.7318, 4.1220, 2.4824, 2.9230], device='cuda:1'), covar=tensor([0.0268, 0.0356, 0.1622, 0.0187, 0.0884, 0.0462, 0.1491, 0.0809], device='cuda:1'), in_proj_covar=tensor([0.0171, 0.0177, 0.0193, 0.0166, 0.0176, 0.0216, 0.0201, 0.0181], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-05-01 20:01:57,590 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=238681.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 20:02:28,268 INFO [train.py:904] (1/8) Epoch 24, batch 5250, loss[loss=0.1775, simple_loss=0.2753, pruned_loss=0.03987, over 15283.00 frames. ], tot_loss[loss=0.1769, simple_loss=0.2678, pruned_loss=0.043, over 3208464.35 frames. ], batch size: 190, lr: 2.81e-03, grad_scale: 8.0 2023-05-01 20:02:51,462 INFO [zipformer.py:625] (1/8) 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] (1/8) 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 20:03:07,613 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=238730.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 20:03:25,268 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=238742.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 20:03:33,768 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=4.70 vs. limit=5.0 2023-05-01 20:03:39,625 INFO [train.py:904] (1/8) Epoch 24, batch 5300, loss[loss=0.1636, simple_loss=0.249, pruned_loss=0.03913, over 16858.00 frames. ], tot_loss[loss=0.1741, simple_loss=0.2643, pruned_loss=0.04193, over 3211404.88 frames. ], batch size: 116, lr: 2.81e-03, grad_scale: 8.0 2023-05-01 20:04:00,406 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=238768.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 20:04:12,326 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=238776.0, num_to_drop=1, layers_to_drop={3} 2023-05-01 20:04:15,071 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=238778.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 20:04:20,393 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=238782.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 20:04:33,776 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=3.07 vs. limit=5.0 2023-05-01 20:04:50,564 INFO [train.py:904] (1/8) Epoch 24, batch 5350, loss[loss=0.1915, simple_loss=0.2822, pruned_loss=0.05037, over 16577.00 frames. ], tot_loss[loss=0.1727, simple_loss=0.2625, pruned_loss=0.04149, over 3210734.49 frames. ], batch size: 75, lr: 2.81e-03, grad_scale: 8.0 2023-05-01 20:04:55,741 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=4.82 vs. limit=5.0 2023-05-01 20:05:21,926 INFO [optim.py:368] (1/8) 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] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=238830.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 20:05:48,262 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([1.9956, 2.2689, 2.3217, 2.6795, 1.6020, 3.1625, 1.7987, 2.6239], device='cuda:1'), covar=tensor([0.1199, 0.0761, 0.1159, 0.0179, 0.0109, 0.0388, 0.1551, 0.0790], device='cuda:1'), in_proj_covar=tensor([0.0169, 0.0178, 0.0196, 0.0194, 0.0206, 0.0216, 0.0205, 0.0195], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-05-01 20:06:01,811 INFO [train.py:904] (1/8) Epoch 24, batch 5400, loss[loss=0.1744, simple_loss=0.2708, pruned_loss=0.03894, over 16294.00 frames. ], tot_loss[loss=0.1751, simple_loss=0.2655, pruned_loss=0.04235, over 3203085.15 frames. ], batch size: 165, lr: 2.81e-03, grad_scale: 8.0 2023-05-01 20:06:14,978 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.7538, 3.1111, 3.2700, 1.9761, 2.8134, 2.2407, 3.3115, 3.3376], device='cuda:1'), covar=tensor([0.0280, 0.0739, 0.0614, 0.1930, 0.0840, 0.0951, 0.0588, 0.0828], device='cuda:1'), in_proj_covar=tensor([0.0157, 0.0166, 0.0168, 0.0154, 0.0146, 0.0130, 0.0144, 0.0178], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:1') 2023-05-01 20:07:13,178 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.4056, 3.2950, 3.7657, 1.8138, 3.9077, 3.9165, 2.9851, 2.8474], device='cuda:1'), covar=tensor([0.0846, 0.0295, 0.0195, 0.1286, 0.0078, 0.0155, 0.0424, 0.0502], device='cuda:1'), in_proj_covar=tensor([0.0148, 0.0109, 0.0100, 0.0139, 0.0082, 0.0126, 0.0129, 0.0131], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0004, 0.0004], device='cuda:1') 2023-05-01 20:07:18,024 INFO [train.py:904] (1/8) Epoch 24, batch 5450, loss[loss=0.2262, simple_loss=0.322, pruned_loss=0.06521, over 16949.00 frames. ], tot_loss[loss=0.1784, simple_loss=0.2688, pruned_loss=0.04397, over 3199659.06 frames. ], batch size: 109, lr: 2.81e-03, grad_scale: 8.0 2023-05-01 20:07:54,283 INFO [optim.py:368] (1/8) 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 20:07:56,195 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.5282, 3.5950, 2.7218, 2.2159, 2.4068, 2.3214, 3.8367, 3.2878], device='cuda:1'), covar=tensor([0.2887, 0.0587, 0.1782, 0.2684, 0.2445, 0.2172, 0.0420, 0.1189], device='cuda:1'), in_proj_covar=tensor([0.0330, 0.0272, 0.0307, 0.0319, 0.0301, 0.0265, 0.0299, 0.0341], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:1') 2023-05-01 20:08:11,004 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=238936.0, num_to_drop=1, layers_to_drop={1} 2023-05-01 20:08:35,279 INFO [train.py:904] (1/8) Epoch 24, batch 5500, loss[loss=0.1964, simple_loss=0.2879, pruned_loss=0.05251, over 16499.00 frames. ], tot_loss[loss=0.1848, simple_loss=0.2748, pruned_loss=0.04745, over 3174897.20 frames. ], batch size: 75, lr: 2.81e-03, grad_scale: 8.0 2023-05-01 20:09:43,457 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 24, batch 5550, loss[loss=0.2236, simple_loss=0.3077, pruned_loss=0.06976, over 16682.00 frames. ], tot_loss[loss=0.1944, simple_loss=0.2824, pruned_loss=0.05317, over 3141109.14 frames. ], batch size: 134, lr: 2.80e-03, grad_scale: 8.0 2023-05-01 20:10:30,505 INFO [optim.py:368] (1/8) 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] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=239037.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 20:11:12,507 INFO [train.py:904] (1/8) Epoch 24, batch 5600, loss[loss=0.2082, simple_loss=0.2902, pruned_loss=0.06312, over 16451.00 frames. ], tot_loss[loss=0.2018, simple_loss=0.2877, pruned_loss=0.05796, over 3075526.56 frames. ], batch size: 146, lr: 2.80e-03, grad_scale: 8.0 2023-05-01 20:11:13,240 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.0330, 4.5872, 4.5582, 3.0115, 3.9159, 4.5331, 3.8278, 2.6814], device='cuda:1'), covar=tensor([0.0464, 0.0039, 0.0040, 0.0403, 0.0118, 0.0112, 0.0101, 0.0411], device='cuda:1'), in_proj_covar=tensor([0.0138, 0.0088, 0.0087, 0.0136, 0.0101, 0.0112, 0.0097, 0.0131], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0003, 0.0004], device='cuda:1') 2023-05-01 20:11:52,355 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=239076.0, num_to_drop=1, layers_to_drop={2} 2023-05-01 20:12:12,228 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([1.6782, 2.7287, 2.4393, 4.4072, 3.1866, 3.9838, 1.6044, 2.8523], device='cuda:1'), covar=tensor([0.1442, 0.0817, 0.1358, 0.0181, 0.0295, 0.0463, 0.1753, 0.0902], device='cuda:1'), in_proj_covar=tensor([0.0168, 0.0178, 0.0196, 0.0195, 0.0206, 0.0217, 0.0205, 0.0195], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-05-01 20:12:36,912 INFO [train.py:904] (1/8) Epoch 24, batch 5650, loss[loss=0.2505, simple_loss=0.3195, pruned_loss=0.09074, over 16474.00 frames. ], tot_loss[loss=0.2084, simple_loss=0.2931, pruned_loss=0.06183, over 3051889.03 frames. ], batch size: 146, lr: 2.80e-03, grad_scale: 8.0 2023-05-01 20:12:54,577 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=4.19 vs. limit=5.0 2023-05-01 20:13:11,110 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=239124.0, num_to_drop=1, layers_to_drop={0} 2023-05-01 20:13:14,764 INFO [optim.py:368] (1/8) 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 20:13:46,699 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([5.0142, 5.6052, 5.7560, 5.4697, 5.5116, 6.0737, 5.5148, 5.2629], device='cuda:1'), covar=tensor([0.0881, 0.1718, 0.2169, 0.1749, 0.2213, 0.0881, 0.1537, 0.2360], device='cuda:1'), in_proj_covar=tensor([0.0411, 0.0602, 0.0661, 0.0496, 0.0660, 0.0692, 0.0519, 0.0659], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-05-01 20:13:55,697 INFO [train.py:904] (1/8) Epoch 24, batch 5700, loss[loss=0.1838, simple_loss=0.2806, pruned_loss=0.04347, over 16871.00 frames. ], tot_loss[loss=0.2102, simple_loss=0.2943, pruned_loss=0.06306, over 3055189.96 frames. ], batch size: 96, lr: 2.80e-03, grad_scale: 8.0 2023-05-01 20:15:14,247 INFO [train.py:904] (1/8) Epoch 24, batch 5750, loss[loss=0.1889, simple_loss=0.288, pruned_loss=0.04487, over 16675.00 frames. ], tot_loss[loss=0.2126, simple_loss=0.2969, pruned_loss=0.06412, over 3040547.45 frames. ], batch size: 89, lr: 2.80e-03, grad_scale: 8.0 2023-05-01 20:15:53,491 INFO [optim.py:368] (1/8) 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] (1/8) Epoch 24, batch 5800, loss[loss=0.1907, simple_loss=0.2791, pruned_loss=0.05113, over 16921.00 frames. ], tot_loss[loss=0.2119, simple_loss=0.2968, pruned_loss=0.06347, over 3020550.56 frames. ], batch size: 109, lr: 2.80e-03, grad_scale: 8.0 2023-05-01 20:17:40,218 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=239292.0, num_to_drop=1, layers_to_drop={0} 2023-05-01 20:17:56,784 INFO [train.py:904] (1/8) Epoch 24, batch 5850, loss[loss=0.2145, simple_loss=0.2974, pruned_loss=0.06577, over 16165.00 frames. ], tot_loss[loss=0.2093, simple_loss=0.2946, pruned_loss=0.06201, over 3023071.62 frames. ], batch size: 165, lr: 2.80e-03, grad_scale: 8.0 2023-05-01 20:18:33,898 INFO [optim.py:368] (1/8) 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 20:18:53,013 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=239337.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 20:19:19,097 INFO [train.py:904] (1/8) Epoch 24, batch 5900, loss[loss=0.1883, simple_loss=0.2787, pruned_loss=0.04899, over 17027.00 frames. ], tot_loss[loss=0.2084, simple_loss=0.2939, pruned_loss=0.06148, over 3028399.70 frames. ], batch size: 55, lr: 2.80e-03, grad_scale: 8.0 2023-05-01 20:19:20,044 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.2837, 3.3294, 2.0063, 3.5726, 2.5626, 3.6276, 2.2206, 2.7399], device='cuda:1'), covar=tensor([0.0303, 0.0403, 0.1868, 0.0323, 0.0850, 0.0628, 0.1645, 0.0792], device='cuda:1'), in_proj_covar=tensor([0.0173, 0.0180, 0.0196, 0.0169, 0.0179, 0.0219, 0.0205, 0.0183], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-05-01 20:19:29,219 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([1.9651, 3.1526, 3.2092, 2.0688, 3.0104, 3.1898, 3.0018, 2.0446], device='cuda:1'), covar=tensor([0.0575, 0.0081, 0.0074, 0.0472, 0.0118, 0.0129, 0.0127, 0.0464], device='cuda:1'), in_proj_covar=tensor([0.0136, 0.0087, 0.0086, 0.0134, 0.0099, 0.0111, 0.0096, 0.0130], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:1') 2023-05-01 20:19:33,205 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.8549, 3.7770, 3.8978, 4.0006, 4.0955, 3.6991, 4.0384, 4.1101], device='cuda:1'), covar=tensor([0.1572, 0.1122, 0.1394, 0.0750, 0.0623, 0.2035, 0.0960, 0.0793], device='cuda:1'), in_proj_covar=tensor([0.0641, 0.0790, 0.0910, 0.0800, 0.0608, 0.0630, 0.0660, 0.0771], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-01 20:20:14,707 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=239385.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 20:20:42,779 INFO [train.py:904] (1/8) Epoch 24, batch 5950, loss[loss=0.1817, simple_loss=0.2822, pruned_loss=0.04058, over 16824.00 frames. ], tot_loss[loss=0.2071, simple_loss=0.2945, pruned_loss=0.05991, over 3044119.19 frames. ], batch size: 102, lr: 2.80e-03, grad_scale: 8.0 2023-05-01 20:21:02,766 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.2768, 3.3192, 1.9628, 3.6532, 2.5307, 3.6899, 2.1547, 2.6952], device='cuda:1'), covar=tensor([0.0326, 0.0430, 0.1741, 0.0257, 0.0865, 0.0610, 0.1633, 0.0809], device='cuda:1'), in_proj_covar=tensor([0.0173, 0.0179, 0.0195, 0.0168, 0.0178, 0.0218, 0.0204, 0.0182], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-05-01 20:21:12,962 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.2871, 3.7102, 3.7588, 2.4078, 3.3979, 3.7649, 3.3835, 2.1234], device='cuda:1'), covar=tensor([0.0563, 0.0067, 0.0059, 0.0443, 0.0132, 0.0118, 0.0120, 0.0483], device='cuda:1'), in_proj_covar=tensor([0.0136, 0.0087, 0.0086, 0.0134, 0.0099, 0.0111, 0.0097, 0.0130], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0003, 0.0004], device='cuda:1') 2023-05-01 20:21:21,253 INFO [optim.py:368] (1/8) Clipping_scale=2.0, grad-norm quartiles 1.785e+02 2.477e+02 3.014e+02 3.724e+02 7.833e+02, threshold=6.029e+02, percent-clipped=2.0 2023-05-01 20:22:03,333 INFO [train.py:904] (1/8) Epoch 24, batch 6000, loss[loss=0.1747, simple_loss=0.2649, pruned_loss=0.04229, over 16505.00 frames. ], tot_loss[loss=0.206, simple_loss=0.2933, pruned_loss=0.05936, over 3062151.36 frames. ], batch size: 75, lr: 2.80e-03, grad_scale: 8.0 2023-05-01 20:22:03,333 INFO [train.py:929] (1/8) Computing validation loss 2023-05-01 20:22:14,268 INFO [train.py:938] (1/8) Epoch 24, validation: loss=0.1493, simple_loss=0.2618, pruned_loss=0.01837, over 944034.00 frames. 2023-05-01 20:22:14,269 INFO [train.py:939] (1/8) Maximum memory allocated so far is 18145MB 2023-05-01 20:23:21,313 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.8007, 1.9335, 2.2170, 3.2578, 1.9192, 2.1691, 2.0856, 2.0583], device='cuda:1'), covar=tensor([0.1924, 0.4615, 0.3194, 0.0840, 0.5397, 0.3401, 0.4311, 0.4415], device='cuda:1'), in_proj_covar=tensor([0.0406, 0.0455, 0.0372, 0.0328, 0.0435, 0.0521, 0.0426, 0.0531], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-01 20:23:32,267 INFO [train.py:904] (1/8) Epoch 24, batch 6050, loss[loss=0.2328, simple_loss=0.3008, pruned_loss=0.08241, over 11624.00 frames. ], tot_loss[loss=0.2042, simple_loss=0.2916, pruned_loss=0.05842, over 3083948.84 frames. ], batch size: 247, lr: 2.80e-03, grad_scale: 8.0 2023-05-01 20:24:09,878 INFO [optim.py:368] (1/8) Clipping_scale=2.0, grad-norm quartiles 1.866e+02 2.713e+02 3.363e+02 4.065e+02 7.660e+02, threshold=6.725e+02, percent-clipped=3.0 2023-05-01 20:24:51,434 INFO [train.py:904] (1/8) Epoch 24, batch 6100, loss[loss=0.191, simple_loss=0.2795, pruned_loss=0.05122, over 16776.00 frames. ], tot_loss[loss=0.2025, simple_loss=0.2909, pruned_loss=0.05703, over 3112938.46 frames. ], batch size: 124, lr: 2.80e-03, grad_scale: 8.0 2023-05-01 20:25:56,192 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=239592.0, num_to_drop=1, layers_to_drop={0} 2023-05-01 20:26:08,454 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.4871, 4.5434, 4.3662, 4.0750, 4.0966, 4.4615, 4.2029, 4.1834], device='cuda:1'), covar=tensor([0.0578, 0.0523, 0.0278, 0.0298, 0.0771, 0.0527, 0.0596, 0.0621], device='cuda:1'), in_proj_covar=tensor([0.0295, 0.0442, 0.0346, 0.0346, 0.0348, 0.0401, 0.0237, 0.0416], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-01 20:26:14,091 INFO [train.py:904] (1/8) Epoch 24, batch 6150, loss[loss=0.2118, simple_loss=0.2915, pruned_loss=0.06603, over 15344.00 frames. ], tot_loss[loss=0.2007, simple_loss=0.2884, pruned_loss=0.05644, over 3107824.26 frames. ], batch size: 190, lr: 2.80e-03, grad_scale: 8.0 2023-05-01 20:26:22,112 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.71 vs. limit=2.0 2023-05-01 20:26:45,380 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=239622.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 20:26:53,236 INFO [optim.py:368] (1/8) 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 20:27:14,513 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=239640.0, num_to_drop=1, layers_to_drop={1} 2023-05-01 20:27:24,052 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.82 vs. limit=2.0 2023-05-01 20:27:28,277 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=2.91 vs. limit=5.0 2023-05-01 20:27:34,956 INFO [train.py:904] (1/8) Epoch 24, batch 6200, loss[loss=0.2186, simple_loss=0.2844, pruned_loss=0.07637, over 11648.00 frames. ], tot_loss[loss=0.1988, simple_loss=0.2862, pruned_loss=0.05566, over 3111805.18 frames. ], batch size: 248, lr: 2.80e-03, grad_scale: 4.0 2023-05-01 20:28:22,521 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 24, batch 6250, loss[loss=0.1699, simple_loss=0.2618, pruned_loss=0.03906, over 17049.00 frames. ], tot_loss[loss=0.1981, simple_loss=0.2855, pruned_loss=0.0554, over 3110654.37 frames. ], batch size: 50, lr: 2.80e-03, grad_scale: 4.0 2023-05-01 20:29:14,005 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.9052, 4.2832, 3.2538, 2.4654, 2.7732, 2.6705, 4.6929, 3.5883], device='cuda:1'), covar=tensor([0.2917, 0.0561, 0.1738, 0.2988, 0.2801, 0.2014, 0.0439, 0.1354], device='cuda:1'), in_proj_covar=tensor([0.0331, 0.0273, 0.0309, 0.0321, 0.0302, 0.0267, 0.0300, 0.0343], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:1') 2023-05-01 20:29:29,230 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([5.0739, 5.1647, 5.5267, 5.4847, 5.5335, 5.2071, 5.1022, 4.8505], device='cuda:1'), covar=tensor([0.0366, 0.0565, 0.0471, 0.0470, 0.0474, 0.0422, 0.1070, 0.0531], device='cuda:1'), in_proj_covar=tensor([0.0422, 0.0471, 0.0459, 0.0421, 0.0505, 0.0478, 0.0562, 0.0382], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:1') 2023-05-01 20:29:29,985 INFO [optim.py:368] (1/8) 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 20:29:43,101 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.5492, 4.3830, 4.5653, 4.7205, 4.8889, 4.4174, 4.8849, 4.9037], device='cuda:1'), covar=tensor([0.1989, 0.1348, 0.1742, 0.0807, 0.0681, 0.1086, 0.0819, 0.0783], device='cuda:1'), in_proj_covar=tensor([0.0644, 0.0791, 0.0912, 0.0802, 0.0612, 0.0633, 0.0664, 0.0772], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-01 20:30:05,921 INFO [train.py:904] (1/8) Epoch 24, batch 6300, loss[loss=0.2044, simple_loss=0.2871, pruned_loss=0.06082, over 16909.00 frames. ], tot_loss[loss=0.1984, simple_loss=0.2856, pruned_loss=0.0556, over 3087869.41 frames. ], batch size: 109, lr: 2.80e-03, grad_scale: 4.0 2023-05-01 20:31:24,178 INFO [train.py:904] (1/8) Epoch 24, batch 6350, loss[loss=0.1702, simple_loss=0.2577, pruned_loss=0.04133, over 16758.00 frames. ], tot_loss[loss=0.2002, simple_loss=0.287, pruned_loss=0.05673, over 3074827.38 frames. ], batch size: 83, lr: 2.80e-03, grad_scale: 4.0 2023-05-01 20:31:30,329 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.5320, 3.5879, 3.3747, 3.0402, 3.2282, 3.5111, 3.3101, 3.3393], device='cuda:1'), covar=tensor([0.0595, 0.0686, 0.0314, 0.0307, 0.0523, 0.0516, 0.1413, 0.0494], device='cuda:1'), in_proj_covar=tensor([0.0292, 0.0437, 0.0343, 0.0343, 0.0346, 0.0397, 0.0235, 0.0412], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-01 20:32:03,927 INFO [optim.py:368] (1/8) 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 20:32:42,129 INFO [train.py:904] (1/8) Epoch 24, batch 6400, loss[loss=0.2145, simple_loss=0.3123, pruned_loss=0.05839, over 16686.00 frames. ], tot_loss[loss=0.2019, simple_loss=0.2878, pruned_loss=0.05795, over 3082035.08 frames. ], batch size: 134, lr: 2.80e-03, grad_scale: 8.0 2023-05-01 20:32:59,859 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.7788, 4.6894, 4.4935, 3.0315, 3.9931, 4.5640, 3.9393, 2.5348], device='cuda:1'), covar=tensor([0.0534, 0.0036, 0.0050, 0.0412, 0.0105, 0.0096, 0.0098, 0.0466], device='cuda:1'), in_proj_covar=tensor([0.0137, 0.0087, 0.0087, 0.0135, 0.0100, 0.0112, 0.0097, 0.0131], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0003, 0.0004], device='cuda:1') 2023-05-01 20:33:56,462 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.15 vs. limit=2.0 2023-05-01 20:33:58,179 INFO [train.py:904] (1/8) Epoch 24, batch 6450, loss[loss=0.1828, simple_loss=0.2797, pruned_loss=0.04297, over 16822.00 frames. ], tot_loss[loss=0.2006, simple_loss=0.2871, pruned_loss=0.05706, over 3084096.40 frames. ], batch size: 83, lr: 2.80e-03, grad_scale: 8.0 2023-05-01 20:34:07,267 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.77 vs. limit=2.0 2023-05-01 20:34:37,420 INFO [optim.py:368] (1/8) 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] (1/8) Epoch 24, batch 6500, loss[loss=0.2246, simple_loss=0.2895, pruned_loss=0.07985, over 11672.00 frames. ], tot_loss[loss=0.1992, simple_loss=0.285, pruned_loss=0.05675, over 3069532.02 frames. ], batch size: 247, lr: 2.80e-03, grad_scale: 8.0 2023-05-01 20:35:55,169 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=239978.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 20:36:39,172 INFO [train.py:904] (1/8) Epoch 24, batch 6550, loss[loss=0.2005, simple_loss=0.3006, pruned_loss=0.05016, over 15249.00 frames. ], tot_loss[loss=0.2015, simple_loss=0.2878, pruned_loss=0.05761, over 3077694.78 frames. ], batch size: 190, lr: 2.80e-03, grad_scale: 8.0 2023-05-01 20:37:16,973 INFO [optim.py:368] (1/8) 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 20:37:36,417 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([5.5805, 5.5853, 5.3593, 4.6847, 5.5071, 2.1166, 5.2195, 4.9597], device='cuda:1'), covar=tensor([0.0054, 0.0049, 0.0176, 0.0376, 0.0069, 0.2642, 0.0092, 0.0236], device='cuda:1'), in_proj_covar=tensor([0.0172, 0.0165, 0.0206, 0.0182, 0.0181, 0.0211, 0.0193, 0.0175], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-01 20:37:54,307 INFO [train.py:904] (1/8) Epoch 24, batch 6600, loss[loss=0.2074, simple_loss=0.2999, pruned_loss=0.05742, over 16722.00 frames. ], tot_loss[loss=0.2035, simple_loss=0.2902, pruned_loss=0.05839, over 3071217.44 frames. ], batch size: 83, lr: 2.80e-03, grad_scale: 8.0 2023-05-01 20:38:05,677 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.6509, 4.0260, 3.5739, 3.8535, 3.5406, 3.6215, 3.6249, 3.9845], device='cuda:1'), covar=tensor([0.3133, 0.1739, 0.3427, 0.1993, 0.2140, 0.3259, 0.2560, 0.2177], device='cuda:1'), in_proj_covar=tensor([0.0693, 0.0831, 0.0690, 0.0646, 0.0532, 0.0534, 0.0702, 0.0654], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:1') 2023-05-01 20:38:27,291 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=3.12 vs. limit=5.0 2023-05-01 20:39:11,703 INFO [train.py:904] (1/8) Epoch 24, batch 6650, loss[loss=0.2086, simple_loss=0.2909, pruned_loss=0.06317, over 16737.00 frames. ], tot_loss[loss=0.2045, simple_loss=0.2906, pruned_loss=0.05917, over 3080551.62 frames. ], batch size: 39, lr: 2.80e-03, grad_scale: 8.0 2023-05-01 20:39:50,364 INFO [optim.py:368] (1/8) 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] (1/8) Epoch 24, batch 6700, loss[loss=0.1968, simple_loss=0.2834, pruned_loss=0.0551, over 16906.00 frames. ], tot_loss[loss=0.2035, simple_loss=0.289, pruned_loss=0.05904, over 3080779.98 frames. ], batch size: 109, lr: 2.80e-03, grad_scale: 8.0 2023-05-01 20:41:04,932 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.5399, 2.9710, 3.1192, 2.0362, 2.7768, 2.0620, 3.1453, 3.2114], device='cuda:1'), covar=tensor([0.0284, 0.0831, 0.0665, 0.2129, 0.0894, 0.1087, 0.0697, 0.0984], device='cuda:1'), in_proj_covar=tensor([0.0157, 0.0166, 0.0170, 0.0155, 0.0147, 0.0132, 0.0144, 0.0179], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-05-01 20:41:37,022 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.7264, 3.5273, 4.2120, 2.0289, 4.3928, 4.3546, 3.0833, 3.2265], device='cuda:1'), covar=tensor([0.0833, 0.0294, 0.0165, 0.1278, 0.0058, 0.0145, 0.0453, 0.0494], device='cuda:1'), in_proj_covar=tensor([0.0148, 0.0109, 0.0100, 0.0138, 0.0082, 0.0128, 0.0129, 0.0131], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0004, 0.0004], device='cuda:1') 2023-05-01 20:41:45,726 INFO [train.py:904] (1/8) Epoch 24, batch 6750, loss[loss=0.1757, simple_loss=0.2686, pruned_loss=0.04144, over 16817.00 frames. ], tot_loss[loss=0.2022, simple_loss=0.2876, pruned_loss=0.05841, over 3109055.07 frames. ], batch size: 102, lr: 2.80e-03, grad_scale: 8.0 2023-05-01 20:42:23,537 INFO [optim.py:368] (1/8) Clipping_scale=2.0, grad-norm quartiles 1.933e+02 2.874e+02 3.326e+02 4.186e+02 6.587e+02, threshold=6.652e+02, percent-clipped=0.0 2023-05-01 20:43:01,407 INFO [train.py:904] (1/8) Epoch 24, batch 6800, loss[loss=0.2533, simple_loss=0.3243, pruned_loss=0.09112, over 11597.00 frames. ], tot_loss[loss=0.2025, simple_loss=0.2882, pruned_loss=0.05845, over 3103524.01 frames. ], batch size: 248, lr: 2.80e-03, grad_scale: 8.0 2023-05-01 20:43:42,277 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 24, batch 6850, loss[loss=0.2083, simple_loss=0.3079, pruned_loss=0.05437, over 16433.00 frames. ], tot_loss[loss=0.2045, simple_loss=0.29, pruned_loss=0.05947, over 3090600.64 frames. ], batch size: 35, lr: 2.80e-03, grad_scale: 4.0 2023-05-01 20:44:56,068 INFO [zipformer.py:625] (1/8) 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] (1/8) 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 20:45:17,552 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=240340.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 20:45:35,453 INFO [train.py:904] (1/8) Epoch 24, batch 6900, loss[loss=0.2688, simple_loss=0.3325, pruned_loss=0.1026, over 11283.00 frames. ], tot_loss[loss=0.2047, simple_loss=0.2918, pruned_loss=0.05881, over 3092334.72 frames. ], batch size: 247, lr: 2.80e-03, grad_scale: 4.0 2023-05-01 20:46:50,843 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 24, batch 6950, loss[loss=0.2051, simple_loss=0.2901, pruned_loss=0.06009, over 15329.00 frames. ], tot_loss[loss=0.2074, simple_loss=0.2934, pruned_loss=0.06069, over 3077711.45 frames. ], batch size: 191, lr: 2.80e-03, grad_scale: 2.0 2023-05-01 20:47:14,499 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([1.7756, 1.4456, 1.7427, 1.6848, 1.7812, 1.8808, 1.6714, 1.8015], device='cuda:1'), covar=tensor([0.0246, 0.0353, 0.0204, 0.0273, 0.0256, 0.0180, 0.0401, 0.0131], device='cuda:1'), in_proj_covar=tensor([0.0192, 0.0194, 0.0181, 0.0184, 0.0200, 0.0160, 0.0198, 0.0157], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-01 20:47:33,340 INFO [optim.py:368] (1/8) 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 20:47:40,788 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.19 vs. limit=2.0 2023-05-01 20:48:07,693 INFO [train.py:904] (1/8) Epoch 24, batch 7000, loss[loss=0.1979, simple_loss=0.283, pruned_loss=0.05639, over 16985.00 frames. ], tot_loss[loss=0.207, simple_loss=0.294, pruned_loss=0.05996, over 3077082.59 frames. ], batch size: 41, lr: 2.80e-03, grad_scale: 2.0 2023-05-01 20:48:33,619 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.33 vs. limit=2.0 2023-05-01 20:48:47,723 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.8702, 3.9219, 2.4333, 4.6992, 3.0777, 4.5264, 2.4884, 3.1277], device='cuda:1'), covar=tensor([0.0290, 0.0399, 0.1832, 0.0245, 0.0869, 0.0632, 0.1711, 0.0952], device='cuda:1'), in_proj_covar=tensor([0.0172, 0.0178, 0.0195, 0.0167, 0.0178, 0.0218, 0.0203, 0.0182], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-05-01 20:49:23,902 INFO [train.py:904] (1/8) Epoch 24, batch 7050, loss[loss=0.213, simple_loss=0.2869, pruned_loss=0.06955, over 11444.00 frames. ], tot_loss[loss=0.2071, simple_loss=0.2944, pruned_loss=0.05993, over 3056885.80 frames. ], batch size: 246, lr: 2.80e-03, grad_scale: 2.0 2023-05-01 20:50:06,655 INFO [optim.py:368] (1/8) 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 20:50:08,290 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.75 vs. limit=2.0 2023-05-01 20:50:15,820 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.3560, 3.2661, 3.2225, 3.4446, 3.4405, 3.2834, 3.4202, 3.4590], device='cuda:1'), covar=tensor([0.1267, 0.1188, 0.1432, 0.0766, 0.0933, 0.2801, 0.1383, 0.1252], device='cuda:1'), in_proj_covar=tensor([0.0635, 0.0784, 0.0902, 0.0789, 0.0606, 0.0626, 0.0658, 0.0763], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-01 20:50:42,273 INFO [train.py:904] (1/8) Epoch 24, batch 7100, loss[loss=0.1904, simple_loss=0.2859, pruned_loss=0.0474, over 16866.00 frames. ], tot_loss[loss=0.2058, simple_loss=0.2927, pruned_loss=0.0594, over 3066730.48 frames. ], batch size: 102, lr: 2.80e-03, grad_scale: 2.0 2023-05-01 20:50:52,495 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.20 vs. limit=2.0 2023-05-01 20:50:56,365 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.8166, 4.0211, 3.0796, 2.4500, 2.7433, 2.6947, 4.4222, 3.4908], device='cuda:1'), covar=tensor([0.2884, 0.0645, 0.1812, 0.2589, 0.2563, 0.1958, 0.0404, 0.1359], device='cuda:1'), in_proj_covar=tensor([0.0330, 0.0272, 0.0307, 0.0319, 0.0300, 0.0265, 0.0299, 0.0341], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:1') 2023-05-01 20:51:59,035 INFO [train.py:904] (1/8) Epoch 24, batch 7150, loss[loss=0.2536, simple_loss=0.3151, pruned_loss=0.09599, over 11689.00 frames. ], tot_loss[loss=0.2042, simple_loss=0.2908, pruned_loss=0.05881, over 3083486.34 frames. ], batch size: 247, lr: 2.80e-03, grad_scale: 2.0 2023-05-01 20:51:59,550 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.3956, 4.4815, 4.3118, 4.0135, 4.0089, 4.4018, 4.1124, 4.1453], device='cuda:1'), covar=tensor([0.0612, 0.0520, 0.0286, 0.0311, 0.0822, 0.0471, 0.0656, 0.0597], device='cuda:1'), in_proj_covar=tensor([0.0290, 0.0436, 0.0340, 0.0340, 0.0345, 0.0394, 0.0235, 0.0411], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-01 20:52:39,163 INFO [optim.py:368] (1/8) 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] (1/8) Epoch 24, batch 7200, loss[loss=0.1872, simple_loss=0.2794, pruned_loss=0.04747, over 17123.00 frames. ], tot_loss[loss=0.2016, simple_loss=0.2889, pruned_loss=0.05712, over 3089018.17 frames. ], batch size: 49, lr: 2.80e-03, grad_scale: 4.0 2023-05-01 20:53:28,447 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.82 vs. limit=2.0 2023-05-01 20:54:20,528 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=240696.0, num_to_drop=1, layers_to_drop={1} 2023-05-01 20:54:32,008 INFO [train.py:904] (1/8) Epoch 24, batch 7250, loss[loss=0.1692, simple_loss=0.259, pruned_loss=0.0397, over 16715.00 frames. ], tot_loss[loss=0.198, simple_loss=0.2858, pruned_loss=0.05512, over 3108612.32 frames. ], batch size: 83, lr: 2.79e-03, grad_scale: 4.0 2023-05-01 20:55:12,201 INFO [optim.py:368] (1/8) Clipping_scale=2.0, grad-norm quartiles 1.572e+02 2.642e+02 3.264e+02 4.366e+02 7.559e+02, threshold=6.528e+02, percent-clipped=1.0 2023-05-01 20:55:45,057 INFO [train.py:904] (1/8) Epoch 24, batch 7300, loss[loss=0.212, simple_loss=0.3029, pruned_loss=0.06053, over 16732.00 frames. ], tot_loss[loss=0.1985, simple_loss=0.2858, pruned_loss=0.05562, over 3090584.56 frames. ], batch size: 124, lr: 2.79e-03, grad_scale: 4.0 2023-05-01 20:57:02,841 INFO [train.py:904] (1/8) Epoch 24, batch 7350, loss[loss=0.1779, simple_loss=0.26, pruned_loss=0.04793, over 17180.00 frames. ], tot_loss[loss=0.2002, simple_loss=0.2867, pruned_loss=0.0569, over 3057083.64 frames. ], batch size: 46, lr: 2.79e-03, grad_scale: 4.0 2023-05-01 20:57:03,394 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([1.9980, 2.2520, 2.2690, 2.7190, 2.0803, 3.0818, 1.8637, 2.6503], device='cuda:1'), covar=tensor([0.1308, 0.0714, 0.1180, 0.0228, 0.0143, 0.0363, 0.1575, 0.0783], device='cuda:1'), in_proj_covar=tensor([0.0169, 0.0177, 0.0197, 0.0193, 0.0205, 0.0216, 0.0204, 0.0194], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-05-01 20:57:19,850 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.74 vs. limit=2.0 2023-05-01 20:57:30,055 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.3651, 3.2170, 2.6286, 2.1511, 2.2191, 2.2117, 3.3721, 2.9653], device='cuda:1'), covar=tensor([0.3206, 0.0751, 0.1927, 0.2889, 0.2781, 0.2384, 0.0568, 0.1427], device='cuda:1'), in_proj_covar=tensor([0.0331, 0.0272, 0.0308, 0.0320, 0.0301, 0.0266, 0.0300, 0.0343], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:1') 2023-05-01 20:57:44,658 INFO [optim.py:368] (1/8) 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] (1/8) Epoch 24, batch 7400, loss[loss=0.2034, simple_loss=0.2917, pruned_loss=0.05754, over 16681.00 frames. ], tot_loss[loss=0.2019, simple_loss=0.2879, pruned_loss=0.0579, over 3043534.68 frames. ], batch size: 62, lr: 2.79e-03, grad_scale: 4.0 2023-05-01 20:58:26,704 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.1581, 3.1554, 1.8862, 3.4542, 2.3238, 3.4805, 1.9951, 2.5830], device='cuda:1'), covar=tensor([0.0344, 0.0437, 0.1827, 0.0259, 0.0919, 0.0580, 0.1680, 0.0899], device='cuda:1'), in_proj_covar=tensor([0.0171, 0.0177, 0.0194, 0.0165, 0.0177, 0.0216, 0.0202, 0.0182], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-05-01 20:59:34,863 INFO [train.py:904] (1/8) Epoch 24, batch 7450, loss[loss=0.2155, simple_loss=0.2987, pruned_loss=0.06621, over 16706.00 frames. ], tot_loss[loss=0.203, simple_loss=0.2889, pruned_loss=0.05857, over 3052024.45 frames. ], batch size: 62, lr: 2.79e-03, grad_scale: 4.0 2023-05-01 21:00:19,561 INFO [optim.py:368] (1/8) 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 21:00:22,560 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=240932.0, num_to_drop=1, layers_to_drop={1} 2023-05-01 21:00:55,598 INFO [train.py:904] (1/8) Epoch 24, batch 7500, loss[loss=0.1929, simple_loss=0.2817, pruned_loss=0.05209, over 16482.00 frames. ], tot_loss[loss=0.2015, simple_loss=0.2883, pruned_loss=0.05739, over 3062425.29 frames. ], batch size: 68, lr: 2.79e-03, grad_scale: 4.0 2023-05-01 21:01:03,254 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([5.0436, 5.0666, 4.8162, 4.1801, 4.9327, 1.8832, 4.6763, 4.5703], device='cuda:1'), covar=tensor([0.0075, 0.0062, 0.0191, 0.0373, 0.0082, 0.2726, 0.0110, 0.0224], device='cuda:1'), in_proj_covar=tensor([0.0171, 0.0163, 0.0205, 0.0181, 0.0179, 0.0211, 0.0192, 0.0173], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-01 21:01:45,361 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=240986.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 21:01:57,405 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=240993.0, num_to_drop=1, layers_to_drop={0} 2023-05-01 21:02:00,892 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=240996.0, num_to_drop=1, layers_to_drop={1} 2023-05-01 21:02:11,648 INFO [train.py:904] (1/8) Epoch 24, batch 7550, loss[loss=0.1956, simple_loss=0.2778, pruned_loss=0.05666, over 15450.00 frames. ], tot_loss[loss=0.2009, simple_loss=0.2873, pruned_loss=0.05729, over 3070699.24 frames. ], batch size: 190, lr: 2.79e-03, grad_scale: 4.0 2023-05-01 21:02:53,683 INFO [optim.py:368] (1/8) 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] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=241044.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 21:03:20,241 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 24, batch 7600, loss[loss=0.1994, simple_loss=0.2856, pruned_loss=0.05658, over 16991.00 frames. ], tot_loss[loss=0.2008, simple_loss=0.2868, pruned_loss=0.05743, over 3076256.95 frames. ], batch size: 55, lr: 2.79e-03, grad_scale: 8.0 2023-05-01 21:04:47,283 INFO [train.py:904] (1/8) Epoch 24, batch 7650, loss[loss=0.2161, simple_loss=0.3017, pruned_loss=0.06519, over 15244.00 frames. ], tot_loss[loss=0.2023, simple_loss=0.2877, pruned_loss=0.05846, over 3068489.76 frames. ], batch size: 190, lr: 2.79e-03, grad_scale: 8.0 2023-05-01 21:05:30,616 INFO [optim.py:368] (1/8) 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] (1/8) Epoch 24, batch 7700, loss[loss=0.2004, simple_loss=0.2925, pruned_loss=0.0542, over 17024.00 frames. ], tot_loss[loss=0.2015, simple_loss=0.2868, pruned_loss=0.05806, over 3080209.60 frames. ], batch size: 50, lr: 2.79e-03, grad_scale: 4.0 2023-05-01 21:06:08,623 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.70 vs. limit=2.0 2023-05-01 21:06:28,184 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.21 vs. limit=2.0 2023-05-01 21:07:23,956 INFO [train.py:904] (1/8) Epoch 24, batch 7750, loss[loss=0.198, simple_loss=0.287, pruned_loss=0.05451, over 16675.00 frames. ], tot_loss[loss=0.2024, simple_loss=0.2877, pruned_loss=0.05858, over 3066239.34 frames. ], batch size: 134, lr: 2.79e-03, grad_scale: 4.0 2023-05-01 21:08:06,387 INFO [optim.py:368] (1/8) 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 21:08:11,967 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.6672, 4.7024, 5.0733, 5.0415, 5.0487, 4.7573, 4.7162, 4.5439], device='cuda:1'), covar=tensor([0.0347, 0.0565, 0.0358, 0.0359, 0.0470, 0.0380, 0.0963, 0.0501], device='cuda:1'), in_proj_covar=tensor([0.0417, 0.0467, 0.0452, 0.0418, 0.0498, 0.0473, 0.0556, 0.0380], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:1') 2023-05-01 21:08:19,124 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=2.87 vs. limit=5.0 2023-05-01 21:08:26,709 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.29 vs. limit=2.0 2023-05-01 21:08:38,514 INFO [train.py:904] (1/8) Epoch 24, batch 7800, loss[loss=0.2025, simple_loss=0.283, pruned_loss=0.06096, over 16439.00 frames. ], tot_loss[loss=0.203, simple_loss=0.2885, pruned_loss=0.05872, over 3091178.73 frames. ], batch size: 146, lr: 2.79e-03, grad_scale: 4.0 2023-05-01 21:08:55,441 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([5.1648, 4.9769, 5.1403, 5.3467, 5.5944, 4.9245, 5.5677, 5.5601], device='cuda:1'), covar=tensor([0.2097, 0.1315, 0.1862, 0.0841, 0.0662, 0.0792, 0.0679, 0.0706], device='cuda:1'), in_proj_covar=tensor([0.0635, 0.0787, 0.0901, 0.0790, 0.0607, 0.0626, 0.0658, 0.0763], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-01 21:09:34,094 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=241288.0, num_to_drop=1, layers_to_drop={0} 2023-05-01 21:09:56,311 INFO [train.py:904] (1/8) Epoch 24, batch 7850, loss[loss=0.1885, simple_loss=0.2847, pruned_loss=0.0462, over 16868.00 frames. ], tot_loss[loss=0.2023, simple_loss=0.2885, pruned_loss=0.05809, over 3093009.82 frames. ], batch size: 96, lr: 2.79e-03, grad_scale: 4.0 2023-05-01 21:10:29,946 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.0693, 3.4809, 3.4000, 2.2424, 3.3405, 3.5738, 3.3225, 1.6756], device='cuda:1'), covar=tensor([0.0702, 0.0094, 0.0103, 0.0565, 0.0125, 0.0141, 0.0122, 0.0720], device='cuda:1'), in_proj_covar=tensor([0.0135, 0.0086, 0.0086, 0.0135, 0.0099, 0.0110, 0.0095, 0.0130], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:1') 2023-05-01 21:10:38,475 INFO [optim.py:368] (1/8) Clipping_scale=2.0, grad-norm quartiles 2.020e+02 2.608e+02 3.095e+02 3.566e+02 5.691e+02, threshold=6.191e+02, percent-clipped=0.0 2023-05-01 21:10:55,479 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 24, batch 7900, loss[loss=0.1956, simple_loss=0.2944, pruned_loss=0.04837, over 16795.00 frames. ], tot_loss[loss=0.2019, simple_loss=0.2882, pruned_loss=0.05782, over 3083588.78 frames. ], batch size: 102, lr: 2.79e-03, grad_scale: 4.0 2023-05-01 21:11:19,426 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.2377, 2.9043, 3.1046, 1.8785, 3.2724, 3.3060, 2.6551, 2.6417], device='cuda:1'), covar=tensor([0.0797, 0.0299, 0.0229, 0.1120, 0.0101, 0.0232, 0.0523, 0.0454], device='cuda:1'), in_proj_covar=tensor([0.0149, 0.0109, 0.0100, 0.0139, 0.0083, 0.0128, 0.0129, 0.0131], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-05-01 21:12:29,004 INFO [train.py:904] (1/8) Epoch 24, batch 7950, loss[loss=0.1788, simple_loss=0.2713, pruned_loss=0.04314, over 16719.00 frames. ], tot_loss[loss=0.2019, simple_loss=0.2883, pruned_loss=0.05772, over 3086620.32 frames. ], batch size: 89, lr: 2.79e-03, grad_scale: 4.0 2023-05-01 21:12:34,457 INFO [zipformer.py:625] (1/8) 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] (1/8) 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 21:13:13,229 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.3613, 2.9564, 2.6944, 2.2688, 2.2855, 2.3120, 2.9600, 2.8865], device='cuda:1'), covar=tensor([0.2514, 0.0723, 0.1590, 0.2620, 0.2406, 0.2221, 0.0531, 0.1280], device='cuda:1'), in_proj_covar=tensor([0.0334, 0.0272, 0.0310, 0.0321, 0.0302, 0.0268, 0.0301, 0.0344], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:1') 2023-05-01 21:13:22,176 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=4.46 vs. limit=5.0 2023-05-01 21:13:40,317 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.9539, 2.7996, 2.8351, 2.1611, 2.7165, 2.1842, 2.7579, 2.9676], device='cuda:1'), covar=tensor([0.0260, 0.0752, 0.0480, 0.1678, 0.0765, 0.0914, 0.0538, 0.0715], device='cuda:1'), in_proj_covar=tensor([0.0155, 0.0165, 0.0167, 0.0154, 0.0145, 0.0130, 0.0143, 0.0177], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:1') 2023-05-01 21:13:46,678 INFO [train.py:904] (1/8) Epoch 24, batch 8000, loss[loss=0.2009, simple_loss=0.2905, pruned_loss=0.0556, over 16762.00 frames. ], tot_loss[loss=0.2033, simple_loss=0.2891, pruned_loss=0.05877, over 3081503.52 frames. ], batch size: 124, lr: 2.79e-03, grad_scale: 8.0 2023-05-01 21:14:09,543 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=241467.0, num_to_drop=1, layers_to_drop={1} 2023-05-01 21:15:04,297 INFO [train.py:904] (1/8) Epoch 24, batch 8050, loss[loss=0.1844, simple_loss=0.2807, pruned_loss=0.04402, over 16846.00 frames. ], tot_loss[loss=0.203, simple_loss=0.2889, pruned_loss=0.05854, over 3073550.80 frames. ], batch size: 102, lr: 2.79e-03, grad_scale: 8.0 2023-05-01 21:15:47,625 INFO [optim.py:368] (1/8) Clipping_scale=2.0, grad-norm quartiles 1.925e+02 2.671e+02 3.058e+02 3.888e+02 6.821e+02, threshold=6.115e+02, percent-clipped=1.0 2023-05-01 21:16:22,138 INFO [train.py:904] (1/8) Epoch 24, batch 8100, loss[loss=0.195, simple_loss=0.2773, pruned_loss=0.0564, over 17006.00 frames. ], tot_loss[loss=0.2022, simple_loss=0.2886, pruned_loss=0.05792, over 3098449.68 frames. ], batch size: 50, lr: 2.79e-03, grad_scale: 8.0 2023-05-01 21:16:50,337 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([1.7175, 2.4806, 2.4304, 3.6366, 2.5875, 3.7685, 1.5098, 2.7461], device='cuda:1'), covar=tensor([0.1345, 0.0812, 0.1247, 0.0226, 0.0206, 0.0452, 0.1700, 0.0851], device='cuda:1'), in_proj_covar=tensor([0.0169, 0.0176, 0.0197, 0.0193, 0.0205, 0.0216, 0.0204, 0.0195], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-05-01 21:17:14,361 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=241588.0, num_to_drop=1, layers_to_drop={1} 2023-05-01 21:17:29,875 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.33 vs. limit=2.0 2023-05-01 21:17:38,324 INFO [train.py:904] (1/8) Epoch 24, batch 8150, loss[loss=0.1582, simple_loss=0.2549, pruned_loss=0.03072, over 16884.00 frames. ], tot_loss[loss=0.2006, simple_loss=0.2866, pruned_loss=0.05733, over 3081392.50 frames. ], batch size: 96, lr: 2.79e-03, grad_scale: 8.0 2023-05-01 21:17:38,719 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.5681, 4.8515, 4.6258, 4.6319, 4.3871, 4.3564, 4.3012, 4.9039], device='cuda:1'), covar=tensor([0.1202, 0.0850, 0.1012, 0.0917, 0.0854, 0.1301, 0.1093, 0.0865], device='cuda:1'), in_proj_covar=tensor([0.0691, 0.0823, 0.0690, 0.0642, 0.0525, 0.0532, 0.0697, 0.0649], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:1') 2023-05-01 21:17:55,931 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.8101, 3.8782, 4.1465, 4.1092, 4.1118, 3.8985, 3.9062, 3.9034], device='cuda:1'), covar=tensor([0.0338, 0.0626, 0.0367, 0.0393, 0.0487, 0.0438, 0.0831, 0.0502], device='cuda:1'), in_proj_covar=tensor([0.0415, 0.0467, 0.0451, 0.0417, 0.0498, 0.0473, 0.0556, 0.0379], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:1') 2023-05-01 21:18:06,800 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=241621.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 21:18:08,044 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.9822, 5.3731, 5.6109, 5.3047, 5.5217, 5.9744, 5.4222, 5.1516], device='cuda:1'), covar=tensor([0.1036, 0.1846, 0.2056, 0.1895, 0.1995, 0.0845, 0.1547, 0.2199], device='cuda:1'), in_proj_covar=tensor([0.0418, 0.0616, 0.0679, 0.0505, 0.0666, 0.0702, 0.0528, 0.0672], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-05-01 21:18:22,058 INFO [optim.py:368] (1/8) 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] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=241636.0, num_to_drop=1, layers_to_drop={0} 2023-05-01 21:18:39,437 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=241642.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 21:18:55,843 INFO [train.py:904] (1/8) Epoch 24, batch 8200, loss[loss=0.1788, simple_loss=0.2789, pruned_loss=0.03939, over 16712.00 frames. ], tot_loss[loss=0.1976, simple_loss=0.2835, pruned_loss=0.05584, over 3103744.70 frames. ], batch size: 89, lr: 2.79e-03, grad_scale: 8.0 2023-05-01 21:19:42,887 INFO [zipformer.py:625] (1/8) 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] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=241690.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 21:20:12,707 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.1605, 3.3445, 3.6486, 2.1006, 3.0376, 2.3444, 3.5409, 3.5114], device='cuda:1'), covar=tensor([0.0285, 0.0856, 0.0522, 0.2157, 0.0845, 0.1024, 0.0623, 0.1009], device='cuda:1'), in_proj_covar=tensor([0.0156, 0.0166, 0.0168, 0.0155, 0.0146, 0.0131, 0.0144, 0.0178], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-05-01 21:20:15,795 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.2208, 2.1262, 2.1504, 3.8669, 2.1531, 2.4586, 2.2294, 2.2710], device='cuda:1'), covar=tensor([0.1359, 0.3948, 0.3350, 0.0577, 0.4412, 0.2960, 0.3967, 0.3815], device='cuda:1'), in_proj_covar=tensor([0.0405, 0.0455, 0.0371, 0.0329, 0.0438, 0.0520, 0.0426, 0.0531], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-01 21:20:16,379 INFO [train.py:904] (1/8) Epoch 24, batch 8250, loss[loss=0.1663, simple_loss=0.2649, pruned_loss=0.0339, over 16783.00 frames. ], tot_loss[loss=0.1947, simple_loss=0.2823, pruned_loss=0.0536, over 3086117.36 frames. ], batch size: 102, lr: 2.79e-03, grad_scale: 8.0 2023-05-01 21:20:36,930 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.5839, 3.6279, 3.4499, 3.0809, 3.1886, 3.5434, 3.3078, 3.3853], device='cuda:1'), covar=tensor([0.0572, 0.0687, 0.0326, 0.0308, 0.0522, 0.0523, 0.1568, 0.0527], device='cuda:1'), in_proj_covar=tensor([0.0290, 0.0436, 0.0339, 0.0339, 0.0343, 0.0394, 0.0235, 0.0410], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-01 21:21:03,234 INFO [optim.py:368] (1/8) 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 21:21:39,005 INFO [train.py:904] (1/8) Epoch 24, batch 8300, loss[loss=0.1678, simple_loss=0.2535, pruned_loss=0.0411, over 12232.00 frames. ], tot_loss[loss=0.1903, simple_loss=0.2793, pruned_loss=0.0506, over 3073366.33 frames. ], batch size: 246, lr: 2.79e-03, grad_scale: 8.0 2023-05-01 21:21:54,702 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=241762.0, num_to_drop=1, layers_to_drop={2} 2023-05-01 21:22:33,122 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.67 vs. limit=2.0 2023-05-01 21:23:02,690 INFO [train.py:904] (1/8) Epoch 24, batch 8350, loss[loss=0.1723, simple_loss=0.2686, pruned_loss=0.03805, over 15264.00 frames. ], tot_loss[loss=0.1882, simple_loss=0.2788, pruned_loss=0.04881, over 3062919.02 frames. ], batch size: 190, lr: 2.79e-03, grad_scale: 8.0 2023-05-01 21:23:03,941 INFO [zipformer.py:625] (1/8) 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] (1/8) 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 21:23:48,761 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=241831.0, num_to_drop=1, layers_to_drop={0} 2023-05-01 21:24:19,719 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.5170, 4.6963, 4.8296, 4.5836, 4.6884, 5.1869, 4.6581, 4.3371], device='cuda:1'), covar=tensor([0.1340, 0.1859, 0.2152, 0.2032, 0.2347, 0.0934, 0.1466, 0.2503], device='cuda:1'), in_proj_covar=tensor([0.0407, 0.0599, 0.0662, 0.0492, 0.0649, 0.0686, 0.0515, 0.0656], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-05-01 21:24:23,521 INFO [train.py:904] (1/8) Epoch 24, batch 8400, loss[loss=0.1731, simple_loss=0.2623, pruned_loss=0.04188, over 15412.00 frames. ], tot_loss[loss=0.1853, simple_loss=0.2766, pruned_loss=0.04699, over 3054372.76 frames. ], batch size: 191, lr: 2.79e-03, grad_scale: 8.0 2023-05-01 21:24:42,770 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=241864.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 21:25:24,478 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.9263, 4.8887, 4.6707, 4.0819, 4.7611, 1.8170, 4.4954, 4.5226], device='cuda:1'), covar=tensor([0.0113, 0.0120, 0.0240, 0.0433, 0.0124, 0.2879, 0.0161, 0.0262], device='cuda:1'), in_proj_covar=tensor([0.0169, 0.0162, 0.0202, 0.0178, 0.0177, 0.0208, 0.0190, 0.0171], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-01 21:25:28,795 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=241892.0, num_to_drop=1, layers_to_drop={0} 2023-05-01 21:25:45,943 INFO [train.py:904] (1/8) Epoch 24, batch 8450, loss[loss=0.1552, simple_loss=0.258, pruned_loss=0.02615, over 17282.00 frames. ], tot_loss[loss=0.1825, simple_loss=0.275, pruned_loss=0.04501, over 3078659.43 frames. ], batch size: 52, lr: 2.79e-03, grad_scale: 8.0 2023-05-01 21:26:31,269 INFO [optim.py:368] (1/8) 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 21:26:44,768 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.2068, 3.1556, 2.7638, 5.0144, 3.7638, 4.3474, 1.9329, 3.1373], device='cuda:1'), covar=tensor([0.1079, 0.0610, 0.0997, 0.0165, 0.0169, 0.0403, 0.1353, 0.0695], device='cuda:1'), in_proj_covar=tensor([0.0168, 0.0174, 0.0194, 0.0190, 0.0202, 0.0213, 0.0203, 0.0194], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-05-01 21:27:07,283 INFO [train.py:904] (1/8) Epoch 24, batch 8500, loss[loss=0.1782, simple_loss=0.2693, pruned_loss=0.04356, over 16395.00 frames. ], tot_loss[loss=0.1787, simple_loss=0.2713, pruned_loss=0.04303, over 3067457.41 frames. ], batch size: 146, lr: 2.79e-03, grad_scale: 8.0 2023-05-01 21:27:47,067 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=241977.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 21:28:33,831 INFO [train.py:904] (1/8) Epoch 24, batch 8550, loss[loss=0.1778, simple_loss=0.2759, pruned_loss=0.03984, over 16702.00 frames. ], tot_loss[loss=0.1773, simple_loss=0.2697, pruned_loss=0.04243, over 3050021.43 frames. ], batch size: 76, lr: 2.79e-03, grad_scale: 8.0 2023-05-01 21:29:26,914 INFO [optim.py:368] (1/8) 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 21:29:58,510 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.8507, 3.7238, 3.9392, 4.0074, 4.0971, 3.6726, 4.0259, 4.1076], device='cuda:1'), covar=tensor([0.1571, 0.1176, 0.1158, 0.0715, 0.0549, 0.1995, 0.0829, 0.0631], device='cuda:1'), in_proj_covar=tensor([0.0629, 0.0779, 0.0892, 0.0785, 0.0601, 0.0620, 0.0653, 0.0758], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-01 21:30:13,226 INFO [train.py:904] (1/8) Epoch 24, batch 8600, loss[loss=0.1589, simple_loss=0.2531, pruned_loss=0.03239, over 16488.00 frames. ], tot_loss[loss=0.1765, simple_loss=0.2697, pruned_loss=0.04162, over 3040822.84 frames. ], batch size: 62, lr: 2.79e-03, grad_scale: 8.0 2023-05-01 21:30:27,624 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.1322, 2.5665, 2.6466, 1.9323, 2.8205, 2.8533, 2.5107, 2.5167], device='cuda:1'), covar=tensor([0.0655, 0.0237, 0.0233, 0.0977, 0.0107, 0.0243, 0.0434, 0.0406], device='cuda:1'), in_proj_covar=tensor([0.0144, 0.0106, 0.0096, 0.0135, 0.0080, 0.0124, 0.0125, 0.0127], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:1') 2023-05-01 21:30:31,476 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=242062.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 21:31:39,738 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.5108, 3.4651, 2.7129, 2.1152, 2.0850, 2.2977, 3.6156, 3.0130], device='cuda:1'), covar=tensor([0.3045, 0.0623, 0.1927, 0.3242, 0.3009, 0.2264, 0.0467, 0.1463], device='cuda:1'), in_proj_covar=tensor([0.0324, 0.0265, 0.0302, 0.0313, 0.0294, 0.0263, 0.0293, 0.0336], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:1') 2023-05-01 21:31:51,604 INFO [train.py:904] (1/8) Epoch 24, batch 8650, loss[loss=0.1515, simple_loss=0.2421, pruned_loss=0.03049, over 12177.00 frames. ], tot_loss[loss=0.1747, simple_loss=0.2683, pruned_loss=0.04055, over 3043803.49 frames. ], batch size: 246, lr: 2.79e-03, grad_scale: 4.0 2023-05-01 21:32:08,637 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.1741, 2.5172, 2.6079, 1.9076, 2.8013, 2.8308, 2.5122, 2.4638], device='cuda:1'), covar=tensor([0.0658, 0.0280, 0.0236, 0.1026, 0.0116, 0.0256, 0.0461, 0.0430], device='cuda:1'), in_proj_covar=tensor([0.0143, 0.0106, 0.0096, 0.0134, 0.0079, 0.0123, 0.0124, 0.0126], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:1') 2023-05-01 21:32:10,578 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=242110.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 21:32:56,838 INFO [optim.py:368] (1/8) 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 21:33:17,281 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.4511, 3.2574, 3.4999, 1.8309, 3.7087, 3.7178, 2.9187, 2.8978], device='cuda:1'), covar=tensor([0.0756, 0.0302, 0.0236, 0.1264, 0.0089, 0.0184, 0.0453, 0.0451], device='cuda:1'), in_proj_covar=tensor([0.0142, 0.0106, 0.0095, 0.0134, 0.0079, 0.0123, 0.0124, 0.0126], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:1') 2023-05-01 21:33:37,175 INFO [train.py:904] (1/8) Epoch 24, batch 8700, loss[loss=0.1711, simple_loss=0.2679, pruned_loss=0.03719, over 16262.00 frames. ], tot_loss[loss=0.172, simple_loss=0.2654, pruned_loss=0.0393, over 3057848.88 frames. ], batch size: 165, lr: 2.79e-03, grad_scale: 4.0 2023-05-01 21:33:38,240 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=242153.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 21:33:43,890 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=2.69 vs. limit=5.0 2023-05-01 21:33:49,873 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=242159.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 21:34:20,819 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.7179, 3.4812, 3.7363, 1.9547, 3.9512, 3.9609, 3.0773, 3.0877], device='cuda:1'), covar=tensor([0.0646, 0.0242, 0.0187, 0.1181, 0.0066, 0.0150, 0.0394, 0.0415], device='cuda:1'), in_proj_covar=tensor([0.0142, 0.0105, 0.0095, 0.0133, 0.0079, 0.0123, 0.0124, 0.0125], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:1') 2023-05-01 21:34:42,494 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=242187.0, num_to_drop=1, layers_to_drop={1} 2023-05-01 21:35:13,449 INFO [train.py:904] (1/8) Epoch 24, batch 8750, loss[loss=0.1842, simple_loss=0.2833, pruned_loss=0.04256, over 16337.00 frames. ], tot_loss[loss=0.1715, simple_loss=0.2652, pruned_loss=0.03894, over 3053024.59 frames. ], batch size: 146, lr: 2.79e-03, grad_scale: 4.0 2023-05-01 21:35:42,258 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=242214.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 21:36:10,046 INFO [zipformer.py:625] (1/8) 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] (1/8) 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 21:36:36,166 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=3.16 vs. limit=5.0 2023-05-01 21:36:50,896 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.6947, 2.6118, 1.7373, 2.8219, 2.1265, 2.8255, 1.9812, 2.3503], device='cuda:1'), covar=tensor([0.0297, 0.0364, 0.1362, 0.0278, 0.0698, 0.0465, 0.1432, 0.0664], device='cuda:1'), in_proj_covar=tensor([0.0167, 0.0172, 0.0189, 0.0161, 0.0173, 0.0210, 0.0199, 0.0177], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0003, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-05-01 21:37:05,081 INFO [train.py:904] (1/8) Epoch 24, batch 8800, loss[loss=0.1625, simple_loss=0.2605, pruned_loss=0.03232, over 15362.00 frames. ], tot_loss[loss=0.1697, simple_loss=0.2636, pruned_loss=0.03792, over 3058664.62 frames. ], batch size: 191, lr: 2.79e-03, grad_scale: 8.0 2023-05-01 21:37:56,143 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=242277.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 21:38:09,330 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([5.1710, 5.4739, 5.3141, 5.2703, 5.0040, 4.9283, 4.8785, 5.5769], device='cuda:1'), covar=tensor([0.1062, 0.0853, 0.0730, 0.0726, 0.0705, 0.0798, 0.1174, 0.0784], device='cuda:1'), in_proj_covar=tensor([0.0677, 0.0811, 0.0674, 0.0630, 0.0516, 0.0525, 0.0684, 0.0637], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:1') 2023-05-01 21:38:17,586 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 24, batch 8850, loss[loss=0.1719, simple_loss=0.2747, pruned_loss=0.03452, over 16769.00 frames. ], tot_loss[loss=0.1711, simple_loss=0.2671, pruned_loss=0.03751, over 3060364.11 frames. ], batch size: 124, lr: 2.79e-03, grad_scale: 8.0 2023-05-01 21:39:03,871 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=242309.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 21:39:30,554 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.32 vs. limit=2.0 2023-05-01 21:39:38,584 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=242325.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 21:39:52,555 INFO [optim.py:368] (1/8) 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] (1/8) Epoch 24, batch 8900, loss[loss=0.1595, simple_loss=0.2594, pruned_loss=0.02976, over 16196.00 frames. ], tot_loss[loss=0.1704, simple_loss=0.2673, pruned_loss=0.03677, over 3062794.11 frames. ], batch size: 165, lr: 2.79e-03, grad_scale: 4.0 2023-05-01 21:41:04,043 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.2632, 2.9770, 3.1179, 1.8937, 3.3003, 3.3294, 2.7349, 2.7434], device='cuda:1'), covar=tensor([0.0728, 0.0280, 0.0198, 0.1076, 0.0087, 0.0180, 0.0455, 0.0428], device='cuda:1'), in_proj_covar=tensor([0.0142, 0.0106, 0.0095, 0.0134, 0.0079, 0.0123, 0.0124, 0.0126], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:1') 2023-05-01 21:41:12,535 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=242370.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 21:42:44,444 INFO [train.py:904] (1/8) Epoch 24, batch 8950, loss[loss=0.1708, simple_loss=0.2577, pruned_loss=0.04196, over 12970.00 frames. ], tot_loss[loss=0.1708, simple_loss=0.2671, pruned_loss=0.03728, over 3068516.80 frames. ], batch size: 248, lr: 2.78e-03, grad_scale: 4.0 2023-05-01 21:43:49,726 INFO [optim.py:368] (1/8) 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 21:44:35,817 INFO [train.py:904] (1/8) Epoch 24, batch 9000, loss[loss=0.1367, simple_loss=0.2309, pruned_loss=0.02119, over 17037.00 frames. ], tot_loss[loss=0.168, simple_loss=0.2638, pruned_loss=0.03612, over 3087841.10 frames. ], batch size: 50, lr: 2.78e-03, grad_scale: 4.0 2023-05-01 21:44:35,818 INFO [train.py:929] (1/8) Computing validation loss 2023-05-01 21:44:45,532 INFO [train.py:938] (1/8) Epoch 24, validation: loss=0.1445, simple_loss=0.2484, pruned_loss=0.02026, over 944034.00 frames. 2023-05-01 21:44:45,532 INFO [train.py:939] (1/8) Maximum memory allocated so far is 18145MB 2023-05-01 21:44:58,942 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=242459.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 21:45:48,510 INFO [zipformer.py:625] (1/8) 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] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=242487.0, num_to_drop=1, layers_to_drop={1} 2023-05-01 21:46:10,868 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.4423, 4.2683, 4.4506, 4.6438, 4.7570, 4.3476, 4.7774, 4.7911], device='cuda:1'), covar=tensor([0.1960, 0.1410, 0.1918, 0.0850, 0.0684, 0.1062, 0.0639, 0.0848], device='cuda:1'), in_proj_covar=tensor([0.0625, 0.0771, 0.0885, 0.0779, 0.0596, 0.0616, 0.0646, 0.0749], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-01 21:46:14,330 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.3513, 1.6886, 2.0006, 2.2530, 2.3121, 2.5405, 1.9064, 2.4502], device='cuda:1'), covar=tensor([0.0235, 0.0559, 0.0357, 0.0371, 0.0372, 0.0237, 0.0555, 0.0185], device='cuda:1'), in_proj_covar=tensor([0.0187, 0.0191, 0.0178, 0.0180, 0.0196, 0.0155, 0.0195, 0.0153], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-01 21:46:30,341 INFO [train.py:904] (1/8) Epoch 24, batch 9050, loss[loss=0.1866, simple_loss=0.2785, pruned_loss=0.04736, over 16201.00 frames. ], tot_loss[loss=0.1686, simple_loss=0.2645, pruned_loss=0.03636, over 3091580.40 frames. ], batch size: 165, lr: 2.78e-03, grad_scale: 4.0 2023-05-01 21:46:40,247 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=242507.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 21:46:44,678 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=242509.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 21:46:55,127 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.7056, 2.6256, 1.9019, 2.8308, 2.1061, 2.8208, 2.1912, 2.4239], device='cuda:1'), covar=tensor([0.0314, 0.0349, 0.1274, 0.0283, 0.0679, 0.0491, 0.1173, 0.0589], device='cuda:1'), in_proj_covar=tensor([0.0167, 0.0173, 0.0190, 0.0160, 0.0174, 0.0210, 0.0199, 0.0178], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-05-01 21:47:30,072 INFO [optim.py:368] (1/8) 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 21:47:35,728 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=242535.0, num_to_drop=1, layers_to_drop={1} 2023-05-01 21:48:00,540 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=242544.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 21:48:17,898 INFO [train.py:904] (1/8) Epoch 24, batch 9100, loss[loss=0.1595, simple_loss=0.2607, pruned_loss=0.02911, over 16317.00 frames. ], tot_loss[loss=0.1681, simple_loss=0.2632, pruned_loss=0.03651, over 3090548.19 frames. ], batch size: 146, lr: 2.78e-03, grad_scale: 4.0 2023-05-01 21:48:23,294 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=242555.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 21:48:56,856 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.2560, 2.2632, 2.1781, 3.9648, 2.0580, 2.5708, 2.2680, 2.3701], device='cuda:1'), covar=tensor([0.1285, 0.3742, 0.3375, 0.0519, 0.4528, 0.2678, 0.3954, 0.3555], device='cuda:1'), in_proj_covar=tensor([0.0399, 0.0447, 0.0368, 0.0322, 0.0432, 0.0510, 0.0419, 0.0521], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-01 21:49:27,725 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=242582.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 21:49:45,692 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=3.00 vs. limit=5.0 2023-05-01 21:50:11,558 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.0005, 4.0644, 4.3142, 4.3184, 4.3142, 4.1122, 4.1003, 4.0791], device='cuda:1'), covar=tensor([0.0343, 0.0584, 0.0450, 0.0402, 0.0391, 0.0385, 0.0755, 0.0435], device='cuda:1'), in_proj_covar=tensor([0.0405, 0.0451, 0.0441, 0.0405, 0.0482, 0.0460, 0.0538, 0.0369], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:1') 2023-05-01 21:50:15,696 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=242602.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 21:50:16,953 INFO [train.py:904] (1/8) Epoch 24, batch 9150, loss[loss=0.1523, simple_loss=0.2514, pruned_loss=0.02657, over 16530.00 frames. ], tot_loss[loss=0.1681, simple_loss=0.264, pruned_loss=0.03612, over 3099807.56 frames. ], batch size: 68, lr: 2.78e-03, grad_scale: 4.0 2023-05-01 21:50:36,998 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=4.21 vs. limit=5.0 2023-05-01 21:50:40,167 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.0675, 2.1051, 2.1154, 3.5457, 2.0737, 2.3697, 2.2322, 2.2239], device='cuda:1'), covar=tensor([0.1361, 0.3776, 0.3474, 0.0655, 0.4520, 0.3004, 0.3784, 0.3805], device='cuda:1'), in_proj_covar=tensor([0.0398, 0.0447, 0.0368, 0.0322, 0.0432, 0.0510, 0.0419, 0.0521], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-01 21:50:46,714 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=242616.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 21:51:21,271 INFO [optim.py:368] (1/8) 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 21:51:34,439 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.1743, 2.2231, 2.2963, 3.7552, 2.1779, 2.4830, 2.3437, 2.3362], device='cuda:1'), covar=tensor([0.1320, 0.3721, 0.3189, 0.0605, 0.4293, 0.2819, 0.3540, 0.3791], device='cuda:1'), in_proj_covar=tensor([0.0398, 0.0446, 0.0368, 0.0321, 0.0430, 0.0508, 0.0418, 0.0520], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-01 21:52:01,739 INFO [train.py:904] (1/8) Epoch 24, batch 9200, loss[loss=0.1493, simple_loss=0.2442, pruned_loss=0.02719, over 16682.00 frames. ], tot_loss[loss=0.1649, simple_loss=0.2595, pruned_loss=0.03518, over 3095873.41 frames. ], batch size: 76, lr: 2.78e-03, grad_scale: 8.0 2023-05-01 21:52:21,716 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=242663.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 21:52:24,772 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=242665.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 21:52:26,551 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.0189, 3.2053, 3.6370, 2.1248, 2.9833, 2.2182, 3.4512, 3.3597], device='cuda:1'), covar=tensor([0.0299, 0.1027, 0.0512, 0.2171, 0.0885, 0.1124, 0.0711, 0.1124], device='cuda:1'), in_proj_covar=tensor([0.0152, 0.0159, 0.0163, 0.0150, 0.0141, 0.0127, 0.0139, 0.0170], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:1') 2023-05-01 21:53:33,801 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.9374, 5.2096, 5.4243, 5.0945, 5.3499, 5.8361, 5.2867, 4.9485], device='cuda:1'), covar=tensor([0.0918, 0.2021, 0.2179, 0.1842, 0.2110, 0.0768, 0.1454, 0.2149], device='cuda:1'), in_proj_covar=tensor([0.0396, 0.0588, 0.0647, 0.0481, 0.0634, 0.0673, 0.0502, 0.0641], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-05-01 21:53:38,106 INFO [train.py:904] (1/8) Epoch 24, batch 9250, loss[loss=0.1608, simple_loss=0.2557, pruned_loss=0.03299, over 16678.00 frames. ], tot_loss[loss=0.1649, simple_loss=0.2594, pruned_loss=0.03523, over 3094308.48 frames. ], batch size: 134, lr: 2.78e-03, grad_scale: 8.0 2023-05-01 21:54:40,871 INFO [optim.py:368] (1/8) 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 21:55:27,519 INFO [train.py:904] (1/8) Epoch 24, batch 9300, loss[loss=0.1554, simple_loss=0.2482, pruned_loss=0.03127, over 16612.00 frames. ], tot_loss[loss=0.1639, simple_loss=0.258, pruned_loss=0.03488, over 3079645.75 frames. ], batch size: 62, lr: 2.78e-03, grad_scale: 8.0 2023-05-01 21:56:13,858 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=242772.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 21:56:44,441 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.7777, 1.9436, 2.2940, 2.7710, 2.6089, 3.0808, 2.1968, 3.0763], device='cuda:1'), covar=tensor([0.0215, 0.0571, 0.0404, 0.0316, 0.0346, 0.0231, 0.0528, 0.0187], device='cuda:1'), in_proj_covar=tensor([0.0187, 0.0189, 0.0177, 0.0179, 0.0195, 0.0155, 0.0194, 0.0152], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-01 21:57:14,573 INFO [train.py:904] (1/8) Epoch 24, batch 9350, loss[loss=0.1779, simple_loss=0.2703, pruned_loss=0.04278, over 16178.00 frames. ], tot_loss[loss=0.1629, simple_loss=0.2569, pruned_loss=0.03449, over 3065852.94 frames. ], batch size: 165, lr: 2.78e-03, grad_scale: 8.0 2023-05-01 21:57:27,074 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=242809.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 21:57:54,737 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.1958, 2.2678, 2.2057, 3.9268, 2.1837, 2.5634, 2.3280, 2.3752], device='cuda:1'), covar=tensor([0.1278, 0.3734, 0.3378, 0.0539, 0.4322, 0.2580, 0.3743, 0.3653], device='cuda:1'), in_proj_covar=tensor([0.0398, 0.0447, 0.0369, 0.0322, 0.0432, 0.0509, 0.0419, 0.0522], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-01 21:58:13,360 INFO [optim.py:368] (1/8) 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 21:58:14,168 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=242833.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 21:58:25,325 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=242839.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 21:58:54,548 INFO [train.py:904] (1/8) Epoch 24, batch 9400, loss[loss=0.1762, simple_loss=0.2813, pruned_loss=0.03551, over 16848.00 frames. ], tot_loss[loss=0.1635, simple_loss=0.2582, pruned_loss=0.03442, over 3085032.29 frames. ], batch size: 116, lr: 2.78e-03, grad_scale: 8.0 2023-05-01 21:59:03,672 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=242857.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 21:59:54,993 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 24, batch 9450, loss[loss=0.1696, simple_loss=0.2637, pruned_loss=0.03778, over 16696.00 frames. ], tot_loss[loss=0.1649, simple_loss=0.2601, pruned_loss=0.03491, over 3074760.32 frames. ], batch size: 134, lr: 2.78e-03, grad_scale: 8.0 2023-05-01 22:00:51,464 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=242911.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 22:01:13,402 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=242921.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 22:01:32,438 INFO [zipformer.py:625] (1/8) 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] (1/8) 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 22:02:16,734 INFO [train.py:904] (1/8) Epoch 24, batch 9500, loss[loss=0.1745, simple_loss=0.2641, pruned_loss=0.04245, over 16682.00 frames. ], tot_loss[loss=0.1644, simple_loss=0.2592, pruned_loss=0.0348, over 3055203.37 frames. ], batch size: 57, lr: 2.78e-03, grad_scale: 8.0 2023-05-01 22:02:30,121 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=242958.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 22:02:43,748 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=242965.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 22:02:45,707 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.74 vs. limit=2.0 2023-05-01 22:02:59,254 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([5.3033, 5.8986, 6.0800, 5.7313, 5.8434, 6.3159, 5.8488, 5.5216], device='cuda:1'), covar=tensor([0.0679, 0.1551, 0.1827, 0.1637, 0.2007, 0.0774, 0.1422, 0.2201], device='cuda:1'), in_proj_covar=tensor([0.0393, 0.0586, 0.0645, 0.0480, 0.0634, 0.0673, 0.0501, 0.0637], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-05-01 22:03:15,730 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=242982.0, num_to_drop=1, layers_to_drop={3} 2023-05-01 22:04:01,164 INFO [train.py:904] (1/8) Epoch 24, batch 9550, loss[loss=0.1638, simple_loss=0.2666, pruned_loss=0.03044, over 16380.00 frames. ], tot_loss[loss=0.1639, simple_loss=0.2588, pruned_loss=0.03452, over 3055602.75 frames. ], batch size: 146, lr: 2.78e-03, grad_scale: 8.0 2023-05-01 22:04:24,735 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=243013.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 22:05:06,080 INFO [optim.py:368] (1/8) 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 22:05:43,247 INFO [train.py:904] (1/8) Epoch 24, batch 9600, loss[loss=0.1826, simple_loss=0.2895, pruned_loss=0.03787, over 16386.00 frames. ], tot_loss[loss=0.1656, simple_loss=0.2605, pruned_loss=0.03531, over 3058642.95 frames. ], batch size: 146, lr: 2.78e-03, grad_scale: 8.0 2023-05-01 22:05:54,104 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([5.1073, 5.1316, 5.5317, 5.5110, 5.5192, 5.2466, 5.1626, 4.9833], device='cuda:1'), covar=tensor([0.0356, 0.0860, 0.0442, 0.0372, 0.0500, 0.0400, 0.0947, 0.0460], device='cuda:1'), in_proj_covar=tensor([0.0402, 0.0448, 0.0438, 0.0403, 0.0480, 0.0456, 0.0532, 0.0367], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:1') 2023-05-01 22:06:01,499 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=3.03 vs. limit=5.0 2023-05-01 22:07:32,098 INFO [train.py:904] (1/8) Epoch 24, batch 9650, loss[loss=0.156, simple_loss=0.2447, pruned_loss=0.03363, over 12276.00 frames. ], tot_loss[loss=0.1666, simple_loss=0.2621, pruned_loss=0.03561, over 3052261.48 frames. ], batch size: 248, lr: 2.78e-03, grad_scale: 8.0 2023-05-01 22:08:29,798 INFO [zipformer.py:625] (1/8) 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] (1/8) 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 22:08:42,589 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.6667, 3.9954, 2.9071, 2.1856, 2.3770, 2.4664, 4.1602, 3.3421], device='cuda:1'), covar=tensor([0.2996, 0.0493, 0.1826, 0.3105, 0.2840, 0.2104, 0.0383, 0.1314], device='cuda:1'), in_proj_covar=tensor([0.0321, 0.0263, 0.0300, 0.0310, 0.0288, 0.0260, 0.0290, 0.0331], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:1') 2023-05-01 22:08:50,202 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 24, batch 9700, loss[loss=0.1566, simple_loss=0.2455, pruned_loss=0.0338, over 12335.00 frames. ], tot_loss[loss=0.1657, simple_loss=0.2608, pruned_loss=0.0353, over 3046342.58 frames. ], batch size: 250, lr: 2.78e-03, grad_scale: 8.0 2023-05-01 22:10:30,929 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=243187.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 22:11:01,229 INFO [train.py:904] (1/8) Epoch 24, batch 9750, loss[loss=0.178, simple_loss=0.273, pruned_loss=0.04156, over 16944.00 frames. ], tot_loss[loss=0.165, simple_loss=0.2596, pruned_loss=0.0352, over 3056767.47 frames. ], batch size: 109, lr: 2.78e-03, grad_scale: 4.0 2023-05-01 22:11:10,863 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.6715, 2.1113, 1.7740, 1.8403, 2.3923, 2.0726, 2.1292, 2.5018], device='cuda:1'), covar=tensor([0.0182, 0.0477, 0.0572, 0.0548, 0.0324, 0.0443, 0.0232, 0.0299], device='cuda:1'), in_proj_covar=tensor([0.0208, 0.0232, 0.0223, 0.0224, 0.0233, 0.0232, 0.0229, 0.0226], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-01 22:11:16,972 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=243211.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 22:11:37,658 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=4.37 vs. limit=5.0 2023-05-01 22:11:48,988 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([1.9767, 1.8301, 1.6469, 1.4743, 1.9790, 1.6314, 1.5630, 1.9248], device='cuda:1'), covar=tensor([0.0193, 0.0350, 0.0465, 0.0437, 0.0275, 0.0336, 0.0181, 0.0239], device='cuda:1'), in_proj_covar=tensor([0.0209, 0.0233, 0.0224, 0.0225, 0.0234, 0.0233, 0.0229, 0.0228], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-01 22:12:03,356 INFO [optim.py:368] (1/8) 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 22:12:37,392 INFO [train.py:904] (1/8) Epoch 24, batch 9800, loss[loss=0.1615, simple_loss=0.264, pruned_loss=0.02955, over 16741.00 frames. ], tot_loss[loss=0.1642, simple_loss=0.2596, pruned_loss=0.03436, over 3077170.76 frames. ], batch size: 134, lr: 2.78e-03, grad_scale: 4.0 2023-05-01 22:12:49,021 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=243258.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 22:12:50,361 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=243259.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 22:13:16,637 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.4109, 2.8972, 3.0912, 1.9556, 2.7607, 2.1637, 2.9067, 3.1126], device='cuda:1'), covar=tensor([0.0369, 0.0925, 0.0585, 0.2124, 0.0910, 0.1044, 0.0879, 0.1043], device='cuda:1'), in_proj_covar=tensor([0.0153, 0.0158, 0.0164, 0.0151, 0.0142, 0.0127, 0.0140, 0.0170], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:1') 2023-05-01 22:13:22,181 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=243277.0, num_to_drop=1, layers_to_drop={2} 2023-05-01 22:13:58,451 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=243293.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 22:14:16,415 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.71 vs. limit=2.0 2023-05-01 22:14:21,694 INFO [train.py:904] (1/8) Epoch 24, batch 9850, loss[loss=0.1802, simple_loss=0.2723, pruned_loss=0.04407, over 16639.00 frames. ], tot_loss[loss=0.1644, simple_loss=0.2603, pruned_loss=0.03422, over 3071749.90 frames. ], batch size: 134, lr: 2.78e-03, grad_scale: 4.0 2023-05-01 22:14:28,522 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=243306.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 22:14:28,680 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([1.9709, 2.7917, 2.6491, 1.9844, 2.5810, 2.8116, 2.6591, 2.0129], device='cuda:1'), covar=tensor([0.0429, 0.0072, 0.0081, 0.0366, 0.0147, 0.0099, 0.0105, 0.0411], device='cuda:1'), in_proj_covar=tensor([0.0134, 0.0084, 0.0084, 0.0132, 0.0097, 0.0107, 0.0093, 0.0128], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-05-01 22:15:22,935 INFO [optim.py:368] (1/8) Clipping_scale=2.0, grad-norm quartiles 1.302e+02 1.939e+02 2.274e+02 2.876e+02 6.181e+02, threshold=4.547e+02, percent-clipped=2.0 2023-05-01 22:16:11,623 INFO [train.py:904] (1/8) Epoch 24, batch 9900, loss[loss=0.165, simple_loss=0.2677, pruned_loss=0.03116, over 16928.00 frames. ], tot_loss[loss=0.1649, simple_loss=0.2609, pruned_loss=0.03439, over 3056598.22 frames. ], batch size: 116, lr: 2.78e-03, grad_scale: 4.0 2023-05-01 22:16:15,151 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=243354.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 22:18:09,536 INFO [train.py:904] (1/8) Epoch 24, batch 9950, loss[loss=0.1679, simple_loss=0.2695, pruned_loss=0.03312, over 15489.00 frames. ], tot_loss[loss=0.166, simple_loss=0.2625, pruned_loss=0.03476, over 3044259.42 frames. ], batch size: 191, lr: 2.78e-03, grad_scale: 4.0 2023-05-01 22:19:11,956 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=243428.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 22:19:24,198 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.21 vs. limit=2.0 2023-05-01 22:19:24,674 INFO [optim.py:368] (1/8) 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 22:19:47,273 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.8562, 3.9102, 4.0358, 3.7574, 3.9211, 4.3292, 3.9934, 3.6597], device='cuda:1'), covar=tensor([0.2022, 0.2164, 0.2886, 0.2398, 0.2799, 0.1653, 0.1699, 0.2680], device='cuda:1'), in_proj_covar=tensor([0.0386, 0.0576, 0.0634, 0.0472, 0.0623, 0.0658, 0.0493, 0.0625], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-05-01 22:20:10,933 INFO [train.py:904] (1/8) Epoch 24, batch 10000, loss[loss=0.1764, simple_loss=0.2755, pruned_loss=0.03867, over 16251.00 frames. ], tot_loss[loss=0.1651, simple_loss=0.2612, pruned_loss=0.03446, over 3050380.24 frames. ], batch size: 165, lr: 2.78e-03, grad_scale: 8.0 2023-05-01 22:20:25,288 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.8088, 5.0653, 4.9397, 4.8935, 4.6462, 4.6451, 4.4075, 5.1472], device='cuda:1'), covar=tensor([0.1079, 0.0812, 0.0824, 0.0781, 0.0721, 0.0954, 0.1162, 0.0825], device='cuda:1'), in_proj_covar=tensor([0.0670, 0.0808, 0.0664, 0.0624, 0.0512, 0.0519, 0.0677, 0.0632], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:1') 2023-05-01 22:20:55,209 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 24, batch 10050, loss[loss=0.1612, simple_loss=0.2525, pruned_loss=0.03492, over 12237.00 frames. ], tot_loss[loss=0.1651, simple_loss=0.2615, pruned_loss=0.0344, over 3066104.91 frames. ], batch size: 248, lr: 2.78e-03, grad_scale: 8.0 2023-05-01 22:22:52,097 INFO [optim.py:368] (1/8) 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 22:23:23,048 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([1.9484, 2.8101, 2.7030, 2.0305, 2.5885, 2.8340, 2.6902, 1.9900], device='cuda:1'), covar=tensor([0.0489, 0.0086, 0.0080, 0.0379, 0.0151, 0.0106, 0.0105, 0.0453], device='cuda:1'), in_proj_covar=tensor([0.0134, 0.0084, 0.0084, 0.0131, 0.0097, 0.0107, 0.0093, 0.0128], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-05-01 22:23:25,157 INFO [train.py:904] (1/8) Epoch 24, batch 10100, loss[loss=0.157, simple_loss=0.239, pruned_loss=0.03752, over 12272.00 frames. ], tot_loss[loss=0.1654, simple_loss=0.2618, pruned_loss=0.03452, over 3075502.66 frames. ], batch size: 248, lr: 2.78e-03, grad_scale: 8.0 2023-05-01 22:24:15,455 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=243577.0, num_to_drop=1, layers_to_drop={2} 2023-05-01 22:25:09,202 INFO [train.py:904] (1/8) Epoch 25, batch 0, loss[loss=0.2214, simple_loss=0.2887, pruned_loss=0.07704, over 16702.00 frames. ], tot_loss[loss=0.2214, simple_loss=0.2887, pruned_loss=0.07704, over 16702.00 frames. ], batch size: 76, lr: 2.72e-03, grad_scale: 8.0 2023-05-01 22:25:09,202 INFO [train.py:929] (1/8) Computing validation loss 2023-05-01 22:25:16,827 INFO [train.py:938] (1/8) Epoch 25, validation: loss=0.1443, simple_loss=0.2477, pruned_loss=0.02048, over 944034.00 frames. 2023-05-01 22:25:16,827 INFO [train.py:939] (1/8) Maximum memory allocated so far is 18145MB 2023-05-01 22:25:48,589 INFO [zipformer.py:625] (1/8) 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] (1/8) 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 22:26:21,442 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=243649.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 22:26:26,906 INFO [train.py:904] (1/8) Epoch 25, batch 50, loss[loss=0.1594, simple_loss=0.2514, pruned_loss=0.03366, over 17171.00 frames. ], tot_loss[loss=0.1797, simple_loss=0.2651, pruned_loss=0.0472, over 760883.11 frames. ], batch size: 46, lr: 2.72e-03, grad_scale: 2.0 2023-05-01 22:26:42,446 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.26 vs. limit=2.0 2023-05-01 22:27:35,800 INFO [train.py:904] (1/8) Epoch 25, batch 100, loss[loss=0.2075, simple_loss=0.2827, pruned_loss=0.06614, over 16704.00 frames. ], tot_loss[loss=0.178, simple_loss=0.2638, pruned_loss=0.04605, over 1329814.23 frames. ], batch size: 134, lr: 2.72e-03, grad_scale: 2.0 2023-05-01 22:27:45,332 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.9645, 4.7038, 4.9991, 5.1766, 5.3929, 4.7623, 5.2930, 5.3523], device='cuda:1'), covar=tensor([0.2065, 0.1403, 0.1938, 0.0859, 0.0584, 0.0869, 0.0652, 0.0710], device='cuda:1'), in_proj_covar=tensor([0.0624, 0.0769, 0.0883, 0.0779, 0.0597, 0.0613, 0.0646, 0.0747], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-01 22:28:22,078 INFO [optim.py:368] (1/8) 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] (1/8) Epoch 25, batch 150, loss[loss=0.1663, simple_loss=0.2514, pruned_loss=0.04066, over 16798.00 frames. ], tot_loss[loss=0.1755, simple_loss=0.2625, pruned_loss=0.0442, over 1776646.44 frames. ], batch size: 102, lr: 2.72e-03, grad_scale: 2.0 2023-05-01 22:28:50,449 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.7310, 4.5791, 4.6105, 4.2835, 4.3533, 4.6469, 4.5216, 4.3988], device='cuda:1'), covar=tensor([0.0641, 0.0957, 0.0349, 0.0344, 0.0912, 0.0544, 0.0419, 0.0680], device='cuda:1'), in_proj_covar=tensor([0.0288, 0.0430, 0.0334, 0.0336, 0.0337, 0.0387, 0.0230, 0.0403], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-01 22:29:55,574 INFO [train.py:904] (1/8) Epoch 25, batch 200, loss[loss=0.184, simple_loss=0.2688, pruned_loss=0.04961, over 16940.00 frames. ], tot_loss[loss=0.1753, simple_loss=0.262, pruned_loss=0.04425, over 2112372.27 frames. ], batch size: 116, lr: 2.72e-03, grad_scale: 2.0 2023-05-01 22:30:40,583 INFO [optim.py:368] (1/8) 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] (1/8) Epoch 25, batch 250, loss[loss=0.1574, simple_loss=0.2526, pruned_loss=0.03113, over 17208.00 frames. ], tot_loss[loss=0.1733, simple_loss=0.2598, pruned_loss=0.04339, over 2381225.51 frames. ], batch size: 44, lr: 2.72e-03, grad_scale: 2.0 2023-05-01 22:31:22,901 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.3389, 3.5223, 4.0034, 2.1711, 3.1737, 2.4967, 3.7985, 3.6395], device='cuda:1'), covar=tensor([0.0303, 0.0983, 0.0496, 0.2145, 0.0838, 0.1025, 0.0670, 0.1197], device='cuda:1'), in_proj_covar=tensor([0.0155, 0.0161, 0.0166, 0.0152, 0.0143, 0.0129, 0.0142, 0.0173], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:1') 2023-05-01 22:32:07,992 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([1.9955, 2.2707, 2.3221, 2.7634, 2.1013, 3.2698, 1.7862, 2.6764], device='cuda:1'), covar=tensor([0.1112, 0.0730, 0.1123, 0.0206, 0.0146, 0.0475, 0.1427, 0.0751], device='cuda:1'), in_proj_covar=tensor([0.0170, 0.0176, 0.0196, 0.0191, 0.0200, 0.0214, 0.0205, 0.0195], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-05-01 22:32:14,684 INFO [train.py:904] (1/8) Epoch 25, batch 300, loss[loss=0.1566, simple_loss=0.239, pruned_loss=0.03706, over 15429.00 frames. ], tot_loss[loss=0.1706, simple_loss=0.2575, pruned_loss=0.04182, over 2585794.38 frames. ], batch size: 190, lr: 2.72e-03, grad_scale: 2.0 2023-05-01 22:33:00,866 INFO [optim.py:368] (1/8) 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 22:33:19,535 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=243949.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 22:33:24,546 INFO [train.py:904] (1/8) Epoch 25, batch 350, loss[loss=0.1704, simple_loss=0.2539, pruned_loss=0.04346, over 16699.00 frames. ], tot_loss[loss=0.1685, simple_loss=0.2547, pruned_loss=0.04114, over 2747964.93 frames. ], batch size: 57, lr: 2.72e-03, grad_scale: 2.0 2023-05-01 22:33:54,456 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.0504, 3.2061, 2.9712, 5.2583, 4.3795, 4.5827, 1.9120, 3.3707], device='cuda:1'), covar=tensor([0.1315, 0.0722, 0.1130, 0.0175, 0.0192, 0.0402, 0.1587, 0.0697], device='cuda:1'), in_proj_covar=tensor([0.0169, 0.0176, 0.0195, 0.0191, 0.0200, 0.0213, 0.0204, 0.0194], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-05-01 22:34:00,968 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.7952, 3.9060, 2.5044, 4.4296, 3.0088, 4.3808, 2.5290, 3.2389], device='cuda:1'), covar=tensor([0.0343, 0.0415, 0.1606, 0.0353, 0.0984, 0.0628, 0.1573, 0.0832], device='cuda:1'), in_proj_covar=tensor([0.0172, 0.0177, 0.0194, 0.0167, 0.0178, 0.0216, 0.0204, 0.0182], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-05-01 22:34:04,695 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.6364, 3.5500, 4.3699, 2.3153, 3.4208, 2.6613, 4.0282, 3.7989], device='cuda:1'), covar=tensor([0.0235, 0.0998, 0.0404, 0.2004, 0.0802, 0.0991, 0.0565, 0.1103], device='cuda:1'), in_proj_covar=tensor([0.0156, 0.0162, 0.0167, 0.0153, 0.0144, 0.0129, 0.0142, 0.0174], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:1') 2023-05-01 22:34:25,843 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 25, batch 400, loss[loss=0.136, simple_loss=0.2192, pruned_loss=0.02646, over 16772.00 frames. ], tot_loss[loss=0.1667, simple_loss=0.2527, pruned_loss=0.04031, over 2874830.57 frames. ], batch size: 39, lr: 2.72e-03, grad_scale: 4.0 2023-05-01 22:34:50,905 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.1430, 3.0414, 2.8670, 5.1506, 4.2833, 4.3028, 1.9217, 3.1842], device='cuda:1'), covar=tensor([0.1216, 0.0760, 0.1135, 0.0194, 0.0235, 0.0561, 0.1511, 0.0808], device='cuda:1'), in_proj_covar=tensor([0.0169, 0.0176, 0.0195, 0.0191, 0.0200, 0.0213, 0.0204, 0.0194], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-05-01 22:34:54,318 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.8598, 4.2818, 3.0734, 2.3693, 2.6225, 2.6102, 4.6380, 3.4924], device='cuda:1'), covar=tensor([0.2950, 0.0639, 0.1868, 0.3068, 0.3206, 0.2201, 0.0362, 0.1562], device='cuda:1'), in_proj_covar=tensor([0.0327, 0.0268, 0.0305, 0.0315, 0.0294, 0.0265, 0.0295, 0.0339], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:1') 2023-05-01 22:35:02,312 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.60 vs. limit=2.0 2023-05-01 22:35:22,991 INFO [optim.py:368] (1/8) 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 22:35:42,262 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=4.13 vs. limit=5.0 2023-05-01 22:35:47,033 INFO [train.py:904] (1/8) Epoch 25, batch 450, loss[loss=0.1682, simple_loss=0.2608, pruned_loss=0.03778, over 17054.00 frames. ], tot_loss[loss=0.1653, simple_loss=0.2515, pruned_loss=0.03953, over 2973437.50 frames. ], batch size: 53, lr: 2.72e-03, grad_scale: 4.0 2023-05-01 22:36:50,467 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.9995, 4.9871, 5.4255, 5.4112, 5.4263, 5.1456, 5.0417, 4.8644], device='cuda:1'), covar=tensor([0.0353, 0.0624, 0.0377, 0.0391, 0.0477, 0.0370, 0.0913, 0.0466], device='cuda:1'), in_proj_covar=tensor([0.0411, 0.0459, 0.0447, 0.0412, 0.0491, 0.0470, 0.0546, 0.0375], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:1') 2023-05-01 22:36:55,203 INFO [train.py:904] (1/8) Epoch 25, batch 500, loss[loss=0.1705, simple_loss=0.2509, pruned_loss=0.04506, over 16845.00 frames. ], tot_loss[loss=0.1646, simple_loss=0.2511, pruned_loss=0.03903, over 3058550.40 frames. ], batch size: 102, lr: 2.72e-03, grad_scale: 4.0 2023-05-01 22:37:18,503 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.4857, 3.3429, 3.5793, 2.5799, 3.3723, 3.7282, 3.4526, 2.2714], device='cuda:1'), covar=tensor([0.0499, 0.0168, 0.0074, 0.0407, 0.0117, 0.0111, 0.0106, 0.0469], device='cuda:1'), in_proj_covar=tensor([0.0138, 0.0088, 0.0087, 0.0136, 0.0100, 0.0111, 0.0096, 0.0132], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0003, 0.0004], device='cuda:1') 2023-05-01 22:37:42,041 INFO [optim.py:368] (1/8) 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 22:38:05,730 INFO [train.py:904] (1/8) Epoch 25, batch 550, loss[loss=0.1687, simple_loss=0.2695, pruned_loss=0.03394, over 16646.00 frames. ], tot_loss[loss=0.1638, simple_loss=0.2505, pruned_loss=0.0385, over 3113423.83 frames. ], batch size: 57, lr: 2.72e-03, grad_scale: 4.0 2023-05-01 22:38:12,386 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.3748, 3.4569, 4.0941, 2.3935, 3.2440, 2.5215, 3.8539, 3.7387], device='cuda:1'), covar=tensor([0.0293, 0.1018, 0.0451, 0.1955, 0.0818, 0.1011, 0.0588, 0.1131], device='cuda:1'), in_proj_covar=tensor([0.0157, 0.0164, 0.0168, 0.0154, 0.0144, 0.0130, 0.0143, 0.0175], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:1') 2023-05-01 22:38:46,647 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([5.2381, 5.2350, 5.0787, 4.5311, 4.6942, 5.1492, 5.0182, 4.6774], device='cuda:1'), covar=tensor([0.0681, 0.0534, 0.0390, 0.0437, 0.1307, 0.0496, 0.0410, 0.0944], device='cuda:1'), in_proj_covar=tensor([0.0299, 0.0446, 0.0346, 0.0349, 0.0350, 0.0402, 0.0239, 0.0418], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-01 22:39:05,126 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.8257, 2.5117, 2.1337, 2.2896, 2.8965, 2.6567, 2.8945, 2.9604], device='cuda:1'), covar=tensor([0.0365, 0.0429, 0.0551, 0.0539, 0.0252, 0.0380, 0.0269, 0.0319], device='cuda:1'), in_proj_covar=tensor([0.0222, 0.0243, 0.0232, 0.0233, 0.0244, 0.0242, 0.0242, 0.0238], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-01 22:39:15,773 INFO [train.py:904] (1/8) Epoch 25, batch 600, loss[loss=0.1559, simple_loss=0.2581, pruned_loss=0.02689, over 17016.00 frames. ], tot_loss[loss=0.1635, simple_loss=0.2502, pruned_loss=0.03847, over 3163026.51 frames. ], batch size: 50, lr: 2.72e-03, grad_scale: 2.0 2023-05-01 22:39:24,756 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.5980, 4.5652, 4.9418, 4.9314, 4.9742, 4.6867, 4.6633, 4.5340], device='cuda:1'), covar=tensor([0.0442, 0.0915, 0.0529, 0.0542, 0.0514, 0.0509, 0.0926, 0.0613], device='cuda:1'), in_proj_covar=tensor([0.0415, 0.0462, 0.0450, 0.0416, 0.0494, 0.0474, 0.0549, 0.0378], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:1') 2023-05-01 22:39:38,208 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.68 vs. limit=2.0 2023-05-01 22:40:02,986 INFO [optim.py:368] (1/8) 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] (1/8) Epoch 25, batch 650, loss[loss=0.1578, simple_loss=0.2557, pruned_loss=0.02998, over 17191.00 frames. ], tot_loss[loss=0.1631, simple_loss=0.2485, pruned_loss=0.03883, over 3204745.02 frames. ], batch size: 46, lr: 2.72e-03, grad_scale: 2.0 2023-05-01 22:40:32,774 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.1983, 2.2129, 2.3765, 3.8784, 2.2514, 2.5783, 2.2795, 2.3941], device='cuda:1'), covar=tensor([0.1531, 0.3902, 0.3130, 0.0720, 0.3933, 0.2646, 0.4032, 0.3316], device='cuda:1'), in_proj_covar=tensor([0.0411, 0.0459, 0.0378, 0.0332, 0.0443, 0.0524, 0.0432, 0.0537], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-01 22:40:53,573 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.8602, 4.0119, 2.8726, 4.5960, 3.1986, 4.5169, 2.8660, 3.3941], device='cuda:1'), covar=tensor([0.0355, 0.0391, 0.1371, 0.0291, 0.0855, 0.0529, 0.1400, 0.0711], device='cuda:1'), in_proj_covar=tensor([0.0173, 0.0179, 0.0195, 0.0169, 0.0179, 0.0218, 0.0205, 0.0183], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-05-01 22:40:53,617 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([1.9117, 2.9650, 2.8288, 5.0682, 4.0529, 4.4780, 1.6973, 3.1803], device='cuda:1'), covar=tensor([0.1378, 0.0856, 0.1233, 0.0210, 0.0259, 0.0393, 0.1728, 0.0825], device='cuda:1'), in_proj_covar=tensor([0.0168, 0.0175, 0.0194, 0.0191, 0.0199, 0.0213, 0.0204, 0.0194], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-05-01 22:41:08,448 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.6791, 3.7805, 2.6521, 4.3088, 2.9574, 4.2730, 2.6265, 3.1520], device='cuda:1'), covar=tensor([0.0359, 0.0418, 0.1487, 0.0383, 0.0919, 0.0614, 0.1484, 0.0801], device='cuda:1'), in_proj_covar=tensor([0.0173, 0.0179, 0.0195, 0.0169, 0.0179, 0.0218, 0.0205, 0.0183], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-05-01 22:41:33,625 INFO [train.py:904] (1/8) Epoch 25, batch 700, loss[loss=0.1508, simple_loss=0.2485, pruned_loss=0.02656, over 17258.00 frames. ], tot_loss[loss=0.163, simple_loss=0.2486, pruned_loss=0.03869, over 3235836.69 frames. ], batch size: 52, lr: 2.72e-03, grad_scale: 2.0 2023-05-01 22:41:50,849 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.8718, 5.1533, 5.2557, 5.0350, 5.0884, 5.7273, 5.2028, 4.9113], device='cuda:1'), covar=tensor([0.1298, 0.2094, 0.2846, 0.2093, 0.2652, 0.1014, 0.1828, 0.2541], device='cuda:1'), in_proj_covar=tensor([0.0413, 0.0612, 0.0677, 0.0502, 0.0665, 0.0698, 0.0525, 0.0665], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-05-01 22:41:55,070 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.3940, 4.4905, 4.5747, 4.3837, 4.4714, 5.0325, 4.5469, 4.2178], device='cuda:1'), covar=tensor([0.1658, 0.2281, 0.2935, 0.2368, 0.3007, 0.1211, 0.1845, 0.2780], device='cuda:1'), in_proj_covar=tensor([0.0413, 0.0613, 0.0677, 0.0503, 0.0666, 0.0699, 0.0525, 0.0665], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-05-01 22:42:16,526 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([5.5223, 5.9289, 5.6945, 5.7113, 5.3054, 5.3621, 5.2997, 6.0441], device='cuda:1'), covar=tensor([0.1464, 0.1024, 0.1038, 0.0921, 0.0955, 0.0657, 0.1240, 0.0929], device='cuda:1'), in_proj_covar=tensor([0.0700, 0.0845, 0.0692, 0.0651, 0.0535, 0.0541, 0.0710, 0.0661], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:1') 2023-05-01 22:42:20,953 INFO [optim.py:368] (1/8) 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] (1/8) Epoch 25, batch 750, loss[loss=0.1599, simple_loss=0.2536, pruned_loss=0.0331, over 17072.00 frames. ], tot_loss[loss=0.1638, simple_loss=0.2493, pruned_loss=0.03914, over 3250897.95 frames. ], batch size: 55, lr: 2.72e-03, grad_scale: 2.0 2023-05-01 22:42:42,745 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=244353.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 22:43:11,797 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.4480, 3.7021, 4.1334, 2.1984, 3.2845, 2.6449, 3.9273, 3.8280], device='cuda:1'), covar=tensor([0.0273, 0.0960, 0.0423, 0.2135, 0.0794, 0.0964, 0.0586, 0.1110], device='cuda:1'), in_proj_covar=tensor([0.0158, 0.0165, 0.0169, 0.0155, 0.0146, 0.0131, 0.0144, 0.0177], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-05-01 22:43:34,925 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([5.4431, 5.3864, 5.2477, 4.6126, 4.8543, 5.3065, 5.2553, 4.8351], device='cuda:1'), covar=tensor([0.0575, 0.0507, 0.0344, 0.0413, 0.1218, 0.0443, 0.0254, 0.0869], device='cuda:1'), in_proj_covar=tensor([0.0303, 0.0451, 0.0351, 0.0353, 0.0355, 0.0407, 0.0242, 0.0424], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002], device='cuda:1') 2023-05-01 22:43:52,150 INFO [train.py:904] (1/8) Epoch 25, batch 800, loss[loss=0.1458, simple_loss=0.2396, pruned_loss=0.02604, over 17194.00 frames. ], tot_loss[loss=0.163, simple_loss=0.2488, pruned_loss=0.03863, over 3263059.46 frames. ], batch size: 46, lr: 2.72e-03, grad_scale: 4.0 2023-05-01 22:44:08,203 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=244414.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 22:44:13,692 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.9782, 3.7775, 4.2010, 2.2559, 4.3629, 4.4385, 3.2870, 3.5854], device='cuda:1'), covar=tensor([0.0737, 0.0272, 0.0232, 0.1183, 0.0091, 0.0220, 0.0454, 0.0371], device='cuda:1'), in_proj_covar=tensor([0.0148, 0.0111, 0.0100, 0.0140, 0.0083, 0.0130, 0.0129, 0.0131], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-05-01 22:44:39,341 INFO [optim.py:368] (1/8) 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] (1/8) Epoch 25, batch 850, loss[loss=0.1746, simple_loss=0.2511, pruned_loss=0.04903, over 12549.00 frames. ], tot_loss[loss=0.1624, simple_loss=0.2481, pruned_loss=0.03836, over 3270745.04 frames. ], batch size: 246, lr: 2.72e-03, grad_scale: 4.0 2023-05-01 22:46:12,665 INFO [train.py:904] (1/8) Epoch 25, batch 900, loss[loss=0.1573, simple_loss=0.2503, pruned_loss=0.03214, over 16745.00 frames. ], tot_loss[loss=0.1614, simple_loss=0.2475, pruned_loss=0.03765, over 3282568.72 frames. ], batch size: 62, lr: 2.72e-03, grad_scale: 4.0 2023-05-01 22:46:34,975 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([5.3750, 5.3728, 5.1342, 4.5993, 5.2141, 2.1370, 4.9606, 5.0885], device='cuda:1'), covar=tensor([0.0099, 0.0107, 0.0240, 0.0421, 0.0116, 0.2711, 0.0149, 0.0197], device='cuda:1'), in_proj_covar=tensor([0.0173, 0.0166, 0.0205, 0.0180, 0.0182, 0.0213, 0.0194, 0.0176], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-01 22:47:00,724 INFO [optim.py:368] (1/8) 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 22:47:03,372 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.39 vs. limit=2.0 2023-05-01 22:47:23,394 INFO [train.py:904] (1/8) Epoch 25, batch 950, loss[loss=0.1783, simple_loss=0.2493, pruned_loss=0.05369, over 16474.00 frames. ], tot_loss[loss=0.1619, simple_loss=0.2479, pruned_loss=0.03797, over 3293799.12 frames. ], batch size: 146, lr: 2.72e-03, grad_scale: 4.0 2023-05-01 22:47:44,343 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.10 vs. limit=2.0 2023-05-01 22:48:33,897 INFO [train.py:904] (1/8) Epoch 25, batch 1000, loss[loss=0.1592, simple_loss=0.2342, pruned_loss=0.04206, over 16861.00 frames. ], tot_loss[loss=0.1618, simple_loss=0.2469, pruned_loss=0.03838, over 3297252.27 frames. ], batch size: 102, lr: 2.72e-03, grad_scale: 4.0 2023-05-01 22:49:21,005 INFO [optim.py:368] (1/8) 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 22:49:42,534 INFO [train.py:904] (1/8) Epoch 25, batch 1050, loss[loss=0.1807, simple_loss=0.2672, pruned_loss=0.04714, over 16448.00 frames. ], tot_loss[loss=0.1615, simple_loss=0.2465, pruned_loss=0.03819, over 3300136.45 frames. ], batch size: 68, lr: 2.72e-03, grad_scale: 4.0 2023-05-01 22:50:53,099 INFO [train.py:904] (1/8) Epoch 25, batch 1100, loss[loss=0.1527, simple_loss=0.2481, pruned_loss=0.02871, over 17164.00 frames. ], tot_loss[loss=0.1609, simple_loss=0.2462, pruned_loss=0.03784, over 3315369.96 frames. ], batch size: 47, lr: 2.71e-03, grad_scale: 4.0 2023-05-01 22:51:01,208 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=244709.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 22:51:13,214 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=244717.0, num_to_drop=1, layers_to_drop={0} 2023-05-01 22:51:14,454 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.9695, 4.5467, 3.2018, 2.4653, 2.8613, 2.7615, 4.8643, 3.7307], device='cuda:1'), covar=tensor([0.2819, 0.0495, 0.1821, 0.2998, 0.2850, 0.2027, 0.0331, 0.1372], device='cuda:1'), in_proj_covar=tensor([0.0330, 0.0271, 0.0309, 0.0318, 0.0299, 0.0268, 0.0299, 0.0344], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:1') 2023-05-01 22:51:40,296 INFO [optim.py:368] (1/8) 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 22:52:02,019 INFO [train.py:904] (1/8) Epoch 25, batch 1150, loss[loss=0.1685, simple_loss=0.2521, pruned_loss=0.04245, over 17053.00 frames. ], tot_loss[loss=0.1607, simple_loss=0.2459, pruned_loss=0.03782, over 3308678.70 frames. ], batch size: 53, lr: 2.71e-03, grad_scale: 4.0 2023-05-01 22:52:37,217 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=244778.0, num_to_drop=1, layers_to_drop={1} 2023-05-01 22:53:11,757 INFO [train.py:904] (1/8) Epoch 25, batch 1200, loss[loss=0.1605, simple_loss=0.2529, pruned_loss=0.03406, over 17080.00 frames. ], tot_loss[loss=0.1596, simple_loss=0.2449, pruned_loss=0.03716, over 3315632.45 frames. ], batch size: 53, lr: 2.71e-03, grad_scale: 8.0 2023-05-01 22:53:57,429 INFO [optim.py:368] (1/8) 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] (1/8) Epoch 25, batch 1250, loss[loss=0.1443, simple_loss=0.2339, pruned_loss=0.02735, over 17201.00 frames. ], tot_loss[loss=0.1595, simple_loss=0.2445, pruned_loss=0.03721, over 3310294.55 frames. ], batch size: 44, lr: 2.71e-03, grad_scale: 8.0 2023-05-01 22:55:06,305 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=244887.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 22:55:12,315 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.5219, 2.3819, 2.3323, 4.3572, 2.4050, 2.8082, 2.4467, 2.5776], device='cuda:1'), covar=tensor([0.1425, 0.3730, 0.3374, 0.0545, 0.4264, 0.2572, 0.3803, 0.3666], device='cuda:1'), in_proj_covar=tensor([0.0416, 0.0464, 0.0382, 0.0336, 0.0446, 0.0531, 0.0436, 0.0543], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-01 22:55:26,306 INFO [train.py:904] (1/8) Epoch 25, batch 1300, loss[loss=0.1793, simple_loss=0.2701, pruned_loss=0.04426, over 17042.00 frames. ], tot_loss[loss=0.1596, simple_loss=0.2449, pruned_loss=0.03711, over 3309493.40 frames. ], batch size: 55, lr: 2.71e-03, grad_scale: 8.0 2023-05-01 22:55:51,384 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.45 vs. limit=2.0 2023-05-01 22:56:12,137 INFO [optim.py:368] (1/8) 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 22:56:27,606 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 25, batch 1350, loss[loss=0.1746, simple_loss=0.2545, pruned_loss=0.04735, over 16708.00 frames. ], tot_loss[loss=0.1602, simple_loss=0.2459, pruned_loss=0.03726, over 3321768.88 frames. ], batch size: 134, lr: 2.71e-03, grad_scale: 8.0 2023-05-01 22:56:36,711 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=244954.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 22:57:43,144 INFO [train.py:904] (1/8) Epoch 25, batch 1400, loss[loss=0.1467, simple_loss=0.2233, pruned_loss=0.03507, over 12173.00 frames. ], tot_loss[loss=0.1605, simple_loss=0.2459, pruned_loss=0.03752, over 3312781.77 frames. ], batch size: 247, lr: 2.71e-03, grad_scale: 8.0 2023-05-01 22:57:51,766 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=245009.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 22:57:59,976 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=245015.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 22:58:28,648 INFO [optim.py:368] (1/8) 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 22:58:51,281 INFO [train.py:904] (1/8) Epoch 25, batch 1450, loss[loss=0.1455, simple_loss=0.2398, pruned_loss=0.02555, over 17200.00 frames. ], tot_loss[loss=0.1595, simple_loss=0.2448, pruned_loss=0.03714, over 3319812.02 frames. ], batch size: 44, lr: 2.71e-03, grad_scale: 8.0 2023-05-01 22:58:56,915 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=245057.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 22:59:00,654 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.6686, 3.2438, 3.7412, 2.1042, 3.7538, 3.8249, 3.1964, 2.8390], device='cuda:1'), covar=tensor([0.0726, 0.0300, 0.0202, 0.1087, 0.0136, 0.0212, 0.0378, 0.0455], device='cuda:1'), in_proj_covar=tensor([0.0148, 0.0111, 0.0100, 0.0139, 0.0084, 0.0130, 0.0129, 0.0131], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-05-01 22:59:01,003 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.34 vs. limit=2.0 2023-05-01 22:59:18,132 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=245073.0, num_to_drop=1, layers_to_drop={2} 2023-05-01 22:59:19,998 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([5.0084, 5.3230, 5.1015, 5.0454, 4.7760, 4.8252, 4.7336, 5.4057], device='cuda:1'), covar=tensor([0.1203, 0.0988, 0.1115, 0.0944, 0.0887, 0.0979, 0.1238, 0.0905], device='cuda:1'), in_proj_covar=tensor([0.0708, 0.0859, 0.0702, 0.0663, 0.0544, 0.0548, 0.0721, 0.0671], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:1') 2023-05-01 23:00:00,082 INFO [train.py:904] (1/8) Epoch 25, batch 1500, loss[loss=0.1793, simple_loss=0.256, pruned_loss=0.0513, over 16302.00 frames. ], tot_loss[loss=0.1599, simple_loss=0.2448, pruned_loss=0.03752, over 3310232.35 frames. ], batch size: 165, lr: 2.71e-03, grad_scale: 8.0 2023-05-01 23:00:46,573 INFO [optim.py:368] (1/8) 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 23:01:08,932 INFO [train.py:904] (1/8) Epoch 25, batch 1550, loss[loss=0.1847, simple_loss=0.2781, pruned_loss=0.04561, over 17051.00 frames. ], tot_loss[loss=0.1615, simple_loss=0.2462, pruned_loss=0.03837, over 3311789.40 frames. ], batch size: 55, lr: 2.71e-03, grad_scale: 8.0 2023-05-01 23:01:34,909 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.76 vs. limit=2.0 2023-05-01 23:02:19,885 INFO [train.py:904] (1/8) Epoch 25, batch 1600, loss[loss=0.1489, simple_loss=0.2435, pruned_loss=0.02715, over 17168.00 frames. ], tot_loss[loss=0.163, simple_loss=0.2485, pruned_loss=0.03874, over 3301254.02 frames. ], batch size: 46, lr: 2.71e-03, grad_scale: 8.0 2023-05-01 23:03:06,948 INFO [optim.py:368] (1/8) 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 23:03:09,128 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([5.0564, 4.7607, 5.0770, 5.2914, 5.5039, 4.7921, 5.4708, 5.4738], device='cuda:1'), covar=tensor([0.1930, 0.1528, 0.1847, 0.0864, 0.0553, 0.0914, 0.0517, 0.0707], device='cuda:1'), in_proj_covar=tensor([0.0678, 0.0836, 0.0963, 0.0846, 0.0644, 0.0662, 0.0694, 0.0812], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-01 23:03:16,801 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=245243.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 23:03:28,986 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.9276, 5.0409, 5.4705, 5.4180, 5.4338, 5.1042, 5.0247, 4.8837], device='cuda:1'), covar=tensor([0.0417, 0.0531, 0.0358, 0.0441, 0.0603, 0.0444, 0.1031, 0.0451], device='cuda:1'), in_proj_covar=tensor([0.0432, 0.0481, 0.0467, 0.0432, 0.0514, 0.0494, 0.0571, 0.0390], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:1') 2023-05-01 23:03:29,752 INFO [train.py:904] (1/8) Epoch 25, batch 1650, loss[loss=0.1543, simple_loss=0.2496, pruned_loss=0.02947, over 17116.00 frames. ], tot_loss[loss=0.1641, simple_loss=0.2494, pruned_loss=0.0394, over 3306059.76 frames. ], batch size: 49, lr: 2.71e-03, grad_scale: 8.0 2023-05-01 23:04:27,462 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.6133, 2.5088, 2.1593, 2.4190, 2.8905, 2.7024, 3.2572, 3.2196], device='cuda:1'), covar=tensor([0.0182, 0.0537, 0.0723, 0.0601, 0.0363, 0.0461, 0.0258, 0.0311], device='cuda:1'), in_proj_covar=tensor([0.0228, 0.0246, 0.0235, 0.0236, 0.0248, 0.0246, 0.0247, 0.0243], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-01 23:04:41,929 INFO [train.py:904] (1/8) Epoch 25, batch 1700, loss[loss=0.1742, simple_loss=0.2669, pruned_loss=0.04079, over 17184.00 frames. ], tot_loss[loss=0.1662, simple_loss=0.2521, pruned_loss=0.04019, over 3297265.06 frames. ], batch size: 46, lr: 2.71e-03, grad_scale: 8.0 2023-05-01 23:04:51,855 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=245310.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 23:04:53,433 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=3.37 vs. limit=5.0 2023-05-01 23:04:59,332 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.3453, 3.5011, 3.8555, 2.1070, 3.1357, 2.5470, 3.6795, 3.7059], device='cuda:1'), covar=tensor([0.0262, 0.0921, 0.0512, 0.2124, 0.0862, 0.0967, 0.0649, 0.1074], device='cuda:1'), in_proj_covar=tensor([0.0158, 0.0166, 0.0169, 0.0155, 0.0146, 0.0131, 0.0144, 0.0178], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:1') 2023-05-01 23:05:07,890 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.2940, 1.6597, 2.0544, 2.1766, 2.3407, 2.3235, 1.8182, 2.3839], device='cuda:1'), covar=tensor([0.0232, 0.0508, 0.0281, 0.0311, 0.0293, 0.0309, 0.0515, 0.0199], device='cuda:1'), in_proj_covar=tensor([0.0194, 0.0196, 0.0183, 0.0187, 0.0203, 0.0161, 0.0199, 0.0160], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-01 23:05:26,240 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([5.0174, 5.3618, 5.0919, 5.0959, 4.8802, 4.8193, 4.7527, 5.4579], device='cuda:1'), covar=tensor([0.1392, 0.1125, 0.1188, 0.1425, 0.0909, 0.1150, 0.1428, 0.1059], device='cuda:1'), in_proj_covar=tensor([0.0710, 0.0861, 0.0704, 0.0665, 0.0546, 0.0548, 0.0723, 0.0672], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:1') 2023-05-01 23:05:30,605 INFO [optim.py:368] (1/8) 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 23:05:52,659 INFO [train.py:904] (1/8) Epoch 25, batch 1750, loss[loss=0.17, simple_loss=0.2585, pruned_loss=0.04074, over 16043.00 frames. ], tot_loss[loss=0.1662, simple_loss=0.2529, pruned_loss=0.03978, over 3301589.75 frames. ], batch size: 35, lr: 2.71e-03, grad_scale: 8.0 2023-05-01 23:06:20,953 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=245373.0, num_to_drop=1, layers_to_drop={1} 2023-05-01 23:07:03,956 INFO [train.py:904] (1/8) Epoch 25, batch 1800, loss[loss=0.165, simple_loss=0.2664, pruned_loss=0.03183, over 16734.00 frames. ], tot_loss[loss=0.1665, simple_loss=0.2535, pruned_loss=0.0397, over 3313082.25 frames. ], batch size: 57, lr: 2.71e-03, grad_scale: 8.0 2023-05-01 23:07:29,357 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=245421.0, num_to_drop=1, layers_to_drop={0} 2023-05-01 23:07:38,480 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.8280, 4.6091, 4.8799, 5.0656, 5.2471, 4.6446, 5.1962, 5.2293], device='cuda:1'), covar=tensor([0.1834, 0.1478, 0.1697, 0.0787, 0.0554, 0.1083, 0.0685, 0.0622], device='cuda:1'), in_proj_covar=tensor([0.0677, 0.0835, 0.0962, 0.0845, 0.0643, 0.0663, 0.0694, 0.0811], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-01 23:07:38,998 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.25 vs. limit=2.0 2023-05-01 23:07:46,289 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.6847, 3.8295, 2.5039, 4.4644, 2.9703, 4.3449, 2.5332, 3.0688], device='cuda:1'), covar=tensor([0.0363, 0.0380, 0.1629, 0.0320, 0.0904, 0.0526, 0.1584, 0.0846], device='cuda:1'), in_proj_covar=tensor([0.0175, 0.0180, 0.0196, 0.0172, 0.0179, 0.0221, 0.0204, 0.0183], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-05-01 23:07:51,078 INFO [optim.py:368] (1/8) 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] (1/8) Epoch 25, batch 1850, loss[loss=0.132, simple_loss=0.2164, pruned_loss=0.02381, over 16859.00 frames. ], tot_loss[loss=0.1664, simple_loss=0.2542, pruned_loss=0.03936, over 3317501.71 frames. ], batch size: 39, lr: 2.71e-03, grad_scale: 8.0 2023-05-01 23:08:23,440 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=245460.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 23:09:12,930 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.0171, 3.2262, 3.3693, 2.1552, 3.0075, 2.3896, 3.5668, 3.5484], device='cuda:1'), covar=tensor([0.0250, 0.0944, 0.0723, 0.2040, 0.0921, 0.1067, 0.0481, 0.0927], device='cuda:1'), in_proj_covar=tensor([0.0160, 0.0167, 0.0170, 0.0156, 0.0147, 0.0132, 0.0145, 0.0179], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-05-01 23:09:23,471 INFO [train.py:904] (1/8) Epoch 25, batch 1900, loss[loss=0.157, simple_loss=0.2444, pruned_loss=0.03482, over 16839.00 frames. ], tot_loss[loss=0.1658, simple_loss=0.2536, pruned_loss=0.03899, over 3323124.34 frames. ], batch size: 96, lr: 2.71e-03, grad_scale: 4.0 2023-05-01 23:09:29,839 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.2511, 4.1942, 4.1887, 3.8954, 3.9725, 4.2416, 3.9129, 4.0188], device='cuda:1'), covar=tensor([0.0654, 0.0762, 0.0285, 0.0283, 0.0623, 0.0499, 0.0697, 0.0610], device='cuda:1'), in_proj_covar=tensor([0.0310, 0.0461, 0.0358, 0.0361, 0.0363, 0.0417, 0.0246, 0.0433], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002], device='cuda:1') 2023-05-01 23:09:37,910 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.17 vs. limit=2.0 2023-05-01 23:09:49,172 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=245521.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 23:10:11,814 INFO [optim.py:368] (1/8) 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 23:10:21,040 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=245543.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 23:10:25,022 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 25, batch 1950, loss[loss=0.1837, simple_loss=0.2599, pruned_loss=0.05375, over 16795.00 frames. ], tot_loss[loss=0.1655, simple_loss=0.2537, pruned_loss=0.03867, over 3300533.09 frames. ], batch size: 124, lr: 2.71e-03, grad_scale: 4.0 2023-05-01 23:10:58,253 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.2725, 3.4524, 3.9261, 2.1617, 3.2212, 2.4743, 3.7339, 3.6661], device='cuda:1'), covar=tensor([0.0299, 0.0986, 0.0507, 0.2127, 0.0857, 0.1059, 0.0669, 0.1170], device='cuda:1'), in_proj_covar=tensor([0.0160, 0.0167, 0.0171, 0.0156, 0.0147, 0.0132, 0.0146, 0.0180], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-05-01 23:11:05,759 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([5.3321, 5.3421, 5.0914, 4.5878, 5.1531, 2.0771, 4.9231, 4.9396], device='cuda:1'), covar=tensor([0.0102, 0.0084, 0.0232, 0.0410, 0.0104, 0.2817, 0.0148, 0.0237], device='cuda:1'), in_proj_covar=tensor([0.0177, 0.0170, 0.0209, 0.0183, 0.0187, 0.0216, 0.0199, 0.0180], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-01 23:11:22,799 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.2677, 3.1948, 2.0397, 3.4272, 2.5074, 3.4120, 2.1488, 2.6738], device='cuda:1'), covar=tensor([0.0302, 0.0461, 0.1577, 0.0331, 0.0821, 0.0689, 0.1483, 0.0754], device='cuda:1'), in_proj_covar=tensor([0.0176, 0.0181, 0.0197, 0.0173, 0.0180, 0.0222, 0.0205, 0.0183], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-05-01 23:11:26,899 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=245591.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 23:11:32,376 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.9331, 2.7825, 2.1254, 2.4479, 3.1315, 2.9399, 3.5789, 3.5281], device='cuda:1'), covar=tensor([0.0165, 0.0489, 0.0718, 0.0558, 0.0321, 0.0388, 0.0241, 0.0260], device='cuda:1'), in_proj_covar=tensor([0.0228, 0.0246, 0.0234, 0.0235, 0.0247, 0.0245, 0.0247, 0.0243], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-01 23:11:41,386 INFO [train.py:904] (1/8) Epoch 25, batch 2000, loss[loss=0.1666, simple_loss=0.2585, pruned_loss=0.03736, over 17136.00 frames. ], tot_loss[loss=0.1655, simple_loss=0.2534, pruned_loss=0.03887, over 3307271.79 frames. ], batch size: 47, lr: 2.71e-03, grad_scale: 4.0 2023-05-01 23:11:49,130 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=245607.0, num_to_drop=1, layers_to_drop={3} 2023-05-01 23:11:53,254 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=245610.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 23:12:04,821 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.25 vs. limit=2.0 2023-05-01 23:12:06,707 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.5990, 4.2073, 4.1699, 2.9335, 3.5995, 4.1567, 3.7565, 2.4577], device='cuda:1'), covar=tensor([0.0591, 0.0100, 0.0074, 0.0438, 0.0161, 0.0142, 0.0131, 0.0527], device='cuda:1'), in_proj_covar=tensor([0.0137, 0.0088, 0.0088, 0.0135, 0.0100, 0.0112, 0.0097, 0.0131], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0003, 0.0004], device='cuda:1') 2023-05-01 23:12:31,549 INFO [optim.py:368] (1/8) Clipping_scale=2.0, grad-norm quartiles 1.417e+02 2.177e+02 2.547e+02 3.013e+02 5.081e+02, threshold=5.093e+02, percent-clipped=1.0 2023-05-01 23:12:50,393 INFO [train.py:904] (1/8) Epoch 25, batch 2050, loss[loss=0.1666, simple_loss=0.2496, pruned_loss=0.04178, over 16704.00 frames. ], tot_loss[loss=0.1656, simple_loss=0.2533, pruned_loss=0.039, over 3312992.49 frames. ], batch size: 76, lr: 2.71e-03, grad_scale: 4.0 2023-05-01 23:12:57,319 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=245658.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 23:13:05,857 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.5894, 3.6673, 2.7659, 2.2678, 2.3257, 2.3093, 3.7318, 3.1553], device='cuda:1'), covar=tensor([0.2921, 0.0656, 0.1957, 0.3126, 0.2916, 0.2313, 0.0539, 0.1604], device='cuda:1'), in_proj_covar=tensor([0.0331, 0.0273, 0.0312, 0.0322, 0.0303, 0.0271, 0.0302, 0.0348], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:1') 2023-05-01 23:13:18,967 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.7770, 4.7021, 4.6903, 4.3643, 4.3990, 4.7218, 4.4633, 4.4761], device='cuda:1'), covar=tensor([0.0755, 0.0977, 0.0330, 0.0322, 0.0844, 0.0589, 0.0519, 0.0658], device='cuda:1'), in_proj_covar=tensor([0.0313, 0.0465, 0.0361, 0.0365, 0.0366, 0.0420, 0.0248, 0.0438], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-05-01 23:13:34,845 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=245686.0, num_to_drop=1, layers_to_drop={0} 2023-05-01 23:13:41,463 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.36 vs. limit=2.0 2023-05-01 23:13:58,098 INFO [train.py:904] (1/8) Epoch 25, batch 2100, loss[loss=0.146, simple_loss=0.2306, pruned_loss=0.03073, over 17156.00 frames. ], tot_loss[loss=0.1665, simple_loss=0.2543, pruned_loss=0.03936, over 3314738.34 frames. ], batch size: 40, lr: 2.71e-03, grad_scale: 4.0 2023-05-01 23:14:11,761 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([5.5464, 5.9019, 5.6488, 5.7253, 5.2949, 5.4147, 5.2825, 6.0125], device='cuda:1'), covar=tensor([0.1552, 0.1099, 0.1215, 0.0991, 0.0951, 0.0735, 0.1251, 0.1004], device='cuda:1'), in_proj_covar=tensor([0.0720, 0.0874, 0.0715, 0.0675, 0.0555, 0.0555, 0.0734, 0.0682], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:1') 2023-05-01 23:14:37,763 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.4359, 4.0547, 4.5492, 2.4021, 4.6879, 4.8054, 3.5839, 3.7594], device='cuda:1'), covar=tensor([0.0606, 0.0246, 0.0209, 0.1087, 0.0082, 0.0172, 0.0417, 0.0364], device='cuda:1'), in_proj_covar=tensor([0.0148, 0.0111, 0.0100, 0.0140, 0.0084, 0.0131, 0.0130, 0.0131], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-05-01 23:14:49,016 INFO [optim.py:368] (1/8) 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 23:14:58,051 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.78 vs. limit=2.0 2023-05-01 23:14:59,025 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=245747.0, num_to_drop=1, layers_to_drop={1} 2023-05-01 23:15:07,070 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.7960, 4.7977, 5.1693, 5.1648, 5.2223, 4.8891, 4.8384, 4.6819], device='cuda:1'), covar=tensor([0.0360, 0.0666, 0.0475, 0.0433, 0.0482, 0.0427, 0.0943, 0.0519], device='cuda:1'), in_proj_covar=tensor([0.0432, 0.0483, 0.0472, 0.0432, 0.0515, 0.0497, 0.0573, 0.0392], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:1') 2023-05-01 23:15:08,434 INFO [train.py:904] (1/8) Epoch 25, batch 2150, loss[loss=0.1551, simple_loss=0.2503, pruned_loss=0.02997, over 16629.00 frames. ], tot_loss[loss=0.167, simple_loss=0.2547, pruned_loss=0.03962, over 3316433.31 frames. ], batch size: 62, lr: 2.71e-03, grad_scale: 4.0 2023-05-01 23:15:25,603 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([5.0740, 5.1151, 5.5384, 5.5601, 5.5886, 5.2175, 5.1719, 4.9800], device='cuda:1'), covar=tensor([0.0333, 0.0599, 0.0423, 0.0402, 0.0431, 0.0424, 0.0911, 0.0465], device='cuda:1'), in_proj_covar=tensor([0.0432, 0.0483, 0.0472, 0.0432, 0.0515, 0.0497, 0.0574, 0.0392], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:1') 2023-05-01 23:16:00,573 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=245792.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 23:16:15,986 INFO [train.py:904] (1/8) Epoch 25, batch 2200, loss[loss=0.1891, simple_loss=0.2671, pruned_loss=0.0555, over 16913.00 frames. ], tot_loss[loss=0.1683, simple_loss=0.2555, pruned_loss=0.04055, over 3316445.72 frames. ], batch size: 109, lr: 2.71e-03, grad_scale: 4.0 2023-05-01 23:16:34,701 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=245816.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 23:17:06,062 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.80 vs. limit=2.0 2023-05-01 23:17:06,934 INFO [optim.py:368] (1/8) 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 23:17:23,664 INFO [train.py:904] (1/8) Epoch 25, batch 2250, loss[loss=0.1528, simple_loss=0.2354, pruned_loss=0.03505, over 17025.00 frames. ], tot_loss[loss=0.1687, simple_loss=0.2561, pruned_loss=0.04064, over 3325897.39 frames. ], batch size: 41, lr: 2.71e-03, grad_scale: 4.0 2023-05-01 23:17:24,118 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=245853.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 23:18:16,609 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=245891.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 23:18:32,161 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=245902.0, num_to_drop=1, layers_to_drop={1} 2023-05-01 23:18:32,939 INFO [train.py:904] (1/8) Epoch 25, batch 2300, loss[loss=0.1817, simple_loss=0.2775, pruned_loss=0.04295, over 16736.00 frames. ], tot_loss[loss=0.1683, simple_loss=0.2556, pruned_loss=0.04047, over 3331825.61 frames. ], batch size: 57, lr: 2.71e-03, grad_scale: 2.0 2023-05-01 23:18:33,446 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.6982, 4.7212, 4.6287, 4.0924, 4.6868, 1.8360, 4.4516, 4.2595], device='cuda:1'), covar=tensor([0.0150, 0.0108, 0.0189, 0.0323, 0.0095, 0.2718, 0.0168, 0.0228], device='cuda:1'), in_proj_covar=tensor([0.0178, 0.0171, 0.0210, 0.0185, 0.0187, 0.0216, 0.0200, 0.0180], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-01 23:19:07,721 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([5.4418, 5.4795, 5.2687, 4.7625, 5.3924, 2.2086, 5.1415, 5.0954], device='cuda:1'), covar=tensor([0.0104, 0.0082, 0.0218, 0.0354, 0.0090, 0.2638, 0.0113, 0.0211], device='cuda:1'), in_proj_covar=tensor([0.0178, 0.0170, 0.0210, 0.0185, 0.0187, 0.0216, 0.0200, 0.0180], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-01 23:19:24,141 INFO [optim.py:368] (1/8) 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 23:19:42,154 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 25, batch 2350, loss[loss=0.1475, simple_loss=0.231, pruned_loss=0.03197, over 16866.00 frames. ], tot_loss[loss=0.1683, simple_loss=0.2556, pruned_loss=0.0405, over 3328656.18 frames. ], batch size: 39, lr: 2.71e-03, grad_scale: 2.0 2023-05-01 23:19:58,716 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([5.2725, 5.7828, 5.9587, 5.5605, 5.7664, 6.3389, 5.8070, 5.4798], device='cuda:1'), covar=tensor([0.0926, 0.1997, 0.2419, 0.2270, 0.2528, 0.0964, 0.1582, 0.2389], device='cuda:1'), in_proj_covar=tensor([0.0428, 0.0638, 0.0696, 0.0515, 0.0691, 0.0718, 0.0539, 0.0686], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-05-01 23:20:14,797 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.8645, 5.0850, 5.2987, 5.0180, 5.0536, 5.7086, 5.1468, 4.8531], device='cuda:1'), covar=tensor([0.1360, 0.2295, 0.3008, 0.2068, 0.2516, 0.1022, 0.1890, 0.2443], device='cuda:1'), in_proj_covar=tensor([0.0428, 0.0638, 0.0696, 0.0516, 0.0691, 0.0718, 0.0539, 0.0686], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-05-01 23:20:54,771 INFO [train.py:904] (1/8) Epoch 25, batch 2400, loss[loss=0.1382, simple_loss=0.2213, pruned_loss=0.0275, over 16783.00 frames. ], tot_loss[loss=0.168, simple_loss=0.2558, pruned_loss=0.04014, over 3333763.69 frames. ], batch size: 39, lr: 2.71e-03, grad_scale: 4.0 2023-05-01 23:21:15,347 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=2.82 vs. limit=5.0 2023-05-01 23:21:37,908 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([5.2016, 5.8536, 5.9739, 5.6262, 5.7397, 6.3214, 5.7887, 5.4210], device='cuda:1'), covar=tensor([0.0957, 0.2041, 0.2517, 0.1976, 0.2597, 0.0932, 0.1636, 0.2382], device='cuda:1'), in_proj_covar=tensor([0.0431, 0.0641, 0.0701, 0.0519, 0.0694, 0.0723, 0.0542, 0.0689], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-05-01 23:21:46,453 INFO [optim.py:368] (1/8) Clipping_scale=2.0, grad-norm quartiles 1.525e+02 2.068e+02 2.493e+02 3.230e+02 1.350e+03, threshold=4.987e+02, percent-clipped=5.0 2023-05-01 23:21:49,244 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=246042.0, num_to_drop=1, layers_to_drop={0} 2023-05-01 23:22:03,693 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.16 vs. limit=2.0 2023-05-01 23:22:04,193 INFO [train.py:904] (1/8) Epoch 25, batch 2450, loss[loss=0.1774, simple_loss=0.2734, pruned_loss=0.04074, over 16643.00 frames. ], tot_loss[loss=0.1683, simple_loss=0.2566, pruned_loss=0.04004, over 3333006.45 frames. ], batch size: 62, lr: 2.71e-03, grad_scale: 4.0 2023-05-01 23:22:35,302 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.8479, 2.1413, 2.5140, 2.7943, 2.6674, 3.3365, 2.2491, 3.3479], device='cuda:1'), covar=tensor([0.0309, 0.0532, 0.0409, 0.0406, 0.0420, 0.0229, 0.0602, 0.0185], device='cuda:1'), in_proj_covar=tensor([0.0197, 0.0198, 0.0185, 0.0189, 0.0205, 0.0163, 0.0202, 0.0163], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-01 23:23:10,959 INFO [train.py:904] (1/8) Epoch 25, batch 2500, loss[loss=0.1594, simple_loss=0.2469, pruned_loss=0.03597, over 15820.00 frames. ], tot_loss[loss=0.1675, simple_loss=0.2558, pruned_loss=0.03964, over 3338572.18 frames. ], batch size: 35, lr: 2.71e-03, grad_scale: 4.0 2023-05-01 23:23:21,501 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=246111.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 23:23:28,029 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=246116.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 23:23:50,633 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([5.1665, 5.1779, 4.9512, 4.4189, 5.0369, 1.7600, 4.7960, 4.7515], device='cuda:1'), covar=tensor([0.0099, 0.0091, 0.0249, 0.0428, 0.0107, 0.3015, 0.0152, 0.0266], device='cuda:1'), in_proj_covar=tensor([0.0178, 0.0171, 0.0210, 0.0185, 0.0188, 0.0217, 0.0200, 0.0180], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-01 23:24:01,582 INFO [optim.py:368] (1/8) 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 23:24:12,432 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=246148.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 23:24:18,465 INFO [train.py:904] (1/8) Epoch 25, batch 2550, loss[loss=0.172, simple_loss=0.2687, pruned_loss=0.03761, over 17050.00 frames. ], tot_loss[loss=0.1681, simple_loss=0.2563, pruned_loss=0.03999, over 3336871.59 frames. ], batch size: 50, lr: 2.71e-03, grad_scale: 4.0 2023-05-01 23:24:24,262 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.9193, 4.6675, 4.9164, 5.1396, 5.3681, 4.6785, 5.3318, 5.3305], device='cuda:1'), covar=tensor([0.2012, 0.1474, 0.2085, 0.0903, 0.0582, 0.1065, 0.0608, 0.0712], device='cuda:1'), in_proj_covar=tensor([0.0681, 0.0841, 0.0970, 0.0850, 0.0648, 0.0670, 0.0695, 0.0816], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-01 23:24:35,189 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=246164.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 23:24:44,846 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=246172.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 23:24:53,985 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.5481, 4.6251, 4.7731, 4.5508, 4.5823, 5.2334, 4.7588, 4.4190], device='cuda:1'), covar=tensor([0.1660, 0.2347, 0.2607, 0.2302, 0.2947, 0.1146, 0.1669, 0.2501], device='cuda:1'), in_proj_covar=tensor([0.0430, 0.0638, 0.0700, 0.0518, 0.0690, 0.0721, 0.0541, 0.0688], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-05-01 23:25:06,127 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.4437, 3.4276, 3.7032, 2.5383, 3.3581, 3.7904, 3.4447, 2.1685], device='cuda:1'), covar=tensor([0.0507, 0.0162, 0.0068, 0.0422, 0.0132, 0.0108, 0.0117, 0.0503], device='cuda:1'), in_proj_covar=tensor([0.0137, 0.0088, 0.0088, 0.0135, 0.0100, 0.0111, 0.0098, 0.0132], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0003, 0.0004], device='cuda:1') 2023-05-01 23:25:17,217 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=246196.0, num_to_drop=1, layers_to_drop={0} 2023-05-01 23:25:26,163 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=246202.0, num_to_drop=1, layers_to_drop={1} 2023-05-01 23:25:26,989 INFO [train.py:904] (1/8) Epoch 25, batch 2600, loss[loss=0.1892, simple_loss=0.2774, pruned_loss=0.0505, over 12198.00 frames. ], tot_loss[loss=0.1689, simple_loss=0.257, pruned_loss=0.04037, over 3317203.37 frames. ], batch size: 246, lr: 2.71e-03, grad_scale: 4.0 2023-05-01 23:25:52,184 INFO [zipformer.py:625] (1/8) 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] (1/8) 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 23:26:28,312 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=246247.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 23:26:33,162 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 25, batch 2650, loss[loss=0.1793, simple_loss=0.2571, pruned_loss=0.05072, over 16439.00 frames. ], tot_loss[loss=0.1686, simple_loss=0.2573, pruned_loss=0.03992, over 3323007.30 frames. ], batch size: 146, lr: 2.71e-03, grad_scale: 4.0 2023-05-01 23:26:42,376 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=246257.0, num_to_drop=1, layers_to_drop={0} 2023-05-01 23:27:16,955 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=246282.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 23:27:32,006 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.6957, 1.8640, 2.2567, 2.6152, 2.6781, 2.6489, 2.0103, 2.9014], device='cuda:1'), covar=tensor([0.0225, 0.0527, 0.0411, 0.0320, 0.0333, 0.0334, 0.0544, 0.0162], device='cuda:1'), in_proj_covar=tensor([0.0197, 0.0197, 0.0186, 0.0190, 0.0205, 0.0163, 0.0202, 0.0163], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-01 23:27:44,599 INFO [train.py:904] (1/8) Epoch 25, batch 2700, loss[loss=0.16, simple_loss=0.2495, pruned_loss=0.03523, over 17235.00 frames. ], tot_loss[loss=0.1684, simple_loss=0.2574, pruned_loss=0.03967, over 3331082.49 frames. ], batch size: 44, lr: 2.71e-03, grad_scale: 4.0 2023-05-01 23:28:34,095 INFO [optim.py:368] (1/8) 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 23:28:36,884 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=246342.0, num_to_drop=1, layers_to_drop={1} 2023-05-01 23:28:52,088 INFO [train.py:904] (1/8) Epoch 25, batch 2750, loss[loss=0.1743, simple_loss=0.26, pruned_loss=0.04427, over 16382.00 frames. ], tot_loss[loss=0.1682, simple_loss=0.2576, pruned_loss=0.03938, over 3336297.04 frames. ], batch size: 146, lr: 2.71e-03, grad_scale: 4.0 2023-05-01 23:29:04,596 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.25 vs. limit=2.0 2023-05-01 23:29:09,267 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.83 vs. limit=2.0 2023-05-01 23:29:17,609 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.7384, 4.9062, 5.0261, 4.8605, 4.8583, 5.5225, 5.0124, 4.7703], device='cuda:1'), covar=tensor([0.1392, 0.2106, 0.2543, 0.2278, 0.2967, 0.1126, 0.1636, 0.2441], device='cuda:1'), in_proj_covar=tensor([0.0429, 0.0638, 0.0700, 0.0519, 0.0692, 0.0722, 0.0541, 0.0689], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-05-01 23:29:23,805 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=2.53 vs. limit=5.0 2023-05-01 23:29:43,588 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=246390.0, num_to_drop=1, layers_to_drop={1} 2023-05-01 23:30:01,769 INFO [train.py:904] (1/8) Epoch 25, batch 2800, loss[loss=0.1581, simple_loss=0.2601, pruned_loss=0.02806, over 17039.00 frames. ], tot_loss[loss=0.1675, simple_loss=0.2573, pruned_loss=0.03882, over 3328011.71 frames. ], batch size: 50, lr: 2.71e-03, grad_scale: 8.0 2023-05-01 23:30:05,364 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.9147, 4.9390, 5.3213, 5.3221, 5.3461, 5.0089, 4.9523, 4.7883], device='cuda:1'), covar=tensor([0.0368, 0.0579, 0.0436, 0.0395, 0.0535, 0.0430, 0.1005, 0.0487], device='cuda:1'), in_proj_covar=tensor([0.0432, 0.0484, 0.0472, 0.0433, 0.0516, 0.0497, 0.0574, 0.0391], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:1') 2023-05-01 23:30:38,815 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.8237, 2.7530, 2.4604, 2.7357, 3.1286, 2.9138, 3.3321, 3.2984], device='cuda:1'), covar=tensor([0.0148, 0.0516, 0.0562, 0.0471, 0.0298, 0.0400, 0.0275, 0.0299], device='cuda:1'), in_proj_covar=tensor([0.0228, 0.0247, 0.0233, 0.0236, 0.0248, 0.0245, 0.0248, 0.0244], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-01 23:30:54,340 INFO [optim.py:368] (1/8) 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 23:31:04,222 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=246448.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 23:31:10,738 INFO [train.py:904] (1/8) Epoch 25, batch 2850, loss[loss=0.1601, simple_loss=0.2533, pruned_loss=0.03345, over 17116.00 frames. ], tot_loss[loss=0.1664, simple_loss=0.2558, pruned_loss=0.03845, over 3332125.61 frames. ], batch size: 49, lr: 2.71e-03, grad_scale: 4.0 2023-05-01 23:31:31,901 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=246467.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 23:32:00,783 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([5.1925, 5.7098, 5.8327, 5.5150, 5.6778, 6.2369, 5.7210, 5.4028], device='cuda:1'), covar=tensor([0.0897, 0.2111, 0.2868, 0.2166, 0.2483, 0.0950, 0.1588, 0.2322], device='cuda:1'), in_proj_covar=tensor([0.0430, 0.0639, 0.0699, 0.0519, 0.0693, 0.0724, 0.0542, 0.0690], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-05-01 23:32:10,939 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=246496.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 23:32:20,911 INFO [train.py:904] (1/8) Epoch 25, batch 2900, loss[loss=0.149, simple_loss=0.2357, pruned_loss=0.03119, over 16873.00 frames. ], tot_loss[loss=0.1667, simple_loss=0.2557, pruned_loss=0.03888, over 3323806.80 frames. ], batch size: 42, lr: 2.70e-03, grad_scale: 4.0 2023-05-01 23:33:14,959 INFO [optim.py:368] (1/8) 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 23:33:25,282 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.0439, 4.0350, 4.3496, 4.3565, 4.4355, 4.1027, 4.1722, 4.0814], device='cuda:1'), covar=tensor([0.0479, 0.0794, 0.0486, 0.0455, 0.0566, 0.0622, 0.0906, 0.0748], device='cuda:1'), in_proj_covar=tensor([0.0434, 0.0487, 0.0474, 0.0434, 0.0517, 0.0499, 0.0577, 0.0394], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:1') 2023-05-01 23:33:25,328 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=246547.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 23:33:32,614 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=246552.0, num_to_drop=1, layers_to_drop={3} 2023-05-01 23:33:33,512 INFO [train.py:904] (1/8) Epoch 25, batch 2950, loss[loss=0.1532, simple_loss=0.2513, pruned_loss=0.02756, over 17112.00 frames. ], tot_loss[loss=0.1664, simple_loss=0.2549, pruned_loss=0.03896, over 3325845.37 frames. ], batch size: 47, lr: 2.70e-03, grad_scale: 4.0 2023-05-01 23:33:47,233 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.2343, 4.2082, 4.1347, 3.6151, 4.1991, 1.8062, 3.9646, 3.5816], device='cuda:1'), covar=tensor([0.0143, 0.0129, 0.0216, 0.0267, 0.0107, 0.2898, 0.0160, 0.0282], device='cuda:1'), in_proj_covar=tensor([0.0178, 0.0171, 0.0211, 0.0186, 0.0189, 0.0217, 0.0201, 0.0181], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-01 23:34:00,348 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.7564, 3.8624, 4.2775, 2.5318, 3.5619, 2.9506, 4.1761, 4.1084], device='cuda:1'), covar=tensor([0.0243, 0.0890, 0.0472, 0.1830, 0.0762, 0.0841, 0.0497, 0.0982], device='cuda:1'), in_proj_covar=tensor([0.0160, 0.0169, 0.0171, 0.0156, 0.0148, 0.0132, 0.0146, 0.0181], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0005], device='cuda:1') 2023-05-01 23:34:06,259 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=246577.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 23:34:32,274 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 25, batch 3000, loss[loss=0.1592, simple_loss=0.2404, pruned_loss=0.03899, over 16741.00 frames. ], tot_loss[loss=0.1668, simple_loss=0.2549, pruned_loss=0.03931, over 3326635.11 frames. ], batch size: 89, lr: 2.70e-03, grad_scale: 4.0 2023-05-01 23:34:42,716 INFO [train.py:929] (1/8) Computing validation loss 2023-05-01 23:34:52,580 INFO [train.py:938] (1/8) Epoch 25, validation: loss=0.1341, simple_loss=0.239, pruned_loss=0.01457, over 944034.00 frames. 2023-05-01 23:34:52,581 INFO [train.py:939] (1/8) Maximum memory allocated so far is 18145MB 2023-05-01 23:34:58,505 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.2588, 4.0542, 4.2056, 4.4379, 4.5161, 4.1844, 4.3805, 4.5201], device='cuda:1'), covar=tensor([0.1631, 0.1448, 0.1782, 0.0881, 0.0804, 0.1292, 0.2881, 0.1028], device='cuda:1'), in_proj_covar=tensor([0.0684, 0.0843, 0.0974, 0.0852, 0.0649, 0.0672, 0.0700, 0.0819], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-01 23:35:46,636 INFO [optim.py:368] (1/8) 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] (1/8) Epoch 25, batch 3050, loss[loss=0.1749, simple_loss=0.2515, pruned_loss=0.04912, over 16430.00 frames. ], tot_loss[loss=0.1669, simple_loss=0.2544, pruned_loss=0.03975, over 3326333.16 frames. ], batch size: 165, lr: 2.70e-03, grad_scale: 4.0 2023-05-01 23:36:38,380 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=246677.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 23:37:13,555 INFO [train.py:904] (1/8) Epoch 25, batch 3100, loss[loss=0.1455, simple_loss=0.2369, pruned_loss=0.02703, over 17215.00 frames. ], tot_loss[loss=0.1664, simple_loss=0.2536, pruned_loss=0.03959, over 3326173.99 frames. ], batch size: 44, lr: 2.70e-03, grad_scale: 4.0 2023-05-01 23:37:26,178 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.3581, 4.0501, 4.5012, 2.3400, 4.7044, 4.7463, 3.5004, 3.8032], device='cuda:1'), covar=tensor([0.0628, 0.0265, 0.0209, 0.1140, 0.0082, 0.0193, 0.0419, 0.0370], device='cuda:1'), in_proj_covar=tensor([0.0148, 0.0112, 0.0101, 0.0141, 0.0084, 0.0132, 0.0131, 0.0132], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-05-01 23:38:04,953 INFO [zipformer.py:625] (1/8) 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] (1/8) Clipping_scale=2.0, grad-norm quartiles 1.354e+02 2.159e+02 2.563e+02 3.074e+02 4.728e+02, threshold=5.126e+02, percent-clipped=0.0 2023-05-01 23:38:25,016 INFO [train.py:904] (1/8) Epoch 25, batch 3150, loss[loss=0.1901, simple_loss=0.2658, pruned_loss=0.0572, over 16848.00 frames. ], tot_loss[loss=0.1663, simple_loss=0.2528, pruned_loss=0.03993, over 3325622.23 frames. ], batch size: 116, lr: 2.70e-03, grad_scale: 4.0 2023-05-01 23:38:44,314 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=246767.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 23:38:53,747 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.9053, 4.4427, 3.1671, 2.4331, 2.6932, 2.6723, 4.9060, 3.6350], device='cuda:1'), covar=tensor([0.2959, 0.0578, 0.1841, 0.2975, 0.3002, 0.2230, 0.0310, 0.1438], device='cuda:1'), in_proj_covar=tensor([0.0334, 0.0276, 0.0313, 0.0324, 0.0306, 0.0273, 0.0304, 0.0351], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:1') 2023-05-01 23:39:08,795 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=3.76 vs. limit=5.0 2023-05-01 23:39:34,228 INFO [train.py:904] (1/8) Epoch 25, batch 3200, loss[loss=0.1345, simple_loss=0.2247, pruned_loss=0.02208, over 16757.00 frames. ], tot_loss[loss=0.1656, simple_loss=0.2522, pruned_loss=0.03944, over 3324852.99 frames. ], batch size: 39, lr: 2.70e-03, grad_scale: 8.0 2023-05-01 23:39:41,728 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.24 vs. limit=2.0 2023-05-01 23:39:51,041 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=246815.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 23:40:27,530 INFO [optim.py:368] (1/8) 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 23:40:42,109 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=246852.0, num_to_drop=1, layers_to_drop={0} 2023-05-01 23:40:42,933 INFO [train.py:904] (1/8) Epoch 25, batch 3250, loss[loss=0.1769, simple_loss=0.2666, pruned_loss=0.04355, over 16510.00 frames. ], tot_loss[loss=0.1658, simple_loss=0.2522, pruned_loss=0.03972, over 3332557.30 frames. ], batch size: 75, lr: 2.70e-03, grad_scale: 8.0 2023-05-01 23:40:49,495 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([5.3238, 5.6248, 5.3911, 5.4708, 5.1503, 5.0333, 4.9740, 5.7582], device='cuda:1'), covar=tensor([0.1353, 0.0914, 0.0989, 0.0845, 0.0813, 0.0851, 0.1280, 0.0919], device='cuda:1'), in_proj_covar=tensor([0.0731, 0.0881, 0.0724, 0.0683, 0.0561, 0.0560, 0.0744, 0.0691], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-01 23:41:16,124 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=246877.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 23:41:49,122 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=246900.0, num_to_drop=1, layers_to_drop={0} 2023-05-01 23:41:53,766 INFO [train.py:904] (1/8) Epoch 25, batch 3300, loss[loss=0.1406, simple_loss=0.2217, pruned_loss=0.02978, over 16985.00 frames. ], tot_loss[loss=0.1676, simple_loss=0.2539, pruned_loss=0.04064, over 3326948.45 frames. ], batch size: 41, lr: 2.70e-03, grad_scale: 8.0 2023-05-01 23:42:24,636 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=246925.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 23:42:34,842 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.26 vs. limit=2.0 2023-05-01 23:42:36,923 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=246934.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 23:42:46,635 INFO [optim.py:368] (1/8) Clipping_scale=2.0, grad-norm quartiles 1.501e+02 2.089e+02 2.401e+02 2.822e+02 3.775e+02, threshold=4.802e+02, percent-clipped=0.0 2023-05-01 23:43:02,687 INFO [train.py:904] (1/8) Epoch 25, batch 3350, loss[loss=0.1528, simple_loss=0.2451, pruned_loss=0.03024, over 16805.00 frames. ], tot_loss[loss=0.1676, simple_loss=0.2545, pruned_loss=0.04035, over 3332245.80 frames. ], batch size: 83, lr: 2.70e-03, grad_scale: 8.0 2023-05-01 23:44:01,740 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=246995.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 23:44:13,546 INFO [train.py:904] (1/8) Epoch 25, batch 3400, loss[loss=0.1892, simple_loss=0.2602, pruned_loss=0.05907, over 16724.00 frames. ], tot_loss[loss=0.1671, simple_loss=0.2538, pruned_loss=0.0402, over 3330770.69 frames. ], batch size: 124, lr: 2.70e-03, grad_scale: 8.0 2023-05-01 23:44:24,413 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.8727, 4.0827, 2.7770, 4.7304, 3.2587, 4.6081, 2.8386, 3.3703], device='cuda:1'), covar=tensor([0.0339, 0.0388, 0.1395, 0.0284, 0.0740, 0.0617, 0.1307, 0.0697], device='cuda:1'), in_proj_covar=tensor([0.0179, 0.0184, 0.0199, 0.0176, 0.0182, 0.0226, 0.0207, 0.0186], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-05-01 23:44:56,896 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=247033.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 23:45:08,919 INFO [optim.py:368] (1/8) 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 23:45:26,156 INFO [train.py:904] (1/8) Epoch 25, batch 3450, loss[loss=0.1531, simple_loss=0.2526, pruned_loss=0.02679, over 17046.00 frames. ], tot_loss[loss=0.1666, simple_loss=0.2528, pruned_loss=0.04025, over 3314938.11 frames. ], batch size: 50, lr: 2.70e-03, grad_scale: 4.0 2023-05-01 23:46:34,425 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=4.77 vs. limit=5.0 2023-05-01 23:46:35,594 INFO [train.py:904] (1/8) Epoch 25, batch 3500, loss[loss=0.1898, simple_loss=0.2653, pruned_loss=0.05712, over 16906.00 frames. ], tot_loss[loss=0.166, simple_loss=0.2524, pruned_loss=0.03978, over 3305620.49 frames. ], batch size: 109, lr: 2.70e-03, grad_scale: 4.0 2023-05-01 23:46:43,775 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([5.3958, 5.7725, 5.5370, 5.5905, 5.2204, 5.2318, 5.1898, 5.9007], device='cuda:1'), covar=tensor([0.1506, 0.0996, 0.0987, 0.0945, 0.0922, 0.0744, 0.1284, 0.0981], device='cuda:1'), in_proj_covar=tensor([0.0730, 0.0880, 0.0722, 0.0682, 0.0559, 0.0560, 0.0742, 0.0690], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-01 23:47:07,327 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.5410, 3.6142, 3.3569, 2.9418, 3.2341, 3.4990, 3.3102, 3.3153], device='cuda:1'), covar=tensor([0.0586, 0.0604, 0.0304, 0.0287, 0.0504, 0.0426, 0.1230, 0.0496], device='cuda:1'), in_proj_covar=tensor([0.0318, 0.0474, 0.0368, 0.0374, 0.0374, 0.0430, 0.0251, 0.0448], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-05-01 23:47:23,730 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.24 vs. limit=2.0 2023-05-01 23:47:27,015 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([5.1276, 5.0166, 4.9017, 3.6993, 4.9807, 1.7783, 4.5631, 4.6612], device='cuda:1'), covar=tensor([0.0170, 0.0161, 0.0298, 0.0849, 0.0163, 0.3713, 0.0244, 0.0411], device='cuda:1'), in_proj_covar=tensor([0.0180, 0.0172, 0.0213, 0.0188, 0.0191, 0.0219, 0.0202, 0.0183], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-01 23:47:30,016 INFO [optim.py:368] (1/8) 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 23:47:45,772 INFO [train.py:904] (1/8) Epoch 25, batch 3550, loss[loss=0.1926, simple_loss=0.2796, pruned_loss=0.05285, over 17029.00 frames. ], tot_loss[loss=0.1648, simple_loss=0.2511, pruned_loss=0.03925, over 3310391.45 frames. ], batch size: 55, lr: 2.70e-03, grad_scale: 4.0 2023-05-01 23:48:55,167 INFO [train.py:904] (1/8) Epoch 25, batch 3600, loss[loss=0.1441, simple_loss=0.2213, pruned_loss=0.03349, over 16679.00 frames. ], tot_loss[loss=0.1647, simple_loss=0.2509, pruned_loss=0.03927, over 3307187.04 frames. ], batch size: 89, lr: 2.70e-03, grad_scale: 8.0 2023-05-01 23:49:35,656 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.8741, 2.7239, 2.8468, 2.1552, 2.7437, 2.2289, 2.8062, 2.8814], device='cuda:1'), covar=tensor([0.0251, 0.0849, 0.0450, 0.1741, 0.0795, 0.0860, 0.0537, 0.0792], device='cuda:1'), in_proj_covar=tensor([0.0161, 0.0170, 0.0171, 0.0156, 0.0148, 0.0133, 0.0147, 0.0182], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0005], device='cuda:1') 2023-05-01 23:49:49,107 INFO [optim.py:368] (1/8) 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 23:50:05,473 INFO [train.py:904] (1/8) Epoch 25, batch 3650, loss[loss=0.1774, simple_loss=0.2561, pruned_loss=0.04933, over 11391.00 frames. ], tot_loss[loss=0.1644, simple_loss=0.2498, pruned_loss=0.03945, over 3290679.08 frames. ], batch size: 247, lr: 2.70e-03, grad_scale: 8.0 2023-05-01 23:50:09,651 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.4448, 2.3728, 2.4031, 4.2641, 2.2724, 2.7425, 2.4509, 2.5346], device='cuda:1'), covar=tensor([0.1333, 0.3632, 0.3213, 0.0574, 0.4242, 0.2797, 0.3653, 0.3880], device='cuda:1'), in_proj_covar=tensor([0.0417, 0.0465, 0.0384, 0.0339, 0.0445, 0.0534, 0.0438, 0.0546], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-01 23:50:27,803 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.36 vs. limit=2.0 2023-05-01 23:50:59,437 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=247290.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 23:51:17,357 INFO [train.py:904] (1/8) Epoch 25, batch 3700, loss[loss=0.1784, simple_loss=0.2491, pruned_loss=0.05381, over 16709.00 frames. ], tot_loss[loss=0.1649, simple_loss=0.2486, pruned_loss=0.04058, over 3286030.45 frames. ], batch size: 134, lr: 2.70e-03, grad_scale: 8.0 2023-05-01 23:51:30,205 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.41 vs. limit=2.0 2023-05-01 23:51:38,875 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.3794, 3.3001, 3.5459, 2.3854, 3.3636, 3.6674, 3.3586, 1.8437], device='cuda:1'), covar=tensor([0.0564, 0.0167, 0.0089, 0.0481, 0.0125, 0.0116, 0.0128, 0.0655], device='cuda:1'), in_proj_covar=tensor([0.0138, 0.0089, 0.0089, 0.0137, 0.0102, 0.0113, 0.0099, 0.0133], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0003, 0.0004], device='cuda:1') 2023-05-01 23:51:43,587 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.27 vs. limit=2.0 2023-05-01 23:52:01,349 INFO [zipformer.py:625] (1/8) 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] (1/8) 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 23:52:30,271 INFO [train.py:904] (1/8) Epoch 25, batch 3750, loss[loss=0.1636, simple_loss=0.2355, pruned_loss=0.04584, over 16904.00 frames. ], tot_loss[loss=0.167, simple_loss=0.2495, pruned_loss=0.0423, over 3280145.80 frames. ], batch size: 90, lr: 2.70e-03, grad_scale: 8.0 2023-05-01 23:53:06,853 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=247379.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 23:53:10,242 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=247381.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 23:53:42,910 INFO [train.py:904] (1/8) Epoch 25, batch 3800, loss[loss=0.1729, simple_loss=0.2498, pruned_loss=0.04796, over 16879.00 frames. ], tot_loss[loss=0.1691, simple_loss=0.2504, pruned_loss=0.04393, over 3280961.00 frames. ], batch size: 116, lr: 2.70e-03, grad_scale: 8.0 2023-05-01 23:54:15,332 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=247426.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 23:54:35,847 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=247440.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 23:54:38,774 INFO [optim.py:368] (1/8) 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 23:54:48,285 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.5329, 3.5049, 2.7321, 2.2069, 2.2201, 2.2943, 3.5560, 3.0549], device='cuda:1'), covar=tensor([0.2923, 0.0605, 0.1817, 0.3050, 0.2990, 0.2250, 0.0561, 0.1667], device='cuda:1'), in_proj_covar=tensor([0.0330, 0.0272, 0.0310, 0.0320, 0.0304, 0.0270, 0.0301, 0.0348], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:1') 2023-05-01 23:54:54,709 INFO [train.py:904] (1/8) Epoch 25, batch 3850, loss[loss=0.1971, simple_loss=0.2805, pruned_loss=0.05681, over 12416.00 frames. ], tot_loss[loss=0.1691, simple_loss=0.2501, pruned_loss=0.04402, over 3284894.51 frames. ], batch size: 247, lr: 2.70e-03, grad_scale: 8.0 2023-05-01 23:55:36,390 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.6389, 2.8263, 3.0824, 2.0460, 2.7568, 2.1710, 3.3151, 3.2238], device='cuda:1'), covar=tensor([0.0255, 0.0998, 0.0638, 0.2013, 0.0932, 0.1050, 0.0533, 0.0864], device='cuda:1'), in_proj_covar=tensor([0.0159, 0.0169, 0.0170, 0.0155, 0.0148, 0.0132, 0.0146, 0.0181], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0005], device='cuda:1') 2023-05-01 23:55:43,353 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=247487.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 23:55:53,625 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.3281, 3.3434, 3.5960, 2.5143, 3.3247, 3.6997, 3.3889, 2.1000], device='cuda:1'), covar=tensor([0.0502, 0.0138, 0.0064, 0.0383, 0.0115, 0.0090, 0.0100, 0.0503], device='cuda:1'), in_proj_covar=tensor([0.0137, 0.0088, 0.0089, 0.0136, 0.0101, 0.0112, 0.0098, 0.0132], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0003, 0.0004], device='cuda:1') 2023-05-01 23:56:02,904 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=247500.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 23:56:06,077 INFO [train.py:904] (1/8) Epoch 25, batch 3900, loss[loss=0.1702, simple_loss=0.2466, pruned_loss=0.04684, over 16863.00 frames. ], tot_loss[loss=0.1688, simple_loss=0.2493, pruned_loss=0.0441, over 3277930.27 frames. ], batch size: 116, lr: 2.70e-03, grad_scale: 8.0 2023-05-01 23:56:28,960 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([1.7877, 1.4047, 1.7241, 1.6531, 1.8063, 2.0240, 1.6700, 1.8815], device='cuda:1'), covar=tensor([0.0302, 0.0423, 0.0252, 0.0335, 0.0318, 0.0189, 0.0428, 0.0138], device='cuda:1'), in_proj_covar=tensor([0.0197, 0.0196, 0.0185, 0.0189, 0.0205, 0.0163, 0.0201, 0.0162], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-01 23:56:39,787 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=247525.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 23:56:57,772 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.95 vs. limit=2.0 2023-05-01 23:57:03,108 INFO [optim.py:368] (1/8) 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] (1/8) Epoch 25, batch 3950, loss[loss=0.1501, simple_loss=0.2285, pruned_loss=0.03584, over 16826.00 frames. ], tot_loss[loss=0.1692, simple_loss=0.2492, pruned_loss=0.04457, over 3275708.77 frames. ], batch size: 102, lr: 2.70e-03, grad_scale: 8.0 2023-05-01 23:57:19,954 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=247554.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 23:57:24,338 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.1533, 2.2235, 2.3260, 3.8434, 2.2343, 2.5038, 2.2903, 2.3930], device='cuda:1'), covar=tensor([0.1623, 0.3737, 0.3131, 0.0653, 0.4013, 0.2678, 0.4026, 0.3120], device='cuda:1'), in_proj_covar=tensor([0.0416, 0.0466, 0.0385, 0.0338, 0.0445, 0.0535, 0.0439, 0.0547], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-01 23:57:30,905 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=247561.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 23:58:05,357 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=247586.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 23:58:12,152 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=247590.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 23:58:30,185 INFO [train.py:904] (1/8) Epoch 25, batch 4000, loss[loss=0.1912, simple_loss=0.2702, pruned_loss=0.05605, over 15387.00 frames. ], tot_loss[loss=0.169, simple_loss=0.2489, pruned_loss=0.04452, over 3275589.09 frames. ], batch size: 190, lr: 2.70e-03, grad_scale: 8.0 2023-05-01 23:58:49,589 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=247615.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 23:59:20,992 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.4660, 3.9493, 3.9812, 2.6637, 3.6357, 4.0576, 3.6167, 2.3703], device='cuda:1'), covar=tensor([0.0550, 0.0080, 0.0056, 0.0410, 0.0098, 0.0095, 0.0105, 0.0450], device='cuda:1'), in_proj_covar=tensor([0.0138, 0.0088, 0.0089, 0.0136, 0.0101, 0.0113, 0.0098, 0.0132], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0003, 0.0004], device='cuda:1') 2023-05-01 23:59:21,866 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=247638.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 23:59:27,468 INFO [optim.py:368] (1/8) 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 23:59:44,159 INFO [train.py:904] (1/8) Epoch 25, batch 4050, loss[loss=0.1654, simple_loss=0.2457, pruned_loss=0.04251, over 16670.00 frames. ], tot_loss[loss=0.169, simple_loss=0.2498, pruned_loss=0.04415, over 3281088.22 frames. ], batch size: 62, lr: 2.70e-03, grad_scale: 8.0 2023-05-02 00:00:59,185 INFO [train.py:904] (1/8) Epoch 25, batch 4100, loss[loss=0.1961, simple_loss=0.2916, pruned_loss=0.05032, over 16836.00 frames. ], tot_loss[loss=0.1696, simple_loss=0.2517, pruned_loss=0.04376, over 3275412.05 frames. ], batch size: 116, lr: 2.70e-03, grad_scale: 8.0 2023-05-02 00:01:20,768 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.16 vs. limit=2.0 2023-05-02 00:01:48,424 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=247735.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 00:01:57,926 INFO [optim.py:368] (1/8) 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] (1/8) Epoch 25, batch 4150, loss[loss=0.1835, simple_loss=0.2833, pruned_loss=0.04178, over 16690.00 frames. ], tot_loss[loss=0.176, simple_loss=0.2589, pruned_loss=0.04657, over 3237903.32 frames. ], batch size: 134, lr: 2.70e-03, grad_scale: 8.0 2023-05-02 00:02:37,670 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.1526, 1.5066, 1.8976, 2.1017, 2.2281, 2.3739, 1.7948, 2.2988], device='cuda:1'), covar=tensor([0.0240, 0.0538, 0.0318, 0.0372, 0.0338, 0.0227, 0.0515, 0.0156], device='cuda:1'), in_proj_covar=tensor([0.0196, 0.0196, 0.0185, 0.0188, 0.0204, 0.0163, 0.0199, 0.0161], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-02 00:02:43,692 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.21 vs. limit=2.0 2023-05-02 00:03:01,855 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=247782.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 00:03:04,939 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=247784.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 00:03:30,059 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([1.8683, 2.7068, 2.5738, 1.8850, 2.5626, 2.7326, 2.6148, 1.8730], device='cuda:1'), covar=tensor([0.0467, 0.0097, 0.0098, 0.0402, 0.0133, 0.0143, 0.0128, 0.0424], device='cuda:1'), in_proj_covar=tensor([0.0137, 0.0087, 0.0088, 0.0135, 0.0101, 0.0111, 0.0098, 0.0131], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0003, 0.0004], device='cuda:1') 2023-05-02 00:03:32,729 INFO [train.py:904] (1/8) Epoch 25, batch 4200, loss[loss=0.2211, simple_loss=0.295, pruned_loss=0.07361, over 11188.00 frames. ], tot_loss[loss=0.1816, simple_loss=0.266, pruned_loss=0.04858, over 3188158.21 frames. ], batch size: 246, lr: 2.70e-03, grad_scale: 8.0 2023-05-02 00:04:05,861 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.16 vs. limit=2.0 2023-05-02 00:04:30,292 INFO [optim.py:368] (1/8) 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 00:04:35,208 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 25, batch 4250, loss[loss=0.1735, simple_loss=0.2543, pruned_loss=0.04634, over 12051.00 frames. ], tot_loss[loss=0.182, simple_loss=0.2686, pruned_loss=0.04774, over 3169495.30 frames. ], batch size: 248, lr: 2.70e-03, grad_scale: 8.0 2023-05-02 00:04:51,426 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=247856.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 00:05:15,531 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([1.7269, 2.3222, 2.2221, 2.9635, 1.9246, 3.4355, 1.5698, 2.6205], device='cuda:1'), covar=tensor([0.1394, 0.0922, 0.1437, 0.0187, 0.0152, 0.0445, 0.1825, 0.1013], device='cuda:1'), in_proj_covar=tensor([0.0169, 0.0178, 0.0197, 0.0197, 0.0206, 0.0218, 0.0206, 0.0197], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-05-02 00:05:28,692 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=247881.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 00:06:02,532 INFO [train.py:904] (1/8) Epoch 25, batch 4300, loss[loss=0.1925, simple_loss=0.2884, pruned_loss=0.04824, over 16689.00 frames. ], tot_loss[loss=0.1817, simple_loss=0.2699, pruned_loss=0.04672, over 3181459.45 frames. ], batch size: 62, lr: 2.70e-03, grad_scale: 8.0 2023-05-02 00:06:13,453 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=247910.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 00:07:01,391 INFO [optim.py:368] (1/8) Clipping_scale=2.0, grad-norm quartiles 1.557e+02 2.163e+02 2.477e+02 3.039e+02 5.024e+02, threshold=4.955e+02, percent-clipped=0.0 2023-05-02 00:07:17,420 INFO [train.py:904] (1/8) Epoch 25, batch 4350, loss[loss=0.1918, simple_loss=0.2869, pruned_loss=0.04838, over 17104.00 frames. ], tot_loss[loss=0.1839, simple_loss=0.2729, pruned_loss=0.04747, over 3168185.08 frames. ], batch size: 47, lr: 2.70e-03, grad_scale: 8.0 2023-05-02 00:07:29,751 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.0617, 2.4376, 2.6471, 1.9372, 2.7463, 2.7933, 2.3922, 2.3676], device='cuda:1'), covar=tensor([0.0734, 0.0269, 0.0238, 0.0923, 0.0119, 0.0235, 0.0473, 0.0423], device='cuda:1'), in_proj_covar=tensor([0.0147, 0.0111, 0.0100, 0.0139, 0.0085, 0.0130, 0.0130, 0.0131], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-05-02 00:07:46,503 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.4005, 3.3345, 3.7461, 1.9083, 3.8853, 3.9561, 2.9019, 2.9783], device='cuda:1'), covar=tensor([0.0889, 0.0285, 0.0219, 0.1178, 0.0088, 0.0130, 0.0508, 0.0443], device='cuda:1'), in_proj_covar=tensor([0.0148, 0.0111, 0.0100, 0.0139, 0.0085, 0.0130, 0.0130, 0.0131], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-05-02 00:08:05,156 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=247984.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 00:08:23,280 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 25, batch 4400, loss[loss=0.185, simple_loss=0.2803, pruned_loss=0.04491, over 16642.00 frames. ], tot_loss[loss=0.1862, simple_loss=0.2753, pruned_loss=0.04856, over 3168776.68 frames. ], batch size: 57, lr: 2.70e-03, grad_scale: 8.0 2023-05-02 00:09:04,084 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([1.8521, 1.4204, 1.7014, 1.7340, 1.7636, 2.0143, 1.5985, 1.8407], device='cuda:1'), covar=tensor([0.0274, 0.0413, 0.0242, 0.0291, 0.0288, 0.0177, 0.0481, 0.0134], device='cuda:1'), in_proj_covar=tensor([0.0195, 0.0195, 0.0184, 0.0188, 0.0203, 0.0161, 0.0199, 0.0161], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-02 00:09:24,792 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=248035.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 00:09:27,637 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.8314, 4.3031, 3.0451, 2.6028, 3.0028, 2.6760, 4.6936, 3.8766], device='cuda:1'), covar=tensor([0.2878, 0.0559, 0.1796, 0.2363, 0.2429, 0.1974, 0.0379, 0.1073], device='cuda:1'), in_proj_covar=tensor([0.0331, 0.0274, 0.0311, 0.0321, 0.0305, 0.0270, 0.0302, 0.0348], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:1') 2023-05-02 00:09:34,454 INFO [optim.py:368] (1/8) 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 00:09:39,339 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 25, batch 4450, loss[loss=0.1916, simple_loss=0.2787, pruned_loss=0.05229, over 12180.00 frames. ], tot_loss[loss=0.1896, simple_loss=0.2792, pruned_loss=0.05004, over 3184993.86 frames. ], batch size: 248, lr: 2.70e-03, grad_scale: 8.0 2023-05-02 00:09:56,469 INFO [zipformer.py:625] (1/8) 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] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=248082.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 00:10:33,207 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=248082.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 00:10:34,921 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=248083.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 00:11:04,508 INFO [train.py:904] (1/8) Epoch 25, batch 4500, loss[loss=0.1864, simple_loss=0.2762, pruned_loss=0.04827, over 16583.00 frames. ], tot_loss[loss=0.1906, simple_loss=0.2799, pruned_loss=0.05064, over 3187261.88 frames. ], batch size: 68, lr: 2.70e-03, grad_scale: 8.0 2023-05-02 00:11:20,233 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.4093, 3.2252, 3.5826, 1.8623, 3.7527, 3.8028, 2.8807, 2.8910], device='cuda:1'), covar=tensor([0.0814, 0.0312, 0.0226, 0.1215, 0.0089, 0.0155, 0.0504, 0.0458], device='cuda:1'), in_proj_covar=tensor([0.0149, 0.0112, 0.0101, 0.0140, 0.0085, 0.0131, 0.0131, 0.0132], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-05-02 00:11:44,502 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=248130.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 00:11:59,981 INFO [zipformer.py:625] (1/8) 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] (1/8) 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 00:12:03,637 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 25, batch 4550, loss[loss=0.2027, simple_loss=0.2897, pruned_loss=0.05786, over 16511.00 frames. ], tot_loss[loss=0.192, simple_loss=0.2808, pruned_loss=0.05155, over 3205554.12 frames. ], batch size: 68, lr: 2.70e-03, grad_scale: 8.0 2023-05-02 00:12:22,860 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=248156.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 00:12:43,449 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.45 vs. limit=2.0 2023-05-02 00:12:50,977 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.1991, 4.0147, 4.1518, 4.3727, 4.4455, 4.1002, 4.4457, 4.5219], device='cuda:1'), covar=tensor([0.1560, 0.1253, 0.1575, 0.0743, 0.0680, 0.1434, 0.0877, 0.0698], device='cuda:1'), in_proj_covar=tensor([0.0662, 0.0815, 0.0939, 0.0824, 0.0632, 0.0649, 0.0676, 0.0790], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-02 00:13:00,048 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=248181.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 00:13:03,842 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=4.84 vs. limit=5.0 2023-05-02 00:13:13,537 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=4.55 vs. limit=5.0 2023-05-02 00:13:32,543 INFO [train.py:904] (1/8) Epoch 25, batch 4600, loss[loss=0.1975, simple_loss=0.2859, pruned_loss=0.05452, over 16629.00 frames. ], tot_loss[loss=0.1929, simple_loss=0.2819, pruned_loss=0.05192, over 3208644.00 frames. ], batch size: 68, lr: 2.70e-03, grad_scale: 8.0 2023-05-02 00:13:34,235 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=248204.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 00:13:42,758 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=248210.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 00:14:00,749 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.1873, 4.0743, 4.2492, 4.3782, 4.4570, 4.0747, 4.4233, 4.5228], device='cuda:1'), covar=tensor([0.1478, 0.1099, 0.1190, 0.0592, 0.0501, 0.1223, 0.0831, 0.0539], device='cuda:1'), in_proj_covar=tensor([0.0660, 0.0814, 0.0937, 0.0822, 0.0630, 0.0649, 0.0675, 0.0789], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-02 00:14:11,937 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=248229.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 00:14:29,190 INFO [optim.py:368] (1/8) 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 00:14:42,933 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.26 vs. limit=2.0 2023-05-02 00:14:46,950 INFO [train.py:904] (1/8) Epoch 25, batch 4650, loss[loss=0.1872, simple_loss=0.2666, pruned_loss=0.0539, over 16691.00 frames. ], tot_loss[loss=0.1926, simple_loss=0.2809, pruned_loss=0.05212, over 3221688.55 frames. ], batch size: 62, lr: 2.70e-03, grad_scale: 8.0 2023-05-02 00:14:54,068 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=248258.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 00:16:01,797 INFO [train.py:904] (1/8) Epoch 25, batch 4700, loss[loss=0.191, simple_loss=0.2867, pruned_loss=0.04764, over 16343.00 frames. ], tot_loss[loss=0.19, simple_loss=0.2781, pruned_loss=0.05094, over 3218736.84 frames. ], batch size: 146, lr: 2.70e-03, grad_scale: 8.0 2023-05-02 00:16:44,278 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=248332.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 00:16:56,009 INFO [zipformer.py:625] (1/8) 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] (1/8) Clipping_scale=2.0, grad-norm quartiles 1.458e+02 1.897e+02 2.230e+02 2.532e+02 4.236e+02, threshold=4.460e+02, percent-clipped=2.0 2023-05-02 00:17:13,836 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 25, batch 4750, loss[loss=0.1884, simple_loss=0.2723, pruned_loss=0.05227, over 11931.00 frames. ], tot_loss[loss=0.1868, simple_loss=0.2748, pruned_loss=0.04936, over 3196763.43 frames. ], batch size: 248, lr: 2.69e-03, grad_scale: 8.0 2023-05-02 00:17:36,321 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.30 vs. limit=2.0 2023-05-02 00:17:59,804 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=248383.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 00:18:13,285 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=248393.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 00:18:29,679 INFO [train.py:904] (1/8) Epoch 25, batch 4800, loss[loss=0.1819, simple_loss=0.2756, pruned_loss=0.04408, over 16745.00 frames. ], tot_loss[loss=0.183, simple_loss=0.2714, pruned_loss=0.04731, over 3204849.01 frames. ], batch size: 134, lr: 2.69e-03, grad_scale: 8.0 2023-05-02 00:19:22,193 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=248438.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 00:19:24,723 INFO [zipformer.py:625] (1/8) 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] (1/8) 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 00:19:31,564 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=248444.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 00:19:43,847 INFO [train.py:904] (1/8) Epoch 25, batch 4850, loss[loss=0.1792, simple_loss=0.2712, pruned_loss=0.04363, over 16690.00 frames. ], tot_loss[loss=0.183, simple_loss=0.2727, pruned_loss=0.04663, over 3189834.09 frames. ], batch size: 89, lr: 2.69e-03, grad_scale: 8.0 2023-05-02 00:20:36,774 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=248488.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 00:20:50,055 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.9791, 5.0741, 4.8630, 4.4519, 4.5104, 4.9584, 4.8234, 4.6370], device='cuda:1'), covar=tensor([0.0611, 0.0435, 0.0305, 0.0331, 0.1068, 0.0524, 0.0331, 0.0675], device='cuda:1'), in_proj_covar=tensor([0.0302, 0.0452, 0.0353, 0.0358, 0.0357, 0.0411, 0.0242, 0.0425], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002], device='cuda:1') 2023-05-02 00:20:54,048 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.6224, 4.6236, 4.5012, 3.6721, 4.5626, 1.7230, 4.2894, 4.1672], device='cuda:1'), covar=tensor([0.0097, 0.0093, 0.0164, 0.0447, 0.0094, 0.2895, 0.0133, 0.0257], device='cuda:1'), in_proj_covar=tensor([0.0174, 0.0167, 0.0207, 0.0184, 0.0184, 0.0213, 0.0197, 0.0178], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-02 00:20:58,427 INFO [train.py:904] (1/8) Epoch 25, batch 4900, loss[loss=0.166, simple_loss=0.261, pruned_loss=0.03556, over 16684.00 frames. ], tot_loss[loss=0.1813, simple_loss=0.2716, pruned_loss=0.04551, over 3173423.29 frames. ], batch size: 89, lr: 2.69e-03, grad_scale: 8.0 2023-05-02 00:21:47,492 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.6160, 4.8858, 4.6678, 4.7088, 4.4573, 4.4107, 4.2973, 4.9370], device='cuda:1'), covar=tensor([0.1159, 0.0816, 0.0866, 0.0726, 0.0772, 0.1203, 0.1138, 0.0805], device='cuda:1'), in_proj_covar=tensor([0.0691, 0.0830, 0.0686, 0.0643, 0.0530, 0.0531, 0.0702, 0.0653], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:1') 2023-05-02 00:21:47,902 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=4.75 vs. limit=5.0 2023-05-02 00:22:03,445 INFO [optim.py:368] (1/8) 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 00:22:20,178 INFO [train.py:904] (1/8) Epoch 25, batch 4950, loss[loss=0.1695, simple_loss=0.2698, pruned_loss=0.03458, over 16716.00 frames. ], tot_loss[loss=0.1803, simple_loss=0.271, pruned_loss=0.04484, over 3192477.73 frames. ], batch size: 89, lr: 2.69e-03, grad_scale: 8.0 2023-05-02 00:23:22,360 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.1590, 2.3503, 2.4227, 3.9076, 2.2407, 2.6679, 2.3995, 2.5091], device='cuda:1'), covar=tensor([0.1440, 0.3350, 0.2828, 0.0570, 0.3863, 0.2392, 0.3482, 0.3182], device='cuda:1'), in_proj_covar=tensor([0.0415, 0.0465, 0.0381, 0.0335, 0.0442, 0.0532, 0.0436, 0.0542], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-02 00:23:31,264 INFO [train.py:904] (1/8) Epoch 25, batch 5000, loss[loss=0.1794, simple_loss=0.275, pruned_loss=0.0419, over 16669.00 frames. ], tot_loss[loss=0.1815, simple_loss=0.273, pruned_loss=0.04506, over 3206155.77 frames. ], batch size: 134, lr: 2.69e-03, grad_scale: 8.0 2023-05-02 00:23:37,421 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([5.3576, 5.3586, 5.6838, 5.6592, 5.7433, 5.3916, 5.2778, 5.0656], device='cuda:1'), covar=tensor([0.0288, 0.0423, 0.0355, 0.0413, 0.0539, 0.0330, 0.1152, 0.0415], device='cuda:1'), in_proj_covar=tensor([0.0415, 0.0465, 0.0456, 0.0416, 0.0500, 0.0476, 0.0554, 0.0379], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:1') 2023-05-02 00:24:25,292 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=248640.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 00:24:27,921 INFO [optim.py:368] (1/8) Clipping_scale=2.0, grad-norm quartiles 1.446e+02 2.126e+02 2.416e+02 2.928e+02 6.487e+02, threshold=4.832e+02, percent-clipped=1.0 2023-05-02 00:24:42,934 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=248652.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 00:24:43,840 INFO [train.py:904] (1/8) Epoch 25, batch 5050, loss[loss=0.1917, simple_loss=0.2787, pruned_loss=0.0523, over 16845.00 frames. ], tot_loss[loss=0.1818, simple_loss=0.2734, pruned_loss=0.04513, over 3206862.14 frames. ], batch size: 42, lr: 2.69e-03, grad_scale: 8.0 2023-05-02 00:25:35,140 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=248688.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 00:25:35,209 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=248688.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 00:25:39,632 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=248691.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 00:25:53,427 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 25, batch 5100, loss[loss=0.1549, simple_loss=0.2455, pruned_loss=0.03214, over 16704.00 frames. ], tot_loss[loss=0.1797, simple_loss=0.2712, pruned_loss=0.04416, over 3207627.27 frames. ], batch size: 57, lr: 2.69e-03, grad_scale: 8.0 2023-05-02 00:26:15,836 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=248715.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 00:26:50,028 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=248738.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 00:26:52,082 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=248739.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 00:26:56,587 INFO [optim.py:368] (1/8) 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 00:27:12,719 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=248752.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 00:27:13,425 INFO [train.py:904] (1/8) Epoch 25, batch 5150, loss[loss=0.1698, simple_loss=0.2708, pruned_loss=0.03433, over 16874.00 frames. ], tot_loss[loss=0.179, simple_loss=0.2714, pruned_loss=0.04332, over 3208225.78 frames. ], batch size: 116, lr: 2.69e-03, grad_scale: 8.0 2023-05-02 00:27:46,950 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=248776.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 00:28:03,431 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=248786.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 00:28:28,598 INFO [train.py:904] (1/8) Epoch 25, batch 5200, loss[loss=0.1898, simple_loss=0.2859, pruned_loss=0.0469, over 16333.00 frames. ], tot_loss[loss=0.1781, simple_loss=0.2706, pruned_loss=0.04276, over 3213655.98 frames. ], batch size: 146, lr: 2.69e-03, grad_scale: 8.0 2023-05-02 00:29:25,194 INFO [optim.py:368] (1/8) 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] (1/8) Epoch 25, batch 5250, loss[loss=0.1739, simple_loss=0.2661, pruned_loss=0.04083, over 16954.00 frames. ], tot_loss[loss=0.1764, simple_loss=0.268, pruned_loss=0.04241, over 3222356.21 frames. ], batch size: 109, lr: 2.69e-03, grad_scale: 8.0 2023-05-02 00:29:48,232 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.4563, 2.7856, 3.0942, 2.0132, 2.8031, 2.1358, 3.0990, 3.0386], device='cuda:1'), covar=tensor([0.0270, 0.0855, 0.0633, 0.1950, 0.0869, 0.1008, 0.0641, 0.0814], device='cuda:1'), in_proj_covar=tensor([0.0160, 0.0168, 0.0171, 0.0156, 0.0148, 0.0132, 0.0146, 0.0180], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0005], device='cuda:1') 2023-05-02 00:30:53,681 INFO [train.py:904] (1/8) Epoch 25, batch 5300, loss[loss=0.187, simple_loss=0.2708, pruned_loss=0.05157, over 15455.00 frames. ], tot_loss[loss=0.1738, simple_loss=0.2646, pruned_loss=0.04151, over 3223531.20 frames. ], batch size: 190, lr: 2.69e-03, grad_scale: 8.0 2023-05-02 00:31:49,261 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.2038, 4.2773, 4.0911, 3.8160, 3.7926, 4.1898, 3.8847, 3.9468], device='cuda:1'), covar=tensor([0.0575, 0.0584, 0.0333, 0.0313, 0.0946, 0.0568, 0.0773, 0.0599], device='cuda:1'), in_proj_covar=tensor([0.0304, 0.0458, 0.0356, 0.0361, 0.0361, 0.0417, 0.0243, 0.0429], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002], device='cuda:1') 2023-05-02 00:31:51,342 INFO [optim.py:368] (1/8) Clipping_scale=2.0, grad-norm quartiles 1.416e+02 1.862e+02 2.129e+02 2.555e+02 4.739e+02, threshold=4.259e+02, percent-clipped=0.0 2023-05-02 00:32:08,016 INFO [train.py:904] (1/8) Epoch 25, batch 5350, loss[loss=0.193, simple_loss=0.2755, pruned_loss=0.05518, over 17074.00 frames. ], tot_loss[loss=0.1725, simple_loss=0.2629, pruned_loss=0.04103, over 3237843.78 frames. ], batch size: 55, lr: 2.69e-03, grad_scale: 8.0 2023-05-02 00:33:00,479 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=248988.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 00:33:05,598 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.20 vs. limit=2.0 2023-05-02 00:33:22,807 INFO [train.py:904] (1/8) Epoch 25, batch 5400, loss[loss=0.1999, simple_loss=0.2863, pruned_loss=0.05676, over 11680.00 frames. ], tot_loss[loss=0.1736, simple_loss=0.2649, pruned_loss=0.04119, over 3234414.35 frames. ], batch size: 246, lr: 2.69e-03, grad_scale: 8.0 2023-05-02 00:33:51,404 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([1.7501, 1.4413, 1.6811, 1.7003, 1.8266, 1.9766, 1.6536, 1.8074], device='cuda:1'), covar=tensor([0.0287, 0.0413, 0.0240, 0.0348, 0.0303, 0.0200, 0.0466, 0.0146], device='cuda:1'), in_proj_covar=tensor([0.0195, 0.0195, 0.0182, 0.0188, 0.0202, 0.0161, 0.0200, 0.0161], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-02 00:34:11,323 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=249036.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 00:34:17,273 INFO [zipformer.py:625] (1/8) 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] (1/8) 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 00:34:29,104 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 25, batch 5450, loss[loss=0.229, simple_loss=0.3136, pruned_loss=0.07213, over 15381.00 frames. ], tot_loss[loss=0.1766, simple_loss=0.2678, pruned_loss=0.0427, over 3218200.24 frames. ], batch size: 191, lr: 2.69e-03, grad_scale: 16.0 2023-05-02 00:35:08,682 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=249071.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 00:35:27,106 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.5543, 4.6196, 4.4251, 4.0964, 4.1304, 4.5145, 4.2738, 4.2421], device='cuda:1'), covar=tensor([0.0590, 0.0764, 0.0301, 0.0321, 0.0841, 0.0616, 0.0537, 0.0618], device='cuda:1'), in_proj_covar=tensor([0.0302, 0.0457, 0.0355, 0.0360, 0.0359, 0.0416, 0.0242, 0.0428], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002], device='cuda:1') 2023-05-02 00:35:29,387 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.79 vs. limit=2.0 2023-05-02 00:35:34,360 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 25, batch 5500, loss[loss=0.1838, simple_loss=0.2755, pruned_loss=0.0461, over 17272.00 frames. ], tot_loss[loss=0.1841, simple_loss=0.2749, pruned_loss=0.04663, over 3201812.15 frames. ], batch size: 52, lr: 2.69e-03, grad_scale: 16.0 2023-05-02 00:36:09,884 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=4.26 vs. limit=5.0 2023-05-02 00:37:00,873 INFO [optim.py:368] (1/8) 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 00:37:17,247 INFO [train.py:904] (1/8) Epoch 25, batch 5550, loss[loss=0.2218, simple_loss=0.3093, pruned_loss=0.06708, over 16183.00 frames. ], tot_loss[loss=0.1922, simple_loss=0.2814, pruned_loss=0.0515, over 3156708.87 frames. ], batch size: 165, lr: 2.69e-03, grad_scale: 16.0 2023-05-02 00:38:41,396 INFO [train.py:904] (1/8) Epoch 25, batch 5600, loss[loss=0.2004, simple_loss=0.2893, pruned_loss=0.05569, over 17253.00 frames. ], tot_loss[loss=0.1991, simple_loss=0.2868, pruned_loss=0.05572, over 3123787.73 frames. ], batch size: 52, lr: 2.69e-03, grad_scale: 16.0 2023-05-02 00:38:53,450 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.1236, 2.3492, 2.3797, 2.7740, 2.0766, 3.2231, 1.8821, 2.7372], device='cuda:1'), covar=tensor([0.1056, 0.0555, 0.1005, 0.0184, 0.0146, 0.0412, 0.1357, 0.0675], device='cuda:1'), in_proj_covar=tensor([0.0169, 0.0178, 0.0196, 0.0195, 0.0205, 0.0215, 0.0206, 0.0196], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-05-02 00:39:02,196 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=249215.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 00:39:29,096 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.65 vs. limit=2.0 2023-05-02 00:39:47,610 INFO [optim.py:368] (1/8) 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 00:39:51,514 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([5.1898, 5.1779, 5.0620, 4.6656, 4.7233, 5.0734, 5.0690, 4.7892], device='cuda:1'), covar=tensor([0.0636, 0.0497, 0.0312, 0.0346, 0.1123, 0.0530, 0.0287, 0.0771], device='cuda:1'), in_proj_covar=tensor([0.0302, 0.0453, 0.0352, 0.0357, 0.0358, 0.0413, 0.0241, 0.0425], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002], device='cuda:1') 2023-05-02 00:40:04,580 INFO [train.py:904] (1/8) Epoch 25, batch 5650, loss[loss=0.2064, simple_loss=0.2963, pruned_loss=0.05827, over 16445.00 frames. ], tot_loss[loss=0.206, simple_loss=0.2918, pruned_loss=0.06011, over 3080408.92 frames. ], batch size: 68, lr: 2.69e-03, grad_scale: 16.0 2023-05-02 00:40:30,519 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.14 vs. limit=2.0 2023-05-02 00:40:37,221 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([1.8277, 2.7040, 2.6303, 1.9052, 2.5535, 2.7400, 2.5949, 1.9792], device='cuda:1'), covar=tensor([0.0479, 0.0098, 0.0102, 0.0425, 0.0152, 0.0144, 0.0130, 0.0418], device='cuda:1'), in_proj_covar=tensor([0.0137, 0.0088, 0.0089, 0.0136, 0.0100, 0.0112, 0.0097, 0.0132], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0003, 0.0004], device='cuda:1') 2023-05-02 00:40:41,262 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=249276.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 00:41:22,856 INFO [train.py:904] (1/8) Epoch 25, batch 5700, loss[loss=0.1934, simple_loss=0.2815, pruned_loss=0.05268, over 17145.00 frames. ], tot_loss[loss=0.2075, simple_loss=0.2928, pruned_loss=0.0611, over 3069318.35 frames. ], batch size: 46, lr: 2.69e-03, grad_scale: 16.0 2023-05-02 00:42:25,335 INFO [optim.py:368] (1/8) 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 00:42:34,397 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 25, batch 5750, loss[loss=0.1925, simple_loss=0.28, pruned_loss=0.0525, over 16702.00 frames. ], tot_loss[loss=0.2104, simple_loss=0.2953, pruned_loss=0.06273, over 3046194.91 frames. ], batch size: 134, lr: 2.69e-03, grad_scale: 16.0 2023-05-02 00:43:13,849 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=249371.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 00:43:17,303 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.1815, 2.1744, 2.7189, 3.1217, 3.0020, 3.6553, 2.2711, 3.6042], device='cuda:1'), covar=tensor([0.0236, 0.0515, 0.0324, 0.0324, 0.0309, 0.0159, 0.0570, 0.0158], device='cuda:1'), in_proj_covar=tensor([0.0192, 0.0193, 0.0180, 0.0186, 0.0200, 0.0159, 0.0197, 0.0159], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-02 00:43:53,995 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=249395.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 00:44:07,472 INFO [train.py:904] (1/8) Epoch 25, batch 5800, loss[loss=0.2098, simple_loss=0.2827, pruned_loss=0.06848, over 12130.00 frames. ], tot_loss[loss=0.2092, simple_loss=0.295, pruned_loss=0.06176, over 3048567.24 frames. ], batch size: 248, lr: 2.69e-03, grad_scale: 16.0 2023-05-02 00:44:32,286 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=249419.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 00:45:09,430 INFO [optim.py:368] (1/8) 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 00:45:26,290 INFO [train.py:904] (1/8) Epoch 25, batch 5850, loss[loss=0.1964, simple_loss=0.288, pruned_loss=0.05234, over 16744.00 frames. ], tot_loss[loss=0.2058, simple_loss=0.2922, pruned_loss=0.05976, over 3056732.67 frames. ], batch size: 83, lr: 2.69e-03, grad_scale: 16.0 2023-05-02 00:46:46,951 INFO [train.py:904] (1/8) Epoch 25, batch 5900, loss[loss=0.1808, simple_loss=0.2792, pruned_loss=0.04117, over 16892.00 frames. ], tot_loss[loss=0.2053, simple_loss=0.292, pruned_loss=0.05927, over 3077326.87 frames. ], batch size: 90, lr: 2.69e-03, grad_scale: 16.0 2023-05-02 00:47:24,308 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.9105, 2.1254, 2.5097, 3.1167, 2.2429, 2.3348, 2.3611, 2.2217], device='cuda:1'), covar=tensor([0.1402, 0.3516, 0.2302, 0.0691, 0.3919, 0.2518, 0.3226, 0.3528], device='cuda:1'), in_proj_covar=tensor([0.0410, 0.0459, 0.0376, 0.0330, 0.0439, 0.0525, 0.0429, 0.0535], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-02 00:47:40,181 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.5239, 1.7051, 2.1630, 2.4080, 2.4572, 2.7553, 1.9500, 2.7092], device='cuda:1'), covar=tensor([0.0224, 0.0567, 0.0359, 0.0365, 0.0348, 0.0209, 0.0569, 0.0145], device='cuda:1'), in_proj_covar=tensor([0.0193, 0.0193, 0.0181, 0.0186, 0.0201, 0.0160, 0.0198, 0.0159], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-02 00:47:52,232 INFO [optim.py:368] (1/8) 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 00:48:08,065 INFO [train.py:904] (1/8) Epoch 25, batch 5950, loss[loss=0.1945, simple_loss=0.2788, pruned_loss=0.05507, over 16202.00 frames. ], tot_loss[loss=0.2044, simple_loss=0.2929, pruned_loss=0.05798, over 3088032.90 frames. ], batch size: 165, lr: 2.69e-03, grad_scale: 16.0 2023-05-02 00:48:36,947 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=249571.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 00:49:09,195 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.8544, 2.7208, 2.8103, 2.1815, 2.6663, 2.1516, 2.7824, 2.9051], device='cuda:1'), covar=tensor([0.0295, 0.0800, 0.0570, 0.1769, 0.0873, 0.0959, 0.0590, 0.0680], device='cuda:1'), in_proj_covar=tensor([0.0158, 0.0167, 0.0169, 0.0155, 0.0147, 0.0131, 0.0145, 0.0179], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:1') 2023-05-02 00:49:28,142 INFO [train.py:904] (1/8) Epoch 25, batch 6000, loss[loss=0.2459, simple_loss=0.3164, pruned_loss=0.08771, over 11269.00 frames. ], tot_loss[loss=0.2027, simple_loss=0.2912, pruned_loss=0.05717, over 3092443.34 frames. ], batch size: 248, lr: 2.69e-03, grad_scale: 16.0 2023-05-02 00:49:28,143 INFO [train.py:929] (1/8) Computing validation loss 2023-05-02 00:49:38,597 INFO [train.py:938] (1/8) Epoch 25, validation: loss=0.1487, simple_loss=0.2613, pruned_loss=0.01811, over 944034.00 frames. 2023-05-02 00:49:38,598 INFO [train.py:939] (1/8) Maximum memory allocated so far is 18145MB 2023-05-02 00:50:17,381 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.5019, 3.5085, 3.4725, 2.7686, 3.3573, 2.1511, 3.1929, 2.8480], device='cuda:1'), covar=tensor([0.0167, 0.0136, 0.0198, 0.0246, 0.0116, 0.2333, 0.0150, 0.0269], device='cuda:1'), in_proj_covar=tensor([0.0174, 0.0167, 0.0206, 0.0184, 0.0184, 0.0213, 0.0196, 0.0177], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-02 00:50:36,538 INFO [optim.py:368] (1/8) Clipping_scale=2.0, grad-norm quartiles 1.971e+02 2.903e+02 3.391e+02 3.955e+02 5.249e+02, threshold=6.782e+02, percent-clipped=0.0 2023-05-02 00:50:54,725 INFO [train.py:904] (1/8) Epoch 25, batch 6050, loss[loss=0.192, simple_loss=0.2898, pruned_loss=0.04711, over 16659.00 frames. ], tot_loss[loss=0.201, simple_loss=0.2891, pruned_loss=0.05644, over 3102509.00 frames. ], batch size: 134, lr: 2.69e-03, grad_scale: 16.0 2023-05-02 00:51:54,553 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.8284, 2.0431, 2.3669, 2.7800, 2.7355, 3.1961, 2.0756, 3.1666], device='cuda:1'), covar=tensor([0.0233, 0.0496, 0.0377, 0.0334, 0.0345, 0.0190, 0.0576, 0.0151], device='cuda:1'), in_proj_covar=tensor([0.0193, 0.0193, 0.0182, 0.0186, 0.0201, 0.0160, 0.0198, 0.0159], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-02 00:52:12,397 INFO [train.py:904] (1/8) Epoch 25, batch 6100, loss[loss=0.1929, simple_loss=0.2811, pruned_loss=0.0523, over 17170.00 frames. ], tot_loss[loss=0.1997, simple_loss=0.2886, pruned_loss=0.05541, over 3109491.15 frames. ], batch size: 46, lr: 2.69e-03, grad_scale: 16.0 2023-05-02 00:53:00,758 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.6463, 1.8475, 2.2626, 2.5699, 2.6113, 2.9498, 1.9721, 2.8359], device='cuda:1'), covar=tensor([0.0227, 0.0539, 0.0339, 0.0340, 0.0316, 0.0174, 0.0551, 0.0146], device='cuda:1'), in_proj_covar=tensor([0.0193, 0.0193, 0.0182, 0.0186, 0.0201, 0.0160, 0.0198, 0.0159], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-02 00:53:15,241 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=5.05 vs. limit=5.0 2023-05-02 00:53:15,555 INFO [optim.py:368] (1/8) 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] (1/8) Epoch 25, batch 6150, loss[loss=0.1907, simple_loss=0.2798, pruned_loss=0.05086, over 16395.00 frames. ], tot_loss[loss=0.1979, simple_loss=0.2863, pruned_loss=0.05475, over 3113936.55 frames. ], batch size: 146, lr: 2.69e-03, grad_scale: 8.0 2023-05-02 00:54:11,177 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.15 vs. limit=2.0 2023-05-02 00:54:45,413 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.7833, 1.9411, 2.3809, 2.7568, 2.7312, 3.1769, 2.0869, 3.0860], device='cuda:1'), covar=tensor([0.0262, 0.0564, 0.0389, 0.0349, 0.0362, 0.0206, 0.0583, 0.0173], device='cuda:1'), in_proj_covar=tensor([0.0194, 0.0194, 0.0182, 0.0187, 0.0202, 0.0161, 0.0199, 0.0160], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-02 00:54:51,553 INFO [train.py:904] (1/8) Epoch 25, batch 6200, loss[loss=0.2064, simple_loss=0.2937, pruned_loss=0.05952, over 16152.00 frames. ], tot_loss[loss=0.1973, simple_loss=0.2851, pruned_loss=0.05472, over 3100774.86 frames. ], batch size: 165, lr: 2.69e-03, grad_scale: 8.0 2023-05-02 00:55:09,960 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=249814.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 00:55:36,573 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.9937, 2.3978, 2.0788, 2.2484, 2.7264, 2.4084, 2.5773, 2.9242], device='cuda:1'), covar=tensor([0.0205, 0.0477, 0.0573, 0.0529, 0.0298, 0.0456, 0.0258, 0.0302], device='cuda:1'), in_proj_covar=tensor([0.0221, 0.0240, 0.0228, 0.0230, 0.0240, 0.0238, 0.0240, 0.0237], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-02 00:55:55,232 INFO [optim.py:368] (1/8) 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 00:56:06,949 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=3.03 vs. limit=5.0 2023-05-02 00:56:10,145 INFO [train.py:904] (1/8) Epoch 25, batch 6250, loss[loss=0.2001, simple_loss=0.2794, pruned_loss=0.0604, over 11906.00 frames. ], tot_loss[loss=0.1972, simple_loss=0.285, pruned_loss=0.05477, over 3107365.08 frames. ], batch size: 248, lr: 2.69e-03, grad_scale: 8.0 2023-05-02 00:56:28,446 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.6710, 3.4197, 3.8303, 1.9485, 3.9644, 4.0113, 3.0978, 3.0547], device='cuda:1'), covar=tensor([0.0764, 0.0305, 0.0208, 0.1240, 0.0093, 0.0164, 0.0441, 0.0442], device='cuda:1'), in_proj_covar=tensor([0.0149, 0.0111, 0.0101, 0.0140, 0.0085, 0.0130, 0.0130, 0.0132], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-05-02 00:56:38,410 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=249871.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 00:56:44,442 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=249875.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 00:56:52,956 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.7411, 3.9845, 2.9888, 2.4552, 2.7178, 2.6844, 4.4120, 3.5445], device='cuda:1'), covar=tensor([0.3132, 0.0759, 0.1976, 0.2919, 0.2795, 0.1991, 0.0458, 0.1410], device='cuda:1'), in_proj_covar=tensor([0.0332, 0.0272, 0.0311, 0.0320, 0.0303, 0.0269, 0.0301, 0.0345], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:1') 2023-05-02 00:56:57,921 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.5110, 2.4664, 2.5090, 4.3938, 2.4002, 2.9179, 2.5378, 2.7147], device='cuda:1'), covar=tensor([0.1300, 0.3627, 0.2839, 0.0461, 0.3898, 0.2362, 0.3588, 0.3035], device='cuda:1'), in_proj_covar=tensor([0.0414, 0.0463, 0.0380, 0.0332, 0.0442, 0.0530, 0.0433, 0.0540], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-02 00:57:26,266 INFO [train.py:904] (1/8) Epoch 25, batch 6300, loss[loss=0.1782, simple_loss=0.2635, pruned_loss=0.04643, over 16473.00 frames. ], tot_loss[loss=0.1965, simple_loss=0.2847, pruned_loss=0.05418, over 3111402.75 frames. ], batch size: 68, lr: 2.69e-03, grad_scale: 8.0 2023-05-02 00:57:52,579 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=249919.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 00:58:29,093 INFO [optim.py:368] (1/8) 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] (1/8) Epoch 25, batch 6350, loss[loss=0.1776, simple_loss=0.2681, pruned_loss=0.04356, over 16649.00 frames. ], tot_loss[loss=0.1976, simple_loss=0.2851, pruned_loss=0.05504, over 3121920.88 frames. ], batch size: 89, lr: 2.69e-03, grad_scale: 8.0 2023-05-02 00:58:48,523 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=249955.0, num_to_drop=1, layers_to_drop={1} 2023-05-02 00:58:54,276 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.15 vs. limit=2.0 2023-05-02 00:58:56,497 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.8176, 3.8523, 3.9495, 3.7898, 3.9032, 4.2709, 3.9340, 3.7304], device='cuda:1'), covar=tensor([0.2046, 0.2324, 0.2858, 0.2396, 0.2474, 0.1764, 0.1723, 0.2492], device='cuda:1'), in_proj_covar=tensor([0.0420, 0.0619, 0.0680, 0.0506, 0.0670, 0.0707, 0.0525, 0.0676], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-05-02 00:59:20,557 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 25, batch 6400, loss[loss=0.1954, simple_loss=0.2785, pruned_loss=0.0562, over 16924.00 frames. ], tot_loss[loss=0.1996, simple_loss=0.286, pruned_loss=0.05665, over 3101319.29 frames. ], batch size: 109, lr: 2.69e-03, grad_scale: 8.0 2023-05-02 01:00:15,707 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.5587, 3.7576, 3.8259, 2.3806, 3.4066, 2.4224, 4.0493, 4.0779], device='cuda:1'), covar=tensor([0.0237, 0.0783, 0.0601, 0.1992, 0.0762, 0.1001, 0.0509, 0.0894], device='cuda:1'), in_proj_covar=tensor([0.0158, 0.0167, 0.0170, 0.0155, 0.0147, 0.0131, 0.0145, 0.0179], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:1') 2023-05-02 01:00:23,824 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=250016.0, num_to_drop=1, layers_to_drop={0} 2023-05-02 01:00:56,451 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=250037.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 01:00:59,053 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.3672, 3.5788, 3.2178, 2.9696, 2.9767, 3.4538, 3.2140, 3.1867], device='cuda:1'), covar=tensor([0.0887, 0.0786, 0.0446, 0.0415, 0.0954, 0.0620, 0.2479, 0.0686], device='cuda:1'), in_proj_covar=tensor([0.0302, 0.0453, 0.0351, 0.0355, 0.0357, 0.0410, 0.0241, 0.0425], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002], device='cuda:1') 2023-05-02 01:01:05,939 INFO [optim.py:368] (1/8) Clipping_scale=2.0, grad-norm quartiles 2.084e+02 2.850e+02 3.461e+02 4.247e+02 7.668e+02, threshold=6.921e+02, percent-clipped=1.0 2023-05-02 01:01:20,199 INFO [train.py:904] (1/8) Epoch 25, batch 6450, loss[loss=0.1913, simple_loss=0.2811, pruned_loss=0.05075, over 16632.00 frames. ], tot_loss[loss=0.1985, simple_loss=0.2856, pruned_loss=0.05574, over 3097214.08 frames. ], batch size: 62, lr: 2.69e-03, grad_scale: 8.0 2023-05-02 01:02:36,762 INFO [train.py:904] (1/8) Epoch 25, batch 6500, loss[loss=0.2024, simple_loss=0.297, pruned_loss=0.05388, over 16712.00 frames. ], tot_loss[loss=0.1971, simple_loss=0.2841, pruned_loss=0.05505, over 3115687.00 frames. ], batch size: 89, lr: 2.69e-03, grad_scale: 8.0 2023-05-02 01:02:39,443 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.23 vs. limit=2.0 2023-05-02 01:03:39,637 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.96 vs. limit=2.0 2023-05-02 01:03:40,562 INFO [optim.py:368] (1/8) 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 01:03:52,680 INFO [train.py:904] (1/8) Epoch 25, batch 6550, loss[loss=0.1847, simple_loss=0.2833, pruned_loss=0.04298, over 16580.00 frames. ], tot_loss[loss=0.1998, simple_loss=0.2869, pruned_loss=0.05636, over 3098657.30 frames. ], batch size: 62, lr: 2.69e-03, grad_scale: 4.0 2023-05-02 01:04:18,234 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=250170.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 01:05:05,603 INFO [train.py:904] (1/8) Epoch 25, batch 6600, loss[loss=0.2484, simple_loss=0.3151, pruned_loss=0.09083, over 11711.00 frames. ], tot_loss[loss=0.2013, simple_loss=0.289, pruned_loss=0.05674, over 3103415.60 frames. ], batch size: 248, lr: 2.68e-03, grad_scale: 4.0 2023-05-02 01:06:04,976 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.5562, 2.2343, 1.9508, 2.0168, 2.4992, 2.1712, 2.2708, 2.6683], device='cuda:1'), covar=tensor([0.0209, 0.0405, 0.0508, 0.0487, 0.0282, 0.0395, 0.0202, 0.0264], device='cuda:1'), in_proj_covar=tensor([0.0218, 0.0237, 0.0227, 0.0228, 0.0238, 0.0237, 0.0238, 0.0235], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-02 01:06:08,495 INFO [optim.py:368] (1/8) 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 01:06:21,900 INFO [train.py:904] (1/8) Epoch 25, batch 6650, loss[loss=0.193, simple_loss=0.2765, pruned_loss=0.05478, over 16529.00 frames. ], tot_loss[loss=0.2022, simple_loss=0.2894, pruned_loss=0.05751, over 3097097.24 frames. ], batch size: 35, lr: 2.68e-03, grad_scale: 4.0 2023-05-02 01:07:37,545 INFO [train.py:904] (1/8) Epoch 25, batch 6700, loss[loss=0.1849, simple_loss=0.2757, pruned_loss=0.04706, over 16731.00 frames. ], tot_loss[loss=0.202, simple_loss=0.2883, pruned_loss=0.0579, over 3066228.66 frames. ], batch size: 39, lr: 2.68e-03, grad_scale: 4.0 2023-05-02 01:07:50,661 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=250311.0, num_to_drop=1, layers_to_drop={3} 2023-05-02 01:08:23,196 INFO [zipformer.py:625] (1/8) 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] (1/8) 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 01:08:53,997 INFO [train.py:904] (1/8) Epoch 25, batch 6750, loss[loss=0.1921, simple_loss=0.2779, pruned_loss=0.05319, over 15380.00 frames. ], tot_loss[loss=0.201, simple_loss=0.2868, pruned_loss=0.05759, over 3082744.80 frames. ], batch size: 190, lr: 2.68e-03, grad_scale: 4.0 2023-05-02 01:10:10,580 INFO [train.py:904] (1/8) Epoch 25, batch 6800, loss[loss=0.2298, simple_loss=0.3055, pruned_loss=0.07705, over 11798.00 frames. ], tot_loss[loss=0.2007, simple_loss=0.2866, pruned_loss=0.05745, over 3081120.65 frames. ], batch size: 248, lr: 2.68e-03, grad_scale: 8.0 2023-05-02 01:11:16,510 INFO [optim.py:368] (1/8) 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 01:11:26,267 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.0978, 2.2421, 2.3367, 2.7345, 1.8180, 3.1496, 1.8182, 2.7260], device='cuda:1'), covar=tensor([0.1181, 0.0686, 0.1140, 0.0192, 0.0118, 0.0383, 0.1576, 0.0750], device='cuda:1'), in_proj_covar=tensor([0.0171, 0.0179, 0.0197, 0.0196, 0.0206, 0.0217, 0.0207, 0.0198], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-05-02 01:11:27,491 INFO [train.py:904] (1/8) Epoch 25, batch 6850, loss[loss=0.2455, simple_loss=0.3111, pruned_loss=0.08993, over 11500.00 frames. ], tot_loss[loss=0.2014, simple_loss=0.2876, pruned_loss=0.05758, over 3084877.98 frames. ], batch size: 248, lr: 2.68e-03, grad_scale: 8.0 2023-05-02 01:11:34,079 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([5.0063, 4.9377, 4.8479, 3.3314, 4.7574, 1.7212, 4.3409, 4.3594], device='cuda:1'), covar=tensor([0.0267, 0.0222, 0.0286, 0.1083, 0.0218, 0.3659, 0.0347, 0.0500], device='cuda:1'), in_proj_covar=tensor([0.0174, 0.0167, 0.0206, 0.0183, 0.0183, 0.0213, 0.0195, 0.0177], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-02 01:11:52,855 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=250470.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 01:11:53,015 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.1425, 2.0134, 2.5741, 3.1280, 2.9041, 3.4479, 2.2487, 3.5292], device='cuda:1'), covar=tensor([0.0201, 0.0607, 0.0416, 0.0279, 0.0335, 0.0185, 0.0639, 0.0142], device='cuda:1'), in_proj_covar=tensor([0.0193, 0.0195, 0.0182, 0.0186, 0.0201, 0.0161, 0.0199, 0.0160], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-02 01:12:43,350 INFO [train.py:904] (1/8) Epoch 25, batch 6900, loss[loss=0.2244, simple_loss=0.3055, pruned_loss=0.07168, over 15289.00 frames. ], tot_loss[loss=0.2028, simple_loss=0.2901, pruned_loss=0.05773, over 3076409.05 frames. ], batch size: 190, lr: 2.68e-03, grad_scale: 8.0 2023-05-02 01:12:51,423 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.6778, 3.7436, 2.2585, 4.3533, 2.8811, 4.2099, 2.4672, 3.0733], device='cuda:1'), covar=tensor([0.0301, 0.0430, 0.1816, 0.0200, 0.0856, 0.0612, 0.1655, 0.0804], device='cuda:1'), in_proj_covar=tensor([0.0173, 0.0180, 0.0196, 0.0168, 0.0177, 0.0219, 0.0203, 0.0181], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-05-02 01:13:06,142 INFO [zipformer.py:625] (1/8) 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] (1/8) 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 01:14:00,192 INFO [train.py:904] (1/8) Epoch 25, batch 6950, loss[loss=0.189, simple_loss=0.2783, pruned_loss=0.04981, over 16799.00 frames. ], tot_loss[loss=0.2059, simple_loss=0.2924, pruned_loss=0.0597, over 3066223.78 frames. ], batch size: 83, lr: 2.68e-03, grad_scale: 8.0 2023-05-02 01:15:18,377 INFO [train.py:904] (1/8) Epoch 25, batch 7000, loss[loss=0.2192, simple_loss=0.3126, pruned_loss=0.06284, over 16740.00 frames. ], tot_loss[loss=0.2044, simple_loss=0.292, pruned_loss=0.05838, over 3084943.29 frames. ], batch size: 124, lr: 2.68e-03, grad_scale: 8.0 2023-05-02 01:15:31,385 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=250611.0, num_to_drop=1, layers_to_drop={2} 2023-05-02 01:16:03,510 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=250632.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 01:16:07,817 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([5.2262, 5.2799, 5.0920, 4.6719, 4.6809, 5.1598, 5.0354, 4.8250], device='cuda:1'), covar=tensor([0.0714, 0.0675, 0.0400, 0.0431, 0.1181, 0.0561, 0.0423, 0.0859], device='cuda:1'), in_proj_covar=tensor([0.0297, 0.0447, 0.0346, 0.0352, 0.0353, 0.0405, 0.0238, 0.0421], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002], device='cuda:1') 2023-05-02 01:16:22,011 INFO [optim.py:368] (1/8) 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] (1/8) Epoch 25, batch 7050, loss[loss=0.1996, simple_loss=0.2861, pruned_loss=0.05651, over 16925.00 frames. ], tot_loss[loss=0.2044, simple_loss=0.2928, pruned_loss=0.05799, over 3100630.67 frames. ], batch size: 109, lr: 2.68e-03, grad_scale: 8.0 2023-05-02 01:16:44,701 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=250659.0, num_to_drop=1, layers_to_drop={0} 2023-05-02 01:16:49,630 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=250662.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 01:17:14,352 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.67 vs. limit=2.0 2023-05-02 01:17:17,436 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=250680.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 01:17:51,104 INFO [train.py:904] (1/8) Epoch 25, batch 7100, loss[loss=0.2084, simple_loss=0.2787, pruned_loss=0.06902, over 11181.00 frames. ], tot_loss[loss=0.2033, simple_loss=0.2912, pruned_loss=0.05774, over 3093171.41 frames. ], batch size: 248, lr: 2.68e-03, grad_scale: 8.0 2023-05-02 01:18:23,424 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=250723.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 01:18:56,843 INFO [optim.py:368] (1/8) 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] (1/8) Epoch 25, batch 7150, loss[loss=0.1713, simple_loss=0.2641, pruned_loss=0.03931, over 16431.00 frames. ], tot_loss[loss=0.2015, simple_loss=0.2891, pruned_loss=0.05697, over 3107813.93 frames. ], batch size: 68, lr: 2.68e-03, grad_scale: 8.0 2023-05-02 01:20:22,681 INFO [train.py:904] (1/8) Epoch 25, batch 7200, loss[loss=0.1876, simple_loss=0.2711, pruned_loss=0.052, over 11676.00 frames. ], tot_loss[loss=0.199, simple_loss=0.287, pruned_loss=0.05552, over 3098059.89 frames. ], batch size: 247, lr: 2.68e-03, grad_scale: 8.0 2023-05-02 01:20:41,932 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([1.7202, 1.7987, 1.6109, 1.4681, 1.9027, 1.5558, 1.5875, 1.8766], device='cuda:1'), covar=tensor([0.0224, 0.0319, 0.0467, 0.0410, 0.0245, 0.0326, 0.0194, 0.0242], device='cuda:1'), in_proj_covar=tensor([0.0217, 0.0238, 0.0227, 0.0228, 0.0238, 0.0236, 0.0236, 0.0235], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-02 01:21:08,268 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=250833.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 01:21:25,481 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.8070, 4.0092, 3.1029, 2.4579, 2.8460, 2.6557, 4.3922, 3.5848], device='cuda:1'), covar=tensor([0.2881, 0.0632, 0.1746, 0.2556, 0.2511, 0.2001, 0.0401, 0.1300], device='cuda:1'), in_proj_covar=tensor([0.0332, 0.0272, 0.0310, 0.0319, 0.0302, 0.0269, 0.0300, 0.0345], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:1') 2023-05-02 01:21:28,104 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=250845.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 01:21:28,839 INFO [optim.py:368] (1/8) 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] (1/8) Epoch 25, batch 7250, loss[loss=0.1622, simple_loss=0.2506, pruned_loss=0.03695, over 16461.00 frames. ], tot_loss[loss=0.1965, simple_loss=0.2845, pruned_loss=0.05426, over 3092818.77 frames. ], batch size: 75, lr: 2.68e-03, grad_scale: 4.0 2023-05-02 01:22:15,796 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.8703, 4.6751, 4.9042, 5.0526, 5.2593, 4.6308, 5.2293, 5.2425], device='cuda:1'), covar=tensor([0.1874, 0.1325, 0.1669, 0.0802, 0.0576, 0.1108, 0.0627, 0.0621], device='cuda:1'), in_proj_covar=tensor([0.0649, 0.0799, 0.0921, 0.0809, 0.0621, 0.0640, 0.0670, 0.0780], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-02 01:22:42,138 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=250894.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 01:22:55,037 INFO [train.py:904] (1/8) Epoch 25, batch 7300, loss[loss=0.1948, simple_loss=0.2811, pruned_loss=0.05426, over 16261.00 frames. ], tot_loss[loss=0.1966, simple_loss=0.2842, pruned_loss=0.05444, over 3098607.80 frames. ], batch size: 35, lr: 2.68e-03, grad_scale: 4.0 2023-05-02 01:23:00,802 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=250906.0, num_to_drop=1, layers_to_drop={3} 2023-05-02 01:23:59,648 INFO [optim.py:368] (1/8) 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 01:24:09,700 INFO [train.py:904] (1/8) Epoch 25, batch 7350, loss[loss=0.2035, simple_loss=0.2829, pruned_loss=0.06205, over 16594.00 frames. ], tot_loss[loss=0.1992, simple_loss=0.2863, pruned_loss=0.05602, over 3066538.98 frames. ], batch size: 62, lr: 2.68e-03, grad_scale: 4.0 2023-05-02 01:25:27,063 INFO [train.py:904] (1/8) Epoch 25, batch 7400, loss[loss=0.1866, simple_loss=0.2867, pruned_loss=0.04326, over 16660.00 frames. ], tot_loss[loss=0.2001, simple_loss=0.2873, pruned_loss=0.05645, over 3078291.50 frames. ], batch size: 62, lr: 2.68e-03, grad_scale: 4.0 2023-05-02 01:25:50,965 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=251018.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 01:26:20,335 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([1.7316, 2.6872, 2.6618, 4.4638, 3.2732, 4.0287, 1.6042, 2.9617], device='cuda:1'), covar=tensor([0.1405, 0.0822, 0.1179, 0.0175, 0.0275, 0.0387, 0.1660, 0.0823], device='cuda:1'), in_proj_covar=tensor([0.0172, 0.0181, 0.0199, 0.0198, 0.0208, 0.0218, 0.0209, 0.0200], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-05-02 01:26:35,984 INFO [optim.py:368] (1/8) 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 01:26:46,722 INFO [train.py:904] (1/8) Epoch 25, batch 7450, loss[loss=0.2031, simple_loss=0.2952, pruned_loss=0.05553, over 16363.00 frames. ], tot_loss[loss=0.201, simple_loss=0.2881, pruned_loss=0.05691, over 3088561.78 frames. ], batch size: 146, lr: 2.68e-03, grad_scale: 4.0 2023-05-02 01:26:58,170 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.14 vs. limit=2.0 2023-05-02 01:27:42,618 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.1367, 2.2246, 2.2063, 3.6315, 2.1284, 2.5621, 2.2703, 2.3307], device='cuda:1'), covar=tensor([0.1414, 0.3394, 0.3071, 0.0645, 0.4157, 0.2395, 0.3592, 0.3388], device='cuda:1'), in_proj_covar=tensor([0.0412, 0.0462, 0.0378, 0.0333, 0.0442, 0.0529, 0.0435, 0.0540], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-02 01:28:06,268 INFO [train.py:904] (1/8) Epoch 25, batch 7500, loss[loss=0.2142, simple_loss=0.3063, pruned_loss=0.06103, over 16545.00 frames. ], tot_loss[loss=0.201, simple_loss=0.2888, pruned_loss=0.05658, over 3090696.45 frames. ], batch size: 75, lr: 2.68e-03, grad_scale: 4.0 2023-05-02 01:28:07,053 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.9266, 2.1793, 2.3189, 3.3819, 2.0925, 2.4170, 2.2677, 2.2816], device='cuda:1'), covar=tensor([0.1533, 0.3441, 0.2936, 0.0698, 0.4302, 0.2608, 0.3462, 0.3301], device='cuda:1'), in_proj_covar=tensor([0.0412, 0.0462, 0.0378, 0.0332, 0.0441, 0.0529, 0.0435, 0.0539], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-02 01:28:11,045 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([1.7417, 2.7275, 2.5779, 4.6416, 3.1711, 4.0691, 1.6574, 3.0139], device='cuda:1'), covar=tensor([0.1486, 0.0940, 0.1337, 0.0174, 0.0316, 0.0436, 0.1827, 0.0885], device='cuda:1'), in_proj_covar=tensor([0.0171, 0.0180, 0.0198, 0.0196, 0.0206, 0.0217, 0.0207, 0.0198], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-05-02 01:29:12,776 INFO [optim.py:368] (1/8) 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 01:29:15,629 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([1.6752, 1.8134, 1.6324, 1.5136, 1.9210, 1.6012, 1.6372, 1.8926], device='cuda:1'), covar=tensor([0.0208, 0.0321, 0.0449, 0.0371, 0.0220, 0.0282, 0.0177, 0.0212], device='cuda:1'), in_proj_covar=tensor([0.0215, 0.0235, 0.0224, 0.0226, 0.0235, 0.0234, 0.0234, 0.0233], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-02 01:29:23,953 INFO [train.py:904] (1/8) Epoch 25, batch 7550, loss[loss=0.1985, simple_loss=0.2742, pruned_loss=0.0614, over 11393.00 frames. ], tot_loss[loss=0.201, simple_loss=0.2881, pruned_loss=0.05695, over 3078853.14 frames. ], batch size: 247, lr: 2.68e-03, grad_scale: 4.0 2023-05-02 01:30:19,159 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=251189.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 01:30:32,693 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=251198.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 01:30:37,924 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=251201.0, num_to_drop=1, layers_to_drop={1} 2023-05-02 01:30:39,948 INFO [train.py:904] (1/8) Epoch 25, batch 7600, loss[loss=0.1885, simple_loss=0.2727, pruned_loss=0.05214, over 16604.00 frames. ], tot_loss[loss=0.1999, simple_loss=0.2867, pruned_loss=0.05657, over 3094837.62 frames. ], batch size: 62, lr: 2.68e-03, grad_scale: 8.0 2023-05-02 01:31:43,657 INFO [optim.py:368] (1/8) 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] (1/8) Epoch 25, batch 7650, loss[loss=0.2437, simple_loss=0.309, pruned_loss=0.08924, over 11447.00 frames. ], tot_loss[loss=0.2015, simple_loss=0.288, pruned_loss=0.05749, over 3082659.03 frames. ], batch size: 246, lr: 2.68e-03, grad_scale: 8.0 2023-05-02 01:32:02,554 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=251259.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 01:33:07,026 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 25, batch 7700, loss[loss=0.1959, simple_loss=0.2852, pruned_loss=0.05326, over 15450.00 frames. ], tot_loss[loss=0.2024, simple_loss=0.2884, pruned_loss=0.05818, over 3077830.26 frames. ], batch size: 191, lr: 2.68e-03, grad_scale: 8.0 2023-05-02 01:33:29,123 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.4721, 4.6779, 4.8197, 4.6126, 4.6661, 5.1801, 4.6408, 4.4084], device='cuda:1'), covar=tensor([0.1449, 0.1737, 0.2173, 0.1892, 0.2290, 0.0929, 0.1629, 0.2415], device='cuda:1'), in_proj_covar=tensor([0.0420, 0.0613, 0.0678, 0.0503, 0.0666, 0.0702, 0.0526, 0.0675], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-05-02 01:33:31,762 INFO [zipformer.py:625] (1/8) 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] (1/8) 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] (1/8) Epoch 25, batch 7750, loss[loss=0.2153, simple_loss=0.3021, pruned_loss=0.06419, over 16426.00 frames. ], tot_loss[loss=0.2024, simple_loss=0.2883, pruned_loss=0.05819, over 3083750.39 frames. ], batch size: 75, lr: 2.68e-03, grad_scale: 8.0 2023-05-02 01:34:36,367 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.8118, 5.1269, 5.3014, 5.0754, 5.1660, 5.6664, 5.1470, 4.9480], device='cuda:1'), covar=tensor([0.1061, 0.1953, 0.2578, 0.1966, 0.2243, 0.0967, 0.1695, 0.2575], device='cuda:1'), in_proj_covar=tensor([0.0421, 0.0614, 0.0679, 0.0504, 0.0668, 0.0703, 0.0526, 0.0677], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-05-02 01:34:39,718 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=251362.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 01:34:45,580 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=251366.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 01:35:40,751 INFO [train.py:904] (1/8) Epoch 25, batch 7800, loss[loss=0.1657, simple_loss=0.2638, pruned_loss=0.03374, over 16841.00 frames. ], tot_loss[loss=0.2043, simple_loss=0.2898, pruned_loss=0.05945, over 3063540.70 frames. ], batch size: 102, lr: 2.68e-03, grad_scale: 8.0 2023-05-02 01:35:42,319 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=251404.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 01:36:45,152 INFO [optim.py:368] (1/8) 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 01:36:55,302 INFO [train.py:904] (1/8) Epoch 25, batch 7850, loss[loss=0.1841, simple_loss=0.2785, pruned_loss=0.04488, over 16827.00 frames. ], tot_loss[loss=0.204, simple_loss=0.2902, pruned_loss=0.05896, over 3067087.13 frames. ], batch size: 39, lr: 2.68e-03, grad_scale: 8.0 2023-05-02 01:37:13,959 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=251465.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 01:37:42,134 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.64 vs. limit=2.0 2023-05-02 01:37:50,394 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=251489.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 01:38:06,772 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=251501.0, num_to_drop=1, layers_to_drop={1} 2023-05-02 01:38:09,238 INFO [train.py:904] (1/8) Epoch 25, batch 7900, loss[loss=0.192, simple_loss=0.2803, pruned_loss=0.0519, over 16352.00 frames. ], tot_loss[loss=0.2031, simple_loss=0.2893, pruned_loss=0.05841, over 3065147.13 frames. ], batch size: 146, lr: 2.68e-03, grad_scale: 8.0 2023-05-02 01:39:04,231 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=251537.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 01:39:17,278 INFO [optim.py:368] (1/8) 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 01:39:22,128 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=251549.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 01:39:23,588 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([1.9329, 3.0508, 3.2667, 1.6503, 3.4009, 3.5863, 2.8169, 2.4604], device='cuda:1'), covar=tensor([0.1189, 0.0272, 0.0227, 0.1378, 0.0118, 0.0184, 0.0469, 0.0642], device='cuda:1'), in_proj_covar=tensor([0.0149, 0.0111, 0.0100, 0.0139, 0.0085, 0.0130, 0.0129, 0.0131], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-05-02 01:39:28,550 INFO [train.py:904] (1/8) Epoch 25, batch 7950, loss[loss=0.2014, simple_loss=0.277, pruned_loss=0.06287, over 17094.00 frames. ], tot_loss[loss=0.203, simple_loss=0.2892, pruned_loss=0.05845, over 3064023.18 frames. ], batch size: 53, lr: 2.68e-03, grad_scale: 8.0 2023-05-02 01:39:30,113 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=251554.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 01:40:46,597 INFO [train.py:904] (1/8) Epoch 25, batch 8000, loss[loss=0.1978, simple_loss=0.2808, pruned_loss=0.05737, over 16403.00 frames. ], tot_loss[loss=0.2033, simple_loss=0.2896, pruned_loss=0.05849, over 3087907.23 frames. ], batch size: 35, lr: 2.68e-03, grad_scale: 8.0 2023-05-02 01:41:44,775 INFO [zipformer.py:625] (1/8) 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] (1/8) 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 01:42:01,886 INFO [train.py:904] (1/8) Epoch 25, batch 8050, loss[loss=0.2031, simple_loss=0.2931, pruned_loss=0.05655, over 16582.00 frames. ], tot_loss[loss=0.2039, simple_loss=0.2899, pruned_loss=0.05891, over 3077820.94 frames. ], batch size: 62, lr: 2.68e-03, grad_scale: 8.0 2023-05-02 01:42:09,015 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=251657.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 01:43:16,986 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 25, batch 8100, loss[loss=0.2082, simple_loss=0.2776, pruned_loss=0.06945, over 11483.00 frames. ], tot_loss[loss=0.2025, simple_loss=0.2888, pruned_loss=0.05805, over 3084666.13 frames. ], batch size: 247, lr: 2.68e-03, grad_scale: 8.0 2023-05-02 01:43:23,384 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.6098, 2.5839, 1.8953, 2.6560, 2.0862, 2.7762, 2.1283, 2.3688], device='cuda:1'), covar=tensor([0.0327, 0.0389, 0.1209, 0.0266, 0.0677, 0.0527, 0.1188, 0.0596], device='cuda:1'), in_proj_covar=tensor([0.0174, 0.0180, 0.0196, 0.0168, 0.0178, 0.0219, 0.0204, 0.0182], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-05-02 01:43:27,270 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=251709.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 01:43:59,649 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.5538, 4.7606, 4.4526, 4.1941, 3.7962, 4.7064, 4.5129, 4.2818], device='cuda:1'), covar=tensor([0.0889, 0.0626, 0.0527, 0.0464, 0.1907, 0.0485, 0.0550, 0.0759], device='cuda:1'), in_proj_covar=tensor([0.0295, 0.0443, 0.0344, 0.0349, 0.0349, 0.0400, 0.0238, 0.0415], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-02 01:44:22,958 INFO [optim.py:368] (1/8) 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] (1/8) Epoch 25, batch 8150, loss[loss=0.1519, simple_loss=0.2362, pruned_loss=0.03383, over 16489.00 frames. ], tot_loss[loss=0.1997, simple_loss=0.2859, pruned_loss=0.05676, over 3113562.87 frames. ], batch size: 68, lr: 2.68e-03, grad_scale: 8.0 2023-05-02 01:44:44,518 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=251760.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 01:45:00,587 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=251770.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 01:45:50,906 INFO [train.py:904] (1/8) Epoch 25, batch 8200, loss[loss=0.2328, simple_loss=0.2976, pruned_loss=0.08403, over 11849.00 frames. ], tot_loss[loss=0.1979, simple_loss=0.2834, pruned_loss=0.05624, over 3110720.36 frames. ], batch size: 246, lr: 2.68e-03, grad_scale: 8.0 2023-05-02 01:46:20,472 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.0711, 1.5785, 1.9530, 2.0761, 2.1945, 2.3385, 1.7778, 2.2390], device='cuda:1'), covar=tensor([0.0259, 0.0514, 0.0306, 0.0356, 0.0364, 0.0222, 0.0531, 0.0186], device='cuda:1'), in_proj_covar=tensor([0.0191, 0.0194, 0.0182, 0.0185, 0.0201, 0.0160, 0.0198, 0.0159], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-02 01:46:59,192 INFO [optim.py:368] (1/8) 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 01:47:10,656 INFO [train.py:904] (1/8) Epoch 25, batch 8250, loss[loss=0.1724, simple_loss=0.266, pruned_loss=0.03941, over 12258.00 frames. ], tot_loss[loss=0.1944, simple_loss=0.2821, pruned_loss=0.05336, over 3100861.64 frames. ], batch size: 246, lr: 2.68e-03, grad_scale: 8.0 2023-05-02 01:47:12,448 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=251854.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 01:48:29,882 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 25, batch 8300, loss[loss=0.1687, simple_loss=0.2681, pruned_loss=0.03467, over 16294.00 frames. ], tot_loss[loss=0.191, simple_loss=0.2802, pruned_loss=0.05089, over 3098390.24 frames. ], batch size: 165, lr: 2.68e-03, grad_scale: 8.0 2023-05-02 01:49:30,128 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.3367, 2.9171, 3.0534, 1.8572, 3.1916, 3.2680, 2.8787, 2.7579], device='cuda:1'), covar=tensor([0.0739, 0.0276, 0.0259, 0.1225, 0.0130, 0.0269, 0.0396, 0.0446], device='cuda:1'), in_proj_covar=tensor([0.0146, 0.0109, 0.0098, 0.0137, 0.0083, 0.0128, 0.0127, 0.0129], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0003], device='cuda:1') 2023-05-02 01:49:41,163 INFO [optim.py:368] (1/8) 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 01:49:52,637 INFO [train.py:904] (1/8) Epoch 25, batch 8350, loss[loss=0.1708, simple_loss=0.2683, pruned_loss=0.0367, over 16169.00 frames. ], tot_loss[loss=0.1894, simple_loss=0.2801, pruned_loss=0.04938, over 3088208.21 frames. ], batch size: 165, lr: 2.68e-03, grad_scale: 8.0 2023-05-02 01:49:53,479 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.25 vs. limit=2.0 2023-05-02 01:49:59,392 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=251957.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 01:51:05,348 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 25, batch 8400, loss[loss=0.1506, simple_loss=0.2544, pruned_loss=0.02346, over 16707.00 frames. ], tot_loss[loss=0.1861, simple_loss=0.2776, pruned_loss=0.04729, over 3084574.83 frames. ], batch size: 89, lr: 2.68e-03, grad_scale: 8.0 2023-05-02 01:51:22,059 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=252005.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 01:52:27,817 INFO [optim.py:368] (1/8) Clipping_scale=2.0, grad-norm quartiles 1.337e+02 2.253e+02 2.668e+02 3.204e+02 6.662e+02, threshold=5.337e+02, percent-clipped=3.0 2023-05-02 01:52:40,176 INFO [train.py:904] (1/8) Epoch 25, batch 8450, loss[loss=0.1579, simple_loss=0.251, pruned_loss=0.03241, over 16611.00 frames. ], tot_loss[loss=0.1831, simple_loss=0.2758, pruned_loss=0.04516, over 3104302.98 frames. ], batch size: 57, lr: 2.67e-03, grad_scale: 8.0 2023-05-02 01:52:51,870 INFO [zipformer.py:625] (1/8) 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] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=252065.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 01:54:01,927 INFO [train.py:904] (1/8) Epoch 25, batch 8500, loss[loss=0.1623, simple_loss=0.2531, pruned_loss=0.03574, over 16473.00 frames. ], tot_loss[loss=0.1795, simple_loss=0.2722, pruned_loss=0.04341, over 3077219.67 frames. ], batch size: 146, lr: 2.67e-03, grad_scale: 8.0 2023-05-02 01:54:11,541 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=252108.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 01:54:11,894 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.73 vs. limit=2.0 2023-05-02 01:54:21,572 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.4864, 3.6694, 2.0061, 3.9694, 2.7182, 3.9520, 2.1921, 2.9124], device='cuda:1'), covar=tensor([0.0274, 0.0319, 0.1786, 0.0246, 0.0776, 0.0479, 0.1630, 0.0681], device='cuda:1'), in_proj_covar=tensor([0.0171, 0.0176, 0.0194, 0.0165, 0.0175, 0.0214, 0.0202, 0.0179], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-05-02 01:54:22,682 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.5763, 4.6089, 4.4407, 4.0826, 4.1286, 4.5179, 4.3024, 4.2548], device='cuda:1'), covar=tensor([0.0574, 0.0745, 0.0325, 0.0361, 0.0896, 0.0611, 0.0557, 0.0686], device='cuda:1'), in_proj_covar=tensor([0.0294, 0.0443, 0.0343, 0.0348, 0.0348, 0.0399, 0.0238, 0.0414], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-02 01:54:31,370 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([1.9636, 2.1617, 2.4039, 1.8731, 2.4976, 2.7546, 2.4409, 2.3196], device='cuda:1'), covar=tensor([0.0837, 0.0255, 0.0230, 0.1101, 0.0131, 0.0320, 0.0414, 0.0500], device='cuda:1'), in_proj_covar=tensor([0.0144, 0.0108, 0.0097, 0.0136, 0.0082, 0.0126, 0.0125, 0.0127], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0004, 0.0003], device='cuda:1') 2023-05-02 01:55:00,642 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([1.9394, 2.7731, 2.4137, 4.3866, 2.7809, 4.1163, 1.6087, 2.9730], device='cuda:1'), covar=tensor([0.1366, 0.0824, 0.1373, 0.0169, 0.0179, 0.0363, 0.1781, 0.0770], device='cuda:1'), in_proj_covar=tensor([0.0170, 0.0177, 0.0195, 0.0194, 0.0204, 0.0214, 0.0206, 0.0196], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-05-02 01:55:15,323 INFO [optim.py:368] (1/8) 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] (1/8) Epoch 25, batch 8550, loss[loss=0.1755, simple_loss=0.2726, pruned_loss=0.03916, over 16795.00 frames. ], tot_loss[loss=0.1774, simple_loss=0.2699, pruned_loss=0.04252, over 3027883.70 frames. ], batch size: 124, lr: 2.67e-03, grad_scale: 8.0 2023-05-02 01:56:54,768 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.7021, 3.1315, 3.4519, 2.0762, 2.8762, 2.2313, 3.3213, 3.3716], device='cuda:1'), covar=tensor([0.0273, 0.0812, 0.0494, 0.2037, 0.0843, 0.1029, 0.0650, 0.0926], device='cuda:1'), in_proj_covar=tensor([0.0155, 0.0163, 0.0166, 0.0152, 0.0144, 0.0129, 0.0142, 0.0175], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:1') 2023-05-02 01:57:09,035 INFO [train.py:904] (1/8) Epoch 25, batch 8600, loss[loss=0.1656, simple_loss=0.2615, pruned_loss=0.03483, over 15385.00 frames. ], tot_loss[loss=0.1763, simple_loss=0.27, pruned_loss=0.04133, over 3042894.57 frames. ], batch size: 190, lr: 2.67e-03, grad_scale: 8.0 2023-05-02 01:58:34,357 INFO [optim.py:368] (1/8) 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 01:58:37,171 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.4146, 3.3501, 3.4930, 3.5521, 3.6013, 3.3219, 3.5541, 3.6395], device='cuda:1'), covar=tensor([0.1319, 0.1009, 0.1047, 0.0643, 0.0561, 0.1912, 0.0798, 0.0793], device='cuda:1'), in_proj_covar=tensor([0.0627, 0.0773, 0.0890, 0.0783, 0.0600, 0.0618, 0.0650, 0.0757], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-02 01:58:48,812 INFO [train.py:904] (1/8) Epoch 25, batch 8650, loss[loss=0.1649, simple_loss=0.2671, pruned_loss=0.03131, over 16287.00 frames. ], tot_loss[loss=0.1738, simple_loss=0.2682, pruned_loss=0.03975, over 3055706.79 frames. ], batch size: 165, lr: 2.67e-03, grad_scale: 8.0 2023-05-02 02:00:23,212 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=252297.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 02:00:33,686 INFO [train.py:904] (1/8) Epoch 25, batch 8700, loss[loss=0.1798, simple_loss=0.2599, pruned_loss=0.04979, over 12410.00 frames. ], tot_loss[loss=0.1711, simple_loss=0.2652, pruned_loss=0.03848, over 3045911.37 frames. ], batch size: 250, lr: 2.67e-03, grad_scale: 8.0 2023-05-02 02:00:54,730 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=4.86 vs. limit=5.0 2023-05-02 02:01:08,448 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.7047, 2.7072, 1.8737, 2.8215, 2.1322, 2.8663, 2.1661, 2.4409], device='cuda:1'), covar=tensor([0.0326, 0.0393, 0.1434, 0.0338, 0.0702, 0.0521, 0.1362, 0.0650], device='cuda:1'), in_proj_covar=tensor([0.0170, 0.0175, 0.0192, 0.0163, 0.0174, 0.0212, 0.0201, 0.0178], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-05-02 02:01:53,003 INFO [zipformer.py:625] (1/8) 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] (1/8) 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 02:02:09,352 INFO [train.py:904] (1/8) Epoch 25, batch 8750, loss[loss=0.1574, simple_loss=0.2488, pruned_loss=0.03303, over 16914.00 frames. ], tot_loss[loss=0.1701, simple_loss=0.2646, pruned_loss=0.03781, over 3061018.92 frames. ], batch size: 42, lr: 2.67e-03, grad_scale: 8.0 2023-05-02 02:02:13,774 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.4238, 2.4846, 2.1816, 2.1934, 2.8139, 2.4286, 2.8500, 3.0606], device='cuda:1'), covar=tensor([0.0196, 0.0581, 0.0625, 0.0591, 0.0341, 0.0487, 0.0282, 0.0321], device='cuda:1'), in_proj_covar=tensor([0.0212, 0.0234, 0.0223, 0.0223, 0.0233, 0.0232, 0.0229, 0.0230], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-02 02:02:41,268 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=252365.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 02:04:01,511 INFO [train.py:904] (1/8) Epoch 25, batch 8800, loss[loss=0.1668, simple_loss=0.2651, pruned_loss=0.03425, over 15524.00 frames. ], tot_loss[loss=0.1682, simple_loss=0.2632, pruned_loss=0.03664, over 3067188.61 frames. ], batch size: 191, lr: 2.67e-03, grad_scale: 8.0 2023-05-02 02:04:21,277 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=252413.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 02:04:30,822 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=252417.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 02:05:07,188 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=252434.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 02:05:28,705 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.22 vs. limit=2.0 2023-05-02 02:05:31,486 INFO [optim.py:368] (1/8) 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 02:05:46,035 INFO [train.py:904] (1/8) Epoch 25, batch 8850, loss[loss=0.1757, simple_loss=0.2767, pruned_loss=0.03739, over 15207.00 frames. ], tot_loss[loss=0.1687, simple_loss=0.2654, pruned_loss=0.03604, over 3062537.93 frames. ], batch size: 190, lr: 2.67e-03, grad_scale: 8.0 2023-05-02 02:06:39,674 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=252478.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 02:07:16,242 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=252495.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 02:07:31,612 INFO [train.py:904] (1/8) Epoch 25, batch 8900, loss[loss=0.1612, simple_loss=0.2641, pruned_loss=0.0291, over 16866.00 frames. ], tot_loss[loss=0.1689, simple_loss=0.2663, pruned_loss=0.0358, over 3075130.23 frames. ], batch size: 90, lr: 2.67e-03, grad_scale: 8.0 2023-05-02 02:09:03,236 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.89 vs. limit=2.0 2023-05-02 02:09:18,928 INFO [optim.py:368] (1/8) 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 02:09:34,565 INFO [train.py:904] (1/8) Epoch 25, batch 8950, loss[loss=0.1743, simple_loss=0.2733, pruned_loss=0.03759, over 16770.00 frames. ], tot_loss[loss=0.1684, simple_loss=0.2653, pruned_loss=0.03576, over 3072651.11 frames. ], batch size: 124, lr: 2.67e-03, grad_scale: 8.0 2023-05-02 02:10:59,693 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=252592.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 02:11:21,907 INFO [train.py:904] (1/8) Epoch 25, batch 9000, loss[loss=0.1541, simple_loss=0.2476, pruned_loss=0.03029, over 17229.00 frames. ], tot_loss[loss=0.1667, simple_loss=0.2627, pruned_loss=0.03537, over 3064302.73 frames. ], batch size: 44, lr: 2.67e-03, grad_scale: 8.0 2023-05-02 02:11:21,907 INFO [train.py:929] (1/8) Computing validation loss 2023-05-02 02:11:31,592 INFO [train.py:938] (1/8) Epoch 25, validation: loss=0.1442, simple_loss=0.248, pruned_loss=0.02014, over 944034.00 frames. 2023-05-02 02:11:31,593 INFO [train.py:939] (1/8) Maximum memory allocated so far is 18145MB 2023-05-02 02:11:53,518 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=252613.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 02:13:01,732 INFO [optim.py:368] (1/8) 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] (1/8) Epoch 25, batch 9050, loss[loss=0.1569, simple_loss=0.248, pruned_loss=0.03291, over 16636.00 frames. ], tot_loss[loss=0.1676, simple_loss=0.2635, pruned_loss=0.03583, over 3066699.97 frames. ], batch size: 134, lr: 2.67e-03, grad_scale: 8.0 2023-05-02 02:13:15,962 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=252653.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 02:13:47,163 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=252668.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 02:13:56,480 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=252674.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 02:14:40,110 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.9067, 5.2158, 5.3669, 5.1838, 5.2413, 5.7586, 5.2144, 4.9143], device='cuda:1'), covar=tensor([0.0914, 0.1831, 0.2113, 0.1917, 0.2409, 0.0921, 0.1656, 0.2562], device='cuda:1'), in_proj_covar=tensor([0.0405, 0.0592, 0.0656, 0.0488, 0.0645, 0.0683, 0.0510, 0.0652], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-05-02 02:14:58,382 INFO [train.py:904] (1/8) Epoch 25, batch 9100, loss[loss=0.1613, simple_loss=0.2511, pruned_loss=0.03573, over 17115.00 frames. ], tot_loss[loss=0.1677, simple_loss=0.2629, pruned_loss=0.03621, over 3088498.21 frames. ], batch size: 49, lr: 2.67e-03, grad_scale: 8.0 2023-05-02 02:16:01,383 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=252729.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 02:16:04,575 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.8700, 2.1146, 2.3266, 3.1466, 2.1423, 2.2642, 2.2906, 2.1976], device='cuda:1'), covar=tensor([0.1442, 0.3777, 0.2999, 0.0763, 0.4589, 0.2887, 0.3600, 0.3945], device='cuda:1'), in_proj_covar=tensor([0.0404, 0.0453, 0.0373, 0.0324, 0.0433, 0.0516, 0.0425, 0.0528], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-02 02:16:11,400 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.72 vs. limit=2.0 2023-05-02 02:16:41,919 INFO [optim.py:368] (1/8) 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 02:16:57,785 INFO [train.py:904] (1/8) Epoch 25, batch 9150, loss[loss=0.1521, simple_loss=0.2433, pruned_loss=0.03044, over 12075.00 frames. ], tot_loss[loss=0.1677, simple_loss=0.2634, pruned_loss=0.03599, over 3081784.49 frames. ], batch size: 248, lr: 2.67e-03, grad_scale: 8.0 2023-05-02 02:17:42,138 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=252773.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 02:17:46,749 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([5.0107, 5.2824, 5.0905, 5.1038, 4.8316, 4.8090, 4.6677, 5.3605], device='cuda:1'), covar=tensor([0.1092, 0.0856, 0.0860, 0.0796, 0.0733, 0.0922, 0.1117, 0.0881], device='cuda:1'), in_proj_covar=tensor([0.0674, 0.0816, 0.0672, 0.0630, 0.0517, 0.0525, 0.0682, 0.0637], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:1') 2023-05-02 02:18:20,292 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=252790.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 02:18:43,278 INFO [train.py:904] (1/8) Epoch 25, batch 9200, loss[loss=0.1567, simple_loss=0.2413, pruned_loss=0.03604, over 12080.00 frames. ], tot_loss[loss=0.1648, simple_loss=0.2591, pruned_loss=0.03531, over 3074583.79 frames. ], batch size: 248, lr: 2.67e-03, grad_scale: 8.0 2023-05-02 02:19:14,683 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.0660, 3.3293, 3.6844, 2.0381, 2.9963, 2.3122, 3.5447, 3.5885], device='cuda:1'), covar=tensor([0.0278, 0.0932, 0.0478, 0.2173, 0.0877, 0.1080, 0.0643, 0.0987], device='cuda:1'), in_proj_covar=tensor([0.0153, 0.0161, 0.0164, 0.0151, 0.0143, 0.0128, 0.0140, 0.0172], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:1') 2023-05-02 02:20:05,202 INFO [optim.py:368] (1/8) 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] (1/8) Epoch 25, batch 9250, loss[loss=0.143, simple_loss=0.2262, pruned_loss=0.02988, over 12251.00 frames. ], tot_loss[loss=0.1647, simple_loss=0.2588, pruned_loss=0.03531, over 3065385.32 frames. ], batch size: 248, lr: 2.67e-03, grad_scale: 16.0 2023-05-02 02:20:28,232 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=252857.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 02:22:11,610 INFO [train.py:904] (1/8) Epoch 25, batch 9300, loss[loss=0.1608, simple_loss=0.2534, pruned_loss=0.03409, over 16248.00 frames. ], tot_loss[loss=0.1637, simple_loss=0.2575, pruned_loss=0.03496, over 3056375.06 frames. ], batch size: 35, lr: 2.67e-03, grad_scale: 8.0 2023-05-02 02:22:47,914 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=252918.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 02:22:58,218 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=3.69 vs. limit=5.0 2023-05-02 02:23:45,074 INFO [optim.py:368] (1/8) 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 02:23:47,295 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=252948.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 02:23:55,314 INFO [train.py:904] (1/8) Epoch 25, batch 9350, loss[loss=0.1596, simple_loss=0.2551, pruned_loss=0.03205, over 16553.00 frames. ], tot_loss[loss=0.1629, simple_loss=0.257, pruned_loss=0.03437, over 3095050.77 frames. ], batch size: 75, lr: 2.67e-03, grad_scale: 8.0 2023-05-02 02:24:29,895 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=252969.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 02:24:51,519 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.6738, 2.6369, 1.8848, 2.8624, 2.1064, 2.8674, 2.1548, 2.4128], device='cuda:1'), covar=tensor([0.0315, 0.0398, 0.1377, 0.0291, 0.0783, 0.0502, 0.1310, 0.0612], device='cuda:1'), in_proj_covar=tensor([0.0168, 0.0172, 0.0190, 0.0160, 0.0173, 0.0208, 0.0198, 0.0176], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0003, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-05-02 02:25:11,248 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.0045, 2.1865, 2.1384, 3.5807, 2.0938, 2.4495, 2.2688, 2.2772], device='cuda:1'), covar=tensor([0.1375, 0.3711, 0.3236, 0.0623, 0.4227, 0.2669, 0.3745, 0.3539], device='cuda:1'), in_proj_covar=tensor([0.0401, 0.0452, 0.0372, 0.0323, 0.0432, 0.0514, 0.0423, 0.0526], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-02 02:25:16,681 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=252993.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 02:25:36,657 INFO [train.py:904] (1/8) Epoch 25, batch 9400, loss[loss=0.1645, simple_loss=0.2624, pruned_loss=0.03327, over 15281.00 frames. ], tot_loss[loss=0.1629, simple_loss=0.2574, pruned_loss=0.03416, over 3085571.37 frames. ], batch size: 191, lr: 2.67e-03, grad_scale: 8.0 2023-05-02 02:26:10,741 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.4058, 3.4821, 3.6538, 3.6483, 3.6586, 3.4667, 3.5509, 3.5540], device='cuda:1'), covar=tensor([0.0840, 0.1806, 0.1555, 0.2127, 0.1898, 0.2096, 0.1524, 0.0944], device='cuda:1'), in_proj_covar=tensor([0.0408, 0.0457, 0.0447, 0.0409, 0.0493, 0.0470, 0.0542, 0.0376], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:1') 2023-05-02 02:26:19,760 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=253024.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 02:26:45,904 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.6392, 1.8901, 2.2500, 2.5622, 2.4452, 3.0069, 2.0531, 2.9460], device='cuda:1'), covar=tensor([0.0293, 0.0601, 0.0472, 0.0437, 0.0453, 0.0242, 0.0604, 0.0192], device='cuda:1'), in_proj_covar=tensor([0.0189, 0.0192, 0.0180, 0.0182, 0.0198, 0.0157, 0.0195, 0.0157], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-02 02:27:05,785 INFO [optim.py:368] (1/8) 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] (1/8) Epoch 25, batch 9450, loss[loss=0.1838, simple_loss=0.2708, pruned_loss=0.04841, over 12261.00 frames. ], tot_loss[loss=0.1638, simple_loss=0.2591, pruned_loss=0.03427, over 3078077.73 frames. ], batch size: 248, lr: 2.67e-03, grad_scale: 8.0 2023-05-02 02:27:19,781 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=253054.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 02:27:57,749 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=253073.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 02:28:31,206 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=253090.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 02:28:31,534 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.16 vs. limit=2.0 2023-05-02 02:28:54,675 INFO [train.py:904] (1/8) Epoch 25, batch 9500, loss[loss=0.1554, simple_loss=0.2523, pruned_loss=0.02928, over 16444.00 frames. ], tot_loss[loss=0.1635, simple_loss=0.2588, pruned_loss=0.03409, over 3080589.95 frames. ], batch size: 68, lr: 2.67e-03, grad_scale: 8.0 2023-05-02 02:29:33,158 INFO [zipformer.py:625] (1/8) 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] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=253138.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 02:30:26,749 INFO [optim.py:368] (1/8) 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 02:30:40,903 INFO [train.py:904] (1/8) Epoch 25, batch 9550, loss[loss=0.1767, simple_loss=0.2771, pruned_loss=0.03815, over 15410.00 frames. ], tot_loss[loss=0.1636, simple_loss=0.2591, pruned_loss=0.0341, over 3095553.61 frames. ], batch size: 191, lr: 2.67e-03, grad_scale: 8.0 2023-05-02 02:32:22,177 INFO [train.py:904] (1/8) Epoch 25, batch 9600, loss[loss=0.1815, simple_loss=0.2794, pruned_loss=0.0418, over 16377.00 frames. ], tot_loss[loss=0.1648, simple_loss=0.2601, pruned_loss=0.03478, over 3080565.42 frames. ], batch size: 146, lr: 2.67e-03, grad_scale: 8.0 2023-05-02 02:32:42,583 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=253213.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 02:33:05,555 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.1531, 3.5869, 3.5940, 2.3776, 3.3026, 3.6542, 3.4300, 2.0788], device='cuda:1'), covar=tensor([0.0608, 0.0051, 0.0059, 0.0446, 0.0116, 0.0080, 0.0082, 0.0518], device='cuda:1'), in_proj_covar=tensor([0.0133, 0.0084, 0.0085, 0.0132, 0.0098, 0.0107, 0.0094, 0.0128], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-05-02 02:33:55,685 INFO [optim.py:368] (1/8) 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 02:33:59,627 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=253248.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 02:34:10,799 INFO [train.py:904] (1/8) Epoch 25, batch 9650, loss[loss=0.1439, simple_loss=0.2458, pruned_loss=0.02101, over 16918.00 frames. ], tot_loss[loss=0.1658, simple_loss=0.2614, pruned_loss=0.03511, over 3056090.91 frames. ], batch size: 102, lr: 2.67e-03, grad_scale: 8.0 2023-05-02 02:34:38,392 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.3383, 3.4417, 2.0597, 3.8756, 2.5110, 3.8159, 2.2804, 2.7539], device='cuda:1'), covar=tensor([0.0368, 0.0416, 0.1735, 0.0230, 0.0988, 0.0528, 0.1562, 0.0834], device='cuda:1'), in_proj_covar=tensor([0.0167, 0.0171, 0.0189, 0.0160, 0.0172, 0.0207, 0.0197, 0.0175], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0003, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-05-02 02:34:52,565 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=253269.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 02:35:46,090 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=253296.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 02:36:00,301 INFO [train.py:904] (1/8) Epoch 25, batch 9700, loss[loss=0.1699, simple_loss=0.2649, pruned_loss=0.03741, over 15307.00 frames. ], tot_loss[loss=0.1656, simple_loss=0.2608, pruned_loss=0.03521, over 3062338.58 frames. ], batch size: 190, lr: 2.67e-03, grad_scale: 8.0 2023-05-02 02:36:27,590 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=253317.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 02:36:42,573 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=253324.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 02:36:56,099 INFO [zipformer.py:625] (1/8) 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] (1/8) 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 02:37:37,142 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 25, batch 9750, loss[loss=0.1594, simple_loss=0.2423, pruned_loss=0.03821, over 12339.00 frames. ], tot_loss[loss=0.165, simple_loss=0.2592, pruned_loss=0.03537, over 3036980.15 frames. ], batch size: 246, lr: 2.67e-03, grad_scale: 4.0 2023-05-02 02:38:18,017 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=4.61 vs. limit=5.0 2023-05-02 02:38:21,277 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=253372.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 02:38:38,961 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=253380.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 02:39:00,979 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=253390.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 02:39:23,686 INFO [train.py:904] (1/8) Epoch 25, batch 9800, loss[loss=0.1711, simple_loss=0.274, pruned_loss=0.0341, over 16922.00 frames. ], tot_loss[loss=0.1645, simple_loss=0.2599, pruned_loss=0.03452, over 3061641.15 frames. ], batch size: 116, lr: 2.67e-03, grad_scale: 4.0 2023-05-02 02:39:53,049 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=253419.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 02:40:06,158 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.20 vs. limit=2.0 2023-05-02 02:40:37,895 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=253441.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 02:40:55,714 INFO [optim.py:368] (1/8) 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 02:41:06,298 INFO [train.py:904] (1/8) Epoch 25, batch 9850, loss[loss=0.1481, simple_loss=0.2532, pruned_loss=0.02153, over 16879.00 frames. ], tot_loss[loss=0.1643, simple_loss=0.2604, pruned_loss=0.03412, over 3056119.50 frames. ], batch size: 102, lr: 2.67e-03, grad_scale: 4.0 2023-05-02 02:42:03,940 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 25, batch 9900, loss[loss=0.1529, simple_loss=0.2408, pruned_loss=0.03249, over 12477.00 frames. ], tot_loss[loss=0.1637, simple_loss=0.2602, pruned_loss=0.03363, over 3062481.54 frames. ], batch size: 250, lr: 2.67e-03, grad_scale: 4.0 2023-05-02 02:43:25,460 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=253513.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 02:43:41,961 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=253521.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 02:44:34,388 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.8400, 3.8269, 4.1285, 4.1123, 4.0938, 3.8877, 3.8918, 3.9126], device='cuda:1'), covar=tensor([0.0457, 0.0954, 0.0550, 0.0475, 0.0595, 0.0622, 0.1004, 0.0603], device='cuda:1'), in_proj_covar=tensor([0.0402, 0.0452, 0.0441, 0.0403, 0.0486, 0.0463, 0.0534, 0.0370], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:1') 2023-05-02 02:44:46,254 INFO [optim.py:368] (1/8) 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 02:44:59,712 INFO [train.py:904] (1/8) Epoch 25, batch 9950, loss[loss=0.162, simple_loss=0.2603, pruned_loss=0.03181, over 16623.00 frames. ], tot_loss[loss=0.1653, simple_loss=0.2623, pruned_loss=0.03413, over 3065430.75 frames. ], batch size: 134, lr: 2.67e-03, grad_scale: 4.0 2023-05-02 02:45:18,705 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=253561.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 02:45:34,696 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=253567.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 02:46:09,544 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([1.9390, 1.7549, 1.5419, 1.4201, 1.8793, 1.6156, 1.5984, 1.9729], device='cuda:1'), covar=tensor([0.0233, 0.0455, 0.0572, 0.0501, 0.0265, 0.0349, 0.0232, 0.0282], device='cuda:1'), in_proj_covar=tensor([0.0212, 0.0236, 0.0225, 0.0225, 0.0234, 0.0235, 0.0228, 0.0230], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-02 02:46:14,385 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=253582.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 02:47:02,500 INFO [train.py:904] (1/8) Epoch 25, batch 10000, loss[loss=0.1638, simple_loss=0.2652, pruned_loss=0.03118, over 16243.00 frames. ], tot_loss[loss=0.1644, simple_loss=0.2612, pruned_loss=0.03383, over 3074844.88 frames. ], batch size: 146, lr: 2.67e-03, grad_scale: 8.0 2023-05-02 02:47:35,261 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.42 vs. limit=2.0 2023-05-02 02:47:39,519 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.5035, 3.9346, 3.9465, 2.7588, 3.5546, 4.0056, 3.6424, 2.3666], device='cuda:1'), covar=tensor([0.0564, 0.0046, 0.0048, 0.0385, 0.0115, 0.0079, 0.0079, 0.0495], device='cuda:1'), in_proj_covar=tensor([0.0132, 0.0084, 0.0085, 0.0131, 0.0098, 0.0107, 0.0094, 0.0128], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-05-02 02:47:53,002 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=253628.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 02:48:22,389 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.70 vs. limit=2.0 2023-05-02 02:48:36,799 INFO [optim.py:368] (1/8) 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 02:48:39,255 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=253649.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 02:48:41,258 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=253650.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 02:48:46,081 INFO [train.py:904] (1/8) Epoch 25, batch 10050, loss[loss=0.1733, simple_loss=0.2671, pruned_loss=0.03976, over 15299.00 frames. ], tot_loss[loss=0.1642, simple_loss=0.2612, pruned_loss=0.03361, over 3094220.21 frames. ], batch size: 191, lr: 2.67e-03, grad_scale: 8.0 2023-05-02 02:49:05,617 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([5.6504, 5.5884, 5.4950, 4.9512, 5.1471, 5.5162, 5.4691, 5.1199], device='cuda:1'), covar=tensor([0.0517, 0.0564, 0.0275, 0.0314, 0.0893, 0.0524, 0.0252, 0.0772], device='cuda:1'), in_proj_covar=tensor([0.0286, 0.0426, 0.0334, 0.0336, 0.0335, 0.0385, 0.0230, 0.0398], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002], device='cuda:1') 2023-05-02 02:49:48,423 INFO [zipformer.py:625] (1/8) 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] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=253697.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 02:50:20,880 INFO [train.py:904] (1/8) Epoch 25, batch 10100, loss[loss=0.1494, simple_loss=0.2421, pruned_loss=0.02831, over 16199.00 frames. ], tot_loss[loss=0.1649, simple_loss=0.2619, pruned_loss=0.03396, over 3099936.46 frames. ], batch size: 165, lr: 2.67e-03, grad_scale: 8.0 2023-05-02 02:50:35,645 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=253711.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 02:51:21,520 INFO [zipformer.py:625] (1/8) 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] (1/8) 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 02:52:07,292 INFO [train.py:904] (1/8) Epoch 26, batch 0, loss[loss=0.1587, simple_loss=0.2444, pruned_loss=0.03649, over 17216.00 frames. ], tot_loss[loss=0.1587, simple_loss=0.2444, pruned_loss=0.03649, over 17216.00 frames. ], batch size: 45, lr: 2.61e-03, grad_scale: 8.0 2023-05-02 02:52:07,292 INFO [train.py:929] (1/8) Computing validation loss 2023-05-02 02:52:14,674 INFO [train.py:938] (1/8) Epoch 26, validation: loss=0.1437, simple_loss=0.2472, pruned_loss=0.02009, over 944034.00 frames. 2023-05-02 02:52:14,674 INFO [train.py:939] (1/8) Maximum memory allocated so far is 18145MB 2023-05-02 02:52:42,098 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.5880, 3.3149, 3.6685, 1.9629, 3.7272, 3.7411, 3.1159, 2.7407], device='cuda:1'), covar=tensor([0.0786, 0.0247, 0.0180, 0.1180, 0.0112, 0.0181, 0.0399, 0.0481], device='cuda:1'), in_proj_covar=tensor([0.0143, 0.0107, 0.0094, 0.0135, 0.0080, 0.0123, 0.0125, 0.0126], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:1') 2023-05-02 02:52:44,976 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=253775.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 02:53:23,661 INFO [train.py:904] (1/8) Epoch 26, batch 50, loss[loss=0.1837, simple_loss=0.2598, pruned_loss=0.0538, over 16479.00 frames. ], tot_loss[loss=0.1771, simple_loss=0.2647, pruned_loss=0.04473, over 753867.50 frames. ], batch size: 75, lr: 2.61e-03, grad_scale: 1.0 2023-05-02 02:53:34,894 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.6409, 4.4266, 4.6486, 4.7964, 4.9354, 4.4400, 4.8770, 4.9052], device='cuda:1'), covar=tensor([0.2296, 0.1750, 0.1912, 0.1085, 0.0799, 0.1309, 0.1481, 0.1791], device='cuda:1'), in_proj_covar=tensor([0.0625, 0.0763, 0.0879, 0.0779, 0.0597, 0.0616, 0.0647, 0.0750], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-02 02:53:47,733 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.23 vs. limit=2.0 2023-05-02 02:54:28,065 INFO [optim.py:368] (1/8) 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] (1/8) Epoch 26, batch 100, loss[loss=0.1563, simple_loss=0.2511, pruned_loss=0.03072, over 17115.00 frames. ], tot_loss[loss=0.1756, simple_loss=0.2638, pruned_loss=0.04368, over 1326773.44 frames. ], batch size: 49, lr: 2.61e-03, grad_scale: 1.0 2023-05-02 02:55:03,715 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=253877.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 02:55:39,315 INFO [train.py:904] (1/8) Epoch 26, batch 150, loss[loss=0.1466, simple_loss=0.2384, pruned_loss=0.02737, over 17194.00 frames. ], tot_loss[loss=0.1724, simple_loss=0.2608, pruned_loss=0.04198, over 1772973.90 frames. ], batch size: 46, lr: 2.61e-03, grad_scale: 1.0 2023-05-02 02:56:07,766 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=253923.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 02:56:11,756 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.8166, 4.0422, 2.6030, 4.5787, 3.0907, 4.5240, 2.8407, 3.3222], device='cuda:1'), covar=tensor([0.0370, 0.0436, 0.1700, 0.0376, 0.0939, 0.0601, 0.1493, 0.0789], device='cuda:1'), in_proj_covar=tensor([0.0171, 0.0175, 0.0194, 0.0165, 0.0176, 0.0213, 0.0202, 0.0180], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-05-02 02:56:46,628 INFO [optim.py:368] (1/8) 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 02:56:49,076 INFO [train.py:904] (1/8) Epoch 26, batch 200, loss[loss=0.1817, simple_loss=0.26, pruned_loss=0.05168, over 16886.00 frames. ], tot_loss[loss=0.1724, simple_loss=0.2608, pruned_loss=0.04196, over 2116551.07 frames. ], batch size: 96, lr: 2.61e-03, grad_scale: 1.0 2023-05-02 02:56:55,779 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.6654, 4.5396, 4.5884, 4.2766, 4.3208, 4.6296, 4.4321, 4.3858], device='cuda:1'), covar=tensor([0.0647, 0.0957, 0.0351, 0.0341, 0.0847, 0.0533, 0.0461, 0.0703], device='cuda:1'), in_proj_covar=tensor([0.0294, 0.0438, 0.0342, 0.0345, 0.0344, 0.0394, 0.0235, 0.0409], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-02 02:57:33,685 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=253985.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 02:58:03,909 INFO [train.py:904] (1/8) Epoch 26, batch 250, loss[loss=0.1718, simple_loss=0.268, pruned_loss=0.03779, over 17042.00 frames. ], tot_loss[loss=0.1729, simple_loss=0.2598, pruned_loss=0.04304, over 2368317.40 frames. ], batch size: 50, lr: 2.61e-03, grad_scale: 1.0 2023-05-02 02:58:07,777 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=254006.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 02:58:42,511 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.3984, 4.2016, 4.4381, 4.5776, 4.7064, 4.2910, 4.5581, 4.6855], device='cuda:1'), covar=tensor([0.1758, 0.1267, 0.1401, 0.0743, 0.0617, 0.1139, 0.2103, 0.0842], device='cuda:1'), in_proj_covar=tensor([0.0637, 0.0779, 0.0898, 0.0793, 0.0606, 0.0628, 0.0662, 0.0763], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-02 02:58:47,155 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=254033.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 02:58:50,820 INFO [zipformer.py:625] (1/8) 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] (1/8) 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] (1/8) Epoch 26, batch 300, loss[loss=0.1537, simple_loss=0.2472, pruned_loss=0.03006, over 17042.00 frames. ], tot_loss[loss=0.1704, simple_loss=0.2573, pruned_loss=0.04178, over 2579937.04 frames. ], batch size: 53, lr: 2.61e-03, grad_scale: 1.0 2023-05-02 02:59:45,464 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=254075.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 02:59:57,060 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 26, batch 350, loss[loss=0.1567, simple_loss=0.2595, pruned_loss=0.0269, over 17274.00 frames. ], tot_loss[loss=0.1675, simple_loss=0.2543, pruned_loss=0.04029, over 2744476.35 frames. ], batch size: 52, lr: 2.61e-03, grad_scale: 1.0 2023-05-02 03:00:24,063 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.0319, 1.9719, 2.6383, 3.0429, 2.7823, 3.4646, 2.0446, 3.5269], device='cuda:1'), covar=tensor([0.0283, 0.0704, 0.0358, 0.0347, 0.0401, 0.0228, 0.0746, 0.0177], device='cuda:1'), in_proj_covar=tensor([0.0191, 0.0195, 0.0183, 0.0185, 0.0202, 0.0159, 0.0198, 0.0159], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-02 03:00:51,986 INFO [zipformer.py:625] (1/8) 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] (1/8) 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] (1/8) Epoch 26, batch 400, loss[loss=0.1599, simple_loss=0.2581, pruned_loss=0.03086, over 17255.00 frames. ], tot_loss[loss=0.1666, simple_loss=0.2526, pruned_loss=0.04031, over 2869596.05 frames. ], batch size: 52, lr: 2.61e-03, grad_scale: 2.0 2023-05-02 03:02:07,820 INFO [zipformer.py:625] (1/8) 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] (1/8) attn_weights_entropy = tensor([3.5769, 3.6224, 3.4358, 3.1194, 3.2823, 3.5548, 3.3325, 3.4282], device='cuda:1'), covar=tensor([0.0535, 0.0565, 0.0320, 0.0285, 0.0539, 0.0435, 0.1413, 0.0466], device='cuda:1'), in_proj_covar=tensor([0.0299, 0.0446, 0.0349, 0.0352, 0.0350, 0.0403, 0.0239, 0.0416], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-02 03:02:44,709 INFO [train.py:904] (1/8) Epoch 26, batch 450, loss[loss=0.1506, simple_loss=0.2451, pruned_loss=0.02799, over 17182.00 frames. ], tot_loss[loss=0.1654, simple_loss=0.2514, pruned_loss=0.03969, over 2966314.68 frames. ], batch size: 43, lr: 2.61e-03, grad_scale: 2.0 2023-05-02 03:03:12,883 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=254223.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 03:03:15,096 INFO [zipformer.py:625] (1/8) 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] (1/8) 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 03:03:52,994 INFO [train.py:904] (1/8) Epoch 26, batch 500, loss[loss=0.1443, simple_loss=0.2294, pruned_loss=0.02966, over 16948.00 frames. ], tot_loss[loss=0.164, simple_loss=0.2498, pruned_loss=0.0391, over 3033954.99 frames. ], batch size: 41, lr: 2.61e-03, grad_scale: 2.0 2023-05-02 03:04:18,987 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=254271.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 03:05:01,743 INFO [train.py:904] (1/8) Epoch 26, batch 550, loss[loss=0.1426, simple_loss=0.2266, pruned_loss=0.02927, over 16964.00 frames. ], tot_loss[loss=0.1635, simple_loss=0.2494, pruned_loss=0.03883, over 3099257.15 frames. ], batch size: 41, lr: 2.61e-03, grad_scale: 2.0 2023-05-02 03:05:05,497 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=254306.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 03:05:47,911 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.09 vs. limit=2.0 2023-05-02 03:06:08,396 INFO [optim.py:368] (1/8) 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 03:06:11,643 INFO [train.py:904] (1/8) Epoch 26, batch 600, loss[loss=0.1589, simple_loss=0.2483, pruned_loss=0.03471, over 17145.00 frames. ], tot_loss[loss=0.1632, simple_loss=0.2485, pruned_loss=0.03894, over 3151326.87 frames. ], batch size: 48, lr: 2.61e-03, grad_scale: 2.0 2023-05-02 03:06:13,034 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=254354.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 03:07:21,734 INFO [train.py:904] (1/8) Epoch 26, batch 650, loss[loss=0.1688, simple_loss=0.2483, pruned_loss=0.04462, over 16442.00 frames. ], tot_loss[loss=0.1621, simple_loss=0.2474, pruned_loss=0.03836, over 3196599.94 frames. ], batch size: 146, lr: 2.61e-03, grad_scale: 2.0 2023-05-02 03:07:22,507 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.83 vs. limit=2.0 2023-05-02 03:07:34,422 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=3.76 vs. limit=5.0 2023-05-02 03:08:28,766 INFO [optim.py:368] (1/8) 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] (1/8) Epoch 26, batch 700, loss[loss=0.1672, simple_loss=0.2533, pruned_loss=0.04056, over 15604.00 frames. ], tot_loss[loss=0.1621, simple_loss=0.248, pruned_loss=0.03806, over 3222978.38 frames. ], batch size: 191, lr: 2.61e-03, grad_scale: 2.0 2023-05-02 03:08:33,882 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.4989, 2.5571, 2.2088, 2.4353, 2.8983, 2.5815, 3.0144, 3.0038], device='cuda:1'), covar=tensor([0.0202, 0.0531, 0.0629, 0.0536, 0.0354, 0.0471, 0.0313, 0.0355], device='cuda:1'), in_proj_covar=tensor([0.0226, 0.0246, 0.0234, 0.0234, 0.0245, 0.0244, 0.0241, 0.0242], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-02 03:09:41,471 INFO [train.py:904] (1/8) Epoch 26, batch 750, loss[loss=0.1927, simple_loss=0.2597, pruned_loss=0.06291, over 16873.00 frames. ], tot_loss[loss=0.1615, simple_loss=0.2477, pruned_loss=0.03769, over 3247346.82 frames. ], batch size: 116, lr: 2.61e-03, grad_scale: 2.0 2023-05-02 03:10:27,491 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=254536.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 03:10:46,993 INFO [optim.py:368] (1/8) 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] (1/8) Epoch 26, batch 800, loss[loss=0.1752, simple_loss=0.2581, pruned_loss=0.04611, over 16630.00 frames. ], tot_loss[loss=0.1617, simple_loss=0.2479, pruned_loss=0.03781, over 3260633.28 frames. ], batch size: 76, lr: 2.61e-03, grad_scale: 4.0 2023-05-02 03:10:50,938 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.6010, 4.6715, 4.7901, 4.6185, 4.6800, 5.2586, 4.7853, 4.4466], device='cuda:1'), covar=tensor([0.1711, 0.2274, 0.2798, 0.2557, 0.2952, 0.1326, 0.1811, 0.2820], device='cuda:1'), in_proj_covar=tensor([0.0423, 0.0622, 0.0691, 0.0512, 0.0677, 0.0713, 0.0535, 0.0681], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-05-02 03:11:19,887 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=254574.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 03:11:51,499 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 26, batch 850, loss[loss=0.1735, simple_loss=0.2517, pruned_loss=0.04767, over 16750.00 frames. ], tot_loss[loss=0.162, simple_loss=0.2481, pruned_loss=0.03791, over 3272493.39 frames. ], batch size: 134, lr: 2.61e-03, grad_scale: 4.0 2023-05-02 03:12:45,132 INFO [zipformer.py:625] (1/8) 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] (1/8) 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 03:13:10,207 INFO [train.py:904] (1/8) Epoch 26, batch 900, loss[loss=0.1546, simple_loss=0.2392, pruned_loss=0.03503, over 16816.00 frames. ], tot_loss[loss=0.1612, simple_loss=0.2473, pruned_loss=0.03755, over 3280038.96 frames. ], batch size: 83, lr: 2.61e-03, grad_scale: 4.0 2023-05-02 03:13:44,830 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.7430, 4.2642, 4.2281, 2.9414, 3.6519, 4.2831, 3.8228, 2.9127], device='cuda:1'), covar=tensor([0.0557, 0.0066, 0.0064, 0.0429, 0.0139, 0.0102, 0.0108, 0.0398], device='cuda:1'), in_proj_covar=tensor([0.0137, 0.0088, 0.0088, 0.0136, 0.0101, 0.0112, 0.0097, 0.0132], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0003, 0.0004], device='cuda:1') 2023-05-02 03:13:58,869 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.0974, 3.1573, 3.3825, 2.1110, 2.8488, 2.3016, 3.4886, 3.5078], device='cuda:1'), covar=tensor([0.0269, 0.0977, 0.0647, 0.2145, 0.0995, 0.1096, 0.0623, 0.1009], device='cuda:1'), in_proj_covar=tensor([0.0160, 0.0168, 0.0170, 0.0157, 0.0148, 0.0132, 0.0146, 0.0181], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0005], device='cuda:1') 2023-05-02 03:14:19,305 INFO [train.py:904] (1/8) Epoch 26, batch 950, loss[loss=0.1428, simple_loss=0.227, pruned_loss=0.02931, over 12261.00 frames. ], tot_loss[loss=0.1607, simple_loss=0.2469, pruned_loss=0.03722, over 3290841.60 frames. ], batch size: 246, lr: 2.61e-03, grad_scale: 4.0 2023-05-02 03:14:34,085 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.9482, 3.4962, 3.8886, 2.2250, 3.9624, 3.9997, 3.2487, 3.0540], device='cuda:1'), covar=tensor([0.0632, 0.0267, 0.0206, 0.1095, 0.0119, 0.0214, 0.0383, 0.0436], device='cuda:1'), in_proj_covar=tensor([0.0148, 0.0111, 0.0099, 0.0140, 0.0085, 0.0129, 0.0129, 0.0131], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-05-02 03:15:24,090 INFO [optim.py:368] (1/8) 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] (1/8) Epoch 26, batch 1000, loss[loss=0.1544, simple_loss=0.227, pruned_loss=0.04087, over 16403.00 frames. ], tot_loss[loss=0.1599, simple_loss=0.2459, pruned_loss=0.03699, over 3296842.52 frames. ], batch size: 146, lr: 2.61e-03, grad_scale: 4.0 2023-05-02 03:15:31,174 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=254755.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 03:16:28,416 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=254797.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 03:16:35,903 INFO [train.py:904] (1/8) Epoch 26, batch 1050, loss[loss=0.1797, simple_loss=0.2774, pruned_loss=0.04103, over 17054.00 frames. ], tot_loss[loss=0.1598, simple_loss=0.246, pruned_loss=0.03679, over 3310720.95 frames. ], batch size: 53, lr: 2.61e-03, grad_scale: 4.0 2023-05-02 03:16:47,338 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.9786, 5.0442, 5.4696, 5.4510, 5.4600, 5.1089, 5.0460, 4.9038], device='cuda:1'), covar=tensor([0.0380, 0.0514, 0.0437, 0.0436, 0.0524, 0.0487, 0.0973, 0.0497], device='cuda:1'), in_proj_covar=tensor([0.0429, 0.0481, 0.0469, 0.0430, 0.0515, 0.0494, 0.0569, 0.0392], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:1') 2023-05-02 03:16:55,049 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=254816.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 03:17:42,909 INFO [zipformer.py:625] (1/8) 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] (1/8) 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] (1/8) Epoch 26, batch 1100, loss[loss=0.1333, simple_loss=0.2197, pruned_loss=0.02344, over 17251.00 frames. ], tot_loss[loss=0.1596, simple_loss=0.2455, pruned_loss=0.03681, over 3307863.49 frames. ], batch size: 45, lr: 2.61e-03, grad_scale: 2.0 2023-05-02 03:17:53,560 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=254858.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 03:18:17,263 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.4613, 4.2896, 4.4945, 4.6499, 4.7524, 4.2807, 4.6147, 4.7363], device='cuda:1'), covar=tensor([0.1651, 0.1191, 0.1432, 0.0675, 0.0581, 0.1192, 0.1781, 0.0806], device='cuda:1'), in_proj_covar=tensor([0.0673, 0.0824, 0.0953, 0.0839, 0.0639, 0.0661, 0.0696, 0.0806], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-02 03:18:36,860 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 26, batch 1150, loss[loss=0.1437, simple_loss=0.2261, pruned_loss=0.03067, over 15632.00 frames. ], tot_loss[loss=0.1587, simple_loss=0.2446, pruned_loss=0.03636, over 3315009.33 frames. ], batch size: 190, lr: 2.61e-03, grad_scale: 2.0 2023-05-02 03:19:04,307 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=254912.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 03:19:29,215 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=254930.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 03:19:46,528 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=3.20 vs. limit=5.0 2023-05-02 03:19:52,757 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.1022, 3.8743, 3.9432, 4.2599, 4.3094, 3.9588, 4.1380, 4.2906], device='cuda:1'), covar=tensor([0.1514, 0.1369, 0.1854, 0.0860, 0.0816, 0.1795, 0.3162, 0.1135], device='cuda:1'), in_proj_covar=tensor([0.0673, 0.0824, 0.0952, 0.0840, 0.0639, 0.0661, 0.0695, 0.0807], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-02 03:19:59,611 INFO [optim.py:368] (1/8) 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] (1/8) Epoch 26, batch 1200, loss[loss=0.1734, simple_loss=0.2653, pruned_loss=0.04077, over 17019.00 frames. ], tot_loss[loss=0.1587, simple_loss=0.2439, pruned_loss=0.03678, over 3297666.57 frames. ], batch size: 55, lr: 2.61e-03, grad_scale: 4.0 2023-05-02 03:21:10,675 INFO [train.py:904] (1/8) Epoch 26, batch 1250, loss[loss=0.1488, simple_loss=0.245, pruned_loss=0.02629, over 17122.00 frames. ], tot_loss[loss=0.1593, simple_loss=0.2442, pruned_loss=0.03716, over 3302698.48 frames. ], batch size: 47, lr: 2.61e-03, grad_scale: 4.0 2023-05-02 03:21:26,774 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.24 vs. limit=2.0 2023-05-02 03:22:19,975 INFO [optim.py:368] (1/8) 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 03:22:21,125 INFO [train.py:904] (1/8) Epoch 26, batch 1300, loss[loss=0.1678, simple_loss=0.2652, pruned_loss=0.0352, over 17258.00 frames. ], tot_loss[loss=0.1589, simple_loss=0.2446, pruned_loss=0.03662, over 3310629.61 frames. ], batch size: 52, lr: 2.61e-03, grad_scale: 4.0 2023-05-02 03:23:12,997 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.15 vs. limit=2.0 2023-05-02 03:23:30,835 INFO [train.py:904] (1/8) Epoch 26, batch 1350, loss[loss=0.1756, simple_loss=0.2504, pruned_loss=0.05042, over 16863.00 frames. ], tot_loss[loss=0.1593, simple_loss=0.2451, pruned_loss=0.03681, over 3316122.35 frames. ], batch size: 96, lr: 2.61e-03, grad_scale: 4.0 2023-05-02 03:23:42,224 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=255111.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 03:24:38,900 INFO [optim.py:368] (1/8) Clipping_scale=2.0, grad-norm quartiles 1.606e+02 2.137e+02 2.466e+02 2.927e+02 7.701e+02, threshold=4.932e+02, percent-clipped=1.0 2023-05-02 03:24:40,852 INFO [train.py:904] (1/8) Epoch 26, batch 1400, loss[loss=0.162, simple_loss=0.2543, pruned_loss=0.03485, over 17033.00 frames. ], tot_loss[loss=0.1583, simple_loss=0.2439, pruned_loss=0.03635, over 3324163.37 frames. ], batch size: 55, lr: 2.61e-03, grad_scale: 4.0 2023-05-02 03:24:41,095 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=255153.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 03:25:12,170 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.2772, 4.0110, 4.4566, 2.4793, 4.7081, 4.7962, 3.4471, 3.7380], device='cuda:1'), covar=tensor([0.0719, 0.0280, 0.0241, 0.1173, 0.0082, 0.0178, 0.0478, 0.0422], device='cuda:1'), in_proj_covar=tensor([0.0150, 0.0113, 0.0101, 0.0141, 0.0086, 0.0131, 0.0131, 0.0133], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-05-02 03:25:29,713 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.89 vs. limit=2.0 2023-05-02 03:25:34,141 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 26, batch 1450, loss[loss=0.1572, simple_loss=0.2293, pruned_loss=0.04255, over 16452.00 frames. ], tot_loss[loss=0.1581, simple_loss=0.2435, pruned_loss=0.03635, over 3332226.27 frames. ], batch size: 75, lr: 2.61e-03, grad_scale: 4.0 2023-05-02 03:25:54,149 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=255207.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 03:26:26,818 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=255230.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 03:26:40,696 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=255240.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 03:26:41,046 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=4.31 vs. limit=5.0 2023-05-02 03:26:56,773 INFO [optim.py:368] (1/8) 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 03:26:57,913 INFO [train.py:904] (1/8) Epoch 26, batch 1500, loss[loss=0.1554, simple_loss=0.2496, pruned_loss=0.03064, over 17254.00 frames. ], tot_loss[loss=0.1591, simple_loss=0.2442, pruned_loss=0.03701, over 3335596.36 frames. ], batch size: 52, lr: 2.61e-03, grad_scale: 4.0 2023-05-02 03:27:09,805 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.8163, 4.8095, 4.7331, 4.1936, 4.7749, 2.0714, 4.5117, 4.4368], device='cuda:1'), covar=tensor([0.0171, 0.0127, 0.0220, 0.0391, 0.0134, 0.2757, 0.0186, 0.0277], device='cuda:1'), in_proj_covar=tensor([0.0179, 0.0172, 0.0210, 0.0185, 0.0186, 0.0218, 0.0199, 0.0180], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-02 03:27:23,589 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=4.42 vs. limit=5.0 2023-05-02 03:27:30,892 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=255278.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 03:28:04,982 INFO [train.py:904] (1/8) Epoch 26, batch 1550, loss[loss=0.1865, simple_loss=0.2547, pruned_loss=0.05914, over 16723.00 frames. ], tot_loss[loss=0.1612, simple_loss=0.2455, pruned_loss=0.03844, over 3328922.70 frames. ], batch size: 89, lr: 2.61e-03, grad_scale: 4.0 2023-05-02 03:29:12,928 INFO [optim.py:368] (1/8) Clipping_scale=2.0, grad-norm quartiles 1.706e+02 2.260e+02 2.709e+02 3.380e+02 1.198e+03, threshold=5.419e+02, percent-clipped=4.0 2023-05-02 03:29:14,103 INFO [train.py:904] (1/8) Epoch 26, batch 1600, loss[loss=0.1874, simple_loss=0.2808, pruned_loss=0.04698, over 16825.00 frames. ], tot_loss[loss=0.1628, simple_loss=0.2479, pruned_loss=0.03882, over 3339319.76 frames. ], batch size: 62, lr: 2.60e-03, grad_scale: 8.0 2023-05-02 03:29:50,873 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([5.2184, 5.7267, 5.8677, 5.6077, 5.6558, 6.2214, 5.7035, 5.3947], device='cuda:1'), covar=tensor([0.0909, 0.2030, 0.2717, 0.1971, 0.2913, 0.1001, 0.1649, 0.2354], device='cuda:1'), in_proj_covar=tensor([0.0427, 0.0632, 0.0701, 0.0519, 0.0686, 0.0725, 0.0544, 0.0690], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-05-02 03:30:03,512 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=255388.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 03:30:23,417 INFO [train.py:904] (1/8) Epoch 26, batch 1650, loss[loss=0.1625, simple_loss=0.2644, pruned_loss=0.0303, over 17119.00 frames. ], tot_loss[loss=0.1633, simple_loss=0.2488, pruned_loss=0.03894, over 3331397.79 frames. ], batch size: 48, lr: 2.60e-03, grad_scale: 8.0 2023-05-02 03:30:35,924 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=255411.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 03:31:28,953 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=255449.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 03:31:29,347 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.73 vs. limit=2.0 2023-05-02 03:31:32,745 INFO [optim.py:368] (1/8) 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] (1/8) Epoch 26, batch 1700, loss[loss=0.1599, simple_loss=0.2598, pruned_loss=0.02998, over 17249.00 frames. ], tot_loss[loss=0.1645, simple_loss=0.2506, pruned_loss=0.03925, over 3330197.47 frames. ], batch size: 52, lr: 2.60e-03, grad_scale: 8.0 2023-05-02 03:31:34,407 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=255453.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 03:31:43,178 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=255459.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 03:31:56,267 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([5.6130, 5.9450, 5.7459, 5.7693, 5.4217, 5.3851, 5.3804, 6.1114], device='cuda:1'), covar=tensor([0.1628, 0.1084, 0.1165, 0.0927, 0.0905, 0.0740, 0.1227, 0.0912], device='cuda:1'), in_proj_covar=tensor([0.0720, 0.0872, 0.0715, 0.0675, 0.0554, 0.0557, 0.0733, 0.0677], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:1') 2023-05-02 03:32:40,141 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=255501.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 03:32:42,741 INFO [train.py:904] (1/8) Epoch 26, batch 1750, loss[loss=0.142, simple_loss=0.2309, pruned_loss=0.02654, over 16975.00 frames. ], tot_loss[loss=0.1659, simple_loss=0.2526, pruned_loss=0.03958, over 3325629.74 frames. ], batch size: 41, lr: 2.60e-03, grad_scale: 8.0 2023-05-02 03:32:47,874 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=255507.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 03:33:49,126 INFO [optim.py:368] (1/8) 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 03:33:49,663 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.71 vs. limit=2.0 2023-05-02 03:33:51,387 INFO [train.py:904] (1/8) Epoch 26, batch 1800, loss[loss=0.1802, simple_loss=0.2713, pruned_loss=0.04457, over 16449.00 frames. ], tot_loss[loss=0.1665, simple_loss=0.2536, pruned_loss=0.03976, over 3319750.91 frames. ], batch size: 146, lr: 2.60e-03, grad_scale: 8.0 2023-05-02 03:33:54,878 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=255555.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 03:34:18,027 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.29 vs. limit=2.0 2023-05-02 03:34:27,945 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.8253, 2.1575, 2.4028, 3.0847, 2.1782, 2.3065, 2.3115, 2.2667], device='cuda:1'), covar=tensor([0.1498, 0.3551, 0.2678, 0.0856, 0.4086, 0.2569, 0.3362, 0.3469], device='cuda:1'), in_proj_covar=tensor([0.0419, 0.0470, 0.0386, 0.0339, 0.0446, 0.0537, 0.0441, 0.0548], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-02 03:34:59,035 INFO [train.py:904] (1/8) Epoch 26, batch 1850, loss[loss=0.1441, simple_loss=0.2377, pruned_loss=0.02525, over 16758.00 frames. ], tot_loss[loss=0.1676, simple_loss=0.2549, pruned_loss=0.04021, over 3316450.87 frames. ], batch size: 39, lr: 2.60e-03, grad_scale: 8.0 2023-05-02 03:36:05,021 INFO [optim.py:368] (1/8) 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] (1/8) Epoch 26, batch 1900, loss[loss=0.1579, simple_loss=0.2366, pruned_loss=0.03963, over 16757.00 frames. ], tot_loss[loss=0.1659, simple_loss=0.2537, pruned_loss=0.03906, over 3317956.15 frames. ], batch size: 124, lr: 2.60e-03, grad_scale: 8.0 2023-05-02 03:37:15,846 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([1.7089, 2.7350, 2.7380, 4.8540, 3.8420, 4.2632, 1.6179, 3.2198], device='cuda:1'), covar=tensor([0.1491, 0.0870, 0.1221, 0.0229, 0.0244, 0.0426, 0.1765, 0.0769], device='cuda:1'), in_proj_covar=tensor([0.0172, 0.0180, 0.0199, 0.0199, 0.0206, 0.0220, 0.0210, 0.0198], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-05-02 03:37:16,451 INFO [train.py:904] (1/8) Epoch 26, batch 1950, loss[loss=0.1875, simple_loss=0.283, pruned_loss=0.04606, over 16737.00 frames. ], tot_loss[loss=0.166, simple_loss=0.2539, pruned_loss=0.03911, over 3315102.80 frames. ], batch size: 57, lr: 2.60e-03, grad_scale: 8.0 2023-05-02 03:37:32,814 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.0566, 3.1589, 3.4562, 2.1226, 2.9870, 2.4071, 3.5260, 3.5173], device='cuda:1'), covar=tensor([0.0262, 0.0939, 0.0635, 0.2066, 0.0871, 0.1062, 0.0590, 0.0881], device='cuda:1'), in_proj_covar=tensor([0.0162, 0.0170, 0.0171, 0.0158, 0.0148, 0.0133, 0.0147, 0.0183], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0005], device='cuda:1') 2023-05-02 03:37:42,785 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.50 vs. limit=2.0 2023-05-02 03:37:55,170 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([5.1004, 5.0291, 4.9565, 4.4546, 4.6274, 5.0163, 4.9389, 4.6437], device='cuda:1'), covar=tensor([0.0644, 0.0570, 0.0375, 0.0422, 0.1177, 0.0497, 0.0397, 0.0801], device='cuda:1'), in_proj_covar=tensor([0.0317, 0.0472, 0.0368, 0.0372, 0.0371, 0.0426, 0.0253, 0.0445], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-05-02 03:38:12,955 INFO [zipformer.py:625] (1/8) 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] (1/8) Clipping_scale=2.0, grad-norm quartiles 1.399e+02 2.052e+02 2.408e+02 2.894e+02 5.578e+02, threshold=4.816e+02, percent-clipped=3.0 2023-05-02 03:38:24,818 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.0544, 3.0899, 3.0800, 5.1695, 4.3818, 4.5191, 1.9171, 3.4799], device='cuda:1'), covar=tensor([0.1286, 0.0771, 0.1070, 0.0184, 0.0208, 0.0409, 0.1569, 0.0709], device='cuda:1'), in_proj_covar=tensor([0.0172, 0.0180, 0.0199, 0.0200, 0.0206, 0.0221, 0.0210, 0.0199], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-05-02 03:38:25,464 INFO [train.py:904] (1/8) Epoch 26, batch 2000, loss[loss=0.1358, simple_loss=0.2189, pruned_loss=0.02633, over 16783.00 frames. ], tot_loss[loss=0.1658, simple_loss=0.2532, pruned_loss=0.03917, over 3317133.59 frames. ], batch size: 39, lr: 2.60e-03, grad_scale: 8.0 2023-05-02 03:38:35,769 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.9165, 4.4617, 2.9209, 2.3458, 2.7662, 2.4924, 4.7574, 3.5777], device='cuda:1'), covar=tensor([0.2891, 0.0533, 0.2080, 0.2960, 0.3077, 0.2275, 0.0355, 0.1493], device='cuda:1'), in_proj_covar=tensor([0.0333, 0.0274, 0.0312, 0.0323, 0.0304, 0.0273, 0.0303, 0.0349], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:1') 2023-05-02 03:38:59,235 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.6404, 1.8403, 2.3263, 2.5406, 2.6147, 2.5560, 1.8706, 2.8220], device='cuda:1'), covar=tensor([0.0215, 0.0549, 0.0386, 0.0344, 0.0334, 0.0387, 0.0573, 0.0199], device='cuda:1'), in_proj_covar=tensor([0.0197, 0.0200, 0.0188, 0.0191, 0.0207, 0.0164, 0.0202, 0.0165], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-02 03:39:35,298 INFO [train.py:904] (1/8) Epoch 26, batch 2050, loss[loss=0.1787, simple_loss=0.2561, pruned_loss=0.05061, over 16708.00 frames. ], tot_loss[loss=0.1655, simple_loss=0.253, pruned_loss=0.03904, over 3317270.48 frames. ], batch size: 134, lr: 2.60e-03, grad_scale: 8.0 2023-05-02 03:40:10,690 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.1109, 3.0531, 2.0582, 3.2371, 2.4166, 3.2868, 2.1704, 2.6188], device='cuda:1'), covar=tensor([0.0345, 0.0482, 0.1572, 0.0420, 0.0876, 0.0673, 0.1453, 0.0819], device='cuda:1'), in_proj_covar=tensor([0.0179, 0.0184, 0.0201, 0.0176, 0.0182, 0.0223, 0.0208, 0.0187], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-05-02 03:40:29,590 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.9855, 2.1819, 2.3347, 3.4576, 2.1573, 2.4358, 2.3177, 2.2919], device='cuda:1'), covar=tensor([0.1493, 0.3417, 0.2957, 0.0820, 0.4073, 0.2426, 0.3407, 0.3399], device='cuda:1'), in_proj_covar=tensor([0.0419, 0.0470, 0.0385, 0.0338, 0.0446, 0.0536, 0.0440, 0.0548], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-02 03:40:44,482 INFO [optim.py:368] (1/8) 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 03:40:45,699 INFO [train.py:904] (1/8) Epoch 26, batch 2100, loss[loss=0.1667, simple_loss=0.2477, pruned_loss=0.04286, over 16858.00 frames. ], tot_loss[loss=0.1664, simple_loss=0.2539, pruned_loss=0.03947, over 3315804.12 frames. ], batch size: 116, lr: 2.60e-03, grad_scale: 8.0 2023-05-02 03:41:54,461 INFO [train.py:904] (1/8) Epoch 26, batch 2150, loss[loss=0.1698, simple_loss=0.2486, pruned_loss=0.04552, over 15517.00 frames. ], tot_loss[loss=0.1667, simple_loss=0.254, pruned_loss=0.0397, over 3318314.44 frames. ], batch size: 190, lr: 2.60e-03, grad_scale: 8.0 2023-05-02 03:42:06,109 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.64 vs. limit=2.0 2023-05-02 03:42:52,683 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([5.1108, 5.0765, 4.9780, 4.5857, 4.6769, 5.0373, 4.8795, 4.6859], device='cuda:1'), covar=tensor([0.0610, 0.0674, 0.0315, 0.0332, 0.0982, 0.0467, 0.0409, 0.0698], device='cuda:1'), in_proj_covar=tensor([0.0319, 0.0474, 0.0369, 0.0374, 0.0372, 0.0427, 0.0254, 0.0446], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-05-02 03:43:04,641 INFO [optim.py:368] (1/8) 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] (1/8) Epoch 26, batch 2200, loss[loss=0.1547, simple_loss=0.2533, pruned_loss=0.02806, over 17136.00 frames. ], tot_loss[loss=0.167, simple_loss=0.2544, pruned_loss=0.03979, over 3323457.43 frames. ], batch size: 49, lr: 2.60e-03, grad_scale: 8.0 2023-05-02 03:43:10,490 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 26, batch 2250, loss[loss=0.1692, simple_loss=0.2613, pruned_loss=0.0386, over 16690.00 frames. ], tot_loss[loss=0.1671, simple_loss=0.2544, pruned_loss=0.03993, over 3316977.87 frames. ], batch size: 62, lr: 2.60e-03, grad_scale: 8.0 2023-05-02 03:44:39,479 INFO [zipformer.py:625] (1/8) 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] (1/8) 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] (1/8) 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] (1/8) 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 03:45:29,155 INFO [train.py:904] (1/8) Epoch 26, batch 2300, loss[loss=0.1583, simple_loss=0.2543, pruned_loss=0.03118, over 16859.00 frames. ], tot_loss[loss=0.1673, simple_loss=0.2547, pruned_loss=0.03995, over 3316324.17 frames. ], batch size: 42, lr: 2.60e-03, grad_scale: 8.0 2023-05-02 03:46:09,908 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=256081.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 03:46:14,189 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=256084.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 03:46:23,797 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=256092.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 03:46:32,856 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([1.9663, 3.1983, 3.3977, 5.2231, 4.5774, 4.4998, 1.8762, 3.7214], device='cuda:1'), covar=tensor([0.1352, 0.0728, 0.0947, 0.0172, 0.0255, 0.0361, 0.1587, 0.0684], device='cuda:1'), in_proj_covar=tensor([0.0172, 0.0181, 0.0200, 0.0200, 0.0207, 0.0221, 0.0210, 0.0199], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-05-02 03:46:34,536 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=256099.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 03:46:39,719 INFO [train.py:904] (1/8) Epoch 26, batch 2350, loss[loss=0.1958, simple_loss=0.2667, pruned_loss=0.06242, over 16758.00 frames. ], tot_loss[loss=0.1672, simple_loss=0.2545, pruned_loss=0.04002, over 3322817.38 frames. ], batch size: 83, lr: 2.60e-03, grad_scale: 8.0 2023-05-02 03:47:19,397 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=256132.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 03:47:32,095 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=256142.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 03:47:36,994 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=256145.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 03:47:45,352 INFO [optim.py:368] (1/8) Clipping_scale=2.0, grad-norm quartiles 1.305e+02 2.073e+02 2.391e+02 2.953e+02 7.650e+02, threshold=4.783e+02, percent-clipped=2.0 2023-05-02 03:47:45,734 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([5.1559, 5.6762, 5.8203, 5.4587, 5.5289, 6.1673, 5.6573, 5.4334], device='cuda:1'), covar=tensor([0.0936, 0.1912, 0.2396, 0.2248, 0.2901, 0.0942, 0.1442, 0.2380], device='cuda:1'), in_proj_covar=tensor([0.0433, 0.0638, 0.0705, 0.0523, 0.0691, 0.0727, 0.0548, 0.0696], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-05-02 03:47:46,525 INFO [train.py:904] (1/8) Epoch 26, batch 2400, loss[loss=0.1652, simple_loss=0.2654, pruned_loss=0.03247, over 17044.00 frames. ], tot_loss[loss=0.1682, simple_loss=0.2555, pruned_loss=0.04049, over 3322715.56 frames. ], batch size: 50, lr: 2.60e-03, grad_scale: 8.0 2023-05-02 03:47:49,959 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.14 vs. limit=2.0 2023-05-02 03:48:00,594 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.6734, 3.7862, 2.8905, 2.2427, 2.4790, 2.4004, 3.9070, 3.3594], device='cuda:1'), covar=tensor([0.2902, 0.0634, 0.1874, 0.3424, 0.3007, 0.2277, 0.0609, 0.1531], device='cuda:1'), in_proj_covar=tensor([0.0332, 0.0274, 0.0311, 0.0322, 0.0304, 0.0272, 0.0302, 0.0350], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:1') 2023-05-02 03:48:24,857 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.1295, 3.9228, 4.4104, 2.2946, 4.6413, 4.6794, 3.3704, 3.6664], device='cuda:1'), covar=tensor([0.0714, 0.0267, 0.0233, 0.1186, 0.0079, 0.0207, 0.0457, 0.0417], device='cuda:1'), in_proj_covar=tensor([0.0150, 0.0113, 0.0101, 0.0141, 0.0086, 0.0132, 0.0131, 0.0133], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-05-02 03:48:42,796 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=256193.0, num_to_drop=1, layers_to_drop={0} 2023-05-02 03:48:55,676 INFO [train.py:904] (1/8) Epoch 26, batch 2450, loss[loss=0.1594, simple_loss=0.2488, pruned_loss=0.03497, over 15944.00 frames. ], tot_loss[loss=0.169, simple_loss=0.2569, pruned_loss=0.04056, over 3323259.79 frames. ], batch size: 35, lr: 2.60e-03, grad_scale: 8.0 2023-05-02 03:50:01,728 INFO [optim.py:368] (1/8) 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 03:50:03,664 INFO [train.py:904] (1/8) Epoch 26, batch 2500, loss[loss=0.1747, simple_loss=0.2689, pruned_loss=0.04029, over 17065.00 frames. ], tot_loss[loss=0.1692, simple_loss=0.2572, pruned_loss=0.04063, over 3325934.63 frames. ], batch size: 53, lr: 2.60e-03, grad_scale: 8.0 2023-05-02 03:50:48,281 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 26, batch 2550, loss[loss=0.1624, simple_loss=0.2615, pruned_loss=0.03161, over 17121.00 frames. ], tot_loss[loss=0.1682, simple_loss=0.2565, pruned_loss=0.03989, over 3327899.89 frames. ], batch size: 49, lr: 2.60e-03, grad_scale: 8.0 2023-05-02 03:51:26,210 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=256312.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 03:52:14,094 INFO [zipformer.py:625] (1/8) 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] (1/8) 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 03:52:23,670 INFO [train.py:904] (1/8) Epoch 26, batch 2600, loss[loss=0.1614, simple_loss=0.2475, pruned_loss=0.03764, over 16785.00 frames. ], tot_loss[loss=0.1684, simple_loss=0.2567, pruned_loss=0.04005, over 3320706.21 frames. ], batch size: 83, lr: 2.60e-03, grad_scale: 8.0 2023-05-02 03:52:50,791 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=256372.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 03:53:05,483 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([5.1893, 5.8155, 6.0616, 5.6173, 5.7175, 6.3309, 5.8292, 5.5689], device='cuda:1'), covar=tensor([0.1009, 0.1891, 0.2245, 0.1829, 0.2445, 0.0868, 0.1251, 0.1916], device='cuda:1'), in_proj_covar=tensor([0.0434, 0.0638, 0.0705, 0.0522, 0.0692, 0.0727, 0.0547, 0.0697], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-05-02 03:53:20,890 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=256394.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 03:53:33,340 INFO [train.py:904] (1/8) Epoch 26, batch 2650, loss[loss=0.178, simple_loss=0.273, pruned_loss=0.04146, over 16525.00 frames. ], tot_loss[loss=0.169, simple_loss=0.2579, pruned_loss=0.04006, over 3320803.46 frames. ], batch size: 68, lr: 2.60e-03, grad_scale: 8.0 2023-05-02 03:54:14,514 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=256433.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 03:54:20,237 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=256437.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 03:54:24,139 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=256440.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 03:54:40,362 INFO [optim.py:368] (1/8) 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] (1/8) Epoch 26, batch 2700, loss[loss=0.197, simple_loss=0.2802, pruned_loss=0.05687, over 16910.00 frames. ], tot_loss[loss=0.1688, simple_loss=0.2583, pruned_loss=0.03963, over 3318857.46 frames. ], batch size: 109, lr: 2.60e-03, grad_scale: 8.0 2023-05-02 03:55:06,537 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=256471.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 03:55:28,872 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=256488.0, num_to_drop=1, layers_to_drop={1} 2023-05-02 03:55:48,614 INFO [train.py:904] (1/8) Epoch 26, batch 2750, loss[loss=0.1917, simple_loss=0.2912, pruned_loss=0.04614, over 17076.00 frames. ], tot_loss[loss=0.1681, simple_loss=0.2582, pruned_loss=0.03899, over 3328642.16 frames. ], batch size: 53, lr: 2.60e-03, grad_scale: 8.0 2023-05-02 03:56:30,003 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=256532.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 03:56:33,699 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.1448, 2.2346, 2.3398, 3.9093, 2.1878, 2.5511, 2.3011, 2.3715], device='cuda:1'), covar=tensor([0.1676, 0.3848, 0.3118, 0.0669, 0.3991, 0.2699, 0.4126, 0.3242], device='cuda:1'), in_proj_covar=tensor([0.0421, 0.0471, 0.0386, 0.0339, 0.0447, 0.0538, 0.0442, 0.0551], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-02 03:56:56,714 INFO [optim.py:368] (1/8) 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 03:56:58,712 INFO [train.py:904] (1/8) Epoch 26, batch 2800, loss[loss=0.1972, simple_loss=0.2659, pruned_loss=0.06428, over 16707.00 frames. ], tot_loss[loss=0.1678, simple_loss=0.2576, pruned_loss=0.03905, over 3331967.16 frames. ], batch size: 134, lr: 2.60e-03, grad_scale: 8.0 2023-05-02 03:57:02,602 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([1.9660, 3.0818, 3.0644, 2.0678, 2.9948, 3.2181, 3.0065, 1.6854], device='cuda:1'), covar=tensor([0.0593, 0.0144, 0.0104, 0.0535, 0.0136, 0.0145, 0.0142, 0.0687], device='cuda:1'), in_proj_covar=tensor([0.0137, 0.0088, 0.0089, 0.0136, 0.0101, 0.0113, 0.0098, 0.0132], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0003, 0.0004], device='cuda:1') 2023-05-02 03:58:07,641 INFO [train.py:904] (1/8) Epoch 26, batch 2850, loss[loss=0.1476, simple_loss=0.2347, pruned_loss=0.03021, over 17015.00 frames. ], tot_loss[loss=0.1672, simple_loss=0.2567, pruned_loss=0.03884, over 3335423.40 frames. ], batch size: 41, lr: 2.60e-03, grad_scale: 8.0 2023-05-02 03:58:21,200 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=256612.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 03:59:00,637 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=256641.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 03:59:09,187 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.0181, 2.1586, 2.7632, 3.0428, 2.9275, 3.5293, 2.4552, 3.5653], device='cuda:1'), covar=tensor([0.0310, 0.0573, 0.0365, 0.0373, 0.0375, 0.0227, 0.0539, 0.0183], device='cuda:1'), in_proj_covar=tensor([0.0197, 0.0199, 0.0188, 0.0191, 0.0206, 0.0166, 0.0203, 0.0165], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-02 03:59:15,039 INFO [optim.py:368] (1/8) 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 03:59:16,876 INFO [train.py:904] (1/8) Epoch 26, batch 2900, loss[loss=0.1519, simple_loss=0.2444, pruned_loss=0.0297, over 17051.00 frames. ], tot_loss[loss=0.1671, simple_loss=0.2555, pruned_loss=0.03936, over 3326514.06 frames. ], batch size: 53, lr: 2.60e-03, grad_scale: 8.0 2023-05-02 03:59:27,184 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=256660.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 03:59:27,447 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([1.8276, 2.9900, 2.9157, 5.0146, 4.0466, 4.3863, 1.7547, 3.3084], device='cuda:1'), covar=tensor([0.1407, 0.0802, 0.1142, 0.0193, 0.0260, 0.0431, 0.1637, 0.0758], device='cuda:1'), in_proj_covar=tensor([0.0171, 0.0181, 0.0200, 0.0201, 0.0207, 0.0220, 0.0209, 0.0198], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-05-02 03:59:42,694 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.69 vs. limit=2.0 2023-05-02 04:00:13,256 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=256694.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 04:00:25,279 INFO [train.py:904] (1/8) Epoch 26, batch 2950, loss[loss=0.1496, simple_loss=0.2474, pruned_loss=0.02593, over 17294.00 frames. ], tot_loss[loss=0.1673, simple_loss=0.2548, pruned_loss=0.03991, over 3326928.04 frames. ], batch size: 52, lr: 2.60e-03, grad_scale: 8.0 2023-05-02 04:00:53,326 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.49 vs. limit=2.0 2023-05-02 04:01:01,151 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=256728.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 04:01:14,552 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=256737.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 04:01:17,981 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=256740.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 04:01:20,853 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=256742.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 04:01:35,074 INFO [optim.py:368] (1/8) 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] (1/8) Epoch 26, batch 3000, loss[loss=0.1957, simple_loss=0.2775, pruned_loss=0.05693, over 16673.00 frames. ], tot_loss[loss=0.1676, simple_loss=0.2545, pruned_loss=0.04032, over 3335306.87 frames. ], batch size: 134, lr: 2.60e-03, grad_scale: 4.0 2023-05-02 04:01:35,090 INFO [train.py:929] (1/8) Computing validation loss 2023-05-02 04:01:44,000 INFO [train.py:938] (1/8) Epoch 26, validation: loss=0.1339, simple_loss=0.2392, pruned_loss=0.01435, over 944034.00 frames. 2023-05-02 04:01:44,001 INFO [train.py:939] (1/8) Maximum memory allocated so far is 18145MB 2023-05-02 04:02:27,414 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=256785.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 04:02:31,514 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=256788.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 04:02:31,626 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=256788.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 04:02:46,208 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.15 vs. limit=2.0 2023-05-02 04:02:53,043 INFO [train.py:904] (1/8) Epoch 26, batch 3050, loss[loss=0.1476, simple_loss=0.2333, pruned_loss=0.03097, over 16994.00 frames. ], tot_loss[loss=0.1671, simple_loss=0.254, pruned_loss=0.04014, over 3334206.50 frames. ], batch size: 41, lr: 2.60e-03, grad_scale: 4.0 2023-05-02 04:03:18,039 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.2553, 2.0954, 1.7266, 1.8320, 2.3101, 2.0599, 2.1833, 2.4176], device='cuda:1'), covar=tensor([0.0303, 0.0404, 0.0516, 0.0470, 0.0247, 0.0327, 0.0232, 0.0305], device='cuda:1'), in_proj_covar=tensor([0.0234, 0.0249, 0.0237, 0.0238, 0.0249, 0.0248, 0.0248, 0.0248], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-02 04:03:27,118 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=256827.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 04:03:27,239 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=256827.0, num_to_drop=1, layers_to_drop={0} 2023-05-02 04:03:38,258 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=256836.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 04:03:38,856 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.33 vs. limit=2.0 2023-05-02 04:04:02,821 INFO [optim.py:368] (1/8) 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 04:04:02,836 INFO [train.py:904] (1/8) Epoch 26, batch 3100, loss[loss=0.1543, simple_loss=0.2326, pruned_loss=0.03796, over 15852.00 frames. ], tot_loss[loss=0.1663, simple_loss=0.2529, pruned_loss=0.03982, over 3333447.16 frames. ], batch size: 35, lr: 2.60e-03, grad_scale: 4.0 2023-05-02 04:04:47,273 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=256883.0, num_to_drop=1, layers_to_drop={1} 2023-05-02 04:04:53,634 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=256888.0, num_to_drop=1, layers_to_drop={2} 2023-05-02 04:05:13,355 INFO [train.py:904] (1/8) Epoch 26, batch 3150, loss[loss=0.1705, simple_loss=0.2584, pruned_loss=0.04126, over 16523.00 frames. ], tot_loss[loss=0.1665, simple_loss=0.253, pruned_loss=0.04006, over 3329474.51 frames. ], batch size: 62, lr: 2.60e-03, grad_scale: 4.0 2023-05-02 04:05:27,405 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.86 vs. limit=2.0 2023-05-02 04:06:05,114 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=256941.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 04:06:09,409 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=256944.0, num_to_drop=1, layers_to_drop={1} 2023-05-02 04:06:21,402 INFO [optim.py:368] (1/8) 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] (1/8) Epoch 26, batch 3200, loss[loss=0.1572, simple_loss=0.2521, pruned_loss=0.03114, over 17169.00 frames. ], tot_loss[loss=0.1658, simple_loss=0.2528, pruned_loss=0.03939, over 3325376.91 frames. ], batch size: 46, lr: 2.60e-03, grad_scale: 8.0 2023-05-02 04:06:36,242 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.6430, 4.6028, 4.5685, 3.9982, 4.6065, 1.8018, 4.3281, 4.1915], device='cuda:1'), covar=tensor([0.0136, 0.0112, 0.0192, 0.0328, 0.0102, 0.2843, 0.0155, 0.0242], device='cuda:1'), in_proj_covar=tensor([0.0180, 0.0174, 0.0212, 0.0188, 0.0188, 0.0218, 0.0201, 0.0182], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-02 04:06:40,301 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.35 vs. limit=2.0 2023-05-02 04:07:11,955 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=256989.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 04:07:29,906 INFO [train.py:904] (1/8) Epoch 26, batch 3250, loss[loss=0.1886, simple_loss=0.2644, pruned_loss=0.05638, over 16706.00 frames. ], tot_loss[loss=0.1657, simple_loss=0.2528, pruned_loss=0.03936, over 3328680.46 frames. ], batch size: 134, lr: 2.60e-03, grad_scale: 8.0 2023-05-02 04:07:43,699 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=257013.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 04:08:03,347 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=257028.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 04:08:04,904 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.3944, 2.3629, 2.4619, 4.2683, 2.3132, 2.7221, 2.4391, 2.5349], device='cuda:1'), covar=tensor([0.1400, 0.3761, 0.3228, 0.0550, 0.4212, 0.2666, 0.3724, 0.3816], device='cuda:1'), in_proj_covar=tensor([0.0421, 0.0471, 0.0385, 0.0339, 0.0446, 0.0538, 0.0441, 0.0551], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-02 04:08:38,567 INFO [optim.py:368] (1/8) 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 04:08:38,583 INFO [train.py:904] (1/8) Epoch 26, batch 3300, loss[loss=0.1669, simple_loss=0.2627, pruned_loss=0.03558, over 16723.00 frames. ], tot_loss[loss=0.167, simple_loss=0.2542, pruned_loss=0.03987, over 3323808.32 frames. ], batch size: 57, lr: 2.60e-03, grad_scale: 8.0 2023-05-02 04:09:07,474 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=257074.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 04:09:09,486 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=257076.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 04:09:48,081 INFO [train.py:904] (1/8) Epoch 26, batch 3350, loss[loss=0.1456, simple_loss=0.2312, pruned_loss=0.02995, over 17223.00 frames. ], tot_loss[loss=0.1664, simple_loss=0.2538, pruned_loss=0.03947, over 3318109.88 frames. ], batch size: 43, lr: 2.60e-03, grad_scale: 8.0 2023-05-02 04:10:20,792 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=257127.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 04:10:26,905 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.65 vs. limit=2.0 2023-05-02 04:10:46,877 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.0153, 3.1491, 3.3332, 2.0768, 2.8277, 2.2868, 3.5609, 3.5145], device='cuda:1'), covar=tensor([0.0263, 0.0987, 0.0660, 0.2013, 0.0895, 0.1078, 0.0508, 0.0859], device='cuda:1'), in_proj_covar=tensor([0.0162, 0.0171, 0.0171, 0.0157, 0.0148, 0.0133, 0.0147, 0.0184], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0005], device='cuda:1') 2023-05-02 04:10:56,577 INFO [optim.py:368] (1/8) 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 04:10:56,593 INFO [train.py:904] (1/8) Epoch 26, batch 3400, loss[loss=0.1447, simple_loss=0.2275, pruned_loss=0.03093, over 15810.00 frames. ], tot_loss[loss=0.1661, simple_loss=0.2539, pruned_loss=0.03919, over 3306274.36 frames. ], batch size: 35, lr: 2.60e-03, grad_scale: 8.0 2023-05-02 04:11:13,585 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=257165.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 04:11:27,519 INFO [zipformer.py:625] (1/8) 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] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=257183.0, num_to_drop=1, layers_to_drop={1} 2023-05-02 04:12:06,429 INFO [train.py:904] (1/8) Epoch 26, batch 3450, loss[loss=0.1374, simple_loss=0.2274, pruned_loss=0.02369, over 16820.00 frames. ], tot_loss[loss=0.1644, simple_loss=0.2522, pruned_loss=0.03834, over 3314802.88 frames. ], batch size: 42, lr: 2.60e-03, grad_scale: 8.0 2023-05-02 04:12:33,413 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=257222.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 04:12:39,900 INFO [zipformer.py:625] (1/8) 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] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=257239.0, num_to_drop=1, layers_to_drop={0} 2023-05-02 04:12:59,750 INFO [zipformer.py:625] (1/8) 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] (1/8) Clipping_scale=2.0, grad-norm quartiles 1.387e+02 2.059e+02 2.361e+02 2.793e+02 7.239e+02, threshold=4.722e+02, percent-clipped=1.0 2023-05-02 04:13:16,473 INFO [train.py:904] (1/8) Epoch 26, batch 3500, loss[loss=0.1786, simple_loss=0.2628, pruned_loss=0.04718, over 15363.00 frames. ], tot_loss[loss=0.1641, simple_loss=0.2518, pruned_loss=0.03818, over 3320405.01 frames. ], batch size: 190, lr: 2.60e-03, grad_scale: 8.0 2023-05-02 04:13:58,922 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=257283.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 04:14:25,700 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=257302.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 04:14:26,373 INFO [train.py:904] (1/8) Epoch 26, batch 3550, loss[loss=0.1521, simple_loss=0.2472, pruned_loss=0.02856, over 17057.00 frames. ], tot_loss[loss=0.1646, simple_loss=0.2519, pruned_loss=0.03862, over 3304812.10 frames. ], batch size: 55, lr: 2.60e-03, grad_scale: 8.0 2023-05-02 04:15:13,148 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.5866, 4.9410, 4.7140, 4.7478, 4.4714, 4.4947, 4.3894, 5.0267], device='cuda:1'), covar=tensor([0.1284, 0.0916, 0.0992, 0.0863, 0.0859, 0.1167, 0.1308, 0.0801], device='cuda:1'), in_proj_covar=tensor([0.0725, 0.0883, 0.0720, 0.0681, 0.0559, 0.0558, 0.0739, 0.0687], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:1') 2023-05-02 04:15:34,859 INFO [train.py:904] (1/8) Epoch 26, batch 3600, loss[loss=0.1504, simple_loss=0.2394, pruned_loss=0.03073, over 17221.00 frames. ], tot_loss[loss=0.163, simple_loss=0.25, pruned_loss=0.03802, over 3309766.07 frames. ], batch size: 44, lr: 2.59e-03, grad_scale: 8.0 2023-05-02 04:15:35,973 INFO [optim.py:368] (1/8) 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 04:15:58,478 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=257369.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 04:16:47,813 INFO [train.py:904] (1/8) Epoch 26, batch 3650, loss[loss=0.1701, simple_loss=0.2412, pruned_loss=0.04956, over 16410.00 frames. ], tot_loss[loss=0.1637, simple_loss=0.2492, pruned_loss=0.03907, over 3290280.29 frames. ], batch size: 68, lr: 2.59e-03, grad_scale: 8.0 2023-05-02 04:17:29,833 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.2668, 4.0295, 4.3780, 2.6215, 4.7201, 4.7721, 3.3225, 3.8211], device='cuda:1'), covar=tensor([0.0658, 0.0298, 0.0281, 0.0981, 0.0092, 0.0184, 0.0466, 0.0352], device='cuda:1'), in_proj_covar=tensor([0.0148, 0.0112, 0.0100, 0.0139, 0.0086, 0.0131, 0.0129, 0.0131], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-05-02 04:17:58,062 INFO [train.py:904] (1/8) Epoch 26, batch 3700, loss[loss=0.1765, simple_loss=0.2552, pruned_loss=0.04894, over 16165.00 frames. ], tot_loss[loss=0.1643, simple_loss=0.2477, pruned_loss=0.04044, over 3281890.72 frames. ], batch size: 164, lr: 2.59e-03, grad_scale: 8.0 2023-05-02 04:17:59,893 INFO [optim.py:368] (1/8) 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 04:18:18,576 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=257467.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 04:18:41,334 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=257483.0, num_to_drop=1, layers_to_drop={1} 2023-05-02 04:18:45,840 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.2934, 4.3142, 4.5955, 4.5745, 4.6207, 4.3587, 4.3635, 4.2471], device='cuda:1'), covar=tensor([0.0349, 0.0614, 0.0378, 0.0388, 0.0485, 0.0443, 0.0839, 0.0730], device='cuda:1'), in_proj_covar=tensor([0.0444, 0.0499, 0.0484, 0.0443, 0.0533, 0.0510, 0.0591, 0.0410], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0003], device='cuda:1') 2023-05-02 04:19:09,128 INFO [train.py:904] (1/8) Epoch 26, batch 3750, loss[loss=0.1897, simple_loss=0.2911, pruned_loss=0.04416, over 16610.00 frames. ], tot_loss[loss=0.1658, simple_loss=0.2482, pruned_loss=0.04166, over 3274557.75 frames. ], batch size: 57, lr: 2.59e-03, grad_scale: 8.0 2023-05-02 04:19:33,996 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=257521.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 04:19:44,804 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=257528.0, num_to_drop=1, layers_to_drop={1} 2023-05-02 04:19:48,977 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=257531.0, num_to_drop=1, layers_to_drop={0} 2023-05-02 04:20:01,202 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=257539.0, num_to_drop=1, layers_to_drop={1} 2023-05-02 04:20:20,023 INFO [train.py:904] (1/8) Epoch 26, batch 3800, loss[loss=0.1779, simple_loss=0.249, pruned_loss=0.05344, over 16852.00 frames. ], tot_loss[loss=0.168, simple_loss=0.2499, pruned_loss=0.04303, over 3275567.63 frames. ], batch size: 116, lr: 2.59e-03, grad_scale: 8.0 2023-05-02 04:20:22,164 INFO [optim.py:368] (1/8) 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 04:20:55,721 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=257578.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 04:20:55,815 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=257578.0, num_to_drop=1, layers_to_drop={1} 2023-05-02 04:21:09,654 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=257587.0, num_to_drop=1, layers_to_drop={1} 2023-05-02 04:21:22,612 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 26, batch 3850, loss[loss=0.176, simple_loss=0.2585, pruned_loss=0.04676, over 16693.00 frames. ], tot_loss[loss=0.1685, simple_loss=0.25, pruned_loss=0.04348, over 3275881.03 frames. ], batch size: 57, lr: 2.59e-03, grad_scale: 8.0 2023-05-02 04:21:54,845 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=3.02 vs. limit=5.0 2023-05-02 04:22:20,355 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=257639.0, num_to_drop=1, layers_to_drop={2} 2023-05-02 04:22:41,230 INFO [train.py:904] (1/8) Epoch 26, batch 3900, loss[loss=0.1524, simple_loss=0.2336, pruned_loss=0.03566, over 16745.00 frames. ], tot_loss[loss=0.1687, simple_loss=0.2499, pruned_loss=0.04379, over 3256102.08 frames. ], batch size: 89, lr: 2.59e-03, grad_scale: 8.0 2023-05-02 04:22:42,469 INFO [optim.py:368] (1/8) 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 04:23:04,479 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=257669.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 04:23:26,313 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.20 vs. limit=2.0 2023-05-02 04:23:51,421 INFO [train.py:904] (1/8) Epoch 26, batch 3950, loss[loss=0.158, simple_loss=0.2382, pruned_loss=0.03888, over 16821.00 frames. ], tot_loss[loss=0.1689, simple_loss=0.2495, pruned_loss=0.04416, over 3271887.50 frames. ], batch size: 102, lr: 2.59e-03, grad_scale: 8.0 2023-05-02 04:24:12,553 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=257717.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 04:24:45,350 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.7977, 4.0689, 2.5870, 2.4706, 2.6038, 2.4016, 4.3493, 3.4072], device='cuda:1'), covar=tensor([0.2986, 0.0753, 0.2326, 0.2743, 0.3105, 0.2359, 0.0588, 0.1389], device='cuda:1'), in_proj_covar=tensor([0.0331, 0.0275, 0.0310, 0.0322, 0.0306, 0.0272, 0.0302, 0.0350], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:1') 2023-05-02 04:25:02,900 INFO [train.py:904] (1/8) Epoch 26, batch 4000, loss[loss=0.1795, simple_loss=0.2629, pruned_loss=0.048, over 12166.00 frames. ], tot_loss[loss=0.1694, simple_loss=0.2495, pruned_loss=0.04469, over 3272885.85 frames. ], batch size: 247, lr: 2.59e-03, grad_scale: 8.0 2023-05-02 04:25:03,988 INFO [optim.py:368] (1/8) 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 04:25:42,209 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.6022, 1.8221, 2.2854, 2.5111, 2.6199, 2.8546, 1.9404, 2.7963], device='cuda:1'), covar=tensor([0.0243, 0.0543, 0.0338, 0.0392, 0.0343, 0.0223, 0.0584, 0.0157], device='cuda:1'), in_proj_covar=tensor([0.0198, 0.0198, 0.0188, 0.0192, 0.0206, 0.0165, 0.0203, 0.0165], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-02 04:26:13,221 INFO [train.py:904] (1/8) Epoch 26, batch 4050, loss[loss=0.1505, simple_loss=0.2346, pruned_loss=0.03315, over 16868.00 frames. ], tot_loss[loss=0.1691, simple_loss=0.2503, pruned_loss=0.04397, over 3271694.17 frames. ], batch size: 109, lr: 2.59e-03, grad_scale: 8.0 2023-05-02 04:26:40,076 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=257821.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 04:26:42,416 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=257823.0, num_to_drop=1, layers_to_drop={1} 2023-05-02 04:27:24,575 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.6570, 4.0588, 3.6650, 3.9373, 3.6270, 3.6185, 3.5864, 4.0374], device='cuda:1'), covar=tensor([0.2899, 0.1742, 0.2733, 0.1664, 0.1833, 0.3365, 0.2721, 0.1915], device='cuda:1'), in_proj_covar=tensor([0.0722, 0.0876, 0.0715, 0.0677, 0.0557, 0.0557, 0.0738, 0.0685], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:1') 2023-05-02 04:27:25,363 INFO [train.py:904] (1/8) Epoch 26, batch 4100, loss[loss=0.2241, simple_loss=0.2964, pruned_loss=0.07593, over 11879.00 frames. ], tot_loss[loss=0.1697, simple_loss=0.2523, pruned_loss=0.04356, over 3261945.98 frames. ], batch size: 246, lr: 2.59e-03, grad_scale: 8.0 2023-05-02 04:27:26,539 INFO [optim.py:368] (1/8) 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 04:27:31,188 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.5718, 4.0063, 3.6242, 3.8796, 3.5263, 3.5900, 3.5451, 3.9809], device='cuda:1'), covar=tensor([0.3060, 0.1706, 0.2681, 0.1737, 0.1990, 0.3427, 0.2673, 0.1990], device='cuda:1'), in_proj_covar=tensor([0.0722, 0.0876, 0.0715, 0.0677, 0.0557, 0.0556, 0.0738, 0.0685], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:1') 2023-05-02 04:27:35,047 INFO [zipformer.py:625] (1/8) 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] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=257869.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 04:28:02,691 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=257878.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 04:28:10,531 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=257883.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 04:28:30,715 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 26, batch 4150, loss[loss=0.2043, simple_loss=0.2924, pruned_loss=0.0581, over 16884.00 frames. ], tot_loss[loss=0.1762, simple_loss=0.2599, pruned_loss=0.04622, over 3234462.11 frames. ], batch size: 109, lr: 2.59e-03, grad_scale: 8.0 2023-05-02 04:29:07,273 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=257920.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 04:29:15,912 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=257926.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 04:29:28,835 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=257934.0, num_to_drop=1, layers_to_drop={1} 2023-05-02 04:29:44,226 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=257944.0, num_to_drop=1, layers_to_drop={0} 2023-05-02 04:29:45,298 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=257945.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 04:29:52,840 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=257950.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 04:29:56,804 INFO [train.py:904] (1/8) Epoch 26, batch 4200, loss[loss=0.221, simple_loss=0.3006, pruned_loss=0.07071, over 11296.00 frames. ], tot_loss[loss=0.1812, simple_loss=0.2664, pruned_loss=0.04798, over 3181925.38 frames. ], batch size: 246, lr: 2.59e-03, grad_scale: 8.0 2023-05-02 04:29:58,474 INFO [optim.py:368] (1/8) 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 04:30:25,233 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.2929, 4.3367, 4.6203, 4.5849, 4.6476, 4.3505, 4.2736, 4.2198], device='cuda:1'), covar=tensor([0.0356, 0.0608, 0.0482, 0.0464, 0.0454, 0.0445, 0.1207, 0.0608], device='cuda:1'), in_proj_covar=tensor([0.0433, 0.0488, 0.0474, 0.0434, 0.0522, 0.0499, 0.0578, 0.0401], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:1') 2023-05-02 04:30:52,167 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.9865, 5.3990, 5.6105, 5.2314, 5.4373, 5.9744, 5.3722, 5.0465], device='cuda:1'), covar=tensor([0.0829, 0.1620, 0.1464, 0.1747, 0.1922, 0.0732, 0.1263, 0.2258], device='cuda:1'), in_proj_covar=tensor([0.0427, 0.0630, 0.0694, 0.0514, 0.0680, 0.0719, 0.0539, 0.0686], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-05-02 04:31:15,317 INFO [train.py:904] (1/8) Epoch 26, batch 4250, loss[loss=0.1671, simple_loss=0.2656, pruned_loss=0.03429, over 15348.00 frames. ], tot_loss[loss=0.1831, simple_loss=0.2699, pruned_loss=0.04811, over 3155361.71 frames. ], batch size: 190, lr: 2.59e-03, grad_scale: 8.0 2023-05-02 04:31:29,155 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=258011.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 04:32:29,617 INFO [train.py:904] (1/8) Epoch 26, batch 4300, loss[loss=0.1919, simple_loss=0.28, pruned_loss=0.05187, over 11922.00 frames. ], tot_loss[loss=0.1829, simple_loss=0.2713, pruned_loss=0.04727, over 3164664.76 frames. ], batch size: 247, lr: 2.59e-03, grad_scale: 8.0 2023-05-02 04:32:31,424 INFO [optim.py:368] (1/8) Clipping_scale=2.0, grad-norm quartiles 1.484e+02 2.134e+02 2.518e+02 2.964e+02 4.474e+02, threshold=5.035e+02, percent-clipped=0.0 2023-05-02 04:33:45,782 INFO [train.py:904] (1/8) Epoch 26, batch 4350, loss[loss=0.2013, simple_loss=0.2906, pruned_loss=0.05599, over 16567.00 frames. ], tot_loss[loss=0.1854, simple_loss=0.2746, pruned_loss=0.04812, over 3174512.81 frames. ], batch size: 62, lr: 2.59e-03, grad_scale: 8.0 2023-05-02 04:34:15,487 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([5.2511, 5.5023, 5.2731, 5.3016, 5.0178, 4.8757, 4.8824, 5.6128], device='cuda:1'), covar=tensor([0.0989, 0.0737, 0.0978, 0.0808, 0.0768, 0.0836, 0.1090, 0.0774], device='cuda:1'), in_proj_covar=tensor([0.0707, 0.0858, 0.0701, 0.0662, 0.0546, 0.0545, 0.0721, 0.0670], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:1') 2023-05-02 04:34:15,527 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=258123.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 04:34:59,000 INFO [train.py:904] (1/8) Epoch 26, batch 4400, loss[loss=0.17, simple_loss=0.2613, pruned_loss=0.03937, over 16648.00 frames. ], tot_loss[loss=0.1879, simple_loss=0.2773, pruned_loss=0.04924, over 3189030.64 frames. ], batch size: 62, lr: 2.59e-03, grad_scale: 8.0 2023-05-02 04:35:00,098 INFO [optim.py:368] (1/8) 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 04:35:25,197 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=258171.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 04:36:11,040 INFO [train.py:904] (1/8) Epoch 26, batch 4450, loss[loss=0.2181, simple_loss=0.2892, pruned_loss=0.07346, over 11604.00 frames. ], tot_loss[loss=0.1912, simple_loss=0.281, pruned_loss=0.05067, over 3188584.37 frames. ], batch size: 247, lr: 2.59e-03, grad_scale: 8.0 2023-05-02 04:36:28,127 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=258215.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 04:36:28,277 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([5.1961, 5.1860, 4.9486, 4.2994, 5.1536, 1.8986, 4.8613, 4.5340], device='cuda:1'), covar=tensor([0.0057, 0.0048, 0.0165, 0.0317, 0.0050, 0.3008, 0.0098, 0.0230], device='cuda:1'), in_proj_covar=tensor([0.0178, 0.0172, 0.0210, 0.0185, 0.0187, 0.0216, 0.0200, 0.0180], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-02 04:36:33,286 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.39 vs. limit=2.0 2023-05-02 04:36:55,327 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=258234.0, num_to_drop=1, layers_to_drop={0} 2023-05-02 04:37:03,324 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=258239.0, num_to_drop=1, layers_to_drop={1} 2023-05-02 04:37:12,417 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.21 vs. limit=2.0 2023-05-02 04:37:22,665 INFO [train.py:904] (1/8) Epoch 26, batch 4500, loss[loss=0.1986, simple_loss=0.2839, pruned_loss=0.05667, over 16553.00 frames. ], tot_loss[loss=0.1925, simple_loss=0.2817, pruned_loss=0.05162, over 3184981.68 frames. ], batch size: 62, lr: 2.59e-03, grad_scale: 8.0 2023-05-02 04:37:23,847 INFO [optim.py:368] (1/8) Clipping_scale=2.0, grad-norm quartiles 1.439e+02 1.900e+02 2.188e+02 2.602e+02 4.588e+02, threshold=4.376e+02, percent-clipped=0.0 2023-05-02 04:38:05,644 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=258282.0, num_to_drop=1, layers_to_drop={1} 2023-05-02 04:38:05,878 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.4020, 3.5650, 3.7737, 2.2855, 3.1658, 2.2110, 3.6564, 3.8030], device='cuda:1'), covar=tensor([0.0227, 0.0754, 0.0535, 0.1999, 0.0857, 0.1088, 0.0564, 0.0840], device='cuda:1'), in_proj_covar=tensor([0.0160, 0.0168, 0.0169, 0.0155, 0.0147, 0.0131, 0.0145, 0.0181], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0005], device='cuda:1') 2023-05-02 04:38:24,305 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.84 vs. limit=2.0 2023-05-02 04:38:35,280 INFO [train.py:904] (1/8) Epoch 26, batch 4550, loss[loss=0.2248, simple_loss=0.3088, pruned_loss=0.07038, over 16877.00 frames. ], tot_loss[loss=0.1941, simple_loss=0.2825, pruned_loss=0.0528, over 3205127.61 frames. ], batch size: 116, lr: 2.59e-03, grad_scale: 8.0 2023-05-02 04:38:39,771 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 26, batch 4600, loss[loss=0.1736, simple_loss=0.2762, pruned_loss=0.03554, over 16761.00 frames. ], tot_loss[loss=0.1942, simple_loss=0.2831, pruned_loss=0.05265, over 3223151.73 frames. ], batch size: 83, lr: 2.59e-03, grad_scale: 8.0 2023-05-02 04:39:50,256 INFO [optim.py:368] (1/8) 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 04:40:03,566 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([5.5764, 5.4937, 5.3790, 4.9893, 5.1124, 5.4242, 5.2986, 5.0812], device='cuda:1'), covar=tensor([0.0430, 0.0390, 0.0218, 0.0264, 0.0775, 0.0306, 0.0233, 0.0597], device='cuda:1'), in_proj_covar=tensor([0.0306, 0.0455, 0.0358, 0.0361, 0.0359, 0.0413, 0.0245, 0.0431], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002], device='cuda:1') 2023-05-02 04:41:03,103 INFO [train.py:904] (1/8) Epoch 26, batch 4650, loss[loss=0.1906, simple_loss=0.2752, pruned_loss=0.05304, over 17027.00 frames. ], tot_loss[loss=0.1943, simple_loss=0.2825, pruned_loss=0.05309, over 3217417.22 frames. ], batch size: 53, lr: 2.59e-03, grad_scale: 8.0 2023-05-02 04:41:07,926 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.9884, 2.1582, 2.2013, 3.5734, 2.0684, 2.4360, 2.3205, 2.2938], device='cuda:1'), covar=tensor([0.1606, 0.3631, 0.3046, 0.0689, 0.4420, 0.2560, 0.3407, 0.3577], device='cuda:1'), in_proj_covar=tensor([0.0417, 0.0467, 0.0381, 0.0335, 0.0444, 0.0535, 0.0439, 0.0546], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-02 04:42:14,223 INFO [train.py:904] (1/8) Epoch 26, batch 4700, loss[loss=0.1946, simple_loss=0.2867, pruned_loss=0.05126, over 15421.00 frames. ], tot_loss[loss=0.1918, simple_loss=0.2798, pruned_loss=0.05187, over 3204838.58 frames. ], batch size: 190, lr: 2.59e-03, grad_scale: 8.0 2023-05-02 04:42:16,009 INFO [optim.py:368] (1/8) 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 04:42:16,469 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=258454.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 04:42:23,068 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.1981, 3.9522, 3.8908, 2.5392, 3.4564, 3.9490, 3.4892, 2.1772], device='cuda:1'), covar=tensor([0.0690, 0.0055, 0.0058, 0.0476, 0.0121, 0.0115, 0.0139, 0.0530], device='cuda:1'), in_proj_covar=tensor([0.0138, 0.0088, 0.0090, 0.0136, 0.0102, 0.0114, 0.0098, 0.0132], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0003, 0.0004], device='cuda:1') 2023-05-02 04:42:45,908 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([1.8157, 2.7555, 2.6092, 1.8956, 2.5604, 2.7394, 2.5816, 1.9171], device='cuda:1'), covar=tensor([0.0540, 0.0103, 0.0092, 0.0435, 0.0154, 0.0152, 0.0155, 0.0437], device='cuda:1'), in_proj_covar=tensor([0.0138, 0.0088, 0.0090, 0.0136, 0.0102, 0.0114, 0.0098, 0.0132], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0003, 0.0004], device='cuda:1') 2023-05-02 04:43:23,145 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([5.2920, 5.2759, 5.1010, 4.3809, 5.1914, 1.9018, 4.9039, 4.7453], device='cuda:1'), covar=tensor([0.0076, 0.0103, 0.0169, 0.0458, 0.0108, 0.2948, 0.0128, 0.0265], device='cuda:1'), in_proj_covar=tensor([0.0177, 0.0172, 0.0210, 0.0184, 0.0186, 0.0215, 0.0199, 0.0179], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-02 04:43:26,791 INFO [train.py:904] (1/8) Epoch 26, batch 4750, loss[loss=0.1775, simple_loss=0.2663, pruned_loss=0.04433, over 15390.00 frames. ], tot_loss[loss=0.187, simple_loss=0.2752, pruned_loss=0.04944, over 3212969.04 frames. ], batch size: 190, lr: 2.59e-03, grad_scale: 8.0 2023-05-02 04:43:38,890 INFO [zipformer.py:625] (1/8) 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] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=258515.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 04:43:45,136 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=258515.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 04:44:21,165 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=258539.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 04:44:41,657 INFO [train.py:904] (1/8) Epoch 26, batch 4800, loss[loss=0.1774, simple_loss=0.2613, pruned_loss=0.04671, over 12026.00 frames. ], tot_loss[loss=0.1831, simple_loss=0.2712, pruned_loss=0.04745, over 3201231.99 frames. ], batch size: 248, lr: 2.59e-03, grad_scale: 8.0 2023-05-02 04:44:43,327 INFO [optim.py:368] (1/8) 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] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=258563.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 04:44:58,733 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.33 vs. limit=2.0 2023-05-02 04:45:11,427 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=258572.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 04:45:34,302 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=258587.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 04:45:57,754 INFO [train.py:904] (1/8) Epoch 26, batch 4850, loss[loss=0.1746, simple_loss=0.2725, pruned_loss=0.03833, over 16416.00 frames. ], tot_loss[loss=0.1814, simple_loss=0.2706, pruned_loss=0.04611, over 3190795.77 frames. ], batch size: 146, lr: 2.59e-03, grad_scale: 8.0 2023-05-02 04:46:03,357 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=258606.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 04:47:14,826 INFO [train.py:904] (1/8) Epoch 26, batch 4900, loss[loss=0.1752, simple_loss=0.2651, pruned_loss=0.04268, over 17056.00 frames. ], tot_loss[loss=0.1798, simple_loss=0.27, pruned_loss=0.04483, over 3187666.31 frames. ], batch size: 50, lr: 2.59e-03, grad_scale: 8.0 2023-05-02 04:47:16,710 INFO [optim.py:368] (1/8) 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] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=258654.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 04:47:20,589 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.84 vs. limit=2.0 2023-05-02 04:47:22,663 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.5646, 4.5989, 4.9248, 4.8646, 4.9028, 4.6090, 4.5835, 4.4497], device='cuda:1'), covar=tensor([0.0274, 0.0528, 0.0305, 0.0392, 0.0404, 0.0365, 0.0809, 0.0478], device='cuda:1'), in_proj_covar=tensor([0.0422, 0.0475, 0.0462, 0.0424, 0.0510, 0.0486, 0.0564, 0.0392], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:1') 2023-05-02 04:47:28,221 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.5549, 3.6551, 2.1200, 4.2547, 2.7159, 4.1799, 2.4064, 2.9378], device='cuda:1'), covar=tensor([0.0308, 0.0375, 0.1829, 0.0166, 0.0876, 0.0460, 0.1515, 0.0822], device='cuda:1'), in_proj_covar=tensor([0.0176, 0.0180, 0.0196, 0.0170, 0.0179, 0.0219, 0.0204, 0.0184], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-05-02 04:47:44,929 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.0815, 3.4155, 3.5579, 2.0638, 3.0042, 2.3592, 3.5837, 3.6495], device='cuda:1'), covar=tensor([0.0249, 0.0791, 0.0619, 0.2086, 0.0862, 0.0920, 0.0608, 0.0891], device='cuda:1'), in_proj_covar=tensor([0.0160, 0.0169, 0.0170, 0.0156, 0.0147, 0.0131, 0.0146, 0.0181], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0005], device='cuda:1') 2023-05-02 04:48:29,883 INFO [train.py:904] (1/8) Epoch 26, batch 4950, loss[loss=0.1721, simple_loss=0.2672, pruned_loss=0.03847, over 16469.00 frames. ], tot_loss[loss=0.1783, simple_loss=0.269, pruned_loss=0.04384, over 3201167.34 frames. ], batch size: 68, lr: 2.59e-03, grad_scale: 8.0 2023-05-02 04:49:41,062 INFO [train.py:904] (1/8) Epoch 26, batch 5000, loss[loss=0.1906, simple_loss=0.2782, pruned_loss=0.05149, over 12045.00 frames. ], tot_loss[loss=0.1789, simple_loss=0.2707, pruned_loss=0.04356, over 3210696.13 frames. ], batch size: 247, lr: 2.59e-03, grad_scale: 8.0 2023-05-02 04:49:42,187 INFO [optim.py:368] (1/8) 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 04:50:54,490 INFO [train.py:904] (1/8) Epoch 26, batch 5050, loss[loss=0.1997, simple_loss=0.2887, pruned_loss=0.05533, over 17060.00 frames. ], tot_loss[loss=0.179, simple_loss=0.2712, pruned_loss=0.04343, over 3220722.68 frames. ], batch size: 53, lr: 2.59e-03, grad_scale: 8.0 2023-05-02 04:51:05,005 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=258810.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 04:52:01,334 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.8644, 3.1020, 3.3743, 2.0311, 2.9156, 2.2377, 3.3706, 3.2942], device='cuda:1'), covar=tensor([0.0246, 0.0832, 0.0583, 0.2010, 0.0835, 0.0961, 0.0597, 0.0873], device='cuda:1'), in_proj_covar=tensor([0.0162, 0.0170, 0.0171, 0.0157, 0.0148, 0.0132, 0.0147, 0.0182], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0005], device='cuda:1') 2023-05-02 04:52:07,189 INFO [train.py:904] (1/8) Epoch 26, batch 5100, loss[loss=0.1645, simple_loss=0.2617, pruned_loss=0.03369, over 16230.00 frames. ], tot_loss[loss=0.1778, simple_loss=0.2698, pruned_loss=0.04291, over 3232893.35 frames. ], batch size: 165, lr: 2.59e-03, grad_scale: 8.0 2023-05-02 04:52:08,930 INFO [optim.py:368] (1/8) 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 04:52:27,840 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=258867.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 04:52:33,430 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.1513, 3.1732, 1.9159, 3.4744, 2.3205, 3.5044, 2.1657, 2.6402], device='cuda:1'), covar=tensor([0.0316, 0.0383, 0.1723, 0.0173, 0.0908, 0.0506, 0.1548, 0.0785], device='cuda:1'), in_proj_covar=tensor([0.0175, 0.0179, 0.0195, 0.0169, 0.0178, 0.0217, 0.0203, 0.0183], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-05-02 04:53:21,028 INFO [train.py:904] (1/8) Epoch 26, batch 5150, loss[loss=0.1878, simple_loss=0.2836, pruned_loss=0.04603, over 16934.00 frames. ], tot_loss[loss=0.1779, simple_loss=0.2702, pruned_loss=0.04274, over 3208269.92 frames. ], batch size: 109, lr: 2.59e-03, grad_scale: 8.0 2023-05-02 04:54:34,709 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.56 vs. limit=2.0 2023-05-02 04:54:35,099 INFO [train.py:904] (1/8) Epoch 26, batch 5200, loss[loss=0.1554, simple_loss=0.2447, pruned_loss=0.03304, over 16590.00 frames. ], tot_loss[loss=0.1768, simple_loss=0.2689, pruned_loss=0.04237, over 3196572.43 frames. ], batch size: 75, lr: 2.59e-03, grad_scale: 8.0 2023-05-02 04:54:36,751 INFO [optim.py:368] (1/8) 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 04:54:52,291 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.4364, 4.6894, 4.4668, 4.5455, 4.2330, 4.2292, 4.1556, 4.7135], device='cuda:1'), covar=tensor([0.1243, 0.0921, 0.1022, 0.0811, 0.0850, 0.1417, 0.1205, 0.1003], device='cuda:1'), in_proj_covar=tensor([0.0698, 0.0847, 0.0693, 0.0652, 0.0538, 0.0536, 0.0714, 0.0662], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:1') 2023-05-02 04:54:53,542 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=258965.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 04:54:59,025 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.9376, 2.0213, 2.2416, 3.5174, 1.9809, 2.2717, 2.1355, 2.1493], device='cuda:1'), covar=tensor([0.1816, 0.4381, 0.3197, 0.0744, 0.5109, 0.3124, 0.4182, 0.3912], device='cuda:1'), in_proj_covar=tensor([0.0418, 0.0470, 0.0382, 0.0336, 0.0445, 0.0536, 0.0440, 0.0549], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-02 04:55:07,744 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.8261, 4.0852, 2.8372, 2.4625, 2.8287, 2.5917, 4.2064, 3.4309], device='cuda:1'), covar=tensor([0.2834, 0.0556, 0.2068, 0.2696, 0.2527, 0.2037, 0.0524, 0.1245], device='cuda:1'), in_proj_covar=tensor([0.0328, 0.0271, 0.0308, 0.0319, 0.0301, 0.0268, 0.0299, 0.0344], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:1') 2023-05-02 04:55:48,687 INFO [train.py:904] (1/8) Epoch 26, batch 5250, loss[loss=0.1648, simple_loss=0.2594, pruned_loss=0.03507, over 16661.00 frames. ], tot_loss[loss=0.1755, simple_loss=0.2665, pruned_loss=0.04225, over 3197902.88 frames. ], batch size: 62, lr: 2.59e-03, grad_scale: 8.0 2023-05-02 04:56:23,027 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=259026.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 04:57:03,177 INFO [train.py:904] (1/8) Epoch 26, batch 5300, loss[loss=0.1548, simple_loss=0.2415, pruned_loss=0.03409, over 16925.00 frames. ], tot_loss[loss=0.1729, simple_loss=0.2634, pruned_loss=0.04125, over 3193733.68 frames. ], batch size: 109, lr: 2.59e-03, grad_scale: 8.0 2023-05-02 04:57:04,410 INFO [optim.py:368] (1/8) 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] (1/8) Epoch 26, batch 5350, loss[loss=0.1822, simple_loss=0.28, pruned_loss=0.0422, over 16653.00 frames. ], tot_loss[loss=0.1714, simple_loss=0.262, pruned_loss=0.04044, over 3206719.97 frames. ], batch size: 134, lr: 2.59e-03, grad_scale: 8.0 2023-05-02 04:58:27,851 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=259110.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 04:59:30,711 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([5.2379, 5.2228, 5.0767, 4.6318, 4.6945, 5.1081, 5.1018, 4.7679], device='cuda:1'), covar=tensor([0.0658, 0.0593, 0.0314, 0.0341, 0.1227, 0.0554, 0.0298, 0.0696], device='cuda:1'), in_proj_covar=tensor([0.0308, 0.0459, 0.0359, 0.0363, 0.0361, 0.0418, 0.0244, 0.0433], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002], device='cuda:1') 2023-05-02 04:59:31,425 INFO [train.py:904] (1/8) Epoch 26, batch 5400, loss[loss=0.1792, simple_loss=0.2742, pruned_loss=0.04212, over 17068.00 frames. ], tot_loss[loss=0.1731, simple_loss=0.2644, pruned_loss=0.04089, over 3195668.14 frames. ], batch size: 50, lr: 2.59e-03, grad_scale: 8.0 2023-05-02 04:59:32,573 INFO [optim.py:368] (1/8) 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 04:59:35,352 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.93 vs. limit=2.0 2023-05-02 04:59:38,341 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=259158.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 04:59:52,964 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=259167.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 05:00:36,729 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.4351, 2.6052, 2.1288, 2.4170, 2.8677, 2.5966, 2.9877, 3.1088], device='cuda:1'), covar=tensor([0.0115, 0.0400, 0.0528, 0.0408, 0.0272, 0.0379, 0.0202, 0.0267], device='cuda:1'), in_proj_covar=tensor([0.0224, 0.0240, 0.0230, 0.0231, 0.0241, 0.0239, 0.0239, 0.0239], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-02 05:00:48,642 INFO [train.py:904] (1/8) Epoch 26, batch 5450, loss[loss=0.1904, simple_loss=0.2729, pruned_loss=0.05399, over 16595.00 frames. ], tot_loss[loss=0.1764, simple_loss=0.2678, pruned_loss=0.04251, over 3177151.23 frames. ], batch size: 57, lr: 2.59e-03, grad_scale: 8.0 2023-05-02 05:01:07,879 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=259215.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 05:01:12,353 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.5264, 4.3699, 4.5248, 4.6935, 4.8868, 4.3957, 4.8617, 4.8835], device='cuda:1'), covar=tensor([0.1882, 0.1275, 0.1738, 0.0798, 0.0581, 0.1105, 0.0627, 0.0714], device='cuda:1'), in_proj_covar=tensor([0.0667, 0.0818, 0.0945, 0.0827, 0.0630, 0.0657, 0.0683, 0.0795], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-02 05:01:24,999 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=259226.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 05:01:29,700 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.4849, 4.5572, 4.3920, 4.0830, 4.0574, 4.4678, 4.2490, 4.1850], device='cuda:1'), covar=tensor([0.0651, 0.0736, 0.0294, 0.0298, 0.0889, 0.0562, 0.0580, 0.0646], device='cuda:1'), in_proj_covar=tensor([0.0307, 0.0459, 0.0359, 0.0362, 0.0361, 0.0418, 0.0244, 0.0433], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002], device='cuda:1') 2023-05-02 05:01:32,569 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=4.86 vs. limit=5.0 2023-05-02 05:02:05,203 INFO [train.py:904] (1/8) Epoch 26, batch 5500, loss[loss=0.1953, simple_loss=0.2898, pruned_loss=0.05043, over 17115.00 frames. ], tot_loss[loss=0.183, simple_loss=0.274, pruned_loss=0.04605, over 3160801.52 frames. ], batch size: 47, lr: 2.59e-03, grad_scale: 8.0 2023-05-02 05:02:07,102 INFO [optim.py:368] (1/8) Clipping_scale=2.0, grad-norm quartiles 1.528e+02 2.367e+02 2.972e+02 3.909e+02 6.600e+02, threshold=5.944e+02, percent-clipped=13.0 2023-05-02 05:02:23,198 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=259264.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 05:02:57,506 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 26, batch 5550, loss[loss=0.2372, simple_loss=0.3158, pruned_loss=0.07936, over 15348.00 frames. ], tot_loss[loss=0.1906, simple_loss=0.2805, pruned_loss=0.05033, over 3148525.67 frames. ], batch size: 190, lr: 2.59e-03, grad_scale: 8.0 2023-05-02 05:03:50,534 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=259321.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 05:03:57,666 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=259325.0, num_to_drop=1, layers_to_drop={1} 2023-05-02 05:04:40,414 INFO [train.py:904] (1/8) Epoch 26, batch 5600, loss[loss=0.1816, simple_loss=0.2668, pruned_loss=0.04824, over 16581.00 frames. ], tot_loss[loss=0.1976, simple_loss=0.2856, pruned_loss=0.05483, over 3110007.77 frames. ], batch size: 57, lr: 2.58e-03, grad_scale: 16.0 2023-05-02 05:04:41,807 INFO [optim.py:368] (1/8) Clipping_scale=2.0, grad-norm quartiles 1.957e+02 3.012e+02 3.705e+02 4.647e+02 1.087e+03, threshold=7.410e+02, percent-clipped=8.0 2023-05-02 05:05:44,995 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.4062, 2.5921, 2.1229, 2.3832, 2.8937, 2.5369, 2.8731, 3.1218], device='cuda:1'), covar=tensor([0.0150, 0.0454, 0.0620, 0.0469, 0.0307, 0.0406, 0.0252, 0.0285], device='cuda:1'), in_proj_covar=tensor([0.0225, 0.0241, 0.0230, 0.0232, 0.0241, 0.0240, 0.0240, 0.0240], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-02 05:06:04,636 INFO [train.py:904] (1/8) Epoch 26, batch 5650, loss[loss=0.1935, simple_loss=0.2875, pruned_loss=0.0497, over 16903.00 frames. ], tot_loss[loss=0.2049, simple_loss=0.291, pruned_loss=0.05939, over 3075883.40 frames. ], batch size: 96, lr: 2.58e-03, grad_scale: 8.0 2023-05-02 05:06:22,253 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=259414.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 05:06:59,071 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.4799, 3.4903, 3.4597, 2.6907, 3.2898, 2.1293, 3.1962, 2.7307], device='cuda:1'), covar=tensor([0.0160, 0.0152, 0.0196, 0.0218, 0.0107, 0.2374, 0.0141, 0.0254], device='cuda:1'), in_proj_covar=tensor([0.0177, 0.0170, 0.0209, 0.0184, 0.0185, 0.0215, 0.0198, 0.0178], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-02 05:07:22,187 INFO [train.py:904] (1/8) Epoch 26, batch 5700, loss[loss=0.2592, simple_loss=0.3183, pruned_loss=0.1, over 11445.00 frames. ], tot_loss[loss=0.2067, simple_loss=0.2921, pruned_loss=0.06062, over 3059940.82 frames. ], batch size: 248, lr: 2.58e-03, grad_scale: 8.0 2023-05-02 05:07:25,089 INFO [optim.py:368] (1/8) 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 05:07:56,984 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=259475.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 05:08:38,092 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.4771, 3.5700, 3.3043, 2.9575, 3.1485, 3.4342, 3.3209, 3.2910], device='cuda:1'), covar=tensor([0.0548, 0.0558, 0.0282, 0.0281, 0.0453, 0.0449, 0.1137, 0.0488], device='cuda:1'), in_proj_covar=tensor([0.0305, 0.0456, 0.0356, 0.0359, 0.0357, 0.0415, 0.0242, 0.0430], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002], device='cuda:1') 2023-05-02 05:08:39,968 INFO [train.py:904] (1/8) Epoch 26, batch 5750, loss[loss=0.1976, simple_loss=0.2886, pruned_loss=0.05335, over 16698.00 frames. ], tot_loss[loss=0.2101, simple_loss=0.2949, pruned_loss=0.0626, over 3019677.47 frames. ], batch size: 89, lr: 2.58e-03, grad_scale: 8.0 2023-05-02 05:10:02,530 INFO [train.py:904] (1/8) Epoch 26, batch 5800, loss[loss=0.2217, simple_loss=0.2897, pruned_loss=0.07684, over 11871.00 frames. ], tot_loss[loss=0.2098, simple_loss=0.2953, pruned_loss=0.06213, over 3006076.82 frames. ], batch size: 248, lr: 2.58e-03, grad_scale: 8.0 2023-05-02 05:10:05,673 INFO [optim.py:368] (1/8) 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 05:10:26,919 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.7004, 4.6812, 4.5065, 3.7911, 4.6217, 1.7590, 4.3810, 4.1335], device='cuda:1'), covar=tensor([0.0126, 0.0111, 0.0222, 0.0372, 0.0101, 0.3099, 0.0137, 0.0293], device='cuda:1'), in_proj_covar=tensor([0.0176, 0.0170, 0.0208, 0.0183, 0.0184, 0.0214, 0.0197, 0.0177], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-02 05:10:47,021 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=259582.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 05:10:50,263 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=4.64 vs. limit=5.0 2023-05-02 05:11:19,029 INFO [train.py:904] (1/8) Epoch 26, batch 5850, loss[loss=0.203, simple_loss=0.2929, pruned_loss=0.05654, over 16790.00 frames. ], tot_loss[loss=0.2063, simple_loss=0.2927, pruned_loss=0.05993, over 3039980.15 frames. ], batch size: 83, lr: 2.58e-03, grad_scale: 8.0 2023-05-02 05:11:44,960 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=259620.0, num_to_drop=1, layers_to_drop={2} 2023-05-02 05:11:46,424 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=259621.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 05:12:39,965 INFO [train.py:904] (1/8) Epoch 26, batch 5900, loss[loss=0.2132, simple_loss=0.3065, pruned_loss=0.05994, over 17112.00 frames. ], tot_loss[loss=0.2059, simple_loss=0.2924, pruned_loss=0.05969, over 3051887.33 frames. ], batch size: 49, lr: 2.58e-03, grad_scale: 8.0 2023-05-02 05:12:43,693 INFO [optim.py:368] (1/8) 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 05:13:08,193 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 26, batch 5950, loss[loss=0.2066, simple_loss=0.3004, pruned_loss=0.05635, over 16754.00 frames. ], tot_loss[loss=0.2047, simple_loss=0.2928, pruned_loss=0.05832, over 3060993.69 frames. ], batch size: 83, lr: 2.58e-03, grad_scale: 8.0 2023-05-02 05:14:05,979 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.8337, 4.6785, 4.8203, 5.0041, 5.1690, 4.6415, 5.1950, 5.1729], device='cuda:1'), covar=tensor([0.1873, 0.1311, 0.1678, 0.0785, 0.0727, 0.1027, 0.0714, 0.0794], device='cuda:1'), in_proj_covar=tensor([0.0664, 0.0814, 0.0939, 0.0824, 0.0628, 0.0654, 0.0680, 0.0794], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-02 05:14:43,092 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.09 vs. limit=2.0 2023-05-02 05:15:12,597 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=259749.0, num_to_drop=1, layers_to_drop={0} 2023-05-02 05:15:18,035 INFO [train.py:904] (1/8) Epoch 26, batch 6000, loss[loss=0.1881, simple_loss=0.2778, pruned_loss=0.04922, over 16219.00 frames. ], tot_loss[loss=0.2047, simple_loss=0.2924, pruned_loss=0.05852, over 3072486.98 frames. ], batch size: 165, lr: 2.58e-03, grad_scale: 8.0 2023-05-02 05:15:18,036 INFO [train.py:929] (1/8) Computing validation loss 2023-05-02 05:15:28,182 INFO [train.py:938] (1/8) Epoch 26, validation: loss=0.1485, simple_loss=0.2607, pruned_loss=0.01818, over 944034.00 frames. 2023-05-02 05:15:28,183 INFO [train.py:939] (1/8) Maximum memory allocated so far is 18145MB 2023-05-02 05:15:30,551 INFO [optim.py:368] (1/8) Clipping_scale=2.0, grad-norm quartiles 1.937e+02 2.681e+02 3.578e+02 4.333e+02 8.656e+02, threshold=7.155e+02, percent-clipped=3.0 2023-05-02 05:15:55,081 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=259770.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 05:16:13,701 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.3097, 3.1136, 3.5073, 1.8822, 3.6175, 3.6653, 2.8253, 2.7376], device='cuda:1'), covar=tensor([0.0865, 0.0308, 0.0216, 0.1204, 0.0096, 0.0204, 0.0489, 0.0483], device='cuda:1'), in_proj_covar=tensor([0.0148, 0.0111, 0.0100, 0.0139, 0.0086, 0.0130, 0.0129, 0.0130], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-05-02 05:16:13,710 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=259782.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 05:16:46,426 INFO [train.py:904] (1/8) Epoch 26, batch 6050, loss[loss=0.2097, simple_loss=0.3007, pruned_loss=0.05935, over 16398.00 frames. ], tot_loss[loss=0.2027, simple_loss=0.2906, pruned_loss=0.05738, over 3085932.93 frames. ], batch size: 146, lr: 2.58e-03, grad_scale: 8.0 2023-05-02 05:16:59,629 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=259810.0, num_to_drop=1, layers_to_drop={2} 2023-05-02 05:17:05,665 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.6779, 4.7278, 5.0675, 5.0232, 5.0438, 4.7605, 4.6934, 4.6167], device='cuda:1'), covar=tensor([0.0370, 0.0680, 0.0438, 0.0447, 0.0475, 0.0459, 0.1090, 0.0576], device='cuda:1'), in_proj_covar=tensor([0.0420, 0.0473, 0.0460, 0.0421, 0.0508, 0.0484, 0.0561, 0.0389], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:1') 2023-05-02 05:17:51,021 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=259843.0, num_to_drop=1, layers_to_drop={1} 2023-05-02 05:18:02,539 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=4.85 vs. limit=5.0 2023-05-02 05:18:05,774 INFO [train.py:904] (1/8) Epoch 26, batch 6100, loss[loss=0.1815, simple_loss=0.2786, pruned_loss=0.04218, over 16861.00 frames. ], tot_loss[loss=0.2014, simple_loss=0.2899, pruned_loss=0.0564, over 3092593.78 frames. ], batch size: 102, lr: 2.58e-03, grad_scale: 8.0 2023-05-02 05:18:09,301 INFO [optim.py:368] (1/8) 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 05:18:27,208 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=259866.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 05:18:52,981 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 26, batch 6150, loss[loss=0.1755, simple_loss=0.27, pruned_loss=0.04052, over 16707.00 frames. ], tot_loss[loss=0.1999, simple_loss=0.2881, pruned_loss=0.05582, over 3095191.30 frames. ], batch size: 124, lr: 2.58e-03, grad_scale: 8.0 2023-05-02 05:19:44,912 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.4067, 3.3715, 3.4198, 3.4936, 3.5225, 3.2767, 3.5048, 3.5720], device='cuda:1'), covar=tensor([0.1219, 0.0910, 0.1112, 0.0665, 0.0748, 0.2353, 0.1086, 0.0866], device='cuda:1'), in_proj_covar=tensor([0.0664, 0.0816, 0.0941, 0.0826, 0.0632, 0.0654, 0.0684, 0.0796], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-02 05:19:48,887 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=259920.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 05:19:54,964 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.2683, 1.6211, 1.9871, 2.1284, 2.2170, 2.4135, 1.8071, 2.3220], device='cuda:1'), covar=tensor([0.0244, 0.0514, 0.0299, 0.0424, 0.0334, 0.0224, 0.0523, 0.0162], device='cuda:1'), in_proj_covar=tensor([0.0195, 0.0197, 0.0185, 0.0190, 0.0205, 0.0163, 0.0202, 0.0163], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-02 05:20:00,754 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=259927.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 05:20:05,850 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=259930.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 05:20:42,382 INFO [train.py:904] (1/8) Epoch 26, batch 6200, loss[loss=0.2023, simple_loss=0.2848, pruned_loss=0.05988, over 11579.00 frames. ], tot_loss[loss=0.1987, simple_loss=0.2863, pruned_loss=0.05559, over 3090192.96 frames. ], batch size: 248, lr: 2.58e-03, grad_scale: 8.0 2023-05-02 05:20:44,632 INFO [optim.py:368] (1/8) 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 05:21:04,922 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=259968.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 05:21:25,538 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.4249, 3.4063, 3.4601, 3.5297, 3.5585, 3.3156, 3.5327, 3.6191], device='cuda:1'), covar=tensor([0.1287, 0.0919, 0.1024, 0.0627, 0.0704, 0.2351, 0.1099, 0.0836], device='cuda:1'), in_proj_covar=tensor([0.0663, 0.0814, 0.0939, 0.0822, 0.0630, 0.0652, 0.0682, 0.0794], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-02 05:22:00,547 INFO [train.py:904] (1/8) Epoch 26, batch 6250, loss[loss=0.2082, simple_loss=0.298, pruned_loss=0.0592, over 15339.00 frames. ], tot_loss[loss=0.1981, simple_loss=0.2858, pruned_loss=0.05524, over 3098250.36 frames. ], batch size: 190, lr: 2.58e-03, grad_scale: 8.0 2023-05-02 05:22:41,232 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=260030.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 05:23:15,623 INFO [train.py:904] (1/8) Epoch 26, batch 6300, loss[loss=0.2099, simple_loss=0.2839, pruned_loss=0.06796, over 11412.00 frames. ], tot_loss[loss=0.1964, simple_loss=0.2847, pruned_loss=0.05406, over 3110769.68 frames. ], batch size: 247, lr: 2.58e-03, grad_scale: 8.0 2023-05-02 05:23:19,600 INFO [optim.py:368] (1/8) Clipping_scale=2.0, grad-norm quartiles 1.731e+02 2.817e+02 3.510e+02 4.108e+02 7.742e+02, threshold=7.020e+02, percent-clipped=4.0 2023-05-02 05:23:32,163 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.7054, 4.9694, 4.7347, 4.7595, 4.5163, 4.4836, 4.4043, 5.0285], device='cuda:1'), covar=tensor([0.1174, 0.0886, 0.1050, 0.0925, 0.0820, 0.1212, 0.1242, 0.0945], device='cuda:1'), in_proj_covar=tensor([0.0707, 0.0852, 0.0702, 0.0658, 0.0541, 0.0540, 0.0719, 0.0667], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:1') 2023-05-02 05:23:44,916 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=260070.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 05:24:17,352 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=260091.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 05:24:35,600 INFO [train.py:904] (1/8) Epoch 26, batch 6350, loss[loss=0.2009, simple_loss=0.2825, pruned_loss=0.05964, over 16979.00 frames. ], tot_loss[loss=0.1969, simple_loss=0.2849, pruned_loss=0.0545, over 3123102.11 frames. ], batch size: 41, lr: 2.58e-03, grad_scale: 8.0 2023-05-02 05:24:37,537 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=2.95 vs. limit=5.0 2023-05-02 05:24:39,480 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=260105.0, num_to_drop=1, layers_to_drop={3} 2023-05-02 05:24:42,653 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.5564, 3.7354, 2.7433, 2.2678, 2.4919, 2.4052, 3.9213, 3.4557], device='cuda:1'), covar=tensor([0.3166, 0.0670, 0.2027, 0.3055, 0.2693, 0.2162, 0.0569, 0.1284], device='cuda:1'), in_proj_covar=tensor([0.0331, 0.0273, 0.0309, 0.0322, 0.0303, 0.0271, 0.0300, 0.0346], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:1') 2023-05-02 05:24:59,053 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=260118.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 05:25:03,806 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.5752, 3.7727, 2.8234, 2.2583, 2.4757, 2.4381, 3.9826, 3.3380], device='cuda:1'), covar=tensor([0.2873, 0.0577, 0.1789, 0.2651, 0.2549, 0.2029, 0.0447, 0.1315], device='cuda:1'), in_proj_covar=tensor([0.0330, 0.0273, 0.0308, 0.0322, 0.0303, 0.0271, 0.0300, 0.0346], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:1') 2023-05-02 05:25:22,923 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.2993, 1.5208, 2.0285, 2.1151, 2.2265, 2.3855, 1.7401, 2.3249], device='cuda:1'), covar=tensor([0.0258, 0.0578, 0.0328, 0.0374, 0.0355, 0.0223, 0.0572, 0.0178], device='cuda:1'), in_proj_covar=tensor([0.0195, 0.0196, 0.0185, 0.0189, 0.0205, 0.0162, 0.0202, 0.0162], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-02 05:25:29,700 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=260138.0, num_to_drop=1, layers_to_drop={2} 2023-05-02 05:25:52,116 INFO [train.py:904] (1/8) Epoch 26, batch 6400, loss[loss=0.2692, simple_loss=0.3362, pruned_loss=0.1011, over 11438.00 frames. ], tot_loss[loss=0.1981, simple_loss=0.285, pruned_loss=0.05559, over 3119787.32 frames. ], batch size: 248, lr: 2.58e-03, grad_scale: 8.0 2023-05-02 05:25:54,619 INFO [optim.py:368] (1/8) 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 05:27:08,966 INFO [train.py:904] (1/8) Epoch 26, batch 6450, loss[loss=0.1825, simple_loss=0.2704, pruned_loss=0.04731, over 16165.00 frames. ], tot_loss[loss=0.197, simple_loss=0.2847, pruned_loss=0.05467, over 3129973.91 frames. ], batch size: 165, lr: 2.58e-03, grad_scale: 8.0 2023-05-02 05:27:39,887 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=260222.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 05:27:43,189 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=260224.0, num_to_drop=1, layers_to_drop={1} 2023-05-02 05:28:28,518 INFO [train.py:904] (1/8) Epoch 26, batch 6500, loss[loss=0.1746, simple_loss=0.263, pruned_loss=0.04317, over 16682.00 frames. ], tot_loss[loss=0.1965, simple_loss=0.2836, pruned_loss=0.0547, over 3120592.08 frames. ], batch size: 76, lr: 2.58e-03, grad_scale: 8.0 2023-05-02 05:28:31,535 INFO [optim.py:368] (1/8) 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 05:29:20,117 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=260285.0, num_to_drop=1, layers_to_drop={0} 2023-05-02 05:29:49,653 INFO [train.py:904] (1/8) Epoch 26, batch 6550, loss[loss=0.2291, simple_loss=0.3299, pruned_loss=0.06411, over 16350.00 frames. ], tot_loss[loss=0.1989, simple_loss=0.2863, pruned_loss=0.05574, over 3123736.12 frames. ], batch size: 146, lr: 2.58e-03, grad_scale: 8.0 2023-05-02 05:31:07,551 INFO [train.py:904] (1/8) Epoch 26, batch 6600, loss[loss=0.249, simple_loss=0.3141, pruned_loss=0.09188, over 11396.00 frames. ], tot_loss[loss=0.2007, simple_loss=0.2884, pruned_loss=0.05655, over 3105123.70 frames. ], batch size: 248, lr: 2.58e-03, grad_scale: 8.0 2023-05-02 05:31:09,946 INFO [optim.py:368] (1/8) 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 05:31:59,445 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 26, batch 6650, loss[loss=0.2481, simple_loss=0.3166, pruned_loss=0.08978, over 11401.00 frames. ], tot_loss[loss=0.2025, simple_loss=0.289, pruned_loss=0.05802, over 3079384.13 frames. ], batch size: 247, lr: 2.58e-03, grad_scale: 8.0 2023-05-02 05:32:30,355 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=260405.0, num_to_drop=1, layers_to_drop={2} 2023-05-02 05:33:21,157 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 26, batch 6700, loss[loss=0.1916, simple_loss=0.2804, pruned_loss=0.0514, over 16558.00 frames. ], tot_loss[loss=0.2017, simple_loss=0.2875, pruned_loss=0.05793, over 3078135.07 frames. ], batch size: 68, lr: 2.58e-03, grad_scale: 8.0 2023-05-02 05:33:43,549 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=260453.0, num_to_drop=1, layers_to_drop={0} 2023-05-02 05:33:45,910 INFO [optim.py:368] (1/8) 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 05:33:46,543 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.6357, 3.8678, 2.8948, 2.1942, 2.5045, 2.4768, 4.1493, 3.3526], device='cuda:1'), covar=tensor([0.3031, 0.0631, 0.1852, 0.2974, 0.2864, 0.2160, 0.0447, 0.1416], device='cuda:1'), in_proj_covar=tensor([0.0331, 0.0274, 0.0309, 0.0322, 0.0303, 0.0271, 0.0301, 0.0346], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:1') 2023-05-02 05:34:34,795 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=260486.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 05:35:01,131 INFO [train.py:904] (1/8) Epoch 26, batch 6750, loss[loss=0.1729, simple_loss=0.2576, pruned_loss=0.04408, over 17124.00 frames. ], tot_loss[loss=0.2, simple_loss=0.2858, pruned_loss=0.05715, over 3089583.41 frames. ], batch size: 48, lr: 2.58e-03, grad_scale: 8.0 2023-05-02 05:35:31,836 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=260522.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 05:36:01,316 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([1.8650, 2.7206, 2.6059, 1.9746, 2.5867, 2.7092, 2.5719, 1.9890], device='cuda:1'), covar=tensor([0.0473, 0.0096, 0.0102, 0.0372, 0.0159, 0.0145, 0.0136, 0.0407], device='cuda:1'), in_proj_covar=tensor([0.0136, 0.0087, 0.0088, 0.0134, 0.0101, 0.0113, 0.0097, 0.0130], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0003, 0.0004], device='cuda:1') 2023-05-02 05:36:19,268 INFO [train.py:904] (1/8) Epoch 26, batch 6800, loss[loss=0.2085, simple_loss=0.3004, pruned_loss=0.0583, over 17137.00 frames. ], tot_loss[loss=0.2001, simple_loss=0.2857, pruned_loss=0.05727, over 3085310.01 frames. ], batch size: 48, lr: 2.58e-03, grad_scale: 8.0 2023-05-02 05:36:21,473 INFO [optim.py:368] (1/8) 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] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=260570.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 05:36:55,017 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.5137, 4.5013, 4.3568, 3.6441, 4.4554, 1.6580, 4.2051, 4.0084], device='cuda:1'), covar=tensor([0.0116, 0.0114, 0.0224, 0.0375, 0.0097, 0.3059, 0.0138, 0.0273], device='cuda:1'), in_proj_covar=tensor([0.0177, 0.0171, 0.0209, 0.0184, 0.0185, 0.0216, 0.0198, 0.0178], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-02 05:37:01,663 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=260580.0, num_to_drop=1, layers_to_drop={1} 2023-05-02 05:37:35,978 INFO [train.py:904] (1/8) Epoch 26, batch 6850, loss[loss=0.1803, simple_loss=0.2886, pruned_loss=0.036, over 16505.00 frames. ], tot_loss[loss=0.2012, simple_loss=0.2873, pruned_loss=0.05751, over 3094842.25 frames. ], batch size: 75, lr: 2.58e-03, grad_scale: 4.0 2023-05-02 05:38:30,025 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.78 vs. limit=2.0 2023-05-02 05:38:50,099 INFO [train.py:904] (1/8) Epoch 26, batch 6900, loss[loss=0.1834, simple_loss=0.2827, pruned_loss=0.04201, over 17096.00 frames. ], tot_loss[loss=0.2027, simple_loss=0.2899, pruned_loss=0.0578, over 3089820.74 frames. ], batch size: 49, lr: 2.58e-03, grad_scale: 4.0 2023-05-02 05:38:53,858 INFO [optim.py:368] (1/8) Clipping_scale=2.0, grad-norm quartiles 1.801e+02 2.543e+02 3.098e+02 3.712e+02 7.299e+02, threshold=6.197e+02, percent-clipped=1.0 2023-05-02 05:39:40,212 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=260686.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 05:40:05,852 INFO [train.py:904] (1/8) Epoch 26, batch 6950, loss[loss=0.2237, simple_loss=0.3073, pruned_loss=0.07008, over 16320.00 frames. ], tot_loss[loss=0.2062, simple_loss=0.2921, pruned_loss=0.06012, over 3071061.71 frames. ], batch size: 165, lr: 2.58e-03, grad_scale: 4.0 2023-05-02 05:40:25,277 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.2280, 2.3308, 2.3625, 3.9727, 2.2954, 2.6691, 2.4072, 2.4445], device='cuda:1'), covar=tensor([0.1405, 0.3448, 0.2976, 0.0560, 0.3993, 0.2491, 0.3515, 0.3326], device='cuda:1'), in_proj_covar=tensor([0.0414, 0.0464, 0.0379, 0.0332, 0.0442, 0.0531, 0.0436, 0.0544], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-02 05:40:30,583 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.3194, 4.0462, 4.0024, 2.6846, 3.6404, 4.0292, 3.6528, 2.4263], device='cuda:1'), covar=tensor([0.0619, 0.0056, 0.0063, 0.0428, 0.0109, 0.0126, 0.0100, 0.0458], device='cuda:1'), in_proj_covar=tensor([0.0137, 0.0088, 0.0089, 0.0135, 0.0101, 0.0114, 0.0097, 0.0131], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0003, 0.0004], device='cuda:1') 2023-05-02 05:40:46,913 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.8505, 4.8760, 4.7076, 4.3328, 4.3445, 4.7945, 4.6254, 4.4771], device='cuda:1'), covar=tensor([0.0616, 0.0555, 0.0350, 0.0361, 0.1109, 0.0509, 0.0454, 0.0727], device='cuda:1'), in_proj_covar=tensor([0.0302, 0.0452, 0.0352, 0.0354, 0.0352, 0.0408, 0.0242, 0.0426], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002], device='cuda:1') 2023-05-02 05:40:48,830 INFO [zipformer.py:625] (1/8) 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] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=260734.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 05:41:21,309 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=260751.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 05:41:23,898 INFO [train.py:904] (1/8) Epoch 26, batch 7000, loss[loss=0.2235, simple_loss=0.3084, pruned_loss=0.06929, over 15562.00 frames. ], tot_loss[loss=0.2046, simple_loss=0.2918, pruned_loss=0.05866, over 3074274.44 frames. ], batch size: 191, lr: 2.58e-03, grad_scale: 2.0 2023-05-02 05:41:29,415 INFO [optim.py:368] (1/8) 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 05:42:24,289 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=260791.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 05:42:39,644 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=4.22 vs. limit=5.0 2023-05-02 05:42:42,009 INFO [train.py:904] (1/8) Epoch 26, batch 7050, loss[loss=0.2117, simple_loss=0.2972, pruned_loss=0.06313, over 17004.00 frames. ], tot_loss[loss=0.2043, simple_loss=0.2926, pruned_loss=0.05801, over 3093823.75 frames. ], batch size: 55, lr: 2.58e-03, grad_scale: 2.0 2023-05-02 05:42:55,673 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=260812.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 05:43:59,435 INFO [train.py:904] (1/8) Epoch 26, batch 7100, loss[loss=0.1951, simple_loss=0.2906, pruned_loss=0.04981, over 16188.00 frames. ], tot_loss[loss=0.2035, simple_loss=0.2914, pruned_loss=0.0578, over 3079991.50 frames. ], batch size: 165, lr: 2.58e-03, grad_scale: 2.0 2023-05-02 05:44:05,368 INFO [optim.py:368] (1/8) 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 05:44:09,460 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.5363, 3.5931, 3.3679, 3.0168, 3.2077, 3.5064, 3.3212, 3.3073], device='cuda:1'), covar=tensor([0.0585, 0.0705, 0.0315, 0.0323, 0.0528, 0.0530, 0.1311, 0.0499], device='cuda:1'), in_proj_covar=tensor([0.0301, 0.0451, 0.0350, 0.0353, 0.0351, 0.0406, 0.0240, 0.0425], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-02 05:44:42,690 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=260880.0, num_to_drop=1, layers_to_drop={1} 2023-05-02 05:44:44,363 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=260881.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 05:45:17,430 INFO [train.py:904] (1/8) Epoch 26, batch 7150, loss[loss=0.1953, simple_loss=0.2779, pruned_loss=0.05638, over 15336.00 frames. ], tot_loss[loss=0.202, simple_loss=0.2892, pruned_loss=0.05735, over 3074850.34 frames. ], batch size: 191, lr: 2.58e-03, grad_scale: 2.0 2023-05-02 05:45:53,427 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=260928.0, num_to_drop=1, layers_to_drop={0} 2023-05-02 05:46:13,591 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 26, batch 7200, loss[loss=0.1786, simple_loss=0.2724, pruned_loss=0.04234, over 16456.00 frames. ], tot_loss[loss=0.1992, simple_loss=0.287, pruned_loss=0.05572, over 3062763.46 frames. ], batch size: 146, lr: 2.58e-03, grad_scale: 4.0 2023-05-02 05:46:35,554 INFO [optim.py:368] (1/8) 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 05:47:28,449 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.97 vs. limit=2.0 2023-05-02 05:47:53,028 INFO [train.py:904] (1/8) Epoch 26, batch 7250, loss[loss=0.2307, simple_loss=0.3068, pruned_loss=0.07728, over 11594.00 frames. ], tot_loss[loss=0.1974, simple_loss=0.2846, pruned_loss=0.05506, over 3034305.77 frames. ], batch size: 247, lr: 2.58e-03, grad_scale: 4.0 2023-05-02 05:49:09,252 INFO [train.py:904] (1/8) Epoch 26, batch 7300, loss[loss=0.2257, simple_loss=0.3121, pruned_loss=0.06962, over 15271.00 frames. ], tot_loss[loss=0.197, simple_loss=0.2842, pruned_loss=0.05489, over 3041641.15 frames. ], batch size: 190, lr: 2.58e-03, grad_scale: 4.0 2023-05-02 05:49:13,650 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.6153, 3.0261, 3.1319, 1.9710, 2.7735, 2.0264, 3.1498, 3.3126], device='cuda:1'), covar=tensor([0.0288, 0.0913, 0.0670, 0.2291, 0.0924, 0.1131, 0.0765, 0.0994], device='cuda:1'), in_proj_covar=tensor([0.0161, 0.0170, 0.0171, 0.0157, 0.0148, 0.0133, 0.0147, 0.0182], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0005], device='cuda:1') 2023-05-02 05:49:15,975 INFO [optim.py:368] (1/8) 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 05:50:00,841 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=261086.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 05:50:26,307 INFO [train.py:904] (1/8) Epoch 26, batch 7350, loss[loss=0.1741, simple_loss=0.2666, pruned_loss=0.04076, over 17047.00 frames. ], tot_loss[loss=0.1987, simple_loss=0.2855, pruned_loss=0.05593, over 3044749.42 frames. ], batch size: 50, lr: 2.58e-03, grad_scale: 4.0 2023-05-02 05:50:32,672 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=261107.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 05:51:18,091 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=261136.0, num_to_drop=1, layers_to_drop={1} 2023-05-02 05:51:24,182 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.9492, 4.1731, 4.0232, 4.0545, 3.7478, 3.7830, 3.8350, 4.1777], device='cuda:1'), covar=tensor([0.1083, 0.0834, 0.0972, 0.0816, 0.0719, 0.1762, 0.0935, 0.0914], device='cuda:1'), in_proj_covar=tensor([0.0695, 0.0832, 0.0688, 0.0643, 0.0532, 0.0533, 0.0705, 0.0653], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:1') 2023-05-02 05:51:34,783 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.3551, 3.2934, 2.6719, 2.1872, 2.2727, 2.3588, 3.3943, 3.0799], device='cuda:1'), covar=tensor([0.3125, 0.0708, 0.1804, 0.2725, 0.2595, 0.2199, 0.0538, 0.1343], device='cuda:1'), in_proj_covar=tensor([0.0333, 0.0274, 0.0310, 0.0323, 0.0304, 0.0272, 0.0301, 0.0348], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:1') 2023-05-02 05:51:43,918 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=4.73 vs. limit=5.0 2023-05-02 05:51:44,996 INFO [train.py:904] (1/8) Epoch 26, batch 7400, loss[loss=0.2018, simple_loss=0.2917, pruned_loss=0.05593, over 16229.00 frames. ], tot_loss[loss=0.2005, simple_loss=0.2871, pruned_loss=0.05697, over 3033719.04 frames. ], batch size: 165, lr: 2.58e-03, grad_scale: 4.0 2023-05-02 05:51:50,753 INFO [optim.py:368] (1/8) 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 05:52:55,019 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=261197.0, num_to_drop=1, layers_to_drop={3} 2023-05-02 05:53:04,382 INFO [train.py:904] (1/8) Epoch 26, batch 7450, loss[loss=0.2058, simple_loss=0.2978, pruned_loss=0.05685, over 16788.00 frames. ], tot_loss[loss=0.2022, simple_loss=0.2882, pruned_loss=0.05813, over 3031400.11 frames. ], batch size: 83, lr: 2.58e-03, grad_scale: 4.0 2023-05-02 05:53:15,232 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.8094, 5.0736, 4.8660, 4.8694, 4.6204, 4.5841, 4.4456, 5.1561], device='cuda:1'), covar=tensor([0.1182, 0.0918, 0.0944, 0.0932, 0.0788, 0.1010, 0.1250, 0.0909], device='cuda:1'), in_proj_covar=tensor([0.0696, 0.0835, 0.0689, 0.0645, 0.0533, 0.0534, 0.0706, 0.0653], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:1') 2023-05-02 05:54:01,619 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=261237.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 05:54:27,096 INFO [train.py:904] (1/8) Epoch 26, batch 7500, loss[loss=0.2014, simple_loss=0.2813, pruned_loss=0.06076, over 16913.00 frames. ], tot_loss[loss=0.2014, simple_loss=0.2884, pruned_loss=0.05725, over 3042129.71 frames. ], batch size: 109, lr: 2.58e-03, grad_scale: 4.0 2023-05-02 05:54:33,503 INFO [optim.py:368] (1/8) 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] (1/8) Epoch 26, batch 7550, loss[loss=0.1896, simple_loss=0.2702, pruned_loss=0.05453, over 17035.00 frames. ], tot_loss[loss=0.2013, simple_loss=0.2874, pruned_loss=0.05766, over 3023871.80 frames. ], batch size: 55, lr: 2.58e-03, grad_scale: 4.0 2023-05-02 05:56:04,347 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([1.5970, 2.4515, 2.2536, 3.6985, 2.6657, 3.8192, 1.3403, 2.7275], device='cuda:1'), covar=tensor([0.1547, 0.0948, 0.1485, 0.0226, 0.0238, 0.0431, 0.2001, 0.0909], device='cuda:1'), in_proj_covar=tensor([0.0172, 0.0180, 0.0200, 0.0199, 0.0209, 0.0218, 0.0209, 0.0198], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-05-02 05:57:01,728 INFO [train.py:904] (1/8) Epoch 26, batch 7600, loss[loss=0.1817, simple_loss=0.2669, pruned_loss=0.04825, over 16974.00 frames. ], tot_loss[loss=0.2002, simple_loss=0.286, pruned_loss=0.05719, over 3040014.38 frames. ], batch size: 55, lr: 2.57e-03, grad_scale: 8.0 2023-05-02 05:57:07,522 INFO [optim.py:368] (1/8) 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 05:57:54,039 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=261386.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 05:57:54,076 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.6360, 3.5065, 3.9460, 1.9629, 4.1382, 4.1759, 3.0481, 3.0559], device='cuda:1'), covar=tensor([0.0813, 0.0297, 0.0235, 0.1284, 0.0084, 0.0167, 0.0436, 0.0485], device='cuda:1'), in_proj_covar=tensor([0.0148, 0.0110, 0.0100, 0.0139, 0.0086, 0.0129, 0.0130, 0.0130], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-05-02 05:58:06,252 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.39 vs. limit=2.0 2023-05-02 05:58:20,640 INFO [train.py:904] (1/8) Epoch 26, batch 7650, loss[loss=0.242, simple_loss=0.3032, pruned_loss=0.09041, over 11605.00 frames. ], tot_loss[loss=0.2007, simple_loss=0.2863, pruned_loss=0.05758, over 3043856.43 frames. ], batch size: 247, lr: 2.57e-03, grad_scale: 4.0 2023-05-02 05:58:26,643 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=261407.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 05:58:54,647 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.5615, 3.6119, 3.8175, 2.1378, 3.2725, 2.5992, 3.8922, 3.8475], device='cuda:1'), covar=tensor([0.0228, 0.0945, 0.0574, 0.2275, 0.0855, 0.0950, 0.0620, 0.1125], device='cuda:1'), in_proj_covar=tensor([0.0159, 0.0169, 0.0170, 0.0156, 0.0147, 0.0132, 0.0146, 0.0181], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0005], device='cuda:1') 2023-05-02 05:59:08,006 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=261434.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 05:59:36,566 INFO [train.py:904] (1/8) Epoch 26, batch 7700, loss[loss=0.1853, simple_loss=0.2741, pruned_loss=0.04829, over 16718.00 frames. ], tot_loss[loss=0.2013, simple_loss=0.2862, pruned_loss=0.05816, over 3052983.47 frames. ], batch size: 76, lr: 2.57e-03, grad_scale: 2.0 2023-05-02 05:59:40,110 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=261455.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 05:59:42,129 INFO [zipformer.py:625] (1/8) 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] (1/8) 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 06:00:36,570 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=261492.0, num_to_drop=1, layers_to_drop={2} 2023-05-02 06:00:52,204 INFO [train.py:904] (1/8) Epoch 26, batch 7750, loss[loss=0.251, simple_loss=0.316, pruned_loss=0.09302, over 11502.00 frames. ], tot_loss[loss=0.2008, simple_loss=0.2863, pruned_loss=0.05761, over 3054678.10 frames. ], batch size: 246, lr: 2.57e-03, grad_scale: 2.0 2023-05-02 06:01:13,807 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=261517.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 06:01:44,949 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=261537.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 06:01:53,480 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.90 vs. limit=2.0 2023-05-02 06:02:09,782 INFO [train.py:904] (1/8) Epoch 26, batch 7800, loss[loss=0.1947, simple_loss=0.2847, pruned_loss=0.05229, over 16421.00 frames. ], tot_loss[loss=0.2013, simple_loss=0.2868, pruned_loss=0.05784, over 3064288.06 frames. ], batch size: 146, lr: 2.57e-03, grad_scale: 2.0 2023-05-02 06:02:19,354 INFO [optim.py:368] (1/8) 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 06:03:00,429 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=261585.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 06:03:13,940 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.3941, 2.9836, 2.7465, 2.3033, 2.2733, 2.3413, 2.9703, 2.8957], device='cuda:1'), covar=tensor([0.2677, 0.0663, 0.1596, 0.2562, 0.2494, 0.2220, 0.0512, 0.1415], device='cuda:1'), in_proj_covar=tensor([0.0334, 0.0274, 0.0311, 0.0323, 0.0305, 0.0273, 0.0301, 0.0349], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:1') 2023-05-02 06:03:28,317 INFO [train.py:904] (1/8) Epoch 26, batch 7850, loss[loss=0.1862, simple_loss=0.2804, pruned_loss=0.04601, over 16708.00 frames. ], tot_loss[loss=0.2006, simple_loss=0.287, pruned_loss=0.05712, over 3065397.22 frames. ], batch size: 76, lr: 2.57e-03, grad_scale: 2.0 2023-05-02 06:04:25,986 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.86 vs. limit=2.0 2023-05-02 06:04:46,515 INFO [train.py:904] (1/8) Epoch 26, batch 7900, loss[loss=0.1927, simple_loss=0.2839, pruned_loss=0.05075, over 16881.00 frames. ], tot_loss[loss=0.1997, simple_loss=0.2866, pruned_loss=0.05645, over 3082706.88 frames. ], batch size: 116, lr: 2.57e-03, grad_scale: 2.0 2023-05-02 06:04:55,564 INFO [optim.py:368] (1/8) Clipping_scale=2.0, grad-norm quartiles 1.825e+02 2.587e+02 3.177e+02 3.708e+02 5.885e+02, threshold=6.353e+02, percent-clipped=0.0 2023-05-02 06:05:17,037 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([1.7249, 2.8023, 2.5704, 4.5670, 3.2501, 4.0555, 1.5747, 2.9312], device='cuda:1'), covar=tensor([0.1418, 0.0847, 0.1266, 0.0195, 0.0366, 0.0440, 0.1767, 0.0884], device='cuda:1'), in_proj_covar=tensor([0.0172, 0.0180, 0.0199, 0.0198, 0.0208, 0.0218, 0.0208, 0.0198], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-05-02 06:06:05,684 INFO [train.py:904] (1/8) Epoch 26, batch 7950, loss[loss=0.1939, simple_loss=0.2759, pruned_loss=0.0559, over 16729.00 frames. ], tot_loss[loss=0.2003, simple_loss=0.2871, pruned_loss=0.05678, over 3092564.38 frames. ], batch size: 124, lr: 2.57e-03, grad_scale: 2.0 2023-05-02 06:07:23,908 INFO [train.py:904] (1/8) Epoch 26, batch 8000, loss[loss=0.2208, simple_loss=0.291, pruned_loss=0.07524, over 11585.00 frames. ], tot_loss[loss=0.2015, simple_loss=0.2877, pruned_loss=0.05761, over 3090299.85 frames. ], batch size: 246, lr: 2.57e-03, grad_scale: 4.0 2023-05-02 06:07:32,670 INFO [optim.py:368] (1/8) Clipping_scale=2.0, grad-norm quartiles 1.840e+02 2.545e+02 3.130e+02 3.720e+02 6.505e+02, threshold=6.260e+02, percent-clipped=1.0 2023-05-02 06:08:23,748 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=261792.0, num_to_drop=1, layers_to_drop={0} 2023-05-02 06:08:40,304 INFO [train.py:904] (1/8) Epoch 26, batch 8050, loss[loss=0.1804, simple_loss=0.274, pruned_loss=0.0434, over 16519.00 frames. ], tot_loss[loss=0.2005, simple_loss=0.2872, pruned_loss=0.05694, over 3111452.26 frames. ], batch size: 75, lr: 2.57e-03, grad_scale: 4.0 2023-05-02 06:08:54,194 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=261812.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 06:09:13,006 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=261824.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 06:09:27,159 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=261833.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 06:09:34,923 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.8149, 1.9801, 2.4290, 3.0857, 2.1669, 2.1908, 2.2203, 2.0980], device='cuda:1'), covar=tensor([0.1718, 0.4267, 0.2681, 0.0856, 0.4804, 0.2929, 0.3670, 0.4169], device='cuda:1'), in_proj_covar=tensor([0.0414, 0.0466, 0.0378, 0.0332, 0.0443, 0.0532, 0.0437, 0.0545], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-02 06:09:37,846 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=261840.0, num_to_drop=1, layers_to_drop={1} 2023-05-02 06:09:57,224 INFO [train.py:904] (1/8) Epoch 26, batch 8100, loss[loss=0.1946, simple_loss=0.2848, pruned_loss=0.05216, over 16861.00 frames. ], tot_loss[loss=0.1989, simple_loss=0.2862, pruned_loss=0.05578, over 3107300.02 frames. ], batch size: 116, lr: 2.57e-03, grad_scale: 4.0 2023-05-02 06:10:06,903 INFO [optim.py:368] (1/8) 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 06:10:46,767 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=261885.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 06:11:00,394 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 26, batch 8150, loss[loss=0.1622, simple_loss=0.2527, pruned_loss=0.03586, over 16910.00 frames. ], tot_loss[loss=0.1975, simple_loss=0.2844, pruned_loss=0.05525, over 3090977.42 frames. ], batch size: 96, lr: 2.57e-03, grad_scale: 4.0 2023-05-02 06:12:33,065 INFO [train.py:904] (1/8) Epoch 26, batch 8200, loss[loss=0.207, simple_loss=0.2832, pruned_loss=0.06539, over 11689.00 frames. ], tot_loss[loss=0.1964, simple_loss=0.2824, pruned_loss=0.05524, over 3071836.69 frames. ], batch size: 246, lr: 2.57e-03, grad_scale: 4.0 2023-05-02 06:12:43,220 INFO [optim.py:368] (1/8) 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 06:13:12,991 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=261977.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 06:13:58,694 INFO [train.py:904] (1/8) Epoch 26, batch 8250, loss[loss=0.1637, simple_loss=0.2559, pruned_loss=0.03571, over 11965.00 frames. ], tot_loss[loss=0.1932, simple_loss=0.2812, pruned_loss=0.05265, over 3060243.80 frames. ], batch size: 248, lr: 2.57e-03, grad_scale: 4.0 2023-05-02 06:14:08,373 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.6783, 3.8402, 2.8715, 2.2351, 2.3228, 2.3884, 4.0362, 3.2484], device='cuda:1'), covar=tensor([0.2837, 0.0512, 0.1779, 0.3010, 0.3090, 0.2247, 0.0361, 0.1368], device='cuda:1'), in_proj_covar=tensor([0.0333, 0.0273, 0.0309, 0.0322, 0.0304, 0.0272, 0.0301, 0.0347], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:1') 2023-05-02 06:14:21,415 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.4141, 4.4102, 4.2283, 3.4739, 4.3277, 1.6361, 4.0839, 3.9716], device='cuda:1'), covar=tensor([0.0131, 0.0123, 0.0229, 0.0387, 0.0128, 0.3209, 0.0161, 0.0303], device='cuda:1'), in_proj_covar=tensor([0.0174, 0.0168, 0.0206, 0.0181, 0.0182, 0.0212, 0.0194, 0.0174], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-02 06:14:57,010 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 26, batch 8300, loss[loss=0.1805, simple_loss=0.2743, pruned_loss=0.04334, over 16621.00 frames. ], tot_loss[loss=0.1894, simple_loss=0.2789, pruned_loss=0.04995, over 3068582.42 frames. ], batch size: 57, lr: 2.57e-03, grad_scale: 4.0 2023-05-02 06:15:30,267 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.9522, 5.0197, 4.7754, 4.3861, 4.4583, 4.8893, 4.7520, 4.5517], device='cuda:1'), covar=tensor([0.0619, 0.0678, 0.0389, 0.0381, 0.1111, 0.0607, 0.0415, 0.0859], device='cuda:1'), in_proj_covar=tensor([0.0302, 0.0453, 0.0352, 0.0354, 0.0350, 0.0407, 0.0242, 0.0425], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002], device='cuda:1') 2023-05-02 06:15:32,830 INFO [optim.py:368] (1/8) Clipping_scale=2.0, grad-norm quartiles 1.532e+02 2.264e+02 2.806e+02 3.280e+02 5.175e+02, threshold=5.611e+02, percent-clipped=0.0 2023-05-02 06:15:33,607 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.21 vs. limit=2.0 2023-05-02 06:16:44,232 INFO [train.py:904] (1/8) Epoch 26, batch 8350, loss[loss=0.1944, simple_loss=0.2901, pruned_loss=0.04937, over 15236.00 frames. ], tot_loss[loss=0.1882, simple_loss=0.279, pruned_loss=0.0487, over 3062895.94 frames. ], batch size: 191, lr: 2.57e-03, grad_scale: 4.0 2023-05-02 06:16:53,226 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.85 vs. limit=2.0 2023-05-02 06:16:59,060 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=262112.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 06:17:11,487 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.90 vs. limit=2.0 2023-05-02 06:17:14,239 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.6585, 2.6355, 1.9262, 2.7913, 2.1863, 2.7849, 2.1952, 2.4270], device='cuda:1'), covar=tensor([0.0323, 0.0360, 0.1166, 0.0394, 0.0643, 0.0522, 0.1255, 0.0643], device='cuda:1'), in_proj_covar=tensor([0.0173, 0.0178, 0.0195, 0.0168, 0.0177, 0.0216, 0.0204, 0.0182], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-05-02 06:18:04,040 INFO [train.py:904] (1/8) Epoch 26, batch 8400, loss[loss=0.1739, simple_loss=0.2714, pruned_loss=0.03819, over 16728.00 frames. ], tot_loss[loss=0.1847, simple_loss=0.2762, pruned_loss=0.04663, over 3052354.15 frames. ], batch size: 124, lr: 2.57e-03, grad_scale: 8.0 2023-05-02 06:18:09,464 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.4305, 3.1086, 2.6312, 2.2743, 2.2203, 2.1824, 3.0227, 2.8567], device='cuda:1'), covar=tensor([0.2792, 0.0881, 0.1895, 0.3210, 0.3133, 0.2786, 0.0635, 0.1569], device='cuda:1'), in_proj_covar=tensor([0.0329, 0.0271, 0.0307, 0.0320, 0.0301, 0.0270, 0.0299, 0.0344], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:1') 2023-05-02 06:18:13,164 INFO [optim.py:368] (1/8) Clipping_scale=2.0, grad-norm quartiles 1.602e+02 2.393e+02 2.671e+02 3.015e+02 4.313e+02, threshold=5.342e+02, percent-clipped=0.0 2023-05-02 06:18:14,934 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=262160.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 06:18:47,965 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=262180.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 06:19:02,232 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=262189.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 06:19:24,256 INFO [train.py:904] (1/8) Epoch 26, batch 8450, loss[loss=0.1627, simple_loss=0.2601, pruned_loss=0.03261, over 16245.00 frames. ], tot_loss[loss=0.1819, simple_loss=0.2743, pruned_loss=0.04474, over 3050989.72 frames. ], batch size: 165, lr: 2.57e-03, grad_scale: 8.0 2023-05-02 06:20:48,803 INFO [train.py:904] (1/8) Epoch 26, batch 8500, loss[loss=0.155, simple_loss=0.2473, pruned_loss=0.03133, over 15383.00 frames. ], tot_loss[loss=0.1782, simple_loss=0.2706, pruned_loss=0.04287, over 3038172.22 frames. ], batch size: 191, lr: 2.57e-03, grad_scale: 8.0 2023-05-02 06:20:49,710 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.50 vs. limit=2.0 2023-05-02 06:20:57,922 INFO [optim.py:368] (1/8) 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 06:21:09,765 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.3462, 4.6701, 4.4926, 4.5056, 4.1812, 4.2053, 4.1839, 4.7371], device='cuda:1'), covar=tensor([0.1372, 0.1037, 0.1048, 0.0904, 0.0896, 0.1542, 0.1289, 0.0997], device='cuda:1'), in_proj_covar=tensor([0.0688, 0.0827, 0.0680, 0.0635, 0.0527, 0.0527, 0.0696, 0.0645], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:1') 2023-05-02 06:22:09,825 INFO [train.py:904] (1/8) Epoch 26, batch 8550, loss[loss=0.16, simple_loss=0.2619, pruned_loss=0.02898, over 16856.00 frames. ], tot_loss[loss=0.1761, simple_loss=0.2683, pruned_loss=0.04199, over 3012025.77 frames. ], batch size: 102, lr: 2.57e-03, grad_scale: 8.0 2023-05-02 06:23:07,464 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=262333.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 06:23:47,964 INFO [train.py:904] (1/8) Epoch 26, batch 8600, loss[loss=0.1854, simple_loss=0.2858, pruned_loss=0.04245, over 16167.00 frames. ], tot_loss[loss=0.1754, simple_loss=0.2688, pruned_loss=0.04106, over 3011514.79 frames. ], batch size: 165, lr: 2.57e-03, grad_scale: 8.0 2023-05-02 06:23:59,713 INFO [optim.py:368] (1/8) Clipping_scale=2.0, grad-norm quartiles 1.426e+02 2.293e+02 2.800e+02 3.467e+02 6.859e+02, threshold=5.600e+02, percent-clipped=2.0 2023-05-02 06:24:34,582 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.3924, 3.5612, 3.6425, 2.3796, 3.3569, 3.6555, 3.4034, 2.2374], device='cuda:1'), covar=tensor([0.0498, 0.0063, 0.0051, 0.0443, 0.0112, 0.0086, 0.0080, 0.0455], device='cuda:1'), in_proj_covar=tensor([0.0135, 0.0086, 0.0087, 0.0132, 0.0099, 0.0111, 0.0096, 0.0128], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:1') 2023-05-02 06:25:27,327 INFO [train.py:904] (1/8) Epoch 26, batch 8650, loss[loss=0.1514, simple_loss=0.255, pruned_loss=0.02388, over 15290.00 frames. ], tot_loss[loss=0.1731, simple_loss=0.2671, pruned_loss=0.03959, over 3023883.02 frames. ], batch size: 192, lr: 2.57e-03, grad_scale: 8.0 2023-05-02 06:25:32,866 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.4199, 3.3378, 3.4963, 3.5518, 3.5837, 3.3114, 3.5578, 3.6323], device='cuda:1'), covar=tensor([0.1366, 0.1005, 0.1033, 0.0651, 0.0642, 0.2329, 0.0832, 0.0762], device='cuda:1'), in_proj_covar=tensor([0.0636, 0.0788, 0.0904, 0.0796, 0.0608, 0.0632, 0.0659, 0.0775], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-02 06:27:13,544 INFO [train.py:904] (1/8) Epoch 26, batch 8700, loss[loss=0.1547, simple_loss=0.2555, pruned_loss=0.02699, over 16306.00 frames. ], tot_loss[loss=0.17, simple_loss=0.2638, pruned_loss=0.03805, over 3033782.24 frames. ], batch size: 165, lr: 2.57e-03, grad_scale: 8.0 2023-05-02 06:27:25,316 INFO [optim.py:368] (1/8) 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 06:27:40,645 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([1.9425, 2.2435, 2.3139, 2.9965, 1.8161, 3.2707, 1.7348, 2.8080], device='cuda:1'), covar=tensor([0.1279, 0.0714, 0.1079, 0.0157, 0.0086, 0.0335, 0.1573, 0.0650], device='cuda:1'), in_proj_covar=tensor([0.0170, 0.0177, 0.0196, 0.0194, 0.0203, 0.0214, 0.0205, 0.0194], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-05-02 06:28:02,758 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=262480.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 06:28:20,324 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 26, batch 8750, loss[loss=0.1818, simple_loss=0.2825, pruned_loss=0.04055, over 16794.00 frames. ], tot_loss[loss=0.1692, simple_loss=0.2634, pruned_loss=0.0375, over 3048338.19 frames. ], batch size: 124, lr: 2.57e-03, grad_scale: 8.0 2023-05-02 06:29:14,090 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.7347, 1.8622, 2.3566, 2.6834, 2.5957, 3.0967, 2.2078, 3.0903], device='cuda:1'), covar=tensor([0.0299, 0.0646, 0.0421, 0.0411, 0.0373, 0.0208, 0.0581, 0.0210], device='cuda:1'), in_proj_covar=tensor([0.0191, 0.0193, 0.0181, 0.0185, 0.0200, 0.0159, 0.0197, 0.0158], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-02 06:29:48,637 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=262528.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 06:30:09,754 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=262537.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 06:30:31,877 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.61 vs. limit=2.0 2023-05-02 06:30:41,532 INFO [train.py:904] (1/8) Epoch 26, batch 8800, loss[loss=0.1889, simple_loss=0.2742, pruned_loss=0.05178, over 12398.00 frames. ], tot_loss[loss=0.168, simple_loss=0.2623, pruned_loss=0.0368, over 3054419.77 frames. ], batch size: 250, lr: 2.57e-03, grad_scale: 8.0 2023-05-02 06:30:52,411 INFO [optim.py:368] (1/8) Clipping_scale=2.0, grad-norm quartiles 1.510e+02 2.093e+02 2.466e+02 2.812e+02 4.660e+02, threshold=4.932e+02, percent-clipped=0.0 2023-05-02 06:32:25,986 INFO [train.py:904] (1/8) Epoch 26, batch 8850, loss[loss=0.1544, simple_loss=0.2687, pruned_loss=0.02005, over 16856.00 frames. ], tot_loss[loss=0.1687, simple_loss=0.2649, pruned_loss=0.03624, over 3060580.47 frames. ], batch size: 90, lr: 2.57e-03, grad_scale: 8.0 2023-05-02 06:33:32,243 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=262633.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 06:33:52,450 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=4.82 vs. limit=5.0 2023-05-02 06:34:13,867 INFO [train.py:904] (1/8) Epoch 26, batch 8900, loss[loss=0.1781, simple_loss=0.2737, pruned_loss=0.04126, over 16746.00 frames. ], tot_loss[loss=0.1678, simple_loss=0.2646, pruned_loss=0.03552, over 3045647.95 frames. ], batch size: 134, lr: 2.57e-03, grad_scale: 8.0 2023-05-02 06:34:26,822 INFO [optim.py:368] (1/8) 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 06:35:19,859 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=262681.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 06:35:23,795 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.9024, 3.2057, 3.4243, 1.9722, 2.8913, 2.1391, 3.3576, 3.3856], device='cuda:1'), covar=tensor([0.0255, 0.0805, 0.0518, 0.2115, 0.0798, 0.1042, 0.0620, 0.0875], device='cuda:1'), in_proj_covar=tensor([0.0156, 0.0164, 0.0165, 0.0152, 0.0143, 0.0129, 0.0142, 0.0175], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:1') 2023-05-02 06:36:18,586 INFO [train.py:904] (1/8) Epoch 26, batch 8950, loss[loss=0.1458, simple_loss=0.2411, pruned_loss=0.02525, over 15185.00 frames. ], tot_loss[loss=0.1684, simple_loss=0.2645, pruned_loss=0.03615, over 3051825.20 frames. ], batch size: 190, lr: 2.57e-03, grad_scale: 8.0 2023-05-02 06:36:50,268 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([1.9413, 1.8207, 1.6390, 1.4491, 1.9835, 1.6271, 1.4900, 1.9099], device='cuda:1'), covar=tensor([0.0214, 0.0387, 0.0512, 0.0438, 0.0301, 0.0335, 0.0187, 0.0277], device='cuda:1'), in_proj_covar=tensor([0.0216, 0.0235, 0.0227, 0.0226, 0.0235, 0.0234, 0.0232, 0.0232], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-02 06:38:00,236 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 26, batch 9000, loss[loss=0.1465, simple_loss=0.2345, pruned_loss=0.02923, over 12271.00 frames. ], tot_loss[loss=0.1657, simple_loss=0.2615, pruned_loss=0.03501, over 3059974.38 frames. ], batch size: 250, lr: 2.57e-03, grad_scale: 8.0 2023-05-02 06:38:08,094 INFO [train.py:929] (1/8) Computing validation loss 2023-05-02 06:38:18,516 INFO [train.py:938] (1/8) Epoch 26, validation: loss=0.1435, simple_loss=0.2475, pruned_loss=0.01976, over 944034.00 frames. 2023-05-02 06:38:18,517 INFO [train.py:939] (1/8) Maximum memory allocated so far is 18145MB 2023-05-02 06:38:30,744 INFO [optim.py:368] (1/8) Clipping_scale=2.0, grad-norm quartiles 1.305e+02 2.045e+02 2.519e+02 3.077e+02 5.140e+02, threshold=5.037e+02, percent-clipped=0.0 2023-05-02 06:39:09,498 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.27 vs. limit=2.0 2023-05-02 06:39:40,116 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.0076, 1.9735, 2.1555, 3.5105, 1.9161, 2.2005, 2.1022, 2.0893], device='cuda:1'), covar=tensor([0.1579, 0.4518, 0.3435, 0.0722, 0.5541, 0.3255, 0.4165, 0.4399], device='cuda:1'), in_proj_covar=tensor([0.0409, 0.0460, 0.0377, 0.0325, 0.0438, 0.0524, 0.0432, 0.0536], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-02 06:40:02,623 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.8213, 2.2641, 1.9672, 2.1618, 2.6100, 2.3640, 2.2875, 2.7474], device='cuda:1'), covar=tensor([0.0183, 0.0517, 0.0618, 0.0534, 0.0319, 0.0447, 0.0226, 0.0308], device='cuda:1'), in_proj_covar=tensor([0.0216, 0.0235, 0.0227, 0.0226, 0.0236, 0.0235, 0.0232, 0.0232], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-02 06:40:03,316 INFO [train.py:904] (1/8) Epoch 26, batch 9050, loss[loss=0.1512, simple_loss=0.2385, pruned_loss=0.03191, over 16722.00 frames. ], tot_loss[loss=0.1666, simple_loss=0.2623, pruned_loss=0.03544, over 3063570.98 frames. ], batch size: 134, lr: 2.57e-03, grad_scale: 8.0 2023-05-02 06:40:20,491 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=262810.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 06:40:21,889 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.9317, 2.7034, 2.8922, 2.1292, 2.7237, 2.1727, 2.7327, 2.9138], device='cuda:1'), covar=tensor([0.0333, 0.0834, 0.0547, 0.1937, 0.0760, 0.1001, 0.0721, 0.0849], device='cuda:1'), in_proj_covar=tensor([0.0156, 0.0164, 0.0165, 0.0152, 0.0143, 0.0129, 0.0142, 0.0174], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:1') 2023-05-02 06:41:33,698 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.9646, 2.1843, 2.3824, 3.2291, 2.1935, 2.3956, 2.3201, 2.2949], device='cuda:1'), covar=tensor([0.1344, 0.3704, 0.2949, 0.0789, 0.4523, 0.2617, 0.3609, 0.3669], device='cuda:1'), in_proj_covar=tensor([0.0409, 0.0459, 0.0376, 0.0325, 0.0438, 0.0523, 0.0431, 0.0535], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-02 06:41:48,709 INFO [train.py:904] (1/8) Epoch 26, batch 9100, loss[loss=0.1674, simple_loss=0.2661, pruned_loss=0.03436, over 16866.00 frames. ], tot_loss[loss=0.1663, simple_loss=0.2614, pruned_loss=0.03554, over 3056453.92 frames. ], batch size: 96, lr: 2.57e-03, grad_scale: 8.0 2023-05-02 06:42:01,224 INFO [optim.py:368] (1/8) Clipping_scale=2.0, grad-norm quartiles 1.532e+02 2.262e+02 2.654e+02 3.227e+02 6.240e+02, threshold=5.309e+02, percent-clipped=4.0 2023-05-02 06:43:45,862 INFO [train.py:904] (1/8) Epoch 26, batch 9150, loss[loss=0.1442, simple_loss=0.233, pruned_loss=0.02769, over 16624.00 frames. ], tot_loss[loss=0.166, simple_loss=0.2619, pruned_loss=0.03502, over 3071351.33 frames. ], batch size: 57, lr: 2.57e-03, grad_scale: 8.0 2023-05-02 06:43:46,561 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.2816, 2.3136, 2.3609, 4.0279, 2.2462, 2.6178, 2.3650, 2.4546], device='cuda:1'), covar=tensor([0.1369, 0.3747, 0.3279, 0.0545, 0.4386, 0.2811, 0.3825, 0.3824], device='cuda:1'), in_proj_covar=tensor([0.0408, 0.0459, 0.0376, 0.0325, 0.0437, 0.0523, 0.0431, 0.0535], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-02 06:43:52,207 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.23 vs. limit=2.0 2023-05-02 06:44:23,935 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=2.99 vs. limit=5.0 2023-05-02 06:44:41,515 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.35 vs. limit=2.0 2023-05-02 06:45:01,323 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.2130, 2.2554, 2.3176, 3.8562, 2.2184, 2.5614, 2.3723, 2.3670], device='cuda:1'), covar=tensor([0.1414, 0.3763, 0.3327, 0.0598, 0.4415, 0.2905, 0.3745, 0.4049], device='cuda:1'), in_proj_covar=tensor([0.0408, 0.0458, 0.0375, 0.0324, 0.0436, 0.0522, 0.0430, 0.0534], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-02 06:45:27,072 INFO [train.py:904] (1/8) Epoch 26, batch 9200, loss[loss=0.1671, simple_loss=0.263, pruned_loss=0.03559, over 15405.00 frames. ], tot_loss[loss=0.1633, simple_loss=0.2579, pruned_loss=0.03438, over 3072092.68 frames. ], batch size: 191, lr: 2.57e-03, grad_scale: 8.0 2023-05-02 06:45:36,599 INFO [optim.py:368] (1/8) Clipping_scale=2.0, grad-norm quartiles 1.418e+02 2.229e+02 2.531e+02 3.050e+02 5.061e+02, threshold=5.062e+02, percent-clipped=0.0 2023-05-02 06:47:01,941 INFO [train.py:904] (1/8) Epoch 26, batch 9250, loss[loss=0.1679, simple_loss=0.2652, pruned_loss=0.0353, over 16387.00 frames. ], tot_loss[loss=0.1635, simple_loss=0.2578, pruned_loss=0.03458, over 3061689.38 frames. ], batch size: 146, lr: 2.57e-03, grad_scale: 8.0 2023-05-02 06:47:41,400 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.1291, 4.1013, 4.4566, 4.4256, 4.4342, 4.1808, 4.1488, 4.1815], device='cuda:1'), covar=tensor([0.0392, 0.0715, 0.0498, 0.0544, 0.0593, 0.0582, 0.0985, 0.0519], device='cuda:1'), in_proj_covar=tensor([0.0411, 0.0466, 0.0452, 0.0415, 0.0500, 0.0476, 0.0547, 0.0380], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:1') 2023-05-02 06:48:37,020 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.4543, 4.5701, 4.3755, 4.0671, 4.0982, 4.4665, 4.2127, 4.1897], device='cuda:1'), covar=tensor([0.0586, 0.0648, 0.0296, 0.0287, 0.0667, 0.0642, 0.0527, 0.0615], device='cuda:1'), in_proj_covar=tensor([0.0296, 0.0439, 0.0343, 0.0345, 0.0341, 0.0397, 0.0236, 0.0413], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-02 06:48:43,242 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.4176, 3.3649, 3.4701, 3.5515, 3.5962, 3.3688, 3.5574, 3.6450], device='cuda:1'), covar=tensor([0.1401, 0.1041, 0.1063, 0.0671, 0.0675, 0.1965, 0.0955, 0.0806], device='cuda:1'), in_proj_covar=tensor([0.0631, 0.0780, 0.0894, 0.0789, 0.0603, 0.0625, 0.0654, 0.0765], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-02 06:48:49,716 INFO [train.py:904] (1/8) Epoch 26, batch 9300, loss[loss=0.1427, simple_loss=0.2394, pruned_loss=0.02295, over 16732.00 frames. ], tot_loss[loss=0.1628, simple_loss=0.2569, pruned_loss=0.03437, over 3054628.61 frames. ], batch size: 83, lr: 2.57e-03, grad_scale: 8.0 2023-05-02 06:49:02,021 INFO [optim.py:368] (1/8) 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 06:49:40,788 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=2.01 vs. limit=2.0 2023-05-02 06:50:20,340 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=263095.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 06:50:33,213 INFO [train.py:904] (1/8) Epoch 26, batch 9350, loss[loss=0.1506, simple_loss=0.2401, pruned_loss=0.03061, over 12313.00 frames. ], tot_loss[loss=0.1626, simple_loss=0.2568, pruned_loss=0.03418, over 3068437.57 frames. ], batch size: 250, lr: 2.57e-03, grad_scale: 8.0 2023-05-02 06:50:38,183 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=263105.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 06:50:51,108 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=263111.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 06:51:01,703 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.5949, 4.5214, 4.3627, 3.6635, 4.4510, 1.8165, 4.1515, 4.1552], device='cuda:1'), covar=tensor([0.0105, 0.0118, 0.0204, 0.0296, 0.0099, 0.2717, 0.0152, 0.0247], device='cuda:1'), in_proj_covar=tensor([0.0171, 0.0165, 0.0201, 0.0176, 0.0179, 0.0210, 0.0190, 0.0170], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-02 06:51:48,094 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=263140.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 06:51:53,903 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.1424, 3.2025, 2.0082, 3.5134, 2.4258, 3.4409, 2.1272, 2.6297], device='cuda:1'), covar=tensor([0.0356, 0.0425, 0.1695, 0.0264, 0.0910, 0.0710, 0.1682, 0.0883], device='cuda:1'), in_proj_covar=tensor([0.0168, 0.0172, 0.0188, 0.0162, 0.0173, 0.0209, 0.0198, 0.0177], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0003, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-05-02 06:52:14,167 INFO [train.py:904] (1/8) Epoch 26, batch 9400, loss[loss=0.1739, simple_loss=0.2787, pruned_loss=0.03454, over 16931.00 frames. ], tot_loss[loss=0.1622, simple_loss=0.2568, pruned_loss=0.03378, over 3063227.02 frames. ], batch size: 116, lr: 2.57e-03, grad_scale: 8.0 2023-05-02 06:52:19,901 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=263156.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 06:52:25,030 INFO [optim.py:368] (1/8) 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 06:52:52,035 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=263172.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 06:53:16,117 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.4842, 3.3573, 2.6957, 2.1843, 2.1243, 2.2987, 3.4835, 2.9129], device='cuda:1'), covar=tensor([0.2911, 0.0702, 0.1808, 0.3073, 0.3302, 0.2346, 0.0406, 0.1568], device='cuda:1'), in_proj_covar=tensor([0.0326, 0.0268, 0.0304, 0.0316, 0.0294, 0.0267, 0.0296, 0.0339], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:1') 2023-05-02 06:53:50,962 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=263201.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 06:53:53,489 INFO [train.py:904] (1/8) Epoch 26, batch 9450, loss[loss=0.1602, simple_loss=0.2579, pruned_loss=0.03122, over 15460.00 frames. ], tot_loss[loss=0.1634, simple_loss=0.2588, pruned_loss=0.03401, over 3057887.76 frames. ], batch size: 191, lr: 2.57e-03, grad_scale: 8.0 2023-05-02 06:54:14,254 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=263214.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 06:55:19,916 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 26, batch 9500, loss[loss=0.1567, simple_loss=0.2557, pruned_loss=0.02888, over 16623.00 frames. ], tot_loss[loss=0.1631, simple_loss=0.2582, pruned_loss=0.03395, over 3066300.78 frames. ], batch size: 62, lr: 2.57e-03, grad_scale: 8.0 2023-05-02 06:55:47,501 INFO [optim.py:368] (1/8) 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 06:55:52,296 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.4895, 2.0242, 1.7230, 1.6753, 2.2888, 1.9360, 1.8442, 2.3832], device='cuda:1'), covar=tensor([0.0253, 0.0517, 0.0726, 0.0640, 0.0378, 0.0454, 0.0262, 0.0336], device='cuda:1'), in_proj_covar=tensor([0.0215, 0.0233, 0.0225, 0.0225, 0.0235, 0.0233, 0.0230, 0.0230], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-02 06:56:12,532 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.24 vs. limit=2.0 2023-05-02 06:56:19,469 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=263275.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 06:57:17,614 INFO [train.py:904] (1/8) Epoch 26, batch 9550, loss[loss=0.1775, simple_loss=0.2595, pruned_loss=0.04771, over 12701.00 frames. ], tot_loss[loss=0.1633, simple_loss=0.2582, pruned_loss=0.03415, over 3060123.53 frames. ], batch size: 246, lr: 2.57e-03, grad_scale: 8.0 2023-05-02 06:57:27,285 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=263307.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 06:58:58,572 INFO [train.py:904] (1/8) Epoch 26, batch 9600, loss[loss=0.1572, simple_loss=0.2623, pruned_loss=0.02608, over 16866.00 frames. ], tot_loss[loss=0.1643, simple_loss=0.259, pruned_loss=0.03479, over 3054809.81 frames. ], batch size: 90, lr: 2.57e-03, grad_scale: 8.0 2023-05-02 06:59:09,908 INFO [optim.py:368] (1/8) 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 06:59:37,508 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.9677, 2.0683, 2.1868, 3.4680, 2.0218, 2.3513, 2.1976, 2.2174], device='cuda:1'), covar=tensor([0.1445, 0.3951, 0.3369, 0.0718, 0.4756, 0.2830, 0.3993, 0.3785], device='cuda:1'), in_proj_covar=tensor([0.0405, 0.0457, 0.0375, 0.0323, 0.0435, 0.0520, 0.0429, 0.0532], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-02 07:00:21,658 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.37 vs. limit=2.0 2023-05-02 07:00:45,649 INFO [train.py:904] (1/8) Epoch 26, batch 9650, loss[loss=0.1534, simple_loss=0.2543, pruned_loss=0.02629, over 16894.00 frames. ], tot_loss[loss=0.1656, simple_loss=0.2606, pruned_loss=0.03523, over 3048125.33 frames. ], batch size: 102, lr: 2.56e-03, grad_scale: 8.0 2023-05-02 07:00:51,808 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=263405.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 07:02:12,676 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.69 vs. limit=2.0 2023-05-02 07:02:28,447 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=263451.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 07:02:31,834 INFO [train.py:904] (1/8) Epoch 26, batch 9700, loss[loss=0.1682, simple_loss=0.2534, pruned_loss=0.0415, over 12190.00 frames. ], tot_loss[loss=0.1651, simple_loss=0.2598, pruned_loss=0.03526, over 3044049.43 frames. ], batch size: 250, lr: 2.56e-03, grad_scale: 16.0 2023-05-02 07:02:33,070 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=263453.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 07:02:42,005 INFO [optim.py:368] (1/8) 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 07:02:58,834 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=263467.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 07:04:01,224 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=263496.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 07:04:13,669 INFO [train.py:904] (1/8) Epoch 26, batch 9750, loss[loss=0.1599, simple_loss=0.2447, pruned_loss=0.03755, over 12307.00 frames. ], tot_loss[loss=0.1646, simple_loss=0.259, pruned_loss=0.0351, over 3066029.98 frames. ], batch size: 246, lr: 2.56e-03, grad_scale: 16.0 2023-05-02 07:04:32,628 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.0107, 2.1385, 2.1115, 3.5101, 2.1128, 2.3935, 2.2354, 2.2755], device='cuda:1'), covar=tensor([0.1389, 0.3830, 0.3428, 0.0654, 0.4388, 0.2811, 0.3826, 0.3672], device='cuda:1'), in_proj_covar=tensor([0.0407, 0.0458, 0.0376, 0.0325, 0.0437, 0.0522, 0.0431, 0.0534], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-02 07:05:51,431 INFO [train.py:904] (1/8) Epoch 26, batch 9800, loss[loss=0.164, simple_loss=0.2558, pruned_loss=0.03614, over 12309.00 frames. ], tot_loss[loss=0.1634, simple_loss=0.2588, pruned_loss=0.03399, over 3066899.36 frames. ], batch size: 248, lr: 2.56e-03, grad_scale: 16.0 2023-05-02 07:06:03,262 INFO [optim.py:368] (1/8) Clipping_scale=2.0, grad-norm quartiles 1.369e+02 2.162e+02 2.515e+02 2.948e+02 4.356e+02, threshold=5.031e+02, percent-clipped=0.0 2023-05-02 07:06:23,987 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=263570.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 07:07:14,560 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.73 vs. limit=2.0 2023-05-02 07:07:32,697 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 26, batch 9850, loss[loss=0.1512, simple_loss=0.2523, pruned_loss=0.02502, over 17277.00 frames. ], tot_loss[loss=0.1639, simple_loss=0.2601, pruned_loss=0.0338, over 3078654.78 frames. ], batch size: 52, lr: 2.56e-03, grad_scale: 16.0 2023-05-02 07:09:24,088 INFO [train.py:904] (1/8) Epoch 26, batch 9900, loss[loss=0.1644, simple_loss=0.2687, pruned_loss=0.03008, over 16363.00 frames. ], tot_loss[loss=0.1635, simple_loss=0.2596, pruned_loss=0.0337, over 3046115.06 frames. ], batch size: 146, lr: 2.56e-03, grad_scale: 16.0 2023-05-02 07:09:36,802 INFO [optim.py:368] (1/8) 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 07:11:20,745 INFO [train.py:904] (1/8) Epoch 26, batch 9950, loss[loss=0.1699, simple_loss=0.2683, pruned_loss=0.03578, over 16806.00 frames. ], tot_loss[loss=0.1651, simple_loss=0.2622, pruned_loss=0.03396, over 3076046.73 frames. ], batch size: 124, lr: 2.56e-03, grad_scale: 16.0 2023-05-02 07:13:18,987 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=263751.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 07:13:21,141 INFO [train.py:904] (1/8) Epoch 26, batch 10000, loss[loss=0.1617, simple_loss=0.2619, pruned_loss=0.03073, over 16823.00 frames. ], tot_loss[loss=0.1635, simple_loss=0.2604, pruned_loss=0.0333, over 3079478.27 frames. ], batch size: 124, lr: 2.56e-03, grad_scale: 16.0 2023-05-02 07:13:34,912 INFO [optim.py:368] (1/8) 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 07:13:40,238 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=2.59 vs. limit=5.0 2023-05-02 07:13:51,364 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=263767.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 07:14:46,345 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.3918, 3.4681, 1.9584, 3.9130, 2.5510, 3.7945, 2.1136, 2.8096], device='cuda:1'), covar=tensor([0.0327, 0.0398, 0.1899, 0.0231, 0.0939, 0.0593, 0.1734, 0.0827], device='cuda:1'), in_proj_covar=tensor([0.0168, 0.0172, 0.0188, 0.0161, 0.0172, 0.0208, 0.0198, 0.0177], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0003, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-05-02 07:14:48,323 INFO [zipformer.py:625] (1/8) 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] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=263799.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 07:15:00,039 INFO [train.py:904] (1/8) Epoch 26, batch 10050, loss[loss=0.159, simple_loss=0.2548, pruned_loss=0.03159, over 16780.00 frames. ], tot_loss[loss=0.1642, simple_loss=0.261, pruned_loss=0.03368, over 3081191.90 frames. ], batch size: 83, lr: 2.56e-03, grad_scale: 16.0 2023-05-02 07:15:06,570 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.73 vs. limit=2.0 2023-05-02 07:15:23,649 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=263815.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 07:15:57,724 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=263833.0, num_to_drop=1, layers_to_drop={0} 2023-05-02 07:16:17,279 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=263844.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 07:16:30,296 INFO [train.py:904] (1/8) Epoch 26, batch 10100, loss[loss=0.1563, simple_loss=0.2495, pruned_loss=0.03152, over 16757.00 frames. ], tot_loss[loss=0.164, simple_loss=0.2608, pruned_loss=0.03361, over 3085964.74 frames. ], batch size: 76, lr: 2.56e-03, grad_scale: 16.0 2023-05-02 07:16:39,590 INFO [optim.py:368] (1/8) 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 07:17:02,901 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=263870.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 07:17:12,460 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.2376, 4.3457, 4.4831, 4.2517, 4.3860, 4.8562, 4.4108, 4.0932], device='cuda:1'), covar=tensor([0.1823, 0.1921, 0.2020, 0.2110, 0.2329, 0.0979, 0.1599, 0.2493], device='cuda:1'), in_proj_covar=tensor([0.0398, 0.0590, 0.0654, 0.0482, 0.0639, 0.0680, 0.0510, 0.0643], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-05-02 07:17:36,319 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=263894.0, num_to_drop=1, layers_to_drop={3} 2023-05-02 07:18:09,004 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=263902.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 07:18:09,864 INFO [train.py:904] (1/8) Epoch 27, batch 0, loss[loss=0.1615, simple_loss=0.2464, pruned_loss=0.0383, over 15919.00 frames. ], tot_loss[loss=0.1615, simple_loss=0.2464, pruned_loss=0.0383, over 15919.00 frames. ], batch size: 35, lr: 2.51e-03, grad_scale: 16.0 2023-05-02 07:18:09,864 INFO [train.py:929] (1/8) Computing validation loss 2023-05-02 07:18:17,113 INFO [train.py:938] (1/8) Epoch 27, validation: loss=0.1434, simple_loss=0.2467, pruned_loss=0.02006, over 944034.00 frames. 2023-05-02 07:18:17,113 INFO [train.py:939] (1/8) Maximum memory allocated so far is 18145MB 2023-05-02 07:18:30,564 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([5.0529, 4.7808, 5.0460, 5.2195, 5.4045, 4.6991, 5.3632, 5.3745], device='cuda:1'), covar=tensor([0.1996, 0.1452, 0.1809, 0.0872, 0.0636, 0.0940, 0.0637, 0.0738], device='cuda:1'), in_proj_covar=tensor([0.0631, 0.0778, 0.0891, 0.0788, 0.0600, 0.0623, 0.0654, 0.0760], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-02 07:18:37,957 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=263918.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 07:18:49,605 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([5.0138, 5.0518, 5.3864, 5.3770, 5.4313, 5.1153, 5.0714, 4.8168], device='cuda:1'), covar=tensor([0.0381, 0.0567, 0.0468, 0.0477, 0.0489, 0.0486, 0.0958, 0.0485], device='cuda:1'), in_proj_covar=tensor([0.0404, 0.0457, 0.0444, 0.0408, 0.0492, 0.0467, 0.0537, 0.0374], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:1') 2023-05-02 07:19:22,226 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=263950.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 07:19:26,173 INFO [train.py:904] (1/8) Epoch 27, batch 50, loss[loss=0.1929, simple_loss=0.2797, pruned_loss=0.05304, over 16104.00 frames. ], tot_loss[loss=0.1735, simple_loss=0.261, pruned_loss=0.04306, over 749097.41 frames. ], batch size: 35, lr: 2.51e-03, grad_scale: 1.0 2023-05-02 07:19:36,200 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=263960.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 07:19:39,887 INFO [optim.py:368] (1/8) 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 07:20:36,410 INFO [train.py:904] (1/8) Epoch 27, batch 100, loss[loss=0.1662, simple_loss=0.2693, pruned_loss=0.03159, over 17277.00 frames. ], tot_loss[loss=0.1732, simple_loss=0.2611, pruned_loss=0.04267, over 1322437.72 frames. ], batch size: 52, lr: 2.51e-03, grad_scale: 1.0 2023-05-02 07:21:00,928 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=264021.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 07:21:02,339 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.16 vs. limit=2.0 2023-05-02 07:21:12,848 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([1.9164, 2.8134, 2.6775, 4.7504, 3.8111, 4.2178, 1.7249, 3.0137], device='cuda:1'), covar=tensor([0.1374, 0.0785, 0.1236, 0.0215, 0.0228, 0.0474, 0.1633, 0.0885], device='cuda:1'), in_proj_covar=tensor([0.0170, 0.0176, 0.0195, 0.0192, 0.0199, 0.0213, 0.0205, 0.0193], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-05-02 07:21:44,757 INFO [train.py:904] (1/8) Epoch 27, batch 150, loss[loss=0.1518, simple_loss=0.2329, pruned_loss=0.03538, over 16701.00 frames. ], tot_loss[loss=0.1712, simple_loss=0.2589, pruned_loss=0.04178, over 1774334.94 frames. ], batch size: 89, lr: 2.51e-03, grad_scale: 1.0 2023-05-02 07:21:57,551 INFO [optim.py:368] (1/8) 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 07:22:41,007 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.17 vs. limit=2.0 2023-05-02 07:22:52,275 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=4.57 vs. limit=5.0 2023-05-02 07:22:53,632 INFO [train.py:904] (1/8) Epoch 27, batch 200, loss[loss=0.1737, simple_loss=0.2574, pruned_loss=0.04496, over 16248.00 frames. ], tot_loss[loss=0.1742, simple_loss=0.2613, pruned_loss=0.04357, over 2117773.91 frames. ], batch size: 165, lr: 2.51e-03, grad_scale: 1.0 2023-05-02 07:22:57,600 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.7360, 2.7230, 2.3669, 2.5373, 3.0380, 2.7865, 3.2481, 3.2577], device='cuda:1'), covar=tensor([0.0196, 0.0503, 0.0592, 0.0570, 0.0348, 0.0480, 0.0330, 0.0335], device='cuda:1'), in_proj_covar=tensor([0.0221, 0.0240, 0.0230, 0.0231, 0.0241, 0.0239, 0.0237, 0.0236], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-02 07:23:45,429 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.4160, 4.4088, 4.7422, 4.7300, 4.7768, 4.5107, 4.5010, 4.4262], device='cuda:1'), covar=tensor([0.0497, 0.0880, 0.0570, 0.0613, 0.0583, 0.0555, 0.0972, 0.0680], device='cuda:1'), in_proj_covar=tensor([0.0410, 0.0464, 0.0451, 0.0414, 0.0498, 0.0474, 0.0545, 0.0379], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:1') 2023-05-02 07:24:00,835 INFO [train.py:904] (1/8) Epoch 27, batch 250, loss[loss=0.153, simple_loss=0.2354, pruned_loss=0.0353, over 16780.00 frames. ], tot_loss[loss=0.1727, simple_loss=0.2598, pruned_loss=0.04283, over 2392100.45 frames. ], batch size: 83, lr: 2.51e-03, grad_scale: 1.0 2023-05-02 07:24:03,003 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.2958, 4.0424, 4.4576, 2.5014, 4.7255, 4.8440, 3.4330, 3.7227], device='cuda:1'), covar=tensor([0.0663, 0.0288, 0.0258, 0.1107, 0.0090, 0.0172, 0.0485, 0.0380], device='cuda:1'), in_proj_covar=tensor([0.0147, 0.0109, 0.0097, 0.0137, 0.0084, 0.0126, 0.0128, 0.0128], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0004, 0.0003], device='cuda:1') 2023-05-02 07:24:14,011 INFO [optim.py:368] (1/8) 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 07:24:50,572 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 27, batch 300, loss[loss=0.1447, simple_loss=0.2332, pruned_loss=0.02806, over 16867.00 frames. ], tot_loss[loss=0.1704, simple_loss=0.2575, pruned_loss=0.04167, over 2585203.54 frames. ], batch size: 39, lr: 2.51e-03, grad_scale: 1.0 2023-05-02 07:25:42,589 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.5064, 3.5342, 3.8578, 2.6606, 3.5237, 3.9242, 3.6669, 2.3970], device='cuda:1'), covar=tensor([0.0530, 0.0360, 0.0068, 0.0404, 0.0131, 0.0113, 0.0111, 0.0458], device='cuda:1'), in_proj_covar=tensor([0.0136, 0.0088, 0.0088, 0.0134, 0.0100, 0.0112, 0.0096, 0.0130], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:1') 2023-05-02 07:25:52,999 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 27, batch 350, loss[loss=0.152, simple_loss=0.2349, pruned_loss=0.03459, over 15437.00 frames. ], tot_loss[loss=0.1672, simple_loss=0.2544, pruned_loss=0.04, over 2754739.96 frames. ], batch size: 190, lr: 2.51e-03, grad_scale: 1.0 2023-05-02 07:26:34,329 INFO [optim.py:368] (1/8) Clipping_scale=2.0, grad-norm quartiles 1.322e+02 2.070e+02 2.475e+02 2.951e+02 5.112e+02, threshold=4.951e+02, percent-clipped=0.0 2023-05-02 07:27:17,369 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=264295.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 07:27:27,025 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.8783, 4.8379, 4.6919, 4.2224, 4.7837, 1.9808, 4.5349, 4.3628], device='cuda:1'), covar=tensor([0.0149, 0.0121, 0.0224, 0.0369, 0.0112, 0.2677, 0.0185, 0.0285], device='cuda:1'), in_proj_covar=tensor([0.0175, 0.0168, 0.0204, 0.0178, 0.0181, 0.0214, 0.0194, 0.0175], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-02 07:27:27,690 INFO [train.py:904] (1/8) Epoch 27, batch 400, loss[loss=0.1828, simple_loss=0.2631, pruned_loss=0.05119, over 16497.00 frames. ], tot_loss[loss=0.1654, simple_loss=0.2523, pruned_loss=0.03927, over 2877053.04 frames. ], batch size: 75, lr: 2.51e-03, grad_scale: 2.0 2023-05-02 07:27:44,538 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=264316.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 07:27:59,260 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.6744, 2.4717, 2.0853, 2.2887, 2.8399, 2.5996, 2.6754, 2.9044], device='cuda:1'), covar=tensor([0.0332, 0.0489, 0.0625, 0.0494, 0.0298, 0.0385, 0.0312, 0.0324], device='cuda:1'), in_proj_covar=tensor([0.0225, 0.0243, 0.0233, 0.0233, 0.0243, 0.0242, 0.0240, 0.0239], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-02 07:28:18,159 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.80 vs. limit=2.0 2023-05-02 07:28:33,908 INFO [train.py:904] (1/8) Epoch 27, batch 450, loss[loss=0.163, simple_loss=0.2628, pruned_loss=0.03158, over 17040.00 frames. ], tot_loss[loss=0.1642, simple_loss=0.251, pruned_loss=0.03867, over 2980248.93 frames. ], batch size: 50, lr: 2.51e-03, grad_scale: 2.0 2023-05-02 07:28:48,386 INFO [optim.py:368] (1/8) 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] (1/8) Epoch 27, batch 500, loss[loss=0.1796, simple_loss=0.2521, pruned_loss=0.05355, over 16439.00 frames. ], tot_loss[loss=0.1623, simple_loss=0.2488, pruned_loss=0.03788, over 3052027.94 frames. ], batch size: 146, lr: 2.51e-03, grad_scale: 2.0 2023-05-02 07:30:51,106 INFO [train.py:904] (1/8) Epoch 27, batch 550, loss[loss=0.1591, simple_loss=0.2487, pruned_loss=0.03479, over 17162.00 frames. ], tot_loss[loss=0.162, simple_loss=0.2488, pruned_loss=0.0376, over 3124991.88 frames. ], batch size: 46, lr: 2.51e-03, grad_scale: 2.0 2023-05-02 07:31:04,229 INFO [optim.py:368] (1/8) 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 07:31:24,260 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.81 vs. limit=2.0 2023-05-02 07:31:26,650 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=264479.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 07:31:40,713 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=264489.0, num_to_drop=1, layers_to_drop={0} 2023-05-02 07:31:55,521 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=264500.0, num_to_drop=1, layers_to_drop={1} 2023-05-02 07:31:59,349 INFO [train.py:904] (1/8) Epoch 27, batch 600, loss[loss=0.1281, simple_loss=0.2149, pruned_loss=0.02064, over 16980.00 frames. ], tot_loss[loss=0.1614, simple_loss=0.2479, pruned_loss=0.03745, over 3167984.59 frames. ], batch size: 41, lr: 2.51e-03, grad_scale: 2.0 2023-05-02 07:32:41,897 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.8181, 4.3428, 4.3110, 3.0502, 3.6220, 4.2885, 3.9257, 2.4868], device='cuda:1'), covar=tensor([0.0496, 0.0074, 0.0059, 0.0389, 0.0150, 0.0107, 0.0103, 0.0534], device='cuda:1'), in_proj_covar=tensor([0.0136, 0.0088, 0.0089, 0.0134, 0.0100, 0.0112, 0.0097, 0.0130], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0003, 0.0004], device='cuda:1') 2023-05-02 07:32:46,461 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=264537.0, num_to_drop=1, layers_to_drop={0} 2023-05-02 07:32:50,481 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=264540.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 07:33:08,796 INFO [train.py:904] (1/8) Epoch 27, batch 650, loss[loss=0.1572, simple_loss=0.253, pruned_loss=0.03069, over 16745.00 frames. ], tot_loss[loss=0.1602, simple_loss=0.2466, pruned_loss=0.03688, over 3203879.94 frames. ], batch size: 57, lr: 2.51e-03, grad_scale: 2.0 2023-05-02 07:33:18,831 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=264561.0, num_to_drop=1, layers_to_drop={3} 2023-05-02 07:33:20,702 INFO [optim.py:368] (1/8) 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 07:33:58,319 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 27, batch 700, loss[loss=0.1532, simple_loss=0.2461, pruned_loss=0.03016, over 17136.00 frames. ], tot_loss[loss=0.1601, simple_loss=0.2463, pruned_loss=0.03692, over 3232279.49 frames. ], batch size: 49, lr: 2.51e-03, grad_scale: 2.0 2023-05-02 07:34:34,335 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=264616.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 07:35:00,695 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.5945, 4.5676, 4.9364, 4.9375, 4.9706, 4.6595, 4.6595, 4.5341], device='cuda:1'), covar=tensor([0.0403, 0.0763, 0.0458, 0.0385, 0.0514, 0.0486, 0.0918, 0.0647], device='cuda:1'), in_proj_covar=tensor([0.0429, 0.0485, 0.0469, 0.0431, 0.0519, 0.0497, 0.0569, 0.0395], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:1') 2023-05-02 07:35:22,117 INFO [train.py:904] (1/8) Epoch 27, batch 750, loss[loss=0.1592, simple_loss=0.2552, pruned_loss=0.03157, over 17127.00 frames. ], tot_loss[loss=0.1609, simple_loss=0.2473, pruned_loss=0.03728, over 3252914.90 frames. ], batch size: 49, lr: 2.51e-03, grad_scale: 2.0 2023-05-02 07:35:35,421 INFO [optim.py:368] (1/8) 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 07:35:35,675 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.2508, 4.3049, 4.4313, 4.2543, 4.3059, 4.8392, 4.3696, 4.0211], device='cuda:1'), covar=tensor([0.1886, 0.2250, 0.3056, 0.2388, 0.3102, 0.1311, 0.1958, 0.2873], device='cuda:1'), in_proj_covar=tensor([0.0422, 0.0625, 0.0694, 0.0513, 0.0679, 0.0717, 0.0539, 0.0682], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-05-02 07:35:37,450 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=264664.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 07:36:29,180 INFO [train.py:904] (1/8) Epoch 27, batch 800, loss[loss=0.1443, simple_loss=0.2295, pruned_loss=0.02959, over 16738.00 frames. ], tot_loss[loss=0.1609, simple_loss=0.247, pruned_loss=0.0374, over 3267768.28 frames. ], batch size: 89, lr: 2.51e-03, grad_scale: 4.0 2023-05-02 07:37:36,919 INFO [train.py:904] (1/8) Epoch 27, batch 850, loss[loss=0.1758, simple_loss=0.2525, pruned_loss=0.04953, over 16443.00 frames. ], tot_loss[loss=0.1608, simple_loss=0.2469, pruned_loss=0.03736, over 3284680.30 frames. ], batch size: 75, lr: 2.51e-03, grad_scale: 4.0 2023-05-02 07:37:51,807 INFO [optim.py:368] (1/8) Clipping_scale=2.0, grad-norm quartiles 1.434e+02 2.032e+02 2.277e+02 2.694e+02 3.998e+02, threshold=4.553e+02, percent-clipped=0.0 2023-05-02 07:37:56,513 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=3.97 vs. limit=5.0 2023-05-02 07:38:11,539 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.9078, 5.0018, 5.3821, 5.3616, 5.3501, 5.0469, 5.0068, 4.8343], device='cuda:1'), covar=tensor([0.0390, 0.0562, 0.0447, 0.0406, 0.0531, 0.0467, 0.1001, 0.0506], device='cuda:1'), in_proj_covar=tensor([0.0432, 0.0489, 0.0472, 0.0434, 0.0523, 0.0500, 0.0574, 0.0398], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:1') 2023-05-02 07:38:44,872 INFO [train.py:904] (1/8) Epoch 27, batch 900, loss[loss=0.1558, simple_loss=0.2413, pruned_loss=0.03511, over 16549.00 frames. ], tot_loss[loss=0.1602, simple_loss=0.2464, pruned_loss=0.03702, over 3295932.56 frames. ], batch size: 75, lr: 2.51e-03, grad_scale: 4.0 2023-05-02 07:38:47,211 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=264804.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 07:39:28,096 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 27, batch 950, loss[loss=0.167, simple_loss=0.2459, pruned_loss=0.04403, over 16719.00 frames. ], tot_loss[loss=0.1604, simple_loss=0.2464, pruned_loss=0.03723, over 3299678.76 frames. ], batch size: 89, lr: 2.51e-03, grad_scale: 4.0 2023-05-02 07:39:55,588 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=264856.0, num_to_drop=1, layers_to_drop={1} 2023-05-02 07:39:55,771 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.8351, 1.9985, 2.5376, 2.8431, 2.6221, 3.3416, 2.1287, 3.3477], device='cuda:1'), covar=tensor([0.0332, 0.0639, 0.0420, 0.0400, 0.0460, 0.0258, 0.0662, 0.0203], device='cuda:1'), in_proj_covar=tensor([0.0195, 0.0196, 0.0184, 0.0190, 0.0205, 0.0163, 0.0203, 0.0162], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-02 07:40:04,149 INFO [optim.py:368] (1/8) 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 07:40:07,762 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=264865.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 07:40:19,905 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.5121, 4.4524, 4.4057, 3.7795, 4.4351, 1.7253, 4.0910, 3.9602], device='cuda:1'), covar=tensor([0.0238, 0.0224, 0.0219, 0.0432, 0.0152, 0.3040, 0.0262, 0.0332], device='cuda:1'), in_proj_covar=tensor([0.0177, 0.0170, 0.0207, 0.0181, 0.0184, 0.0216, 0.0197, 0.0178], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-02 07:40:41,674 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=264890.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 07:40:57,199 INFO [train.py:904] (1/8) Epoch 27, batch 1000, loss[loss=0.1519, simple_loss=0.2505, pruned_loss=0.0267, over 17253.00 frames. ], tot_loss[loss=0.1603, simple_loss=0.2457, pruned_loss=0.03739, over 3312837.80 frames. ], batch size: 52, lr: 2.51e-03, grad_scale: 4.0 2023-05-02 07:41:46,843 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=264938.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 07:42:07,408 INFO [train.py:904] (1/8) Epoch 27, batch 1050, loss[loss=0.1774, simple_loss=0.2547, pruned_loss=0.05007, over 16736.00 frames. ], tot_loss[loss=0.1599, simple_loss=0.2454, pruned_loss=0.03717, over 3314256.72 frames. ], batch size: 102, lr: 2.51e-03, grad_scale: 4.0 2023-05-02 07:42:19,716 INFO [optim.py:368] (1/8) 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 07:43:16,444 INFO [train.py:904] (1/8) Epoch 27, batch 1100, loss[loss=0.1566, simple_loss=0.2533, pruned_loss=0.02996, over 17025.00 frames. ], tot_loss[loss=0.1596, simple_loss=0.2454, pruned_loss=0.03689, over 3307099.43 frames. ], batch size: 50, lr: 2.51e-03, grad_scale: 4.0 2023-05-02 07:43:49,351 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.6570, 3.9001, 2.6386, 4.5005, 3.1088, 4.4649, 2.5864, 3.2633], device='cuda:1'), covar=tensor([0.0389, 0.0409, 0.1555, 0.0353, 0.0874, 0.0566, 0.1659, 0.0825], device='cuda:1'), in_proj_covar=tensor([0.0177, 0.0182, 0.0197, 0.0174, 0.0181, 0.0221, 0.0207, 0.0185], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-05-02 07:44:25,482 INFO [train.py:904] (1/8) Epoch 27, batch 1150, loss[loss=0.1393, simple_loss=0.231, pruned_loss=0.02379, over 17199.00 frames. ], tot_loss[loss=0.1587, simple_loss=0.2447, pruned_loss=0.03637, over 3310775.45 frames. ], batch size: 46, lr: 2.51e-03, grad_scale: 4.0 2023-05-02 07:44:39,251 INFO [optim.py:368] (1/8) 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] (1/8) Epoch 27, batch 1200, loss[loss=0.1553, simple_loss=0.2367, pruned_loss=0.03693, over 16769.00 frames. ], tot_loss[loss=0.1576, simple_loss=0.2434, pruned_loss=0.03585, over 3313268.12 frames. ], batch size: 124, lr: 2.51e-03, grad_scale: 8.0 2023-05-02 07:45:35,805 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.9667, 5.4343, 5.5368, 5.2580, 5.3192, 5.9590, 5.4709, 5.1712], device='cuda:1'), covar=tensor([0.1091, 0.2141, 0.2801, 0.2022, 0.2812, 0.1052, 0.1584, 0.2226], device='cuda:1'), in_proj_covar=tensor([0.0426, 0.0631, 0.0701, 0.0518, 0.0687, 0.0724, 0.0545, 0.0689], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-05-02 07:46:19,322 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=265135.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 07:46:42,960 INFO [train.py:904] (1/8) Epoch 27, batch 1250, loss[loss=0.1621, simple_loss=0.2393, pruned_loss=0.04246, over 16402.00 frames. ], tot_loss[loss=0.1585, simple_loss=0.2441, pruned_loss=0.03645, over 3306796.13 frames. ], batch size: 68, lr: 2.51e-03, grad_scale: 8.0 2023-05-02 07:46:48,325 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=265156.0, num_to_drop=1, layers_to_drop={1} 2023-05-02 07:46:53,495 INFO [zipformer.py:625] (1/8) 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] (1/8) 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] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=265183.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 07:47:53,025 INFO [train.py:904] (1/8) Epoch 27, batch 1300, loss[loss=0.1389, simple_loss=0.2264, pruned_loss=0.02569, over 17200.00 frames. ], tot_loss[loss=0.159, simple_loss=0.2451, pruned_loss=0.03646, over 3320799.23 frames. ], batch size: 44, lr: 2.51e-03, grad_scale: 8.0 2023-05-02 07:47:54,315 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=265204.0, num_to_drop=1, layers_to_drop={0} 2023-05-02 07:47:54,374 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.6388, 4.7861, 5.0749, 5.0408, 5.1110, 4.8656, 4.7275, 4.6504], device='cuda:1'), covar=tensor([0.0558, 0.0932, 0.0683, 0.0713, 0.0680, 0.0654, 0.1237, 0.0576], device='cuda:1'), in_proj_covar=tensor([0.0432, 0.0488, 0.0471, 0.0433, 0.0523, 0.0500, 0.0575, 0.0398], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:1') 2023-05-02 07:49:00,603 INFO [train.py:904] (1/8) Epoch 27, batch 1350, loss[loss=0.1371, simple_loss=0.2301, pruned_loss=0.02207, over 17158.00 frames. ], tot_loss[loss=0.1593, simple_loss=0.2453, pruned_loss=0.03662, over 3325261.90 frames. ], batch size: 46, lr: 2.51e-03, grad_scale: 8.0 2023-05-02 07:49:14,442 INFO [optim.py:368] (1/8) Clipping_scale=2.0, grad-norm quartiles 1.465e+02 2.174e+02 2.461e+02 3.019e+02 8.065e+02, threshold=4.923e+02, percent-clipped=2.0 2023-05-02 07:49:28,370 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.2899, 2.4379, 2.3711, 4.1215, 2.3143, 2.7166, 2.4535, 2.5174], device='cuda:1'), covar=tensor([0.1489, 0.3658, 0.3401, 0.0635, 0.4240, 0.2579, 0.3871, 0.3680], device='cuda:1'), in_proj_covar=tensor([0.0421, 0.0473, 0.0387, 0.0337, 0.0448, 0.0540, 0.0444, 0.0553], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-02 07:49:35,244 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=265278.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 07:49:39,499 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.7354, 3.3597, 3.8075, 1.9971, 3.9484, 3.9307, 3.1972, 2.9129], device='cuda:1'), covar=tensor([0.0749, 0.0315, 0.0228, 0.1265, 0.0110, 0.0237, 0.0422, 0.0467], device='cuda:1'), in_proj_covar=tensor([0.0150, 0.0112, 0.0101, 0.0140, 0.0086, 0.0131, 0.0131, 0.0131], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-05-02 07:50:07,766 INFO [train.py:904] (1/8) Epoch 27, batch 1400, loss[loss=0.175, simple_loss=0.2717, pruned_loss=0.03914, over 16688.00 frames. ], tot_loss[loss=0.16, simple_loss=0.2459, pruned_loss=0.03704, over 3331473.17 frames. ], batch size: 57, lr: 2.51e-03, grad_scale: 8.0 2023-05-02 07:50:27,669 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.3767, 3.7337, 3.6763, 1.8233, 2.9137, 2.1371, 3.7249, 3.8916], device='cuda:1'), covar=tensor([0.0274, 0.0888, 0.0667, 0.2556, 0.1079, 0.1299, 0.0682, 0.1113], device='cuda:1'), in_proj_covar=tensor([0.0161, 0.0170, 0.0171, 0.0158, 0.0148, 0.0133, 0.0147, 0.0182], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0005], device='cuda:1') 2023-05-02 07:50:46,833 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=4.83 vs. limit=5.0 2023-05-02 07:50:56,868 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=265339.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 07:51:15,551 INFO [train.py:904] (1/8) Epoch 27, batch 1450, loss[loss=0.1526, simple_loss=0.2396, pruned_loss=0.03282, over 17224.00 frames. ], tot_loss[loss=0.1592, simple_loss=0.2448, pruned_loss=0.03686, over 3334639.21 frames. ], batch size: 45, lr: 2.51e-03, grad_scale: 8.0 2023-05-02 07:51:29,668 INFO [optim.py:368] (1/8) 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 07:51:35,753 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.24 vs. limit=2.0 2023-05-02 07:51:57,966 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.2843, 2.6967, 2.2514, 2.4197, 3.0173, 2.7178, 3.0106, 3.1185], device='cuda:1'), covar=tensor([0.0244, 0.0459, 0.0560, 0.0513, 0.0296, 0.0408, 0.0307, 0.0298], device='cuda:1'), in_proj_covar=tensor([0.0232, 0.0248, 0.0237, 0.0237, 0.0249, 0.0247, 0.0246, 0.0245], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-02 07:52:24,304 INFO [train.py:904] (1/8) Epoch 27, batch 1500, loss[loss=0.1551, simple_loss=0.2512, pruned_loss=0.0295, over 16661.00 frames. ], tot_loss[loss=0.1596, simple_loss=0.2448, pruned_loss=0.03717, over 3334261.47 frames. ], batch size: 62, lr: 2.51e-03, grad_scale: 8.0 2023-05-02 07:53:26,489 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.4522, 4.3776, 4.3811, 4.0837, 4.1391, 4.3893, 4.1586, 4.2136], device='cuda:1'), covar=tensor([0.0628, 0.0915, 0.0322, 0.0328, 0.0751, 0.0607, 0.0729, 0.0615], device='cuda:1'), in_proj_covar=tensor([0.0320, 0.0476, 0.0371, 0.0374, 0.0370, 0.0431, 0.0255, 0.0446], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-05-02 07:53:36,975 INFO [train.py:904] (1/8) Epoch 27, batch 1550, loss[loss=0.1453, simple_loss=0.2379, pruned_loss=0.02632, over 16845.00 frames. ], tot_loss[loss=0.1614, simple_loss=0.2461, pruned_loss=0.03833, over 3318542.51 frames. ], batch size: 42, lr: 2.51e-03, grad_scale: 8.0 2023-05-02 07:53:46,033 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=265460.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 07:53:49,836 INFO [optim.py:368] (1/8) 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 07:54:04,945 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.4337, 3.0845, 3.4308, 1.9399, 3.5313, 3.5593, 2.9583, 2.7363], device='cuda:1'), covar=tensor([0.0827, 0.0294, 0.0209, 0.1173, 0.0124, 0.0250, 0.0425, 0.0450], device='cuda:1'), in_proj_covar=tensor([0.0149, 0.0111, 0.0100, 0.0140, 0.0086, 0.0131, 0.0130, 0.0130], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-05-02 07:54:43,792 INFO [train.py:904] (1/8) Epoch 27, batch 1600, loss[loss=0.1815, simple_loss=0.2816, pruned_loss=0.04064, over 17056.00 frames. ], tot_loss[loss=0.163, simple_loss=0.2479, pruned_loss=0.03903, over 3316187.46 frames. ], batch size: 53, lr: 2.51e-03, grad_scale: 8.0 2023-05-02 07:54:51,419 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=265508.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 07:55:21,273 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=265530.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 07:55:34,139 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.7322, 4.8906, 5.0639, 4.8393, 4.9314, 5.5110, 5.0501, 4.6710], device='cuda:1'), covar=tensor([0.1582, 0.2406, 0.2463, 0.2427, 0.2871, 0.1251, 0.1965, 0.2972], device='cuda:1'), in_proj_covar=tensor([0.0432, 0.0639, 0.0708, 0.0526, 0.0696, 0.0734, 0.0552, 0.0700], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-05-02 07:55:51,085 INFO [train.py:904] (1/8) Epoch 27, batch 1650, loss[loss=0.1748, simple_loss=0.2557, pruned_loss=0.04693, over 16836.00 frames. ], tot_loss[loss=0.1644, simple_loss=0.2495, pruned_loss=0.03966, over 3313382.41 frames. ], batch size: 96, lr: 2.51e-03, grad_scale: 8.0 2023-05-02 07:55:59,747 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([5.4787, 5.8723, 5.6419, 5.7175, 5.3160, 5.3667, 5.2374, 6.0419], device='cuda:1'), covar=tensor([0.1691, 0.1168, 0.0995, 0.0980, 0.0977, 0.0780, 0.1556, 0.1011], device='cuda:1'), in_proj_covar=tensor([0.0722, 0.0871, 0.0712, 0.0671, 0.0554, 0.0549, 0.0738, 0.0682], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:1') 2023-05-02 07:56:04,414 INFO [optim.py:368] (1/8) 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 07:56:20,668 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.7679, 4.0478, 2.8173, 2.3157, 2.7606, 2.4460, 4.2814, 3.4736], device='cuda:1'), covar=tensor([0.3202, 0.0726, 0.2157, 0.3180, 0.2791, 0.2400, 0.0580, 0.1507], device='cuda:1'), in_proj_covar=tensor([0.0332, 0.0274, 0.0311, 0.0325, 0.0303, 0.0274, 0.0303, 0.0349], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:1') 2023-05-02 07:56:43,840 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=265591.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 07:57:00,225 INFO [train.py:904] (1/8) Epoch 27, batch 1700, loss[loss=0.1991, simple_loss=0.2756, pruned_loss=0.06132, over 16351.00 frames. ], tot_loss[loss=0.1653, simple_loss=0.251, pruned_loss=0.03978, over 3321102.02 frames. ], batch size: 165, lr: 2.51e-03, grad_scale: 8.0 2023-05-02 07:57:01,903 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.9670, 4.7258, 5.0100, 5.1749, 5.3740, 4.7472, 5.3917, 5.3964], device='cuda:1'), covar=tensor([0.1908, 0.1477, 0.1715, 0.0794, 0.0569, 0.0927, 0.0514, 0.0609], device='cuda:1'), in_proj_covar=tensor([0.0687, 0.0846, 0.0976, 0.0857, 0.0649, 0.0676, 0.0710, 0.0827], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-02 07:57:16,031 INFO [zipformer.py:625] (1/8) 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] (1/8) attn_weights_entropy = tensor([5.0509, 5.6606, 5.8556, 5.5152, 5.6667, 6.2266, 5.6485, 5.3541], device='cuda:1'), covar=tensor([0.1008, 0.2214, 0.2359, 0.1980, 0.2331, 0.0946, 0.1607, 0.2271], device='cuda:1'), in_proj_covar=tensor([0.0430, 0.0638, 0.0706, 0.0524, 0.0693, 0.0731, 0.0550, 0.0696], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-05-02 07:57:42,660 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 27, batch 1750, loss[loss=0.1481, simple_loss=0.2441, pruned_loss=0.02601, over 17102.00 frames. ], tot_loss[loss=0.1652, simple_loss=0.2516, pruned_loss=0.03939, over 3331711.05 frames. ], batch size: 47, lr: 2.51e-03, grad_scale: 8.0 2023-05-02 07:58:14,310 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.84 vs. limit=2.0 2023-05-02 07:58:22,124 INFO [optim.py:368] (1/8) 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 07:58:38,431 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=265675.0, num_to_drop=1, layers_to_drop={3} 2023-05-02 07:59:15,661 INFO [train.py:904] (1/8) Epoch 27, batch 1800, loss[loss=0.1454, simple_loss=0.2407, pruned_loss=0.02506, over 17191.00 frames. ], tot_loss[loss=0.1658, simple_loss=0.2531, pruned_loss=0.0393, over 3318857.27 frames. ], batch size: 46, lr: 2.51e-03, grad_scale: 8.0 2023-05-02 07:59:32,448 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=265715.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 08:00:07,674 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.93 vs. limit=2.0 2023-05-02 08:00:23,534 INFO [scaling.py:679] (1/8) 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] (1/8) Epoch 27, batch 1850, loss[loss=0.1562, simple_loss=0.2445, pruned_loss=0.03394, over 16716.00 frames. ], tot_loss[loss=0.1658, simple_loss=0.2536, pruned_loss=0.039, over 3325851.00 frames. ], batch size: 89, lr: 2.50e-03, grad_scale: 8.0 2023-05-02 08:00:37,844 INFO [optim.py:368] (1/8) 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 08:00:55,825 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=265776.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 08:01:33,281 INFO [train.py:904] (1/8) Epoch 27, batch 1900, loss[loss=0.1806, simple_loss=0.2675, pruned_loss=0.04682, over 12102.00 frames. ], tot_loss[loss=0.165, simple_loss=0.253, pruned_loss=0.0385, over 3319986.79 frames. ], batch size: 248, lr: 2.50e-03, grad_scale: 8.0 2023-05-02 08:02:42,098 INFO [train.py:904] (1/8) Epoch 27, batch 1950, loss[loss=0.16, simple_loss=0.2577, pruned_loss=0.0312, over 17127.00 frames. ], tot_loss[loss=0.1643, simple_loss=0.2525, pruned_loss=0.03798, over 3328276.75 frames. ], batch size: 49, lr: 2.50e-03, grad_scale: 8.0 2023-05-02 08:02:54,907 INFO [optim.py:368] (1/8) 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 08:02:58,389 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.8472, 3.7619, 3.8774, 3.6993, 3.8381, 4.2963, 3.8567, 3.5747], device='cuda:1'), covar=tensor([0.2177, 0.2524, 0.2585, 0.2663, 0.3059, 0.1936, 0.1823, 0.2802], device='cuda:1'), in_proj_covar=tensor([0.0432, 0.0640, 0.0707, 0.0525, 0.0696, 0.0734, 0.0551, 0.0699], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-05-02 08:03:07,600 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=265872.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 08:03:26,216 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 27, batch 2000, loss[loss=0.1675, simple_loss=0.2585, pruned_loss=0.03819, over 16670.00 frames. ], tot_loss[loss=0.1646, simple_loss=0.2526, pruned_loss=0.03835, over 3328272.92 frames. ], batch size: 62, lr: 2.50e-03, grad_scale: 8.0 2023-05-02 08:04:10,503 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=265917.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 08:04:33,625 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=265933.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 08:04:34,881 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=265934.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 08:05:01,216 INFO [train.py:904] (1/8) Epoch 27, batch 2050, loss[loss=0.1516, simple_loss=0.2569, pruned_loss=0.02317, over 17259.00 frames. ], tot_loss[loss=0.1649, simple_loss=0.2524, pruned_loss=0.03869, over 3321408.80 frames. ], batch size: 45, lr: 2.50e-03, grad_scale: 16.0 2023-05-02 08:05:14,204 INFO [optim.py:368] (1/8) 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 08:05:24,938 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=265970.0, num_to_drop=1, layers_to_drop={1} 2023-05-02 08:05:36,781 INFO [zipformer.py:625] (1/8) 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] (1/8) 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] (1/8) Epoch 27, batch 2100, loss[loss=0.2, simple_loss=0.2853, pruned_loss=0.05731, over 16735.00 frames. ], tot_loss[loss=0.1659, simple_loss=0.2532, pruned_loss=0.03934, over 3314282.60 frames. ], batch size: 124, lr: 2.50e-03, grad_scale: 16.0 2023-05-02 08:07:16,608 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.9013, 4.3716, 2.7300, 2.3886, 2.4554, 2.3922, 4.6718, 3.4511], device='cuda:1'), covar=tensor([0.3156, 0.0711, 0.2456, 0.3360, 0.3565, 0.2649, 0.0507, 0.1656], device='cuda:1'), in_proj_covar=tensor([0.0334, 0.0276, 0.0313, 0.0326, 0.0305, 0.0275, 0.0305, 0.0352], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:1') 2023-05-02 08:07:22,091 INFO [train.py:904] (1/8) Epoch 27, batch 2150, loss[loss=0.1817, simple_loss=0.2844, pruned_loss=0.03954, over 17198.00 frames. ], tot_loss[loss=0.1665, simple_loss=0.2541, pruned_loss=0.03943, over 3319758.14 frames. ], batch size: 52, lr: 2.50e-03, grad_scale: 8.0 2023-05-02 08:07:37,030 INFO [optim.py:368] (1/8) 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 08:07:46,310 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=266071.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 08:08:32,966 INFO [train.py:904] (1/8) Epoch 27, batch 2200, loss[loss=0.1515, simple_loss=0.2412, pruned_loss=0.03088, over 17218.00 frames. ], tot_loss[loss=0.166, simple_loss=0.2536, pruned_loss=0.03924, over 3321036.59 frames. ], batch size: 45, lr: 2.50e-03, grad_scale: 8.0 2023-05-02 08:09:41,316 INFO [train.py:904] (1/8) Epoch 27, batch 2250, loss[loss=0.1601, simple_loss=0.2525, pruned_loss=0.03391, over 16833.00 frames. ], tot_loss[loss=0.1664, simple_loss=0.2533, pruned_loss=0.03973, over 3319684.09 frames. ], batch size: 102, lr: 2.50e-03, grad_scale: 8.0 2023-05-02 08:09:56,594 INFO [optim.py:368] (1/8) 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 08:10:27,374 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=266186.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 08:10:51,414 INFO [train.py:904] (1/8) Epoch 27, batch 2300, loss[loss=0.2013, simple_loss=0.2814, pruned_loss=0.0606, over 11645.00 frames. ], tot_loss[loss=0.1664, simple_loss=0.2538, pruned_loss=0.03949, over 3318802.43 frames. ], batch size: 246, lr: 2.50e-03, grad_scale: 8.0 2023-05-02 08:11:06,843 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=266214.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 08:11:23,555 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=266226.0, num_to_drop=1, layers_to_drop={0} 2023-05-02 08:11:23,923 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=3.13 vs. limit=5.0 2023-05-02 08:11:25,873 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=266228.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 08:11:34,296 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 27, batch 2350, loss[loss=0.1797, simple_loss=0.2574, pruned_loss=0.05101, over 16488.00 frames. ], tot_loss[loss=0.1671, simple_loss=0.2544, pruned_loss=0.03985, over 3320567.89 frames. ], batch size: 146, lr: 2.50e-03, grad_scale: 8.0 2023-05-02 08:12:14,159 INFO [optim.py:368] (1/8) Clipping_scale=2.0, grad-norm quartiles 1.417e+02 2.140e+02 2.442e+02 2.999e+02 5.112e+02, threshold=4.884e+02, percent-clipped=1.0 2023-05-02 08:12:22,264 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=266270.0, num_to_drop=1, layers_to_drop={2} 2023-05-02 08:12:24,868 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([1.9877, 2.3269, 2.3336, 2.6259, 2.0507, 3.1935, 1.8094, 2.7178], device='cuda:1'), covar=tensor([0.1136, 0.0740, 0.1085, 0.0199, 0.0133, 0.0382, 0.1412, 0.0735], device='cuda:1'), in_proj_covar=tensor([0.0173, 0.0180, 0.0199, 0.0200, 0.0206, 0.0219, 0.0208, 0.0198], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-05-02 08:12:27,931 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=266273.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 08:12:30,370 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=266275.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 08:12:46,361 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=266287.0, num_to_drop=1, layers_to_drop={0} 2023-05-02 08:13:07,765 INFO [train.py:904] (1/8) Epoch 27, batch 2400, loss[loss=0.1575, simple_loss=0.2587, pruned_loss=0.02815, over 17294.00 frames. ], tot_loss[loss=0.1681, simple_loss=0.2556, pruned_loss=0.04027, over 3307843.46 frames. ], batch size: 52, lr: 2.50e-03, grad_scale: 8.0 2023-05-02 08:13:28,958 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=266318.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 08:14:07,096 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=266346.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 08:14:16,214 INFO [train.py:904] (1/8) Epoch 27, batch 2450, loss[loss=0.176, simple_loss=0.2583, pruned_loss=0.04683, over 16427.00 frames. ], tot_loss[loss=0.1689, simple_loss=0.2571, pruned_loss=0.04037, over 3307387.93 frames. ], batch size: 146, lr: 2.50e-03, grad_scale: 8.0 2023-05-02 08:14:31,082 INFO [optim.py:368] (1/8) 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 08:14:41,913 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=266371.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 08:15:26,444 INFO [train.py:904] (1/8) Epoch 27, batch 2500, loss[loss=0.1795, simple_loss=0.2733, pruned_loss=0.04283, over 16517.00 frames. ], tot_loss[loss=0.1683, simple_loss=0.2569, pruned_loss=0.03986, over 3309745.89 frames. ], batch size: 68, lr: 2.50e-03, grad_scale: 8.0 2023-05-02 08:15:31,477 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.9281, 4.9448, 5.3520, 5.3179, 5.3497, 5.0170, 4.9392, 4.7538], device='cuda:1'), covar=tensor([0.0362, 0.0638, 0.0375, 0.0416, 0.0501, 0.0425, 0.1073, 0.0506], device='cuda:1'), in_proj_covar=tensor([0.0440, 0.0496, 0.0478, 0.0441, 0.0530, 0.0508, 0.0583, 0.0405], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:1') 2023-05-02 08:15:31,608 INFO [zipformer.py:625] (1/8) 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] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=266419.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 08:16:35,710 INFO [train.py:904] (1/8) Epoch 27, batch 2550, loss[loss=0.2101, simple_loss=0.289, pruned_loss=0.06559, over 11965.00 frames. ], tot_loss[loss=0.169, simple_loss=0.2577, pruned_loss=0.04017, over 3297862.41 frames. ], batch size: 247, lr: 2.50e-03, grad_scale: 8.0 2023-05-02 08:16:51,124 INFO [optim.py:368] (1/8) 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 08:16:51,499 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=266464.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 08:17:45,263 INFO [train.py:904] (1/8) Epoch 27, batch 2600, loss[loss=0.1456, simple_loss=0.2392, pruned_loss=0.02597, over 16853.00 frames. ], tot_loss[loss=0.169, simple_loss=0.2576, pruned_loss=0.04014, over 3292255.78 frames. ], batch size: 42, lr: 2.50e-03, grad_scale: 8.0 2023-05-02 08:17:48,575 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([5.5836, 5.9817, 5.7404, 5.8284, 5.3859, 5.3578, 5.3953, 6.1178], device='cuda:1'), covar=tensor([0.1599, 0.1018, 0.1028, 0.0914, 0.0908, 0.0693, 0.1382, 0.0943], device='cuda:1'), in_proj_covar=tensor([0.0724, 0.0876, 0.0713, 0.0674, 0.0555, 0.0550, 0.0740, 0.0686], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:1') 2023-05-02 08:18:07,346 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.78 vs. limit=2.0 2023-05-02 08:18:16,598 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=266525.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 08:18:20,994 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=266528.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 08:18:55,439 INFO [train.py:904] (1/8) Epoch 27, batch 2650, loss[loss=0.1474, simple_loss=0.2339, pruned_loss=0.03044, over 16774.00 frames. ], tot_loss[loss=0.1696, simple_loss=0.2584, pruned_loss=0.04045, over 3299358.38 frames. ], batch size: 39, lr: 2.50e-03, grad_scale: 8.0 2023-05-02 08:19:08,882 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([1.7334, 2.7225, 2.4685, 4.0935, 3.2259, 4.0278, 1.6158, 2.8630], device='cuda:1'), covar=tensor([0.1492, 0.0752, 0.1266, 0.0189, 0.0146, 0.0378, 0.1674, 0.0910], device='cuda:1'), in_proj_covar=tensor([0.0172, 0.0180, 0.0199, 0.0200, 0.0206, 0.0219, 0.0208, 0.0197], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-05-02 08:19:11,568 INFO [optim.py:368] (1/8) 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] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=266570.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 08:19:21,017 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.0973, 3.0943, 3.1585, 2.1966, 2.8997, 3.2635, 3.0083, 2.0042], device='cuda:1'), covar=tensor([0.0509, 0.0129, 0.0081, 0.0433, 0.0164, 0.0116, 0.0126, 0.0498], device='cuda:1'), in_proj_covar=tensor([0.0138, 0.0090, 0.0091, 0.0136, 0.0103, 0.0116, 0.0099, 0.0132], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0003, 0.0004], device='cuda:1') 2023-05-02 08:19:24,561 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=266573.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 08:19:29,419 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=266576.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 08:19:37,429 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=266582.0, num_to_drop=1, layers_to_drop={3} 2023-05-02 08:20:05,798 INFO [train.py:904] (1/8) Epoch 27, batch 2700, loss[loss=0.1596, simple_loss=0.2502, pruned_loss=0.03451, over 15445.00 frames. ], tot_loss[loss=0.1691, simple_loss=0.2582, pruned_loss=0.03998, over 3302726.60 frames. ], batch size: 191, lr: 2.50e-03, grad_scale: 8.0 2023-05-02 08:20:31,906 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=266621.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 08:21:15,402 INFO [train.py:904] (1/8) Epoch 27, batch 2750, loss[loss=0.1612, simple_loss=0.2511, pruned_loss=0.03565, over 16863.00 frames. ], tot_loss[loss=0.1685, simple_loss=0.2579, pruned_loss=0.03952, over 3315305.45 frames. ], batch size: 42, lr: 2.50e-03, grad_scale: 8.0 2023-05-02 08:21:29,197 INFO [optim.py:368] (1/8) 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 08:21:51,268 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=266679.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 08:22:22,657 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 27, batch 2800, loss[loss=0.1783, simple_loss=0.2584, pruned_loss=0.04912, over 16853.00 frames. ], tot_loss[loss=0.168, simple_loss=0.2575, pruned_loss=0.03927, over 3325171.44 frames. ], batch size: 109, lr: 2.50e-03, grad_scale: 8.0 2023-05-02 08:22:25,081 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=266704.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 08:22:32,472 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.0195, 3.7022, 4.2683, 2.2389, 4.3830, 4.5571, 3.2839, 3.5219], device='cuda:1'), covar=tensor([0.0700, 0.0296, 0.0245, 0.1112, 0.0092, 0.0192, 0.0432, 0.0387], device='cuda:1'), in_proj_covar=tensor([0.0149, 0.0111, 0.0101, 0.0139, 0.0086, 0.0132, 0.0130, 0.0130], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-05-02 08:22:38,169 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.6466, 3.4935, 3.9604, 2.0368, 4.0416, 4.0875, 3.1690, 3.0989], device='cuda:1'), covar=tensor([0.0853, 0.0281, 0.0217, 0.1219, 0.0097, 0.0223, 0.0434, 0.0460], device='cuda:1'), in_proj_covar=tensor([0.0149, 0.0111, 0.0101, 0.0139, 0.0086, 0.0132, 0.0131, 0.0130], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-05-02 08:23:14,822 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=266740.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 08:23:31,358 INFO [train.py:904] (1/8) Epoch 27, batch 2850, loss[loss=0.1425, simple_loss=0.2445, pruned_loss=0.0203, over 17111.00 frames. ], tot_loss[loss=0.1675, simple_loss=0.2569, pruned_loss=0.03902, over 3332567.22 frames. ], batch size: 47, lr: 2.50e-03, grad_scale: 8.0 2023-05-02 08:23:48,196 INFO [optim.py:368] (1/8) 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 08:23:49,868 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=266765.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 08:23:59,951 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.40 vs. limit=2.0 2023-05-02 08:24:13,941 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.0104, 4.4032, 3.2122, 2.4845, 2.7229, 2.6537, 4.8721, 3.6615], device='cuda:1'), covar=tensor([0.2917, 0.0610, 0.1828, 0.3067, 0.3114, 0.2240, 0.0357, 0.1499], device='cuda:1'), in_proj_covar=tensor([0.0335, 0.0277, 0.0314, 0.0327, 0.0306, 0.0276, 0.0305, 0.0354], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:1') 2023-05-02 08:24:40,562 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.0069, 3.6845, 4.2873, 2.1334, 4.3703, 4.5370, 3.2616, 3.5207], device='cuda:1'), covar=tensor([0.0745, 0.0276, 0.0220, 0.1256, 0.0089, 0.0201, 0.0468, 0.0399], device='cuda:1'), in_proj_covar=tensor([0.0149, 0.0111, 0.0102, 0.0139, 0.0086, 0.0132, 0.0131, 0.0130], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-05-02 08:24:41,243 INFO [train.py:904] (1/8) Epoch 27, batch 2900, loss[loss=0.1919, simple_loss=0.2777, pruned_loss=0.05308, over 17037.00 frames. ], tot_loss[loss=0.1672, simple_loss=0.2556, pruned_loss=0.03944, over 3335491.53 frames. ], batch size: 53, lr: 2.50e-03, grad_scale: 8.0 2023-05-02 08:25:04,248 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 27, batch 2950, loss[loss=0.1456, simple_loss=0.2397, pruned_loss=0.02579, over 17123.00 frames. ], tot_loss[loss=0.167, simple_loss=0.2549, pruned_loss=0.03954, over 3336313.89 frames. ], batch size: 47, lr: 2.50e-03, grad_scale: 8.0 2023-05-02 08:26:03,566 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.6458, 4.7669, 4.9397, 4.6693, 4.7983, 5.3678, 4.8687, 4.5223], device='cuda:1'), covar=tensor([0.1669, 0.2176, 0.2420, 0.2457, 0.2714, 0.1099, 0.1775, 0.2712], device='cuda:1'), in_proj_covar=tensor([0.0437, 0.0644, 0.0714, 0.0529, 0.0704, 0.0739, 0.0555, 0.0705], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-05-02 08:26:04,412 INFO [optim.py:368] (1/8) 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 08:26:14,816 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=266870.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 08:26:27,751 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=4.84 vs. limit=5.0 2023-05-02 08:26:29,654 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 27, batch 3000, loss[loss=0.1742, simple_loss=0.2569, pruned_loss=0.04573, over 15430.00 frames. ], tot_loss[loss=0.1673, simple_loss=0.2549, pruned_loss=0.03986, over 3334034.10 frames. ], batch size: 190, lr: 2.50e-03, grad_scale: 8.0 2023-05-02 08:26:58,265 INFO [train.py:929] (1/8) Computing validation loss 2023-05-02 08:27:07,043 INFO [train.py:938] (1/8) Epoch 27, validation: loss=0.1336, simple_loss=0.2386, pruned_loss=0.01429, over 944034.00 frames. 2023-05-02 08:27:07,044 INFO [train.py:939] (1/8) Maximum memory allocated so far is 18145MB 2023-05-02 08:27:27,842 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=266918.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 08:27:30,360 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=266920.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 08:27:45,279 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=266930.0, num_to_drop=1, layers_to_drop={0} 2023-05-02 08:27:57,263 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=4.08 vs. limit=5.0 2023-05-02 08:27:59,962 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.85 vs. limit=2.0 2023-05-02 08:28:16,682 INFO [train.py:904] (1/8) Epoch 27, batch 3050, loss[loss=0.1838, simple_loss=0.2644, pruned_loss=0.05162, over 16665.00 frames. ], tot_loss[loss=0.1672, simple_loss=0.2546, pruned_loss=0.03989, over 3332496.29 frames. ], batch size: 76, lr: 2.50e-03, grad_scale: 8.0 2023-05-02 08:28:30,545 INFO [optim.py:368] (1/8) 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 08:28:55,879 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=266981.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 08:29:09,084 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.8931, 4.3149, 3.0246, 2.4482, 2.5908, 2.6178, 4.6737, 3.5147], device='cuda:1'), covar=tensor([0.2947, 0.0580, 0.1903, 0.3132, 0.3269, 0.2236, 0.0376, 0.1554], device='cuda:1'), in_proj_covar=tensor([0.0335, 0.0277, 0.0314, 0.0327, 0.0306, 0.0276, 0.0306, 0.0354], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:1') 2023-05-02 08:29:24,833 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 27, batch 3100, loss[loss=0.1821, simple_loss=0.2704, pruned_loss=0.04689, over 16738.00 frames. ], tot_loss[loss=0.1668, simple_loss=0.2543, pruned_loss=0.03961, over 3324464.25 frames. ], batch size: 57, lr: 2.50e-03, grad_scale: 8.0 2023-05-02 08:29:43,999 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=4.75 vs. limit=5.0 2023-05-02 08:29:59,749 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([1.9306, 1.4811, 1.7678, 1.7969, 1.8487, 2.0242, 1.7458, 1.9310], device='cuda:1'), covar=tensor([0.0280, 0.0418, 0.0238, 0.0331, 0.0290, 0.0219, 0.0438, 0.0162], device='cuda:1'), in_proj_covar=tensor([0.0202, 0.0201, 0.0189, 0.0196, 0.0211, 0.0169, 0.0207, 0.0168], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0002], device='cuda:1') 2023-05-02 08:30:11,819 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=267035.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 08:30:21,310 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.8320, 2.8617, 2.5229, 2.7815, 3.1130, 2.9205, 3.4295, 3.3758], device='cuda:1'), covar=tensor([0.0198, 0.0453, 0.0577, 0.0463, 0.0352, 0.0425, 0.0251, 0.0332], device='cuda:1'), in_proj_covar=tensor([0.0236, 0.0249, 0.0237, 0.0237, 0.0251, 0.0250, 0.0249, 0.0248], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-02 08:30:33,230 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 27, batch 3150, loss[loss=0.1826, simple_loss=0.2658, pruned_loss=0.04969, over 12172.00 frames. ], tot_loss[loss=0.1655, simple_loss=0.2529, pruned_loss=0.03906, over 3312176.93 frames. ], batch size: 247, lr: 2.50e-03, grad_scale: 8.0 2023-05-02 08:30:45,721 INFO [zipformer.py:625] (1/8) 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] (1/8) 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 08:30:52,468 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.7313, 4.3838, 3.0682, 2.3130, 2.6223, 2.6890, 4.6813, 3.4967], device='cuda:1'), covar=tensor([0.3277, 0.0546, 0.2037, 0.3395, 0.3350, 0.2048, 0.0413, 0.1689], device='cuda:1'), in_proj_covar=tensor([0.0334, 0.0277, 0.0313, 0.0327, 0.0306, 0.0276, 0.0305, 0.0354], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:1') 2023-05-02 08:31:14,507 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.81 vs. limit=2.0 2023-05-02 08:31:15,317 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=267082.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 08:31:37,584 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=267098.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 08:31:43,510 INFO [train.py:904] (1/8) Epoch 27, batch 3200, loss[loss=0.1549, simple_loss=0.2455, pruned_loss=0.0321, over 16006.00 frames. ], tot_loss[loss=0.1645, simple_loss=0.252, pruned_loss=0.0385, over 3309357.04 frames. ], batch size: 35, lr: 2.50e-03, grad_scale: 8.0 2023-05-02 08:32:07,998 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=267120.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 08:32:19,880 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=3.80 vs. limit=5.0 2023-05-02 08:32:39,047 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=267143.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 08:32:41,718 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=4.85 vs. limit=5.0 2023-05-02 08:32:50,799 INFO [train.py:904] (1/8) Epoch 27, batch 3250, loss[loss=0.173, simple_loss=0.2646, pruned_loss=0.04064, over 17043.00 frames. ], tot_loss[loss=0.165, simple_loss=0.2523, pruned_loss=0.03887, over 3313359.68 frames. ], batch size: 53, lr: 2.50e-03, grad_scale: 8.0 2023-05-02 08:33:01,467 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=267159.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 08:33:07,328 INFO [optim.py:368] (1/8) 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 08:33:12,825 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=267168.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 08:33:18,014 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=4.65 vs. limit=5.0 2023-05-02 08:33:41,140 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.8647, 2.7888, 2.4285, 2.6773, 3.0952, 2.8955, 3.4528, 3.3592], device='cuda:1'), covar=tensor([0.0173, 0.0497, 0.0596, 0.0541, 0.0361, 0.0442, 0.0291, 0.0336], device='cuda:1'), in_proj_covar=tensor([0.0236, 0.0248, 0.0237, 0.0237, 0.0251, 0.0249, 0.0250, 0.0248], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-02 08:34:00,088 INFO [train.py:904] (1/8) Epoch 27, batch 3300, loss[loss=0.1521, simple_loss=0.2549, pruned_loss=0.0247, over 17050.00 frames. ], tot_loss[loss=0.1653, simple_loss=0.2528, pruned_loss=0.03897, over 3315013.54 frames. ], batch size: 50, lr: 2.50e-03, grad_scale: 8.0 2023-05-02 08:34:31,624 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.58 vs. limit=2.0 2023-05-02 08:35:07,428 INFO [train.py:904] (1/8) Epoch 27, batch 3350, loss[loss=0.2101, simple_loss=0.2879, pruned_loss=0.06613, over 15566.00 frames. ], tot_loss[loss=0.1655, simple_loss=0.2532, pruned_loss=0.03884, over 3318437.71 frames. ], batch size: 190, lr: 2.50e-03, grad_scale: 8.0 2023-05-02 08:35:22,678 INFO [optim.py:368] (1/8) 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 08:35:31,698 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([5.2926, 5.9434, 6.0649, 5.7768, 5.8928, 6.3844, 5.9055, 5.6033], device='cuda:1'), covar=tensor([0.0928, 0.1996, 0.2257, 0.2103, 0.2519, 0.0889, 0.1468, 0.2293], device='cuda:1'), in_proj_covar=tensor([0.0437, 0.0647, 0.0714, 0.0528, 0.0706, 0.0742, 0.0554, 0.0704], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-05-02 08:35:38,144 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=267276.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 08:36:03,834 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.2603, 2.8875, 3.1732, 1.9066, 3.2260, 3.2314, 2.8096, 2.6889], device='cuda:1'), covar=tensor([0.0833, 0.0272, 0.0247, 0.1093, 0.0142, 0.0272, 0.0456, 0.0431], device='cuda:1'), in_proj_covar=tensor([0.0150, 0.0112, 0.0102, 0.0140, 0.0087, 0.0133, 0.0131, 0.0131], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-05-02 08:36:13,898 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=3.34 vs. limit=5.0 2023-05-02 08:36:14,882 INFO [train.py:904] (1/8) Epoch 27, batch 3400, loss[loss=0.1328, simple_loss=0.219, pruned_loss=0.02331, over 16248.00 frames. ], tot_loss[loss=0.1655, simple_loss=0.253, pruned_loss=0.039, over 3321661.41 frames. ], batch size: 36, lr: 2.50e-03, grad_scale: 8.0 2023-05-02 08:36:44,204 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.23 vs. limit=2.0 2023-05-02 08:36:58,129 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=267335.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 08:37:02,352 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([5.2831, 5.2699, 5.0044, 4.4193, 5.0941, 1.9915, 4.8156, 4.7567], device='cuda:1'), covar=tensor([0.0093, 0.0092, 0.0244, 0.0452, 0.0122, 0.3047, 0.0174, 0.0290], device='cuda:1'), in_proj_covar=tensor([0.0183, 0.0177, 0.0215, 0.0188, 0.0192, 0.0221, 0.0205, 0.0184], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-02 08:37:14,617 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.34 vs. limit=2.0 2023-05-02 08:37:21,843 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.5057, 4.8366, 4.6723, 4.6703, 4.4048, 4.3994, 4.3924, 4.9029], device='cuda:1'), covar=tensor([0.1383, 0.0889, 0.0992, 0.0909, 0.0844, 0.1369, 0.1162, 0.0845], device='cuda:1'), in_proj_covar=tensor([0.0736, 0.0890, 0.0726, 0.0685, 0.0564, 0.0558, 0.0751, 0.0696], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:1') 2023-05-02 08:37:22,625 INFO [train.py:904] (1/8) Epoch 27, batch 3450, loss[loss=0.1638, simple_loss=0.2379, pruned_loss=0.04488, over 16236.00 frames. ], tot_loss[loss=0.1637, simple_loss=0.251, pruned_loss=0.0382, over 3324245.35 frames. ], batch size: 165, lr: 2.50e-03, grad_scale: 8.0 2023-05-02 08:37:31,756 INFO [zipformer.py:625] (1/8) 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] (1/8) 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 08:38:02,851 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=267383.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 08:38:20,944 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=2.93 vs. limit=5.0 2023-05-02 08:38:30,913 INFO [train.py:904] (1/8) Epoch 27, batch 3500, loss[loss=0.1528, simple_loss=0.2465, pruned_loss=0.02953, over 17187.00 frames. ], tot_loss[loss=0.1624, simple_loss=0.2495, pruned_loss=0.03769, over 3326363.50 frames. ], batch size: 46, lr: 2.50e-03, grad_scale: 8.0 2023-05-02 08:38:39,646 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=267408.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 08:39:19,831 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=267438.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 08:39:40,533 INFO [train.py:904] (1/8) Epoch 27, batch 3550, loss[loss=0.1717, simple_loss=0.2607, pruned_loss=0.04134, over 16766.00 frames. ], tot_loss[loss=0.1618, simple_loss=0.2488, pruned_loss=0.03742, over 3317699.17 frames. ], batch size: 57, lr: 2.50e-03, grad_scale: 8.0 2023-05-02 08:39:41,927 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=267454.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 08:39:53,962 INFO [optim.py:368] (1/8) 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 08:40:36,684 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 27, batch 3600, loss[loss=0.1639, simple_loss=0.2436, pruned_loss=0.04214, over 16386.00 frames. ], tot_loss[loss=0.1622, simple_loss=0.2489, pruned_loss=0.03774, over 3321598.15 frames. ], batch size: 145, lr: 2.50e-03, grad_scale: 8.0 2023-05-02 08:42:00,667 INFO [train.py:904] (1/8) Epoch 27, batch 3650, loss[loss=0.1711, simple_loss=0.2624, pruned_loss=0.03987, over 16625.00 frames. ], tot_loss[loss=0.1627, simple_loss=0.2483, pruned_loss=0.0385, over 3305947.46 frames. ], batch size: 57, lr: 2.50e-03, grad_scale: 8.0 2023-05-02 08:42:03,428 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=267555.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 08:42:04,791 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.7459, 3.7989, 2.5421, 4.4588, 3.0427, 4.3530, 2.7904, 3.2749], device='cuda:1'), covar=tensor([0.0304, 0.0430, 0.1660, 0.0245, 0.0850, 0.0551, 0.1408, 0.0762], device='cuda:1'), in_proj_covar=tensor([0.0179, 0.0184, 0.0199, 0.0177, 0.0182, 0.0224, 0.0206, 0.0186], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-05-02 08:42:07,895 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.5317, 3.1689, 3.5320, 1.9973, 3.6063, 3.6386, 3.1030, 2.8316], device='cuda:1'), covar=tensor([0.0771, 0.0305, 0.0227, 0.1132, 0.0131, 0.0220, 0.0428, 0.0444], device='cuda:1'), in_proj_covar=tensor([0.0149, 0.0112, 0.0102, 0.0139, 0.0087, 0.0132, 0.0131, 0.0131], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-05-02 08:42:16,686 INFO [optim.py:368] (1/8) 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 08:42:23,334 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.1694, 3.0316, 2.7532, 4.3475, 3.6823, 4.1797, 1.6952, 3.1304], device='cuda:1'), covar=tensor([0.1218, 0.0659, 0.1081, 0.0193, 0.0176, 0.0409, 0.1613, 0.0769], device='cuda:1'), in_proj_covar=tensor([0.0172, 0.0180, 0.0200, 0.0202, 0.0208, 0.0220, 0.0209, 0.0198], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-05-02 08:42:36,599 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=267576.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 08:43:14,284 INFO [train.py:904] (1/8) Epoch 27, batch 3700, loss[loss=0.1613, simple_loss=0.2369, pruned_loss=0.04283, over 16659.00 frames. ], tot_loss[loss=0.1627, simple_loss=0.2462, pruned_loss=0.03966, over 3287640.60 frames. ], batch size: 89, lr: 2.50e-03, grad_scale: 8.0 2023-05-02 08:43:45,314 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 27, batch 3750, loss[loss=0.1579, simple_loss=0.235, pruned_loss=0.04046, over 16790.00 frames. ], tot_loss[loss=0.1647, simple_loss=0.2475, pruned_loss=0.04096, over 3278267.53 frames. ], batch size: 83, lr: 2.50e-03, grad_scale: 8.0 2023-05-02 08:44:41,908 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([5.0234, 5.6208, 5.7753, 5.3624, 5.4598, 6.0717, 5.6023, 5.2906], device='cuda:1'), covar=tensor([0.0875, 0.1629, 0.1882, 0.1931, 0.2515, 0.0882, 0.1267, 0.2332], device='cuda:1'), in_proj_covar=tensor([0.0437, 0.0645, 0.0712, 0.0527, 0.0704, 0.0739, 0.0554, 0.0703], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-05-02 08:44:42,769 INFO [optim.py:368] (1/8) 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 08:44:49,715 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.2960, 3.3895, 3.5464, 2.3276, 3.0698, 2.5536, 3.8172, 3.8689], device='cuda:1'), covar=tensor([0.0232, 0.0937, 0.0630, 0.1976, 0.0891, 0.1041, 0.0446, 0.0746], device='cuda:1'), in_proj_covar=tensor([0.0162, 0.0173, 0.0171, 0.0158, 0.0149, 0.0133, 0.0148, 0.0184], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0005], device='cuda:1') 2023-05-02 08:45:20,738 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.7029, 4.8391, 4.9916, 4.8058, 4.8031, 5.4194, 4.9081, 4.6471], device='cuda:1'), covar=tensor([0.1586, 0.2032, 0.2284, 0.2192, 0.2804, 0.1037, 0.1717, 0.2642], device='cuda:1'), in_proj_covar=tensor([0.0438, 0.0647, 0.0713, 0.0528, 0.0706, 0.0740, 0.0555, 0.0704], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-05-02 08:45:40,198 INFO [train.py:904] (1/8) Epoch 27, batch 3800, loss[loss=0.1562, simple_loss=0.2347, pruned_loss=0.03884, over 16706.00 frames. ], tot_loss[loss=0.1665, simple_loss=0.2486, pruned_loss=0.04219, over 3281511.73 frames. ], batch size: 89, lr: 2.50e-03, grad_scale: 8.0 2023-05-02 08:46:31,885 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=267738.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 08:46:52,914 INFO [train.py:904] (1/8) Epoch 27, batch 3850, loss[loss=0.1749, simple_loss=0.2551, pruned_loss=0.0474, over 17023.00 frames. ], tot_loss[loss=0.1677, simple_loss=0.2495, pruned_loss=0.04297, over 3274219.61 frames. ], batch size: 41, lr: 2.50e-03, grad_scale: 8.0 2023-05-02 08:46:54,306 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=267754.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 08:47:08,877 INFO [optim.py:368] (1/8) 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 08:47:39,701 INFO [zipformer.py:625] (1/8) 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] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=267802.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 08:48:03,540 INFO [train.py:904] (1/8) Epoch 27, batch 3900, loss[loss=0.1608, simple_loss=0.2447, pruned_loss=0.03845, over 16488.00 frames. ], tot_loss[loss=0.1679, simple_loss=0.2489, pruned_loss=0.04346, over 3273404.39 frames. ], batch size: 68, lr: 2.50e-03, grad_scale: 8.0 2023-05-02 08:49:11,680 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=267850.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 08:49:14,860 INFO [train.py:904] (1/8) Epoch 27, batch 3950, loss[loss=0.1666, simple_loss=0.2504, pruned_loss=0.0414, over 17115.00 frames. ], tot_loss[loss=0.1682, simple_loss=0.2485, pruned_loss=0.04397, over 3276002.10 frames. ], batch size: 47, lr: 2.49e-03, grad_scale: 8.0 2023-05-02 08:49:32,326 INFO [optim.py:368] (1/8) Clipping_scale=2.0, grad-norm quartiles 1.702e+02 2.233e+02 2.530e+02 3.152e+02 5.510e+02, threshold=5.059e+02, percent-clipped=1.0 2023-05-02 08:49:44,324 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=4.04 vs. limit=5.0 2023-05-02 08:49:55,564 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.2841, 4.1286, 4.3363, 4.4723, 4.5565, 4.1370, 4.3459, 4.5572], device='cuda:1'), covar=tensor([0.1749, 0.1314, 0.1400, 0.0725, 0.0725, 0.1361, 0.3034, 0.0934], device='cuda:1'), in_proj_covar=tensor([0.0699, 0.0861, 0.1000, 0.0870, 0.0662, 0.0692, 0.0722, 0.0841], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-02 08:50:28,770 INFO [train.py:904] (1/8) Epoch 27, batch 4000, loss[loss=0.1706, simple_loss=0.256, pruned_loss=0.04261, over 16708.00 frames. ], tot_loss[loss=0.1687, simple_loss=0.2488, pruned_loss=0.04432, over 3278340.18 frames. ], batch size: 89, lr: 2.49e-03, grad_scale: 8.0 2023-05-02 08:51:04,388 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.5806, 3.8479, 2.8663, 2.3137, 2.5114, 2.5144, 4.1266, 3.3807], device='cuda:1'), covar=tensor([0.3062, 0.0574, 0.1783, 0.2828, 0.2749, 0.2060, 0.0417, 0.1270], device='cuda:1'), in_proj_covar=tensor([0.0335, 0.0279, 0.0315, 0.0329, 0.0309, 0.0277, 0.0307, 0.0355], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:1') 2023-05-02 08:51:24,075 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.0038, 4.1070, 4.2932, 4.2731, 4.3167, 4.0646, 4.0883, 4.0857], device='cuda:1'), covar=tensor([0.0328, 0.0572, 0.0462, 0.0484, 0.0473, 0.0469, 0.0743, 0.0533], device='cuda:1'), in_proj_covar=tensor([0.0444, 0.0502, 0.0482, 0.0445, 0.0532, 0.0510, 0.0585, 0.0405], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0003], device='cuda:1') 2023-05-02 08:51:42,546 INFO [train.py:904] (1/8) Epoch 27, batch 4050, loss[loss=0.1506, simple_loss=0.241, pruned_loss=0.03012, over 17114.00 frames. ], tot_loss[loss=0.1685, simple_loss=0.2496, pruned_loss=0.04365, over 3287389.11 frames. ], batch size: 48, lr: 2.49e-03, grad_scale: 8.0 2023-05-02 08:51:58,298 INFO [optim.py:368] (1/8) 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 08:52:48,359 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([5.1024, 5.4398, 4.9676, 5.3294, 4.9406, 4.6677, 4.9430, 5.4752], device='cuda:1'), covar=tensor([0.2304, 0.1341, 0.2228, 0.1278, 0.1466, 0.1450, 0.2319, 0.1682], device='cuda:1'), in_proj_covar=tensor([0.0732, 0.0885, 0.0720, 0.0682, 0.0562, 0.0557, 0.0749, 0.0695], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:1') 2023-05-02 08:52:59,882 INFO [train.py:904] (1/8) Epoch 27, batch 4100, loss[loss=0.166, simple_loss=0.253, pruned_loss=0.03945, over 16555.00 frames. ], tot_loss[loss=0.169, simple_loss=0.2515, pruned_loss=0.04328, over 3281664.04 frames. ], batch size: 62, lr: 2.49e-03, grad_scale: 8.0 2023-05-02 08:53:30,487 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.14 vs. limit=2.0 2023-05-02 08:54:16,165 INFO [train.py:904] (1/8) Epoch 27, batch 4150, loss[loss=0.1886, simple_loss=0.2807, pruned_loss=0.04825, over 16751.00 frames. ], tot_loss[loss=0.175, simple_loss=0.2585, pruned_loss=0.04575, over 3233560.25 frames. ], batch size: 83, lr: 2.49e-03, grad_scale: 16.0 2023-05-02 08:54:20,126 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.16 vs. limit=2.0 2023-05-02 08:54:33,171 INFO [optim.py:368] (1/8) 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 08:55:32,120 INFO [train.py:904] (1/8) Epoch 27, batch 4200, loss[loss=0.1821, simple_loss=0.2827, pruned_loss=0.04074, over 16825.00 frames. ], tot_loss[loss=0.1793, simple_loss=0.2648, pruned_loss=0.04693, over 3195208.82 frames. ], batch size: 116, lr: 2.49e-03, grad_scale: 16.0 2023-05-02 08:56:40,949 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=268150.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 08:56:44,663 INFO [train.py:904] (1/8) Epoch 27, batch 4250, loss[loss=0.1956, simple_loss=0.2731, pruned_loss=0.05905, over 12483.00 frames. ], tot_loss[loss=0.1815, simple_loss=0.2691, pruned_loss=0.04697, over 3178253.26 frames. ], batch size: 248, lr: 2.49e-03, grad_scale: 16.0 2023-05-02 08:56:57,782 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.52 vs. limit=2.0 2023-05-02 08:57:00,874 INFO [optim.py:368] (1/8) 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 08:57:50,362 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=268198.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 08:57:57,371 INFO [train.py:904] (1/8) Epoch 27, batch 4300, loss[loss=0.2029, simple_loss=0.2988, pruned_loss=0.05345, over 16719.00 frames. ], tot_loss[loss=0.1809, simple_loss=0.2702, pruned_loss=0.04582, over 3191593.92 frames. ], batch size: 124, lr: 2.49e-03, grad_scale: 16.0 2023-05-02 08:58:03,966 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.22 vs. limit=2.0 2023-05-02 08:58:10,898 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=268212.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 08:58:32,293 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([5.4796, 5.4407, 5.3592, 4.9407, 5.0119, 5.3644, 5.2711, 5.0640], device='cuda:1'), covar=tensor([0.0544, 0.0376, 0.0268, 0.0309, 0.0981, 0.0376, 0.0250, 0.0607], device='cuda:1'), in_proj_covar=tensor([0.0317, 0.0475, 0.0369, 0.0375, 0.0370, 0.0429, 0.0252, 0.0443], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002], device='cuda:1') 2023-05-02 08:58:39,883 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.72 vs. limit=2.0 2023-05-02 08:59:12,216 INFO [train.py:904] (1/8) Epoch 27, batch 4350, loss[loss=0.1947, simple_loss=0.2868, pruned_loss=0.05134, over 16384.00 frames. ], tot_loss[loss=0.1835, simple_loss=0.2733, pruned_loss=0.04687, over 3188441.71 frames. ], batch size: 68, lr: 2.49e-03, grad_scale: 16.0 2023-05-02 08:59:27,930 INFO [optim.py:368] (1/8) 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 08:59:43,238 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=268273.0, num_to_drop=1, layers_to_drop={0} 2023-05-02 08:59:45,593 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 27, batch 4400, loss[loss=0.1848, simple_loss=0.2761, pruned_loss=0.04668, over 16842.00 frames. ], tot_loss[loss=0.1852, simple_loss=0.2751, pruned_loss=0.04768, over 3212376.85 frames. ], batch size: 116, lr: 2.49e-03, grad_scale: 8.0 2023-05-02 09:01:14,924 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=268336.0, num_to_drop=1, layers_to_drop={2} 2023-05-02 09:01:37,962 INFO [train.py:904] (1/8) Epoch 27, batch 4450, loss[loss=0.2066, simple_loss=0.3013, pruned_loss=0.05599, over 16774.00 frames. ], tot_loss[loss=0.1885, simple_loss=0.2789, pruned_loss=0.04904, over 3221864.91 frames. ], batch size: 83, lr: 2.49e-03, grad_scale: 8.0 2023-05-02 09:01:55,123 INFO [optim.py:368] (1/8) 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 09:02:50,811 INFO [train.py:904] (1/8) Epoch 27, batch 4500, loss[loss=0.1947, simple_loss=0.2834, pruned_loss=0.05297, over 17114.00 frames. ], tot_loss[loss=0.1895, simple_loss=0.2793, pruned_loss=0.0498, over 3234559.22 frames. ], batch size: 49, lr: 2.49e-03, grad_scale: 8.0 2023-05-02 09:04:03,668 INFO [train.py:904] (1/8) Epoch 27, batch 4550, loss[loss=0.2011, simple_loss=0.2849, pruned_loss=0.05859, over 17268.00 frames. ], tot_loss[loss=0.1909, simple_loss=0.2802, pruned_loss=0.05075, over 3259733.53 frames. ], batch size: 52, lr: 2.49e-03, grad_scale: 8.0 2023-05-02 09:04:20,726 INFO [optim.py:368] (1/8) 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 09:05:15,491 INFO [train.py:904] (1/8) Epoch 27, batch 4600, loss[loss=0.2007, simple_loss=0.2814, pruned_loss=0.06002, over 16991.00 frames. ], tot_loss[loss=0.1918, simple_loss=0.2813, pruned_loss=0.05117, over 3272270.59 frames. ], batch size: 41, lr: 2.49e-03, grad_scale: 8.0 2023-05-02 09:06:20,205 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.7679, 3.7883, 3.9200, 3.6417, 3.8034, 4.2497, 3.8502, 3.5366], device='cuda:1'), covar=tensor([0.2228, 0.2429, 0.2383, 0.2440, 0.2742, 0.1917, 0.1696, 0.2691], device='cuda:1'), in_proj_covar=tensor([0.0431, 0.0636, 0.0697, 0.0519, 0.0691, 0.0728, 0.0544, 0.0693], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-05-02 09:06:23,484 INFO [train.py:904] (1/8) Epoch 27, batch 4650, loss[loss=0.1896, simple_loss=0.2755, pruned_loss=0.05186, over 17011.00 frames. ], tot_loss[loss=0.1919, simple_loss=0.2805, pruned_loss=0.0517, over 3260212.53 frames. ], batch size: 55, lr: 2.49e-03, grad_scale: 8.0 2023-05-02 09:06:35,463 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.4914, 3.2038, 3.5421, 1.8209, 3.6760, 3.6982, 3.0032, 2.8270], device='cuda:1'), covar=tensor([0.0874, 0.0334, 0.0254, 0.1363, 0.0110, 0.0198, 0.0465, 0.0541], device='cuda:1'), in_proj_covar=tensor([0.0150, 0.0112, 0.0102, 0.0140, 0.0087, 0.0132, 0.0131, 0.0132], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-05-02 09:06:40,832 INFO [optim.py:368] (1/8) 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 09:06:45,258 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=268568.0, num_to_drop=1, layers_to_drop={2} 2023-05-02 09:06:54,326 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([5.6862, 5.9277, 5.6971, 5.8214, 5.4595, 5.1456, 5.4317, 6.1273], device='cuda:1'), covar=tensor([0.1209, 0.0769, 0.1035, 0.0772, 0.0781, 0.0677, 0.1117, 0.0757], device='cuda:1'), in_proj_covar=tensor([0.0706, 0.0857, 0.0697, 0.0658, 0.0545, 0.0539, 0.0720, 0.0673], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:1') 2023-05-02 09:07:06,883 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.4505, 3.5100, 2.1765, 4.0821, 2.7442, 3.9817, 2.3698, 2.8225], device='cuda:1'), covar=tensor([0.0349, 0.0428, 0.1836, 0.0146, 0.0891, 0.0506, 0.1506, 0.0870], device='cuda:1'), in_proj_covar=tensor([0.0177, 0.0182, 0.0197, 0.0173, 0.0180, 0.0222, 0.0203, 0.0184], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-05-02 09:07:32,859 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 27, batch 4700, loss[loss=0.1753, simple_loss=0.2656, pruned_loss=0.04246, over 16459.00 frames. ], tot_loss[loss=0.1896, simple_loss=0.2778, pruned_loss=0.05068, over 3252603.27 frames. ], batch size: 68, lr: 2.49e-03, grad_scale: 8.0 2023-05-02 09:08:15,800 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=268631.0, num_to_drop=1, layers_to_drop={1} 2023-05-02 09:08:38,429 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=2.68 vs. limit=5.0 2023-05-02 09:08:45,864 INFO [train.py:904] (1/8) Epoch 27, batch 4750, loss[loss=0.1686, simple_loss=0.2568, pruned_loss=0.04018, over 17056.00 frames. ], tot_loss[loss=0.1862, simple_loss=0.2745, pruned_loss=0.0489, over 3229657.53 frames. ], batch size: 53, lr: 2.49e-03, grad_scale: 8.0 2023-05-02 09:08:59,589 INFO [zipformer.py:625] (1/8) 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] (1/8) 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 09:09:33,074 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.4414, 3.3762, 3.8667, 1.8220, 3.9645, 3.9967, 3.0473, 2.8500], device='cuda:1'), covar=tensor([0.0924, 0.0323, 0.0196, 0.1371, 0.0090, 0.0159, 0.0445, 0.0561], device='cuda:1'), in_proj_covar=tensor([0.0150, 0.0112, 0.0102, 0.0141, 0.0087, 0.0132, 0.0131, 0.0132], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-05-02 09:09:59,302 INFO [train.py:904] (1/8) Epoch 27, batch 4800, loss[loss=0.1688, simple_loss=0.2647, pruned_loss=0.03643, over 16736.00 frames. ], tot_loss[loss=0.1819, simple_loss=0.2707, pruned_loss=0.04656, over 3238783.59 frames. ], batch size: 124, lr: 2.49e-03, grad_scale: 8.0 2023-05-02 09:11:14,170 INFO [train.py:904] (1/8) Epoch 27, batch 4850, loss[loss=0.1699, simple_loss=0.2525, pruned_loss=0.04366, over 17034.00 frames. ], tot_loss[loss=0.1813, simple_loss=0.271, pruned_loss=0.04584, over 3202773.42 frames. ], batch size: 53, lr: 2.49e-03, grad_scale: 8.0 2023-05-02 09:11:23,088 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.10 vs. limit=2.0 2023-05-02 09:11:31,502 INFO [optim.py:368] (1/8) Clipping_scale=2.0, grad-norm quartiles 1.342e+02 1.812e+02 2.075e+02 2.444e+02 6.308e+02, threshold=4.151e+02, percent-clipped=1.0 2023-05-02 09:12:01,062 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.8766, 2.1824, 2.2332, 3.3776, 2.0987, 2.3680, 2.2352, 2.2972], device='cuda:1'), covar=tensor([0.1616, 0.3586, 0.3075, 0.0700, 0.4221, 0.2798, 0.4051, 0.3189], device='cuda:1'), in_proj_covar=tensor([0.0419, 0.0471, 0.0382, 0.0335, 0.0444, 0.0538, 0.0441, 0.0551], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-02 09:12:27,704 INFO [train.py:904] (1/8) Epoch 27, batch 4900, loss[loss=0.1585, simple_loss=0.2478, pruned_loss=0.03458, over 17096.00 frames. ], tot_loss[loss=0.1801, simple_loss=0.2704, pruned_loss=0.04486, over 3186880.72 frames. ], batch size: 49, lr: 2.49e-03, grad_scale: 8.0 2023-05-02 09:13:14,678 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.1086, 2.0906, 1.7884, 1.8900, 2.3269, 2.0701, 1.8752, 2.4054], device='cuda:1'), covar=tensor([0.0232, 0.0466, 0.0613, 0.0521, 0.0291, 0.0391, 0.0241, 0.0330], device='cuda:1'), in_proj_covar=tensor([0.0228, 0.0242, 0.0232, 0.0232, 0.0244, 0.0242, 0.0241, 0.0241], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-02 09:13:37,967 INFO [train.py:904] (1/8) Epoch 27, batch 4950, loss[loss=0.1797, simple_loss=0.2797, pruned_loss=0.03989, over 16713.00 frames. ], tot_loss[loss=0.1786, simple_loss=0.2694, pruned_loss=0.04383, over 3188308.74 frames. ], batch size: 134, lr: 2.49e-03, grad_scale: 8.0 2023-05-02 09:13:49,277 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.68 vs. limit=2.0 2023-05-02 09:13:54,418 INFO [optim.py:368] (1/8) 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 09:13:58,963 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=268868.0, num_to_drop=1, layers_to_drop={1} 2023-05-02 09:14:36,308 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=268894.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 09:14:50,661 INFO [train.py:904] (1/8) Epoch 27, batch 5000, loss[loss=0.2099, simple_loss=0.2945, pruned_loss=0.06265, over 12051.00 frames. ], tot_loss[loss=0.18, simple_loss=0.2714, pruned_loss=0.04431, over 3191394.81 frames. ], batch size: 246, lr: 2.49e-03, grad_scale: 8.0 2023-05-02 09:15:09,314 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=268916.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 09:15:29,367 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=268931.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 09:15:41,578 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.49 vs. limit=2.0 2023-05-02 09:16:00,154 INFO [train.py:904] (1/8) Epoch 27, batch 5050, loss[loss=0.162, simple_loss=0.2582, pruned_loss=0.03294, over 16546.00 frames. ], tot_loss[loss=0.1803, simple_loss=0.2724, pruned_loss=0.0441, over 3218468.64 frames. ], batch size: 75, lr: 2.49e-03, grad_scale: 8.0 2023-05-02 09:16:03,743 INFO [zipformer.py:625] (1/8) 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] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=268957.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 09:16:17,238 INFO [optim.py:368] (1/8) 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 09:16:37,073 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=268979.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 09:16:54,173 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 27, batch 5100, loss[loss=0.157, simple_loss=0.2517, pruned_loss=0.03118, over 16468.00 frames. ], tot_loss[loss=0.179, simple_loss=0.271, pruned_loss=0.04354, over 3217012.66 frames. ], batch size: 75, lr: 2.49e-03, grad_scale: 8.0 2023-05-02 09:17:44,602 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.46 vs. limit=2.0 2023-05-02 09:18:18,284 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([1.7238, 2.4387, 2.1883, 3.2422, 1.8527, 3.5531, 1.4717, 2.7458], device='cuda:1'), covar=tensor([0.1450, 0.0810, 0.1342, 0.0151, 0.0097, 0.0349, 0.1856, 0.0834], device='cuda:1'), in_proj_covar=tensor([0.0171, 0.0179, 0.0199, 0.0199, 0.0206, 0.0217, 0.0208, 0.0197], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-05-02 09:18:22,279 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 27, batch 5150, loss[loss=0.1929, simple_loss=0.2937, pruned_loss=0.04608, over 15478.00 frames. ], tot_loss[loss=0.1781, simple_loss=0.2708, pruned_loss=0.04274, over 3205112.77 frames. ], batch size: 190, lr: 2.49e-03, grad_scale: 8.0 2023-05-02 09:18:41,483 INFO [optim.py:368] (1/8) 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 09:19:00,918 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([5.0200, 4.8272, 5.0521, 5.2665, 5.4703, 4.8744, 5.4323, 5.4347], device='cuda:1'), covar=tensor([0.2097, 0.1465, 0.2009, 0.0830, 0.0571, 0.0926, 0.0651, 0.0750], device='cuda:1'), in_proj_covar=tensor([0.0661, 0.0812, 0.0943, 0.0828, 0.0626, 0.0655, 0.0683, 0.0796], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-02 09:19:36,079 INFO [train.py:904] (1/8) Epoch 27, batch 5200, loss[loss=0.1655, simple_loss=0.252, pruned_loss=0.03949, over 16481.00 frames. ], tot_loss[loss=0.1765, simple_loss=0.2686, pruned_loss=0.04218, over 3216517.49 frames. ], batch size: 75, lr: 2.49e-03, grad_scale: 8.0 2023-05-02 09:20:19,496 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([5.0398, 4.9934, 4.9130, 4.1274, 4.9686, 2.0040, 4.6729, 4.5931], device='cuda:1'), covar=tensor([0.0113, 0.0116, 0.0186, 0.0509, 0.0113, 0.2667, 0.0161, 0.0260], device='cuda:1'), in_proj_covar=tensor([0.0177, 0.0173, 0.0209, 0.0184, 0.0186, 0.0216, 0.0199, 0.0179], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-02 09:20:41,466 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([1.6945, 2.5103, 2.4676, 3.5903, 2.3670, 3.8437, 1.5099, 2.7661], device='cuda:1'), covar=tensor([0.1337, 0.0798, 0.1157, 0.0149, 0.0125, 0.0377, 0.1657, 0.0832], device='cuda:1'), in_proj_covar=tensor([0.0171, 0.0179, 0.0199, 0.0200, 0.0206, 0.0217, 0.0208, 0.0198], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-05-02 09:20:47,204 INFO [train.py:904] (1/8) Epoch 27, batch 5250, loss[loss=0.1582, simple_loss=0.2415, pruned_loss=0.03743, over 17182.00 frames. ], tot_loss[loss=0.1749, simple_loss=0.2661, pruned_loss=0.04183, over 3232136.73 frames. ], batch size: 46, lr: 2.49e-03, grad_scale: 8.0 2023-05-02 09:21:04,386 INFO [optim.py:368] (1/8) Clipping_scale=2.0, grad-norm quartiles 1.499e+02 1.922e+02 2.287e+02 2.667e+02 4.356e+02, threshold=4.574e+02, percent-clipped=2.0 2023-05-02 09:22:00,556 INFO [train.py:904] (1/8) Epoch 27, batch 5300, loss[loss=0.1629, simple_loss=0.2454, pruned_loss=0.0402, over 16861.00 frames. ], tot_loss[loss=0.1721, simple_loss=0.2626, pruned_loss=0.0408, over 3228674.84 frames. ], batch size: 109, lr: 2.49e-03, grad_scale: 8.0 2023-05-02 09:22:55,459 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=269241.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 09:23:08,329 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 27, batch 5350, loss[loss=0.1785, simple_loss=0.2561, pruned_loss=0.05044, over 12065.00 frames. ], tot_loss[loss=0.1707, simple_loss=0.2609, pruned_loss=0.04025, over 3233117.66 frames. ], batch size: 246, lr: 2.49e-03, grad_scale: 8.0 2023-05-02 09:23:18,810 INFO [zipformer.py:625] (1/8) 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] (1/8) 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 09:23:32,110 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.66 vs. limit=2.0 2023-05-02 09:24:01,741 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([1.6729, 2.6380, 2.4172, 4.4206, 3.0135, 3.9485, 1.5374, 2.7684], device='cuda:1'), covar=tensor([0.1429, 0.0831, 0.1349, 0.0136, 0.0240, 0.0385, 0.1751, 0.0912], device='cuda:1'), in_proj_covar=tensor([0.0171, 0.0180, 0.0200, 0.0200, 0.0206, 0.0218, 0.0209, 0.0198], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-05-02 09:24:24,490 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 27, batch 5400, loss[loss=0.162, simple_loss=0.2564, pruned_loss=0.0338, over 16757.00 frames. ], tot_loss[loss=0.172, simple_loss=0.2628, pruned_loss=0.04056, over 3224992.51 frames. ], batch size: 83, lr: 2.49e-03, grad_scale: 8.0 2023-05-02 09:24:28,350 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=269305.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 09:24:47,796 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.29 vs. limit=2.0 2023-05-02 09:24:53,307 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.4825, 3.3277, 3.7701, 1.8057, 3.9194, 3.9065, 2.9967, 2.9014], device='cuda:1'), covar=tensor([0.0882, 0.0321, 0.0192, 0.1435, 0.0075, 0.0155, 0.0464, 0.0522], device='cuda:1'), in_proj_covar=tensor([0.0151, 0.0112, 0.0102, 0.0141, 0.0087, 0.0131, 0.0131, 0.0132], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-05-02 09:25:31,282 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=269347.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 09:25:40,781 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 27, batch 5450, loss[loss=0.2166, simple_loss=0.307, pruned_loss=0.06314, over 16213.00 frames. ], tot_loss[loss=0.1757, simple_loss=0.2665, pruned_loss=0.04241, over 3198795.76 frames. ], batch size: 165, lr: 2.49e-03, grad_scale: 8.0 2023-05-02 09:25:58,794 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.7345, 4.0545, 3.1435, 2.4738, 2.8217, 2.6914, 4.5309, 3.5947], device='cuda:1'), covar=tensor([0.3056, 0.0653, 0.1755, 0.2822, 0.2781, 0.2020, 0.0419, 0.1297], device='cuda:1'), in_proj_covar=tensor([0.0334, 0.0276, 0.0313, 0.0326, 0.0305, 0.0275, 0.0305, 0.0351], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:1') 2023-05-02 09:26:01,148 INFO [optim.py:368] (1/8) 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] (1/8) Epoch 27, batch 5500, loss[loss=0.1954, simple_loss=0.292, pruned_loss=0.04939, over 16893.00 frames. ], tot_loss[loss=0.1834, simple_loss=0.2736, pruned_loss=0.0466, over 3155291.63 frames. ], batch size: 96, lr: 2.49e-03, grad_scale: 8.0 2023-05-02 09:27:16,701 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=269413.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 09:28:00,145 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 27, batch 5550, loss[loss=0.2649, simple_loss=0.3266, pruned_loss=0.1016, over 10963.00 frames. ], tot_loss[loss=0.1912, simple_loss=0.2805, pruned_loss=0.05099, over 3155574.14 frames. ], batch size: 248, lr: 2.49e-03, grad_scale: 8.0 2023-05-02 09:28:38,496 INFO [optim.py:368] (1/8) 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 09:29:37,442 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=269502.0, num_to_drop=1, layers_to_drop={2} 2023-05-02 09:29:38,093 INFO [train.py:904] (1/8) Epoch 27, batch 5600, loss[loss=0.1903, simple_loss=0.2829, pruned_loss=0.04885, over 16783.00 frames. ], tot_loss[loss=0.1969, simple_loss=0.2847, pruned_loss=0.05456, over 3135160.59 frames. ], batch size: 124, lr: 2.49e-03, grad_scale: 8.0 2023-05-02 09:30:56,804 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 27, batch 5650, loss[loss=0.1959, simple_loss=0.2868, pruned_loss=0.05251, over 16670.00 frames. ], tot_loss[loss=0.2016, simple_loss=0.2885, pruned_loss=0.05736, over 3103057.78 frames. ], batch size: 76, lr: 2.49e-03, grad_scale: 8.0 2023-05-02 09:31:20,024 INFO [optim.py:368] (1/8) Clipping_scale=2.0, grad-norm quartiles 2.426e+02 3.478e+02 4.295e+02 5.160e+02 1.255e+03, threshold=8.591e+02, percent-clipped=5.0 2023-05-02 09:31:36,981 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.6081, 4.8834, 4.4582, 4.7201, 4.4460, 4.3318, 4.4403, 4.8954], device='cuda:1'), covar=tensor([0.2145, 0.1395, 0.2434, 0.1459, 0.1615, 0.2207, 0.2349, 0.1814], device='cuda:1'), in_proj_covar=tensor([0.0702, 0.0853, 0.0695, 0.0655, 0.0540, 0.0536, 0.0717, 0.0669], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:1') 2023-05-02 09:32:11,862 INFO [zipformer.py:625] (1/8) 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] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=269598.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 09:32:20,977 INFO [train.py:904] (1/8) Epoch 27, batch 5700, loss[loss=0.2071, simple_loss=0.2928, pruned_loss=0.0607, over 16737.00 frames. ], tot_loss[loss=0.2029, simple_loss=0.2894, pruned_loss=0.05826, over 3114458.83 frames. ], batch size: 57, lr: 2.49e-03, grad_scale: 8.0 2023-05-02 09:33:31,920 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=269647.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 09:33:41,283 INFO [train.py:904] (1/8) Epoch 27, batch 5750, loss[loss=0.2008, simple_loss=0.2856, pruned_loss=0.05804, over 15360.00 frames. ], tot_loss[loss=0.2079, simple_loss=0.2933, pruned_loss=0.06123, over 3059809.16 frames. ], batch size: 190, lr: 2.49e-03, grad_scale: 8.0 2023-05-02 09:33:59,182 INFO [optim.py:368] (1/8) 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 09:34:49,727 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 27, batch 5800, loss[loss=0.1925, simple_loss=0.284, pruned_loss=0.0505, over 15381.00 frames. ], tot_loss[loss=0.2073, simple_loss=0.2934, pruned_loss=0.06061, over 3049572.96 frames. ], batch size: 190, lr: 2.49e-03, grad_scale: 4.0 2023-05-02 09:35:11,014 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=269708.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 09:36:21,260 INFO [train.py:904] (1/8) Epoch 27, batch 5850, loss[loss=0.2272, simple_loss=0.2924, pruned_loss=0.08103, over 11631.00 frames. ], tot_loss[loss=0.2043, simple_loss=0.2911, pruned_loss=0.05878, over 3069950.12 frames. ], batch size: 247, lr: 2.49e-03, grad_scale: 4.0 2023-05-02 09:36:40,939 INFO [optim.py:368] (1/8) 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 09:36:48,097 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.19 vs. limit=2.0 2023-05-02 09:36:52,895 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=269773.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 09:37:34,883 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=269797.0, num_to_drop=1, layers_to_drop={3} 2023-05-02 09:37:45,792 INFO [train.py:904] (1/8) Epoch 27, batch 5900, loss[loss=0.1808, simple_loss=0.2679, pruned_loss=0.04688, over 15493.00 frames. ], tot_loss[loss=0.2037, simple_loss=0.2907, pruned_loss=0.05838, over 3088476.31 frames. ], batch size: 190, lr: 2.49e-03, grad_scale: 4.0 2023-05-02 09:38:19,327 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.61 vs. limit=2.0 2023-05-02 09:38:38,108 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=269834.0, num_to_drop=1, layers_to_drop={1} 2023-05-02 09:38:40,744 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=269836.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 09:39:06,826 INFO [train.py:904] (1/8) Epoch 27, batch 5950, loss[loss=0.1993, simple_loss=0.2949, pruned_loss=0.05187, over 16786.00 frames. ], tot_loss[loss=0.203, simple_loss=0.2914, pruned_loss=0.05728, over 3092239.06 frames. ], batch size: 83, lr: 2.49e-03, grad_scale: 4.0 2023-05-02 09:39:27,599 INFO [optim.py:368] (1/8) 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 09:40:02,616 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.4416, 4.2745, 4.4809, 4.6321, 4.7882, 4.3538, 4.7701, 4.7915], device='cuda:1'), covar=tensor([0.1968, 0.1382, 0.1540, 0.0717, 0.0593, 0.1050, 0.0693, 0.0724], device='cuda:1'), in_proj_covar=tensor([0.0664, 0.0815, 0.0946, 0.0824, 0.0627, 0.0659, 0.0685, 0.0799], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-02 09:40:10,011 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=4.25 vs. limit=5.0 2023-05-02 09:40:17,027 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=269897.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 09:40:17,104 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=269897.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 09:40:24,921 INFO [train.py:904] (1/8) Epoch 27, batch 6000, loss[loss=0.1655, simple_loss=0.2558, pruned_loss=0.03761, over 16699.00 frames. ], tot_loss[loss=0.2023, simple_loss=0.2908, pruned_loss=0.05686, over 3102735.00 frames. ], batch size: 83, lr: 2.49e-03, grad_scale: 8.0 2023-05-02 09:40:24,922 INFO [train.py:929] (1/8) Computing validation loss 2023-05-02 09:40:35,105 INFO [train.py:938] (1/8) Epoch 27, validation: loss=0.148, simple_loss=0.2603, pruned_loss=0.01783, over 944034.00 frames. 2023-05-02 09:40:35,106 INFO [train.py:939] (1/8) Maximum memory allocated so far is 18145MB 2023-05-02 09:41:37,530 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=269945.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 09:41:52,114 INFO [train.py:904] (1/8) Epoch 27, batch 6050, loss[loss=0.197, simple_loss=0.2923, pruned_loss=0.05089, over 16872.00 frames. ], tot_loss[loss=0.2005, simple_loss=0.2895, pruned_loss=0.05579, over 3128327.47 frames. ], batch size: 116, lr: 2.49e-03, grad_scale: 8.0 2023-05-02 09:42:12,241 INFO [optim.py:368] (1/8) 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 09:43:13,684 INFO [train.py:904] (1/8) Epoch 27, batch 6100, loss[loss=0.1734, simple_loss=0.2592, pruned_loss=0.04383, over 16603.00 frames. ], tot_loss[loss=0.1996, simple_loss=0.289, pruned_loss=0.05507, over 3126886.50 frames. ], batch size: 57, lr: 2.49e-03, grad_scale: 8.0 2023-05-02 09:43:22,359 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=270008.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 09:44:33,192 INFO [train.py:904] (1/8) Epoch 27, batch 6150, loss[loss=0.1859, simple_loss=0.2654, pruned_loss=0.05324, over 16610.00 frames. ], tot_loss[loss=0.1982, simple_loss=0.2871, pruned_loss=0.0547, over 3135044.43 frames. ], batch size: 62, lr: 2.48e-03, grad_scale: 8.0 2023-05-02 09:44:37,804 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=270056.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 09:44:53,893 INFO [optim.py:368] (1/8) 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 09:45:15,807 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([1.7021, 2.9058, 2.8264, 5.0964, 3.9177, 4.3540, 1.7794, 3.1125], device='cuda:1'), covar=tensor([0.1441, 0.0819, 0.1246, 0.0165, 0.0417, 0.0433, 0.1667, 0.0886], device='cuda:1'), in_proj_covar=tensor([0.0172, 0.0180, 0.0201, 0.0201, 0.0208, 0.0219, 0.0209, 0.0198], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-05-02 09:45:42,289 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=270097.0, num_to_drop=1, layers_to_drop={2} 2023-05-02 09:45:51,123 INFO [train.py:904] (1/8) Epoch 27, batch 6200, loss[loss=0.1934, simple_loss=0.2953, pruned_loss=0.04573, over 16753.00 frames. ], tot_loss[loss=0.1972, simple_loss=0.2854, pruned_loss=0.05454, over 3128562.91 frames. ], batch size: 83, lr: 2.48e-03, grad_scale: 8.0 2023-05-02 09:46:33,640 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=270129.0, num_to_drop=1, layers_to_drop={2} 2023-05-02 09:46:50,107 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([1.7621, 1.8309, 1.6760, 1.4963, 1.9642, 1.6140, 1.6317, 1.9909], device='cuda:1'), covar=tensor([0.0208, 0.0311, 0.0443, 0.0393, 0.0228, 0.0291, 0.0179, 0.0249], device='cuda:1'), in_proj_covar=tensor([0.0226, 0.0240, 0.0230, 0.0231, 0.0242, 0.0240, 0.0239, 0.0239], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-02 09:46:57,436 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 27, batch 6250, loss[loss=0.2312, simple_loss=0.3001, pruned_loss=0.08116, over 11827.00 frames. ], tot_loss[loss=0.1966, simple_loss=0.2848, pruned_loss=0.05418, over 3123268.69 frames. ], batch size: 248, lr: 2.48e-03, grad_scale: 8.0 2023-05-02 09:47:29,202 INFO [optim.py:368] (1/8) 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 09:47:37,782 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.3739, 4.3511, 4.2603, 3.4292, 4.2901, 1.7393, 4.0532, 3.7871], device='cuda:1'), covar=tensor([0.0153, 0.0137, 0.0221, 0.0348, 0.0116, 0.2967, 0.0159, 0.0304], device='cuda:1'), in_proj_covar=tensor([0.0177, 0.0173, 0.0210, 0.0185, 0.0186, 0.0215, 0.0199, 0.0178], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-02 09:48:08,658 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=270192.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 09:48:25,153 INFO [train.py:904] (1/8) Epoch 27, batch 6300, loss[loss=0.1822, simple_loss=0.2704, pruned_loss=0.04705, over 17000.00 frames. ], tot_loss[loss=0.1952, simple_loss=0.2839, pruned_loss=0.05321, over 3140225.06 frames. ], batch size: 55, lr: 2.48e-03, grad_scale: 8.0 2023-05-02 09:48:29,483 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.5853, 3.4523, 3.9829, 1.9751, 4.1506, 4.1311, 3.1050, 3.0383], device='cuda:1'), covar=tensor([0.0852, 0.0312, 0.0220, 0.1272, 0.0073, 0.0174, 0.0465, 0.0494], device='cuda:1'), in_proj_covar=tensor([0.0150, 0.0111, 0.0102, 0.0140, 0.0086, 0.0132, 0.0131, 0.0132], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-05-02 09:48:35,698 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.5667, 3.4897, 4.0255, 2.1089, 4.1466, 4.1356, 3.0973, 3.0399], device='cuda:1'), covar=tensor([0.0853, 0.0295, 0.0184, 0.1156, 0.0079, 0.0179, 0.0464, 0.0476], device='cuda:1'), in_proj_covar=tensor([0.0150, 0.0111, 0.0102, 0.0140, 0.0086, 0.0132, 0.0131, 0.0132], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-05-02 09:49:45,986 INFO [train.py:904] (1/8) Epoch 27, batch 6350, loss[loss=0.246, simple_loss=0.3075, pruned_loss=0.0922, over 11652.00 frames. ], tot_loss[loss=0.1971, simple_loss=0.2849, pruned_loss=0.05463, over 3117602.60 frames. ], batch size: 248, lr: 2.48e-03, grad_scale: 8.0 2023-05-02 09:50:05,608 INFO [optim.py:368] (1/8) 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 09:50:28,094 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 27, batch 6400, loss[loss=0.1902, simple_loss=0.2737, pruned_loss=0.05337, over 16623.00 frames. ], tot_loss[loss=0.198, simple_loss=0.2851, pruned_loss=0.05544, over 3124263.16 frames. ], batch size: 76, lr: 2.48e-03, grad_scale: 8.0 2023-05-02 09:52:01,057 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=270341.0, num_to_drop=1, layers_to_drop={1} 2023-05-02 09:52:19,089 INFO [train.py:904] (1/8) Epoch 27, batch 6450, loss[loss=0.1771, simple_loss=0.2742, pruned_loss=0.04001, over 16885.00 frames. ], tot_loss[loss=0.1973, simple_loss=0.2849, pruned_loss=0.05486, over 3126822.49 frames. ], batch size: 96, lr: 2.48e-03, grad_scale: 8.0 2023-05-02 09:52:39,047 INFO [optim.py:368] (1/8) 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 09:53:12,953 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.3043, 4.4403, 4.2333, 3.9385, 3.7830, 4.3654, 4.0632, 3.9958], device='cuda:1'), covar=tensor([0.0739, 0.0682, 0.0393, 0.0418, 0.1072, 0.0559, 0.0859, 0.0777], device='cuda:1'), in_proj_covar=tensor([0.0304, 0.0459, 0.0356, 0.0360, 0.0356, 0.0413, 0.0244, 0.0427], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002], device='cuda:1') 2023-05-02 09:53:18,541 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.29 vs. limit=2.0 2023-05-02 09:53:26,991 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 27, batch 6500, loss[loss=0.1849, simple_loss=0.2735, pruned_loss=0.04817, over 16450.00 frames. ], tot_loss[loss=0.1957, simple_loss=0.2832, pruned_loss=0.05407, over 3142242.24 frames. ], batch size: 68, lr: 2.48e-03, grad_scale: 8.0 2023-05-02 09:53:59,369 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.0167, 3.5861, 3.5536, 2.2001, 3.3156, 3.5484, 3.3550, 1.9581], device='cuda:1'), covar=tensor([0.0658, 0.0066, 0.0083, 0.0527, 0.0125, 0.0142, 0.0118, 0.0575], device='cuda:1'), in_proj_covar=tensor([0.0139, 0.0090, 0.0091, 0.0137, 0.0102, 0.0116, 0.0099, 0.0131], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0003, 0.0004], device='cuda:1') 2023-05-02 09:54:17,103 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=270429.0, num_to_drop=1, layers_to_drop={2} 2023-05-02 09:54:21,258 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([5.1681, 5.7345, 5.9015, 5.6098, 5.7356, 6.2058, 5.6384, 5.3808], device='cuda:1'), covar=tensor([0.0842, 0.1784, 0.2220, 0.1711, 0.2103, 0.0908, 0.1534, 0.2189], device='cuda:1'), in_proj_covar=tensor([0.0426, 0.0630, 0.0690, 0.0512, 0.0682, 0.0719, 0.0540, 0.0689], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-05-02 09:54:33,539 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=270439.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 09:54:43,242 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.20 vs. limit=2.0 2023-05-02 09:54:58,115 INFO [train.py:904] (1/8) Epoch 27, batch 6550, loss[loss=0.2012, simple_loss=0.3042, pruned_loss=0.04914, over 15318.00 frames. ], tot_loss[loss=0.1984, simple_loss=0.2861, pruned_loss=0.05531, over 3117274.78 frames. ], batch size: 190, lr: 2.48e-03, grad_scale: 8.0 2023-05-02 09:55:02,925 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.5757, 2.4984, 1.8976, 2.6533, 2.0640, 2.7439, 2.1078, 2.3634], device='cuda:1'), covar=tensor([0.0322, 0.0380, 0.1222, 0.0274, 0.0692, 0.0499, 0.1241, 0.0621], device='cuda:1'), in_proj_covar=tensor([0.0175, 0.0180, 0.0195, 0.0171, 0.0179, 0.0219, 0.0202, 0.0183], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-05-02 09:55:04,723 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=270457.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 09:55:17,871 INFO [optim.py:368] (1/8) 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] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=270477.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 09:55:57,735 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=270492.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 09:55:59,535 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.9771, 4.0462, 4.3192, 4.2887, 4.3013, 4.0626, 4.0537, 4.0671], device='cuda:1'), covar=tensor([0.0382, 0.0649, 0.0409, 0.0417, 0.0494, 0.0482, 0.0845, 0.0494], device='cuda:1'), in_proj_covar=tensor([0.0425, 0.0476, 0.0462, 0.0423, 0.0510, 0.0487, 0.0560, 0.0389], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:1') 2023-05-02 09:56:09,970 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 27, batch 6600, loss[loss=0.2061, simple_loss=0.2947, pruned_loss=0.0588, over 16701.00 frames. ], tot_loss[loss=0.2005, simple_loss=0.2888, pruned_loss=0.05616, over 3099126.36 frames. ], batch size: 124, lr: 2.48e-03, grad_scale: 8.0 2023-05-02 09:56:30,672 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.0424, 2.3171, 1.9042, 2.2147, 2.7142, 2.3880, 2.6715, 2.9597], device='cuda:1'), covar=tensor([0.0227, 0.0586, 0.0764, 0.0611, 0.0358, 0.0499, 0.0273, 0.0305], device='cuda:1'), in_proj_covar=tensor([0.0224, 0.0239, 0.0229, 0.0231, 0.0241, 0.0239, 0.0238, 0.0238], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-02 09:57:09,701 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 27, batch 6650, loss[loss=0.2512, simple_loss=0.3108, pruned_loss=0.09577, over 11519.00 frames. ], tot_loss[loss=0.2013, simple_loss=0.2889, pruned_loss=0.05692, over 3093949.62 frames. ], batch size: 246, lr: 2.48e-03, grad_scale: 8.0 2023-05-02 09:57:50,323 INFO [optim.py:368] (1/8) 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 09:58:00,606 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.35 vs. limit=2.0 2023-05-02 09:58:46,096 INFO [train.py:904] (1/8) Epoch 27, batch 6700, loss[loss=0.1956, simple_loss=0.2786, pruned_loss=0.05634, over 16283.00 frames. ], tot_loss[loss=0.2009, simple_loss=0.2876, pruned_loss=0.05704, over 3091325.72 frames. ], batch size: 165, lr: 2.48e-03, grad_scale: 8.0 2023-05-02 09:59:01,539 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=270613.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 09:59:26,233 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.82 vs. limit=2.0 2023-05-02 09:59:35,300 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=270636.0, num_to_drop=1, layers_to_drop={1} 2023-05-02 10:00:01,086 INFO [train.py:904] (1/8) Epoch 27, batch 6750, loss[loss=0.1912, simple_loss=0.2788, pruned_loss=0.05177, over 16479.00 frames. ], tot_loss[loss=0.2002, simple_loss=0.2866, pruned_loss=0.05693, over 3100916.44 frames. ], batch size: 146, lr: 2.48e-03, grad_scale: 8.0 2023-05-02 10:00:20,167 INFO [optim.py:368] (1/8) 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 10:00:32,921 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=270674.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 10:00:46,945 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.3286, 3.4785, 3.6275, 3.6034, 3.6174, 3.4606, 3.4682, 3.5324], device='cuda:1'), covar=tensor([0.0437, 0.0800, 0.0495, 0.0464, 0.0543, 0.0638, 0.0868, 0.0575], device='cuda:1'), in_proj_covar=tensor([0.0428, 0.0480, 0.0466, 0.0426, 0.0513, 0.0491, 0.0564, 0.0392], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:1') 2023-05-02 10:01:15,211 INFO [train.py:904] (1/8) Epoch 27, batch 6800, loss[loss=0.2344, simple_loss=0.3137, pruned_loss=0.07757, over 17105.00 frames. ], tot_loss[loss=0.1998, simple_loss=0.2867, pruned_loss=0.05646, over 3125237.21 frames. ], batch size: 47, lr: 2.48e-03, grad_scale: 8.0 2023-05-02 10:01:53,375 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=270726.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 10:02:33,362 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=270752.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 10:02:34,159 INFO [train.py:904] (1/8) Epoch 27, batch 6850, loss[loss=0.2035, simple_loss=0.2997, pruned_loss=0.05365, over 16763.00 frames. ], tot_loss[loss=0.202, simple_loss=0.2884, pruned_loss=0.05775, over 3105288.34 frames. ], batch size: 124, lr: 2.48e-03, grad_scale: 8.0 2023-05-02 10:02:53,221 INFO [optim.py:368] (1/8) 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 10:03:25,313 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=270787.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 10:03:37,581 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=270795.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 10:03:49,486 INFO [train.py:904] (1/8) Epoch 27, batch 6900, loss[loss=0.2251, simple_loss=0.3079, pruned_loss=0.07109, over 16377.00 frames. ], tot_loss[loss=0.2013, simple_loss=0.29, pruned_loss=0.05627, over 3134606.98 frames. ], batch size: 146, lr: 2.48e-03, grad_scale: 8.0 2023-05-02 10:03:59,891 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.71 vs. limit=2.0 2023-05-02 10:05:07,698 INFO [train.py:904] (1/8) Epoch 27, batch 6950, loss[loss=0.1848, simple_loss=0.2699, pruned_loss=0.04987, over 16768.00 frames. ], tot_loss[loss=0.2028, simple_loss=0.2909, pruned_loss=0.05739, over 3130134.85 frames. ], batch size: 83, lr: 2.48e-03, grad_scale: 8.0 2023-05-02 10:05:28,445 INFO [optim.py:368] (1/8) Clipping_scale=2.0, grad-norm quartiles 1.875e+02 2.915e+02 3.513e+02 4.432e+02 8.204e+02, threshold=7.026e+02, percent-clipped=5.0 2023-05-02 10:05:46,658 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.5753, 3.4181, 3.9425, 1.9991, 4.0268, 4.1058, 3.0109, 2.9914], device='cuda:1'), covar=tensor([0.0828, 0.0311, 0.0194, 0.1229, 0.0092, 0.0186, 0.0465, 0.0520], device='cuda:1'), in_proj_covar=tensor([0.0149, 0.0110, 0.0101, 0.0139, 0.0086, 0.0131, 0.0130, 0.0131], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-05-02 10:06:06,094 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.7401, 4.5886, 4.4469, 2.9930, 3.9402, 4.4739, 4.0043, 2.6436], device='cuda:1'), covar=tensor([0.0542, 0.0054, 0.0063, 0.0428, 0.0105, 0.0137, 0.0094, 0.0470], device='cuda:1'), in_proj_covar=tensor([0.0139, 0.0090, 0.0090, 0.0136, 0.0101, 0.0115, 0.0098, 0.0131], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0003, 0.0004], device='cuda:1') 2023-05-02 10:06:23,846 INFO [train.py:904] (1/8) Epoch 27, batch 7000, loss[loss=0.2093, simple_loss=0.3087, pruned_loss=0.05498, over 16814.00 frames. ], tot_loss[loss=0.2018, simple_loss=0.2908, pruned_loss=0.05644, over 3140881.18 frames. ], batch size: 83, lr: 2.48e-03, grad_scale: 8.0 2023-05-02 10:06:45,945 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=270917.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 10:07:15,569 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=270936.0, num_to_drop=1, layers_to_drop={1} 2023-05-02 10:07:40,697 INFO [train.py:904] (1/8) Epoch 27, batch 7050, loss[loss=0.1769, simple_loss=0.2679, pruned_loss=0.04294, over 16491.00 frames. ], tot_loss[loss=0.2019, simple_loss=0.2915, pruned_loss=0.05617, over 3136265.71 frames. ], batch size: 75, lr: 2.48e-03, grad_scale: 4.0 2023-05-02 10:08:00,424 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.2292, 2.9553, 3.2678, 1.6925, 3.3681, 3.4706, 2.7547, 2.6055], device='cuda:1'), covar=tensor([0.0868, 0.0336, 0.0220, 0.1390, 0.0116, 0.0211, 0.0482, 0.0534], device='cuda:1'), in_proj_covar=tensor([0.0150, 0.0110, 0.0102, 0.0139, 0.0086, 0.0131, 0.0130, 0.0131], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-05-02 10:08:01,040 INFO [optim.py:368] (1/8) 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 10:08:05,078 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=270969.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 10:08:18,497 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=270978.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 10:08:27,518 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=270984.0, num_to_drop=1, layers_to_drop={1} 2023-05-02 10:08:58,235 INFO [train.py:904] (1/8) Epoch 27, batch 7100, loss[loss=0.2051, simple_loss=0.2946, pruned_loss=0.05781, over 16454.00 frames. ], tot_loss[loss=0.2022, simple_loss=0.2906, pruned_loss=0.0569, over 3082261.68 frames. ], batch size: 146, lr: 2.48e-03, grad_scale: 4.0 2023-05-02 10:09:59,925 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.8189, 3.8882, 3.9211, 3.7276, 3.8779, 4.2404, 3.8931, 3.6512], device='cuda:1'), covar=tensor([0.1964, 0.2200, 0.2563, 0.2354, 0.2533, 0.1668, 0.1838, 0.2668], device='cuda:1'), in_proj_covar=tensor([0.0426, 0.0634, 0.0695, 0.0515, 0.0684, 0.0724, 0.0544, 0.0693], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-05-02 10:10:09,692 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.0096, 2.2305, 2.2738, 3.5717, 2.1435, 2.5482, 2.3158, 2.3779], device='cuda:1'), covar=tensor([0.1541, 0.3484, 0.3089, 0.0655, 0.4286, 0.2442, 0.3547, 0.3378], device='cuda:1'), in_proj_covar=tensor([0.0417, 0.0469, 0.0381, 0.0332, 0.0443, 0.0537, 0.0441, 0.0548], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-02 10:10:15,251 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 27, batch 7150, loss[loss=0.1841, simple_loss=0.2739, pruned_loss=0.0471, over 16405.00 frames. ], tot_loss[loss=0.2013, simple_loss=0.2887, pruned_loss=0.05696, over 3080084.90 frames. ], batch size: 68, lr: 2.48e-03, grad_scale: 4.0 2023-05-02 10:10:37,965 INFO [optim.py:368] (1/8) 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 10:10:42,974 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.8522, 2.7546, 2.8658, 2.1651, 2.6972, 2.1876, 2.7163, 2.9340], device='cuda:1'), covar=tensor([0.0270, 0.0737, 0.0512, 0.1739, 0.0780, 0.0872, 0.0593, 0.0695], device='cuda:1'), in_proj_covar=tensor([0.0161, 0.0172, 0.0171, 0.0157, 0.0149, 0.0133, 0.0147, 0.0182], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0005], device='cuda:1') 2023-05-02 10:10:59,670 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=271082.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 10:11:19,037 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=271095.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 10:11:28,069 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=271100.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 10:11:31,639 INFO [train.py:904] (1/8) Epoch 27, batch 7200, loss[loss=0.1943, simple_loss=0.2935, pruned_loss=0.04757, over 16686.00 frames. ], tot_loss[loss=0.1982, simple_loss=0.2862, pruned_loss=0.05507, over 3103615.93 frames. ], batch size: 134, lr: 2.48e-03, grad_scale: 8.0 2023-05-02 10:12:02,278 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.3318, 3.4958, 3.8568, 2.2348, 3.2624, 2.2593, 3.8151, 3.8102], device='cuda:1'), covar=tensor([0.0220, 0.0816, 0.0503, 0.2114, 0.0795, 0.1054, 0.0533, 0.0911], device='cuda:1'), in_proj_covar=tensor([0.0161, 0.0172, 0.0171, 0.0157, 0.0149, 0.0133, 0.0147, 0.0182], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0005], device='cuda:1') 2023-05-02 10:12:37,362 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 27, batch 7250, loss[loss=0.2161, simple_loss=0.291, pruned_loss=0.0706, over 11480.00 frames. ], tot_loss[loss=0.1959, simple_loss=0.2838, pruned_loss=0.05398, over 3094132.97 frames. ], batch size: 248, lr: 2.48e-03, grad_scale: 8.0 2023-05-02 10:13:16,037 INFO [optim.py:368] (1/8) 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 10:13:26,702 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.0851, 2.4459, 2.5401, 1.9340, 2.6574, 2.7632, 2.4221, 2.3687], device='cuda:1'), covar=tensor([0.0720, 0.0299, 0.0248, 0.0992, 0.0139, 0.0278, 0.0509, 0.0493], device='cuda:1'), in_proj_covar=tensor([0.0149, 0.0110, 0.0101, 0.0139, 0.0086, 0.0131, 0.0130, 0.0131], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-05-02 10:13:42,620 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.5338, 3.5202, 3.4736, 2.7018, 3.3436, 2.0956, 3.1553, 2.7522], device='cuda:1'), covar=tensor([0.0177, 0.0150, 0.0184, 0.0249, 0.0108, 0.2492, 0.0160, 0.0288], device='cuda:1'), in_proj_covar=tensor([0.0177, 0.0172, 0.0211, 0.0185, 0.0185, 0.0216, 0.0199, 0.0177], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-02 10:14:06,636 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.6943, 2.9406, 3.2541, 2.0451, 2.7725, 2.0494, 3.2207, 3.3073], device='cuda:1'), covar=tensor([0.0281, 0.0973, 0.0601, 0.2221, 0.0972, 0.1147, 0.0724, 0.0957], device='cuda:1'), in_proj_covar=tensor([0.0161, 0.0172, 0.0171, 0.0157, 0.0149, 0.0133, 0.0147, 0.0182], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0005], device='cuda:1') 2023-05-02 10:14:10,772 INFO [train.py:904] (1/8) Epoch 27, batch 7300, loss[loss=0.2159, simple_loss=0.2841, pruned_loss=0.07386, over 10950.00 frames. ], tot_loss[loss=0.1957, simple_loss=0.2833, pruned_loss=0.05408, over 3079632.38 frames. ], batch size: 246, lr: 2.48e-03, grad_scale: 8.0 2023-05-02 10:15:02,686 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.2522, 3.2851, 2.0250, 3.7181, 2.5288, 3.7006, 2.2628, 2.6788], device='cuda:1'), covar=tensor([0.0356, 0.0447, 0.1824, 0.0189, 0.0902, 0.0561, 0.1574, 0.0869], device='cuda:1'), in_proj_covar=tensor([0.0175, 0.0181, 0.0197, 0.0171, 0.0180, 0.0220, 0.0203, 0.0184], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-05-02 10:15:29,751 INFO [train.py:904] (1/8) Epoch 27, batch 7350, loss[loss=0.1872, simple_loss=0.271, pruned_loss=0.05168, over 16650.00 frames. ], tot_loss[loss=0.1969, simple_loss=0.2842, pruned_loss=0.05482, over 3081080.50 frames. ], batch size: 62, lr: 2.48e-03, grad_scale: 8.0 2023-05-02 10:15:42,302 INFO [zipformer.py:625] (1/8) 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] (1/8) Clipping_scale=2.0, grad-norm quartiles 1.854e+02 2.875e+02 3.303e+02 4.103e+02 7.171e+02, threshold=6.606e+02, percent-clipped=3.0 2023-05-02 10:15:54,298 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=271269.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 10:16:01,220 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=271273.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 10:16:41,030 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.9996, 5.3673, 5.5839, 5.3445, 5.3523, 5.9047, 5.4457, 5.1924], device='cuda:1'), covar=tensor([0.0917, 0.1764, 0.2110, 0.1862, 0.2255, 0.0769, 0.1574, 0.2414], device='cuda:1'), in_proj_covar=tensor([0.0423, 0.0630, 0.0690, 0.0512, 0.0683, 0.0722, 0.0543, 0.0689], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-05-02 10:16:48,723 INFO [train.py:904] (1/8) Epoch 27, batch 7400, loss[loss=0.1855, simple_loss=0.2789, pruned_loss=0.04608, over 16465.00 frames. ], tot_loss[loss=0.1982, simple_loss=0.2851, pruned_loss=0.05564, over 3076811.27 frames. ], batch size: 68, lr: 2.48e-03, grad_scale: 8.0 2023-05-02 10:16:56,588 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.6777, 4.8210, 5.0155, 4.8059, 4.8496, 5.3730, 4.8713, 4.6409], device='cuda:1'), covar=tensor([0.1163, 0.1944, 0.2318, 0.1974, 0.2430, 0.1005, 0.1741, 0.2587], device='cuda:1'), in_proj_covar=tensor([0.0423, 0.0630, 0.0691, 0.0513, 0.0684, 0.0722, 0.0543, 0.0690], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-05-02 10:17:10,654 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=271317.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 10:17:18,302 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=271322.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 10:18:06,477 INFO [train.py:904] (1/8) Epoch 27, batch 7450, loss[loss=0.1946, simple_loss=0.2888, pruned_loss=0.05021, over 16893.00 frames. ], tot_loss[loss=0.1994, simple_loss=0.2862, pruned_loss=0.05627, over 3095628.16 frames. ], batch size: 96, lr: 2.48e-03, grad_scale: 8.0 2023-05-02 10:18:30,891 INFO [optim.py:368] (1/8) 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 10:18:45,200 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=271375.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 10:18:55,950 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=271382.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 10:19:30,450 INFO [train.py:904] (1/8) Epoch 27, batch 7500, loss[loss=0.175, simple_loss=0.273, pruned_loss=0.03847, over 16705.00 frames. ], tot_loss[loss=0.1984, simple_loss=0.2863, pruned_loss=0.05521, over 3113080.93 frames. ], batch size: 89, lr: 2.48e-03, grad_scale: 8.0 2023-05-02 10:20:08,067 INFO [zipformer.py:625] (1/8) 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] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=271430.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 10:20:22,926 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=271436.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 10:20:30,945 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.6083, 4.6047, 4.4142, 3.6473, 4.5225, 1.6973, 4.2454, 3.9847], device='cuda:1'), covar=tensor([0.0111, 0.0099, 0.0206, 0.0363, 0.0090, 0.3077, 0.0128, 0.0322], device='cuda:1'), in_proj_covar=tensor([0.0176, 0.0171, 0.0209, 0.0183, 0.0184, 0.0214, 0.0197, 0.0176], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-02 10:20:49,667 INFO [train.py:904] (1/8) Epoch 27, batch 7550, loss[loss=0.2085, simple_loss=0.2929, pruned_loss=0.06203, over 16712.00 frames. ], tot_loss[loss=0.1982, simple_loss=0.2854, pruned_loss=0.05554, over 3100803.89 frames. ], batch size: 124, lr: 2.48e-03, grad_scale: 8.0 2023-05-02 10:21:11,191 INFO [optim.py:368] (1/8) 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 10:21:36,699 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=4.41 vs. limit=5.0 2023-05-02 10:21:41,018 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=271487.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 10:21:53,355 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.4254, 4.4851, 4.3105, 4.0218, 4.0454, 4.4208, 4.1240, 4.1464], device='cuda:1'), covar=tensor([0.0614, 0.0518, 0.0292, 0.0320, 0.0794, 0.0499, 0.0619, 0.0640], device='cuda:1'), in_proj_covar=tensor([0.0302, 0.0456, 0.0354, 0.0356, 0.0352, 0.0410, 0.0243, 0.0423], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-02 10:22:05,495 INFO [train.py:904] (1/8) Epoch 27, batch 7600, loss[loss=0.2404, simple_loss=0.3064, pruned_loss=0.0872, over 11273.00 frames. ], tot_loss[loss=0.1987, simple_loss=0.2852, pruned_loss=0.05616, over 3102576.30 frames. ], batch size: 246, lr: 2.48e-03, grad_scale: 8.0 2023-05-02 10:22:40,592 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=3.76 vs. limit=5.0 2023-05-02 10:22:42,227 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.55 vs. limit=2.0 2023-05-02 10:23:22,939 INFO [train.py:904] (1/8) Epoch 27, batch 7650, loss[loss=0.1706, simple_loss=0.2629, pruned_loss=0.03912, over 17273.00 frames. ], tot_loss[loss=0.1999, simple_loss=0.2857, pruned_loss=0.05707, over 3068160.28 frames. ], batch size: 52, lr: 2.48e-03, grad_scale: 8.0 2023-05-02 10:23:44,879 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.7308, 3.7500, 3.8426, 2.3113, 3.2884, 2.5939, 4.0564, 4.0325], device='cuda:1'), covar=tensor([0.0226, 0.0845, 0.0578, 0.2140, 0.0867, 0.0992, 0.0573, 0.1021], device='cuda:1'), in_proj_covar=tensor([0.0160, 0.0171, 0.0171, 0.0157, 0.0149, 0.0133, 0.0147, 0.0182], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0005], device='cuda:1') 2023-05-02 10:23:45,490 INFO [optim.py:368] (1/8) 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 10:23:55,692 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=271573.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 10:24:43,700 INFO [train.py:904] (1/8) Epoch 27, batch 7700, loss[loss=0.1744, simple_loss=0.2684, pruned_loss=0.04019, over 15423.00 frames. ], tot_loss[loss=0.2002, simple_loss=0.2857, pruned_loss=0.05738, over 3060030.60 frames. ], batch size: 191, lr: 2.48e-03, grad_scale: 8.0 2023-05-02 10:25:06,081 INFO [zipformer.py:625] (1/8) 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] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=271621.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 10:26:02,416 INFO [train.py:904] (1/8) Epoch 27, batch 7750, loss[loss=0.1946, simple_loss=0.292, pruned_loss=0.04859, over 17185.00 frames. ], tot_loss[loss=0.2005, simple_loss=0.2864, pruned_loss=0.05728, over 3068305.24 frames. ], batch size: 44, lr: 2.48e-03, grad_scale: 8.0 2023-05-02 10:26:24,500 INFO [optim.py:368] (1/8) Clipping_scale=2.0, grad-norm quartiles 1.915e+02 2.896e+02 3.348e+02 4.032e+02 8.871e+02, threshold=6.696e+02, percent-clipped=2.0 2023-05-02 10:27:20,090 INFO [train.py:904] (1/8) Epoch 27, batch 7800, loss[loss=0.1873, simple_loss=0.2775, pruned_loss=0.04852, over 16868.00 frames. ], tot_loss[loss=0.2015, simple_loss=0.2876, pruned_loss=0.05777, over 3086786.14 frames. ], batch size: 116, lr: 2.48e-03, grad_scale: 8.0 2023-05-02 10:28:01,251 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.8014, 3.4914, 4.1103, 2.0166, 4.2209, 4.2662, 3.1540, 3.1514], device='cuda:1'), covar=tensor([0.0787, 0.0348, 0.0209, 0.1267, 0.0079, 0.0176, 0.0429, 0.0482], device='cuda:1'), in_proj_covar=tensor([0.0148, 0.0110, 0.0101, 0.0138, 0.0086, 0.0130, 0.0130, 0.0131], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-05-02 10:28:04,234 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 27, batch 7850, loss[loss=0.2106, simple_loss=0.3055, pruned_loss=0.05782, over 16730.00 frames. ], tot_loss[loss=0.2027, simple_loss=0.2888, pruned_loss=0.05828, over 3063177.06 frames. ], batch size: 89, lr: 2.48e-03, grad_scale: 4.0 2023-05-02 10:28:41,651 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.21 vs. limit=2.0 2023-05-02 10:28:57,992 INFO [optim.py:368] (1/8) 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 10:29:20,925 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=271782.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 10:29:52,537 INFO [train.py:904] (1/8) Epoch 27, batch 7900, loss[loss=0.2116, simple_loss=0.2973, pruned_loss=0.06295, over 16362.00 frames. ], tot_loss[loss=0.2019, simple_loss=0.2881, pruned_loss=0.05783, over 3061965.08 frames. ], batch size: 146, lr: 2.48e-03, grad_scale: 4.0 2023-05-02 10:30:47,942 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.3884, 3.5157, 2.1556, 3.8472, 2.6790, 3.8296, 2.2405, 2.8196], device='cuda:1'), covar=tensor([0.0317, 0.0398, 0.1643, 0.0270, 0.0791, 0.0649, 0.1572, 0.0848], device='cuda:1'), in_proj_covar=tensor([0.0176, 0.0181, 0.0198, 0.0172, 0.0181, 0.0221, 0.0204, 0.0185], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-05-02 10:31:12,163 INFO [train.py:904] (1/8) Epoch 27, batch 7950, loss[loss=0.2095, simple_loss=0.291, pruned_loss=0.06404, over 16395.00 frames. ], tot_loss[loss=0.2037, simple_loss=0.2892, pruned_loss=0.05914, over 3048848.32 frames. ], batch size: 146, lr: 2.48e-03, grad_scale: 4.0 2023-05-02 10:31:15,293 INFO [zipformer.py:625] (1/8) 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] (1/8) Clipping_scale=2.0, grad-norm quartiles 1.948e+02 2.764e+02 3.233e+02 3.778e+02 5.723e+02, threshold=6.466e+02, percent-clipped=0.0 2023-05-02 10:32:29,929 INFO [train.py:904] (1/8) Epoch 27, batch 8000, loss[loss=0.2076, simple_loss=0.3009, pruned_loss=0.05717, over 16200.00 frames. ], tot_loss[loss=0.2042, simple_loss=0.2895, pruned_loss=0.05944, over 3040969.71 frames. ], batch size: 165, lr: 2.48e-03, grad_scale: 8.0 2023-05-02 10:32:48,352 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([5.7842, 6.1074, 5.8027, 5.9371, 5.5195, 5.4088, 5.4649, 6.2569], device='cuda:1'), covar=tensor([0.1254, 0.0794, 0.0984, 0.0824, 0.0801, 0.0696, 0.1351, 0.0817], device='cuda:1'), in_proj_covar=tensor([0.0703, 0.0850, 0.0699, 0.0656, 0.0536, 0.0539, 0.0715, 0.0667], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:1') 2023-05-02 10:32:49,737 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=271915.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 10:32:52,976 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=271917.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 10:33:19,504 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=271934.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 10:33:47,184 INFO [train.py:904] (1/8) Epoch 27, batch 8050, loss[loss=0.2267, simple_loss=0.3007, pruned_loss=0.07632, over 11727.00 frames. ], tot_loss[loss=0.2028, simple_loss=0.2887, pruned_loss=0.05848, over 3064013.29 frames. ], batch size: 248, lr: 2.48e-03, grad_scale: 8.0 2023-05-02 10:34:05,654 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=271965.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 10:34:09,131 INFO [optim.py:368] (1/8) 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 10:34:49,599 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=271995.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 10:35:04,681 INFO [train.py:904] (1/8) Epoch 27, batch 8100, loss[loss=0.1933, simple_loss=0.2753, pruned_loss=0.0556, over 16353.00 frames. ], tot_loss[loss=0.2021, simple_loss=0.2884, pruned_loss=0.05793, over 3068180.33 frames. ], batch size: 35, lr: 2.48e-03, grad_scale: 8.0 2023-05-02 10:35:05,326 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=2.76 vs. limit=5.0 2023-05-02 10:35:42,649 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.14 vs. limit=2.0 2023-05-02 10:35:45,792 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=272031.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 10:36:17,760 INFO [train.py:904] (1/8) Epoch 27, batch 8150, loss[loss=0.1873, simple_loss=0.2745, pruned_loss=0.0501, over 16995.00 frames. ], tot_loss[loss=0.1998, simple_loss=0.2861, pruned_loss=0.05678, over 3087874.70 frames. ], batch size: 55, lr: 2.48e-03, grad_scale: 8.0 2023-05-02 10:36:20,480 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=4.89 vs. limit=5.0 2023-05-02 10:36:39,765 INFO [optim.py:368] (1/8) 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 10:36:57,159 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=272079.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 10:37:01,572 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=272082.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 10:37:07,762 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.8746, 2.7201, 2.8468, 2.1337, 2.7122, 2.1932, 2.7492, 2.9406], device='cuda:1'), covar=tensor([0.0272, 0.0885, 0.0516, 0.1846, 0.0825, 0.0897, 0.0582, 0.0763], device='cuda:1'), in_proj_covar=tensor([0.0160, 0.0170, 0.0170, 0.0156, 0.0148, 0.0133, 0.0146, 0.0181], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0005], device='cuda:1') 2023-05-02 10:37:19,523 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.83 vs. limit=2.0 2023-05-02 10:37:32,132 INFO [train.py:904] (1/8) Epoch 27, batch 8200, loss[loss=0.1784, simple_loss=0.2721, pruned_loss=0.04232, over 16233.00 frames. ], tot_loss[loss=0.1967, simple_loss=0.2826, pruned_loss=0.05543, over 3096164.52 frames. ], batch size: 165, lr: 2.48e-03, grad_scale: 8.0 2023-05-02 10:37:34,580 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.2068, 4.3140, 4.4551, 4.1652, 4.3511, 4.8209, 4.3139, 4.0009], device='cuda:1'), covar=tensor([0.1831, 0.2196, 0.2379, 0.2238, 0.2443, 0.1100, 0.1659, 0.2540], device='cuda:1'), in_proj_covar=tensor([0.0425, 0.0630, 0.0692, 0.0514, 0.0684, 0.0721, 0.0543, 0.0689], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-05-02 10:37:38,483 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.8737, 2.7634, 2.8598, 2.1755, 2.7266, 2.2031, 2.7573, 2.9777], device='cuda:1'), covar=tensor([0.0270, 0.0783, 0.0527, 0.1786, 0.0797, 0.0903, 0.0521, 0.0649], device='cuda:1'), in_proj_covar=tensor([0.0159, 0.0170, 0.0170, 0.0156, 0.0148, 0.0132, 0.0146, 0.0181], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0005], device='cuda:1') 2023-05-02 10:37:48,896 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([1.6279, 2.5311, 2.3467, 3.5750, 2.0980, 3.7882, 1.4116, 2.7894], device='cuda:1'), covar=tensor([0.1654, 0.0839, 0.1385, 0.0241, 0.0152, 0.0451, 0.2063, 0.0869], device='cuda:1'), in_proj_covar=tensor([0.0173, 0.0180, 0.0201, 0.0201, 0.0209, 0.0219, 0.0210, 0.0199], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-05-02 10:37:55,566 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=4.11 vs. limit=5.0 2023-05-02 10:38:12,562 INFO [zipformer.py:625] (1/8) 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] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=272130.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 10:38:53,589 INFO [train.py:904] (1/8) Epoch 27, batch 8250, loss[loss=0.1752, simple_loss=0.2711, pruned_loss=0.0396, over 16887.00 frames. ], tot_loss[loss=0.1931, simple_loss=0.2811, pruned_loss=0.05257, over 3080376.48 frames. ], batch size: 116, lr: 2.48e-03, grad_scale: 8.0 2023-05-02 10:39:14,776 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=3.16 vs. limit=5.0 2023-05-02 10:39:19,036 INFO [optim.py:368] (1/8) 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 10:39:40,172 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=272181.0, num_to_drop=1, layers_to_drop={0} 2023-05-02 10:39:50,771 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=272188.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 10:40:00,275 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.2305, 4.0421, 4.2958, 4.4201, 4.5764, 4.1069, 4.5010, 4.5743], device='cuda:1'), covar=tensor([0.1854, 0.1362, 0.1503, 0.0825, 0.0606, 0.1387, 0.0824, 0.0829], device='cuda:1'), in_proj_covar=tensor([0.0654, 0.0797, 0.0929, 0.0808, 0.0622, 0.0647, 0.0679, 0.0791], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-02 10:40:14,329 INFO [train.py:904] (1/8) Epoch 27, batch 8300, loss[loss=0.1689, simple_loss=0.2679, pruned_loss=0.03492, over 16751.00 frames. ], tot_loss[loss=0.1889, simple_loss=0.2782, pruned_loss=0.04977, over 3053059.96 frames. ], batch size: 124, lr: 2.47e-03, grad_scale: 8.0 2023-05-02 10:40:26,347 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=272210.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 10:40:46,133 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=272223.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 10:41:15,171 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=272242.0, num_to_drop=1, layers_to_drop={3} 2023-05-02 10:41:22,911 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=272247.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 10:41:31,411 INFO [train.py:904] (1/8) Epoch 27, batch 8350, loss[loss=0.1857, simple_loss=0.2806, pruned_loss=0.04541, over 16453.00 frames. ], tot_loss[loss=0.187, simple_loss=0.2779, pruned_loss=0.04802, over 3049473.93 frames. ], batch size: 146, lr: 2.47e-03, grad_scale: 8.0 2023-05-02 10:41:52,015 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.9024, 2.1428, 2.3465, 3.1492, 2.1307, 2.2893, 2.2611, 2.2236], device='cuda:1'), covar=tensor([0.1359, 0.3756, 0.2962, 0.0768, 0.4736, 0.2931, 0.3862, 0.3747], device='cuda:1'), in_proj_covar=tensor([0.0409, 0.0462, 0.0375, 0.0327, 0.0437, 0.0527, 0.0434, 0.0539], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-02 10:41:54,866 INFO [optim.py:368] (1/8) 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 10:42:06,056 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([1.9912, 2.3430, 2.3410, 2.9619, 1.7592, 3.2781, 1.7727, 2.8023], device='cuda:1'), covar=tensor([0.1300, 0.0657, 0.1088, 0.0203, 0.0080, 0.0380, 0.1587, 0.0694], device='cuda:1'), in_proj_covar=tensor([0.0172, 0.0179, 0.0200, 0.0200, 0.0207, 0.0218, 0.0209, 0.0198], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-05-02 10:42:20,366 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=272284.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 10:42:30,777 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 27, batch 8400, loss[loss=0.1648, simple_loss=0.2604, pruned_loss=0.03455, over 15315.00 frames. ], tot_loss[loss=0.1832, simple_loss=0.275, pruned_loss=0.04569, over 3046490.04 frames. ], batch size: 190, lr: 2.47e-03, grad_scale: 8.0 2023-05-02 10:42:59,287 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=272308.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 10:43:41,173 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.45 vs. limit=2.0 2023-05-02 10:44:09,888 INFO [train.py:904] (1/8) Epoch 27, batch 8450, loss[loss=0.1691, simple_loss=0.2578, pruned_loss=0.04022, over 12221.00 frames. ], tot_loss[loss=0.181, simple_loss=0.2734, pruned_loss=0.04427, over 3048652.28 frames. ], batch size: 248, lr: 2.47e-03, grad_scale: 8.0 2023-05-02 10:44:14,057 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.94 vs. limit=2.0 2023-05-02 10:44:34,140 INFO [optim.py:368] (1/8) 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 10:44:40,766 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.9764, 4.9189, 4.7091, 4.0882, 4.8153, 1.8907, 4.5943, 4.5456], device='cuda:1'), covar=tensor([0.0103, 0.0116, 0.0225, 0.0417, 0.0116, 0.2928, 0.0144, 0.0252], device='cuda:1'), in_proj_covar=tensor([0.0176, 0.0171, 0.0209, 0.0182, 0.0184, 0.0214, 0.0197, 0.0174], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-02 10:44:42,763 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=272373.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 10:45:32,409 INFO [train.py:904] (1/8) Epoch 27, batch 8500, loss[loss=0.173, simple_loss=0.2655, pruned_loss=0.0402, over 16408.00 frames. ], tot_loss[loss=0.1771, simple_loss=0.2699, pruned_loss=0.04218, over 3056033.35 frames. ], batch size: 146, lr: 2.47e-03, grad_scale: 8.0 2023-05-02 10:45:55,066 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.2926, 4.5591, 4.8324, 4.7870, 4.7885, 4.5198, 4.2562, 4.4028], device='cuda:1'), covar=tensor([0.0793, 0.1018, 0.0710, 0.0744, 0.0845, 0.0844, 0.1763, 0.0861], device='cuda:1'), in_proj_covar=tensor([0.0423, 0.0477, 0.0462, 0.0424, 0.0508, 0.0486, 0.0559, 0.0390], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:1') 2023-05-02 10:46:15,865 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.52 vs. limit=2.0 2023-05-02 10:46:24,080 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 27, batch 8550, loss[loss=0.1712, simple_loss=0.2532, pruned_loss=0.04463, over 11849.00 frames. ], tot_loss[loss=0.1752, simple_loss=0.2679, pruned_loss=0.04131, over 3046878.82 frames. ], batch size: 247, lr: 2.47e-03, grad_scale: 8.0 2023-05-02 10:47:25,001 INFO [optim.py:368] (1/8) Clipping_scale=2.0, grad-norm quartiles 1.466e+02 2.247e+02 2.591e+02 3.219e+02 6.393e+02, threshold=5.181e+02, percent-clipped=2.0 2023-05-02 10:47:55,108 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=272483.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 10:48:12,587 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.6957, 3.0610, 3.4351, 2.0365, 2.8730, 2.2457, 3.2326, 3.2804], device='cuda:1'), covar=tensor([0.0337, 0.0960, 0.0506, 0.2216, 0.0844, 0.1033, 0.0689, 0.1015], device='cuda:1'), in_proj_covar=tensor([0.0157, 0.0167, 0.0167, 0.0154, 0.0146, 0.0131, 0.0144, 0.0178], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:1') 2023-05-02 10:48:35,726 INFO [train.py:904] (1/8) Epoch 27, batch 8600, loss[loss=0.1893, simple_loss=0.2909, pruned_loss=0.04389, over 15430.00 frames. ], tot_loss[loss=0.1746, simple_loss=0.2683, pruned_loss=0.04041, over 3047127.93 frames. ], batch size: 191, lr: 2.47e-03, grad_scale: 8.0 2023-05-02 10:48:50,698 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=272510.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 10:49:32,318 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.1669, 2.2050, 2.2514, 3.8225, 2.1476, 2.5452, 2.3244, 2.3632], device='cuda:1'), covar=tensor([0.1478, 0.4004, 0.3352, 0.0580, 0.4424, 0.2777, 0.4076, 0.3593], device='cuda:1'), in_proj_covar=tensor([0.0409, 0.0462, 0.0375, 0.0326, 0.0436, 0.0526, 0.0433, 0.0538], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-02 10:49:43,719 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=272537.0, num_to_drop=1, layers_to_drop={3} 2023-05-02 10:50:13,935 INFO [train.py:904] (1/8) Epoch 27, batch 8650, loss[loss=0.1667, simple_loss=0.2555, pruned_loss=0.03894, over 12134.00 frames. ], tot_loss[loss=0.1721, simple_loss=0.2664, pruned_loss=0.03892, over 3041793.63 frames. ], batch size: 247, lr: 2.47e-03, grad_scale: 8.0 2023-05-02 10:50:26,980 INFO [zipformer.py:625] (1/8) 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] (1/8) Clipping_scale=2.0, grad-norm quartiles 1.411e+02 2.181e+02 2.564e+02 3.046e+02 5.835e+02, threshold=5.129e+02, percent-clipped=3.0 2023-05-02 10:51:13,417 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=4.77 vs. limit=5.0 2023-05-02 10:51:15,178 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=272579.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 10:51:36,598 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 27, batch 8700, loss[loss=0.1617, simple_loss=0.2555, pruned_loss=0.03401, over 16165.00 frames. ], tot_loss[loss=0.1699, simple_loss=0.2641, pruned_loss=0.03782, over 3050134.57 frames. ], batch size: 165, lr: 2.47e-03, grad_scale: 8.0 2023-05-02 10:52:01,200 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=272603.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 10:52:21,005 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=272613.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 10:52:36,773 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.6988, 4.7426, 5.0556, 5.0477, 5.0236, 4.7582, 4.7110, 4.6404], device='cuda:1'), covar=tensor([0.0404, 0.0755, 0.0496, 0.0528, 0.0620, 0.0594, 0.1022, 0.0501], device='cuda:1'), in_proj_covar=tensor([0.0422, 0.0475, 0.0461, 0.0424, 0.0508, 0.0486, 0.0557, 0.0390], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:1') 2023-05-02 10:52:45,562 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.2184, 3.6806, 3.5866, 2.3758, 3.3409, 3.6791, 3.5151, 1.9738], device='cuda:1'), covar=tensor([0.0582, 0.0061, 0.0070, 0.0477, 0.0125, 0.0104, 0.0094, 0.0632], device='cuda:1'), in_proj_covar=tensor([0.0135, 0.0087, 0.0088, 0.0133, 0.0099, 0.0112, 0.0095, 0.0128], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:1') 2023-05-02 10:53:06,582 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 27, batch 8750, loss[loss=0.1739, simple_loss=0.2727, pruned_loss=0.03753, over 16216.00 frames. ], tot_loss[loss=0.1688, simple_loss=0.2637, pruned_loss=0.03699, over 3073885.47 frames. ], batch size: 165, lr: 2.47e-03, grad_scale: 4.0 2023-05-02 10:54:15,880 INFO [optim.py:368] (1/8) 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 10:54:28,844 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=272674.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 10:54:54,297 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([1.9171, 2.1945, 2.2872, 2.8543, 1.7271, 3.2124, 1.6908, 2.7696], device='cuda:1'), covar=tensor([0.1338, 0.0831, 0.1185, 0.0173, 0.0092, 0.0344, 0.1724, 0.0753], device='cuda:1'), in_proj_covar=tensor([0.0171, 0.0177, 0.0198, 0.0196, 0.0204, 0.0215, 0.0207, 0.0196], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-05-02 10:55:27,070 INFO [train.py:904] (1/8) Epoch 27, batch 8800, loss[loss=0.1671, simple_loss=0.26, pruned_loss=0.03708, over 15259.00 frames. ], tot_loss[loss=0.1664, simple_loss=0.2617, pruned_loss=0.03553, over 3080467.00 frames. ], batch size: 190, lr: 2.47e-03, grad_scale: 8.0 2023-05-02 10:55:59,422 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.7711, 2.7265, 2.3291, 2.5524, 3.0111, 2.7578, 3.0915, 3.2438], device='cuda:1'), covar=tensor([0.0145, 0.0523, 0.0640, 0.0522, 0.0383, 0.0493, 0.0342, 0.0321], device='cuda:1'), in_proj_covar=tensor([0.0219, 0.0238, 0.0227, 0.0229, 0.0239, 0.0237, 0.0235, 0.0234], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-02 10:56:22,578 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=272729.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 10:57:12,471 INFO [train.py:904] (1/8) Epoch 27, batch 8850, loss[loss=0.1658, simple_loss=0.2752, pruned_loss=0.02822, over 15508.00 frames. ], tot_loss[loss=0.167, simple_loss=0.2636, pruned_loss=0.03521, over 3050747.83 frames. ], batch size: 191, lr: 2.47e-03, grad_scale: 8.0 2023-05-02 10:57:27,286 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.68 vs. limit=2.0 2023-05-02 10:57:46,548 INFO [optim.py:368] (1/8) 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 10:58:17,230 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=272783.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 10:58:31,984 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.44 vs. limit=2.0 2023-05-02 10:58:57,009 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.93 vs. limit=2.0 2023-05-02 10:58:57,763 INFO [train.py:904] (1/8) Epoch 27, batch 8900, loss[loss=0.1575, simple_loss=0.2519, pruned_loss=0.03152, over 12482.00 frames. ], tot_loss[loss=0.1666, simple_loss=0.264, pruned_loss=0.0346, over 3061198.45 frames. ], batch size: 247, lr: 2.47e-03, grad_scale: 8.0 2023-05-02 11:00:01,529 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=272831.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 11:00:20,721 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=272837.0, num_to_drop=1, layers_to_drop={2} 2023-05-02 11:01:00,399 INFO [train.py:904] (1/8) Epoch 27, batch 8950, loss[loss=0.1757, simple_loss=0.2649, pruned_loss=0.04318, over 15330.00 frames. ], tot_loss[loss=0.167, simple_loss=0.2639, pruned_loss=0.03508, over 3075107.61 frames. ], batch size: 190, lr: 2.47e-03, grad_scale: 8.0 2023-05-02 11:01:35,337 INFO [optim.py:368] (1/8) Clipping_scale=2.0, grad-norm quartiles 1.412e+02 1.963e+02 2.469e+02 3.081e+02 5.293e+02, threshold=4.938e+02, percent-clipped=1.0 2023-05-02 11:01:56,761 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=272879.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 11:02:12,418 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=272885.0, num_to_drop=1, layers_to_drop={0} 2023-05-02 11:02:33,719 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.9129, 3.6887, 4.0697, 2.1774, 4.2525, 4.2851, 3.2451, 3.3137], device='cuda:1'), covar=tensor([0.0708, 0.0236, 0.0178, 0.1187, 0.0067, 0.0136, 0.0398, 0.0441], device='cuda:1'), in_proj_covar=tensor([0.0145, 0.0107, 0.0098, 0.0135, 0.0083, 0.0125, 0.0126, 0.0127], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0004, 0.0003], device='cuda:1') 2023-05-02 11:02:49,712 INFO [train.py:904] (1/8) Epoch 27, batch 9000, loss[loss=0.1589, simple_loss=0.2488, pruned_loss=0.03454, over 16327.00 frames. ], tot_loss[loss=0.1646, simple_loss=0.2608, pruned_loss=0.03415, over 3064194.87 frames. ], batch size: 146, lr: 2.47e-03, grad_scale: 8.0 2023-05-02 11:02:49,713 INFO [train.py:929] (1/8) Computing validation loss 2023-05-02 11:02:59,737 INFO [train.py:938] (1/8) Epoch 27, validation: loss=0.1436, simple_loss=0.2474, pruned_loss=0.01989, over 944034.00 frames. 2023-05-02 11:02:59,738 INFO [train.py:939] (1/8) Maximum memory allocated so far is 18145MB 2023-05-02 11:03:00,798 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=272903.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 11:03:37,531 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.2947, 1.6027, 1.9955, 2.2080, 2.2760, 2.5143, 1.8101, 2.4470], device='cuda:1'), covar=tensor([0.0278, 0.0618, 0.0367, 0.0386, 0.0422, 0.0243, 0.0590, 0.0173], device='cuda:1'), in_proj_covar=tensor([0.0193, 0.0196, 0.0183, 0.0186, 0.0203, 0.0162, 0.0199, 0.0161], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-02 11:03:51,478 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=272927.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 11:04:35,283 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.28 vs. limit=2.0 2023-05-02 11:04:42,535 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 27, batch 9050, loss[loss=0.1863, simple_loss=0.2804, pruned_loss=0.04614, over 12398.00 frames. ], tot_loss[loss=0.165, simple_loss=0.2612, pruned_loss=0.03441, over 3072781.91 frames. ], batch size: 248, lr: 2.47e-03, grad_scale: 8.0 2023-05-02 11:05:15,894 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.4413, 3.0416, 2.6969, 2.3009, 2.1987, 2.3368, 3.0750, 2.8382], device='cuda:1'), covar=tensor([0.2662, 0.0644, 0.1746, 0.2824, 0.2749, 0.2325, 0.0465, 0.1539], device='cuda:1'), in_proj_covar=tensor([0.0327, 0.0269, 0.0306, 0.0319, 0.0297, 0.0270, 0.0297, 0.0341], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:1') 2023-05-02 11:05:18,775 INFO [optim.py:368] (1/8) 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 11:05:19,379 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=272969.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 11:06:32,228 INFO [train.py:904] (1/8) Epoch 27, batch 9100, loss[loss=0.1764, simple_loss=0.2618, pruned_loss=0.04548, over 12957.00 frames. ], tot_loss[loss=0.1654, simple_loss=0.2608, pruned_loss=0.03501, over 3066162.26 frames. ], batch size: 248, lr: 2.47e-03, grad_scale: 8.0 2023-05-02 11:07:36,035 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=273029.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 11:07:49,401 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.29 vs. limit=2.0 2023-05-02 11:08:32,208 INFO [train.py:904] (1/8) Epoch 27, batch 9150, loss[loss=0.1454, simple_loss=0.2426, pruned_loss=0.02412, over 16399.00 frames. ], tot_loss[loss=0.1655, simple_loss=0.2616, pruned_loss=0.03469, over 3067686.46 frames. ], batch size: 68, lr: 2.47e-03, grad_scale: 8.0 2023-05-02 11:09:08,127 INFO [optim.py:368] (1/8) 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 11:09:27,707 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=273077.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 11:10:18,351 INFO [train.py:904] (1/8) Epoch 27, batch 9200, loss[loss=0.1666, simple_loss=0.2602, pruned_loss=0.03651, over 16238.00 frames. ], tot_loss[loss=0.1628, simple_loss=0.2577, pruned_loss=0.03397, over 3070542.63 frames. ], batch size: 165, lr: 2.47e-03, grad_scale: 8.0 2023-05-02 11:10:46,414 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.5890, 3.8258, 2.8747, 2.2123, 2.3413, 2.5110, 4.1178, 3.2069], device='cuda:1'), covar=tensor([0.3123, 0.0590, 0.1878, 0.3286, 0.3318, 0.2234, 0.0372, 0.1559], device='cuda:1'), in_proj_covar=tensor([0.0329, 0.0270, 0.0308, 0.0321, 0.0298, 0.0271, 0.0299, 0.0343], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:1') 2023-05-02 11:11:53,368 INFO [train.py:904] (1/8) Epoch 27, batch 9250, loss[loss=0.1496, simple_loss=0.2371, pruned_loss=0.03107, over 12396.00 frames. ], tot_loss[loss=0.1627, simple_loss=0.2575, pruned_loss=0.03398, over 3061553.62 frames. ], batch size: 248, lr: 2.47e-03, grad_scale: 8.0 2023-05-02 11:11:56,649 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.9799, 4.2673, 4.1198, 4.1323, 3.7874, 3.8389, 3.8954, 4.2608], device='cuda:1'), covar=tensor([0.1161, 0.1008, 0.1041, 0.0842, 0.0869, 0.1753, 0.1026, 0.1039], device='cuda:1'), in_proj_covar=tensor([0.0693, 0.0840, 0.0689, 0.0646, 0.0529, 0.0531, 0.0703, 0.0657], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:1') 2023-05-02 11:11:56,740 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.5944, 3.5605, 3.5218, 2.7032, 3.3983, 2.0705, 3.2731, 2.8125], device='cuda:1'), covar=tensor([0.0166, 0.0191, 0.0205, 0.0232, 0.0118, 0.2577, 0.0164, 0.0268], device='cuda:1'), in_proj_covar=tensor([0.0173, 0.0168, 0.0204, 0.0178, 0.0181, 0.0211, 0.0193, 0.0171], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-02 11:12:12,582 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.8771, 3.8646, 4.1488, 4.1325, 4.1270, 3.9230, 3.9266, 3.9900], device='cuda:1'), covar=tensor([0.0709, 0.1506, 0.1039, 0.1440, 0.1281, 0.1588, 0.1392, 0.0662], device='cuda:1'), in_proj_covar=tensor([0.0417, 0.0469, 0.0454, 0.0419, 0.0501, 0.0480, 0.0550, 0.0385], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:1') 2023-05-02 11:12:25,532 INFO [optim.py:368] (1/8) 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 11:12:59,935 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.2710, 2.9408, 3.0654, 1.8231, 3.2096, 3.3149, 2.8504, 2.7108], device='cuda:1'), covar=tensor([0.0802, 0.0284, 0.0271, 0.1264, 0.0117, 0.0207, 0.0430, 0.0466], device='cuda:1'), in_proj_covar=tensor([0.0143, 0.0106, 0.0097, 0.0134, 0.0082, 0.0124, 0.0125, 0.0126], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:1') 2023-05-02 11:13:43,821 INFO [train.py:904] (1/8) Epoch 27, batch 9300, loss[loss=0.1426, simple_loss=0.2401, pruned_loss=0.02252, over 16670.00 frames. ], tot_loss[loss=0.1609, simple_loss=0.2555, pruned_loss=0.03318, over 3057059.36 frames. ], batch size: 76, lr: 2.47e-03, grad_scale: 8.0 2023-05-02 11:14:27,530 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.7487, 4.5399, 4.7736, 4.9350, 5.0988, 4.5938, 5.1222, 5.1028], device='cuda:1'), covar=tensor([0.2019, 0.1442, 0.1735, 0.0781, 0.0616, 0.1080, 0.0615, 0.0828], device='cuda:1'), in_proj_covar=tensor([0.0643, 0.0784, 0.0909, 0.0797, 0.0612, 0.0637, 0.0667, 0.0778], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-02 11:15:09,359 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.3515, 2.4878, 2.1747, 2.2684, 2.8516, 2.5259, 2.8536, 3.0274], device='cuda:1'), covar=tensor([0.0191, 0.0473, 0.0572, 0.0539, 0.0336, 0.0461, 0.0268, 0.0304], device='cuda:1'), in_proj_covar=tensor([0.0217, 0.0237, 0.0225, 0.0228, 0.0237, 0.0235, 0.0232, 0.0232], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-02 11:15:28,949 INFO [train.py:904] (1/8) Epoch 27, batch 9350, loss[loss=0.1641, simple_loss=0.2572, pruned_loss=0.03554, over 15338.00 frames. ], tot_loss[loss=0.1608, simple_loss=0.2551, pruned_loss=0.03328, over 3057608.69 frames. ], batch size: 191, lr: 2.47e-03, grad_scale: 8.0 2023-05-02 11:16:02,982 INFO [optim.py:368] (1/8) 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 11:16:03,987 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=273269.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 11:17:10,294 INFO [train.py:904] (1/8) Epoch 27, batch 9400, loss[loss=0.1755, simple_loss=0.2731, pruned_loss=0.03893, over 16736.00 frames. ], tot_loss[loss=0.1609, simple_loss=0.2551, pruned_loss=0.03332, over 3045371.53 frames. ], batch size: 134, lr: 2.47e-03, grad_scale: 8.0 2023-05-02 11:17:39,074 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=273317.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 11:17:50,961 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.9005, 3.2458, 3.5418, 2.0223, 2.9828, 2.1767, 3.3842, 3.3863], device='cuda:1'), covar=tensor([0.0286, 0.0902, 0.0499, 0.2244, 0.0824, 0.1084, 0.0661, 0.1027], device='cuda:1'), in_proj_covar=tensor([0.0155, 0.0164, 0.0164, 0.0152, 0.0143, 0.0129, 0.0141, 0.0174], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:1') 2023-05-02 11:18:52,055 INFO [train.py:904] (1/8) Epoch 27, batch 9450, loss[loss=0.1397, simple_loss=0.2383, pruned_loss=0.02059, over 16756.00 frames. ], tot_loss[loss=0.1623, simple_loss=0.2574, pruned_loss=0.03357, over 3066899.60 frames. ], batch size: 83, lr: 2.47e-03, grad_scale: 8.0 2023-05-02 11:19:21,968 INFO [optim.py:368] (1/8) 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 11:19:52,761 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=2.70 vs. limit=5.0 2023-05-02 11:20:27,705 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.8816, 3.8443, 4.0082, 3.7223, 3.9683, 4.3452, 4.0110, 3.6163], device='cuda:1'), covar=tensor([0.1931, 0.2384, 0.2337, 0.2599, 0.2508, 0.1550, 0.1649, 0.2626], device='cuda:1'), in_proj_covar=tensor([0.0404, 0.0602, 0.0661, 0.0490, 0.0652, 0.0694, 0.0521, 0.0653], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-05-02 11:20:32,362 INFO [train.py:904] (1/8) Epoch 27, batch 9500, loss[loss=0.1643, simple_loss=0.263, pruned_loss=0.03284, over 16315.00 frames. ], tot_loss[loss=0.162, simple_loss=0.2571, pruned_loss=0.03343, over 3053176.17 frames. ], batch size: 146, lr: 2.47e-03, grad_scale: 8.0 2023-05-02 11:20:49,206 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=273409.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 11:21:38,356 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 27, batch 9550, loss[loss=0.1598, simple_loss=0.2647, pruned_loss=0.02748, over 15450.00 frames. ], tot_loss[loss=0.1615, simple_loss=0.2565, pruned_loss=0.03328, over 3048689.62 frames. ], batch size: 191, lr: 2.47e-03, grad_scale: 8.0 2023-05-02 11:22:51,234 INFO [optim.py:368] (1/8) 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 11:22:53,719 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=273470.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 11:23:33,155 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.4184, 2.3850, 2.2313, 4.0748, 2.2845, 2.7406, 2.4275, 2.5310], device='cuda:1'), covar=tensor([0.1192, 0.3689, 0.3361, 0.0494, 0.4223, 0.2487, 0.3688, 0.3439], device='cuda:1'), in_proj_covar=tensor([0.0409, 0.0462, 0.0378, 0.0326, 0.0437, 0.0526, 0.0435, 0.0538], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-02 11:23:37,420 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.22 vs. limit=2.0 2023-05-02 11:23:46,429 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=273496.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 11:23:48,309 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.68 vs. limit=2.0 2023-05-02 11:23:57,863 INFO [train.py:904] (1/8) Epoch 27, batch 9600, loss[loss=0.1732, simple_loss=0.2772, pruned_loss=0.03461, over 15449.00 frames. ], tot_loss[loss=0.1626, simple_loss=0.2577, pruned_loss=0.03375, over 3035358.36 frames. ], batch size: 191, lr: 2.47e-03, grad_scale: 8.0 2023-05-02 11:24:54,717 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.4251, 3.2111, 3.6509, 1.9787, 3.7199, 3.8143, 2.9770, 2.8797], device='cuda:1'), covar=tensor([0.0800, 0.0297, 0.0197, 0.1114, 0.0091, 0.0179, 0.0450, 0.0465], device='cuda:1'), in_proj_covar=tensor([0.0144, 0.0106, 0.0097, 0.0134, 0.0083, 0.0124, 0.0125, 0.0126], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0004, 0.0003], device='cuda:1') 2023-05-02 11:25:44,214 INFO [train.py:904] (1/8) Epoch 27, batch 9650, loss[loss=0.1639, simple_loss=0.2623, pruned_loss=0.03276, over 16827.00 frames. ], tot_loss[loss=0.1639, simple_loss=0.2597, pruned_loss=0.034, over 3049597.54 frames. ], batch size: 116, lr: 2.47e-03, grad_scale: 8.0 2023-05-02 11:26:23,245 INFO [optim.py:368] (1/8) 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 11:27:30,812 INFO [train.py:904] (1/8) Epoch 27, batch 9700, loss[loss=0.1845, simple_loss=0.2686, pruned_loss=0.05015, over 12508.00 frames. ], tot_loss[loss=0.1629, simple_loss=0.2581, pruned_loss=0.03385, over 3033923.74 frames. ], batch size: 250, lr: 2.47e-03, grad_scale: 8.0 2023-05-02 11:29:12,471 INFO [train.py:904] (1/8) Epoch 27, batch 9750, loss[loss=0.1745, simple_loss=0.2714, pruned_loss=0.03884, over 16779.00 frames. ], tot_loss[loss=0.1632, simple_loss=0.2577, pruned_loss=0.03432, over 3039339.43 frames. ], batch size: 124, lr: 2.47e-03, grad_scale: 8.0 2023-05-02 11:29:42,115 INFO [optim.py:368] (1/8) 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 11:30:08,781 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.3637, 1.6502, 2.1085, 2.3690, 2.3780, 2.6035, 1.8717, 2.5475], device='cuda:1'), covar=tensor([0.0322, 0.0661, 0.0387, 0.0385, 0.0409, 0.0265, 0.0628, 0.0180], device='cuda:1'), in_proj_covar=tensor([0.0192, 0.0194, 0.0181, 0.0185, 0.0201, 0.0160, 0.0198, 0.0159], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-02 11:30:48,403 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.38 vs. limit=2.0 2023-05-02 11:30:51,288 INFO [train.py:904] (1/8) Epoch 27, batch 9800, loss[loss=0.1805, simple_loss=0.2839, pruned_loss=0.03857, over 16367.00 frames. ], tot_loss[loss=0.1625, simple_loss=0.2577, pruned_loss=0.03364, over 3049871.55 frames. ], batch size: 165, lr: 2.47e-03, grad_scale: 4.0 2023-05-02 11:31:27,381 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.35 vs. limit=2.0 2023-05-02 11:32:36,071 INFO [train.py:904] (1/8) Epoch 27, batch 9850, loss[loss=0.163, simple_loss=0.2666, pruned_loss=0.02964, over 16869.00 frames. ], tot_loss[loss=0.1625, simple_loss=0.2587, pruned_loss=0.03313, over 3047559.32 frames. ], batch size: 90, lr: 2.47e-03, grad_scale: 4.0 2023-05-02 11:33:00,287 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=273765.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 11:33:11,080 INFO [optim.py:368] (1/8) 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] (1/8) 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] (1/8) Epoch 27, batch 9900, loss[loss=0.1634, simple_loss=0.2513, pruned_loss=0.03772, over 12666.00 frames. ], tot_loss[loss=0.163, simple_loss=0.2595, pruned_loss=0.03331, over 3051071.44 frames. ], batch size: 250, lr: 2.47e-03, grad_scale: 4.0 2023-05-02 11:36:22,572 INFO [train.py:904] (1/8) Epoch 27, batch 9950, loss[loss=0.1609, simple_loss=0.2654, pruned_loss=0.02819, over 16234.00 frames. ], tot_loss[loss=0.1647, simple_loss=0.2623, pruned_loss=0.03355, over 3071708.76 frames. ], batch size: 165, lr: 2.47e-03, grad_scale: 4.0 2023-05-02 11:37:03,259 INFO [optim.py:368] (1/8) 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] (1/8) Epoch 27, batch 10000, loss[loss=0.1813, simple_loss=0.2632, pruned_loss=0.04971, over 12487.00 frames. ], tot_loss[loss=0.1639, simple_loss=0.2611, pruned_loss=0.03332, over 3088764.94 frames. ], batch size: 250, lr: 2.47e-03, grad_scale: 8.0 2023-05-02 11:38:32,894 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.48 vs. limit=2.0 2023-05-02 11:38:40,617 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.7260, 3.8360, 2.3648, 4.2719, 2.9160, 4.1999, 2.6310, 3.1271], device='cuda:1'), covar=tensor([0.0272, 0.0297, 0.1594, 0.0247, 0.0824, 0.0414, 0.1359, 0.0683], device='cuda:1'), in_proj_covar=tensor([0.0167, 0.0172, 0.0189, 0.0162, 0.0172, 0.0208, 0.0198, 0.0176], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0003, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-05-02 11:40:06,640 INFO [train.py:904] (1/8) Epoch 27, batch 10050, loss[loss=0.1744, simple_loss=0.2632, pruned_loss=0.04278, over 12092.00 frames. ], tot_loss[loss=0.1643, simple_loss=0.2616, pruned_loss=0.03351, over 3089039.39 frames. ], batch size: 248, lr: 2.47e-03, grad_scale: 8.0 2023-05-02 11:40:39,173 INFO [optim.py:368] (1/8) 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 11:41:40,641 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 27, batch 10100, loss[loss=0.1825, simple_loss=0.2721, pruned_loss=0.04645, over 16386.00 frames. ], tot_loss[loss=0.1644, simple_loss=0.2615, pruned_loss=0.03368, over 3077992.14 frames. ], batch size: 146, lr: 2.47e-03, grad_scale: 8.0 2023-05-02 11:42:11,175 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.3857, 1.6698, 2.0341, 2.2998, 2.3268, 2.5119, 1.8805, 2.4618], device='cuda:1'), covar=tensor([0.0330, 0.0637, 0.0420, 0.0398, 0.0439, 0.0273, 0.0616, 0.0233], device='cuda:1'), in_proj_covar=tensor([0.0194, 0.0196, 0.0184, 0.0187, 0.0203, 0.0161, 0.0200, 0.0161], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-02 11:43:27,165 INFO [train.py:904] (1/8) Epoch 28, batch 0, loss[loss=0.1875, simple_loss=0.2751, pruned_loss=0.04992, over 16622.00 frames. ], tot_loss[loss=0.1875, simple_loss=0.2751, pruned_loss=0.04992, over 16622.00 frames. ], batch size: 57, lr: 2.42e-03, grad_scale: 8.0 2023-05-02 11:43:27,166 INFO [train.py:929] (1/8) Computing validation loss 2023-05-02 11:43:34,593 INFO [train.py:938] (1/8) Epoch 28, validation: loss=0.1434, simple_loss=0.2464, pruned_loss=0.02022, over 944034.00 frames. 2023-05-02 11:43:34,594 INFO [train.py:939] (1/8) Maximum memory allocated so far is 18145MB 2023-05-02 11:43:48,092 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=274062.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 11:43:52,119 INFO [zipformer.py:625] (1/8) 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] (1/8) 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] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=274091.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 11:44:43,857 INFO [train.py:904] (1/8) Epoch 28, batch 50, loss[loss=0.1732, simple_loss=0.2589, pruned_loss=0.04375, over 17245.00 frames. ], tot_loss[loss=0.1731, simple_loss=0.2628, pruned_loss=0.04171, over 761796.25 frames. ], batch size: 45, lr: 2.42e-03, grad_scale: 2.0 2023-05-02 11:44:59,063 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=274113.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 11:45:17,216 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.0925, 3.0262, 3.0087, 5.2013, 4.2489, 4.5368, 1.8679, 3.3924], device='cuda:1'), covar=tensor([0.1333, 0.0819, 0.1173, 0.0186, 0.0230, 0.0397, 0.1647, 0.0774], device='cuda:1'), in_proj_covar=tensor([0.0171, 0.0177, 0.0198, 0.0195, 0.0200, 0.0214, 0.0207, 0.0196], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-05-02 11:45:24,071 INFO [zipformer.py:625] (1/8) 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] (1/8) 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] (1/8) Epoch 28, batch 100, loss[loss=0.1923, simple_loss=0.2782, pruned_loss=0.05315, over 16426.00 frames. ], tot_loss[loss=0.1721, simple_loss=0.2599, pruned_loss=0.04214, over 1327525.55 frames. ], batch size: 146, lr: 2.42e-03, grad_scale: 2.0 2023-05-02 11:46:20,295 INFO [optim.py:368] (1/8) 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 11:46:26,292 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=5.10 vs. limit=5.0 2023-05-02 11:46:47,998 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=274192.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 11:46:52,649 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.75 vs. limit=2.0 2023-05-02 11:47:02,021 INFO [train.py:904] (1/8) Epoch 28, batch 150, loss[loss=0.1745, simple_loss=0.2685, pruned_loss=0.0403, over 17037.00 frames. ], tot_loss[loss=0.1722, simple_loss=0.2605, pruned_loss=0.04195, over 1775654.49 frames. ], batch size: 55, lr: 2.42e-03, grad_scale: 2.0 2023-05-02 11:47:24,541 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.3734, 2.6913, 2.3032, 2.4426, 2.9744, 2.6577, 3.0497, 3.1311], device='cuda:1'), covar=tensor([0.0266, 0.0502, 0.0578, 0.0527, 0.0328, 0.0486, 0.0291, 0.0317], device='cuda:1'), in_proj_covar=tensor([0.0224, 0.0242, 0.0231, 0.0232, 0.0243, 0.0240, 0.0237, 0.0237], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-02 11:48:08,590 INFO [train.py:904] (1/8) Epoch 28, batch 200, loss[loss=0.172, simple_loss=0.2485, pruned_loss=0.04778, over 15925.00 frames. ], tot_loss[loss=0.1734, simple_loss=0.2611, pruned_loss=0.04287, over 2119897.96 frames. ], batch size: 35, lr: 2.42e-03, grad_scale: 2.0 2023-05-02 11:48:29,628 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=274269.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 11:48:33,858 INFO [optim.py:368] (1/8) 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] (1/8) Epoch 28, batch 250, loss[loss=0.1626, simple_loss=0.2563, pruned_loss=0.03449, over 17105.00 frames. ], tot_loss[loss=0.1726, simple_loss=0.2594, pruned_loss=0.04291, over 2382002.09 frames. ], batch size: 53, lr: 2.42e-03, grad_scale: 2.0 2023-05-02 11:49:16,544 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=274303.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 11:49:38,689 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.8237, 4.9804, 5.1011, 4.8716, 4.9119, 5.5334, 4.9900, 4.6675], device='cuda:1'), covar=tensor([0.1217, 0.2118, 0.2637, 0.2380, 0.2729, 0.1080, 0.1819, 0.2726], device='cuda:1'), in_proj_covar=tensor([0.0413, 0.0619, 0.0681, 0.0503, 0.0671, 0.0712, 0.0534, 0.0669], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-05-02 11:49:41,592 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=4.65 vs. limit=5.0 2023-05-02 11:49:52,014 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=274330.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 11:50:23,017 INFO [train.py:904] (1/8) Epoch 28, batch 300, loss[loss=0.1507, simple_loss=0.2428, pruned_loss=0.02925, over 17241.00 frames. ], tot_loss[loss=0.168, simple_loss=0.2552, pruned_loss=0.04038, over 2600524.60 frames. ], batch size: 44, lr: 2.42e-03, grad_scale: 1.0 2023-05-02 11:50:29,899 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=274357.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 11:50:39,340 INFO [zipformer.py:625] (1/8) 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] (1/8) 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 11:51:07,352 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.8362, 4.8444, 4.7083, 4.2356, 4.7645, 1.9466, 4.5507, 4.4802], device='cuda:1'), covar=tensor([0.0135, 0.0127, 0.0218, 0.0334, 0.0120, 0.2772, 0.0154, 0.0235], device='cuda:1'), in_proj_covar=tensor([0.0175, 0.0169, 0.0205, 0.0178, 0.0183, 0.0213, 0.0195, 0.0173], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-02 11:51:25,289 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.6796, 3.4289, 3.8504, 2.0782, 3.9026, 3.9188, 3.2645, 2.8640], device='cuda:1'), covar=tensor([0.0785, 0.0279, 0.0201, 0.1191, 0.0120, 0.0221, 0.0406, 0.0504], device='cuda:1'), in_proj_covar=tensor([0.0148, 0.0108, 0.0100, 0.0138, 0.0085, 0.0128, 0.0128, 0.0129], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-05-02 11:51:31,010 INFO [train.py:904] (1/8) Epoch 28, batch 350, loss[loss=0.1392, simple_loss=0.2258, pruned_loss=0.02635, over 17241.00 frames. ], tot_loss[loss=0.1658, simple_loss=0.2528, pruned_loss=0.03935, over 2772187.20 frames. ], batch size: 43, lr: 2.42e-03, grad_scale: 1.0 2023-05-02 11:52:01,187 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.43 vs. limit=2.0 2023-05-02 11:52:37,187 INFO [train.py:904] (1/8) Epoch 28, batch 400, loss[loss=0.1819, simple_loss=0.2701, pruned_loss=0.04681, over 16777.00 frames. ], tot_loss[loss=0.1649, simple_loss=0.2522, pruned_loss=0.03874, over 2895962.32 frames. ], batch size: 57, lr: 2.42e-03, grad_scale: 2.0 2023-05-02 11:53:03,891 INFO [optim.py:368] (1/8) 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 11:53:24,164 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=274487.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 11:53:44,157 INFO [train.py:904] (1/8) Epoch 28, batch 450, loss[loss=0.1518, simple_loss=0.2386, pruned_loss=0.03251, over 16494.00 frames. ], tot_loss[loss=0.1642, simple_loss=0.2514, pruned_loss=0.03849, over 2992072.25 frames. ], batch size: 68, lr: 2.42e-03, grad_scale: 1.0 2023-05-02 11:54:35,127 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=274540.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 11:54:53,414 INFO [train.py:904] (1/8) Epoch 28, batch 500, loss[loss=0.1606, simple_loss=0.2539, pruned_loss=0.03363, over 17187.00 frames. ], tot_loss[loss=0.1624, simple_loss=0.2496, pruned_loss=0.0376, over 3055674.00 frames. ], batch size: 46, lr: 2.42e-03, grad_scale: 1.0 2023-05-02 11:54:55,023 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([5.1562, 5.1575, 5.6422, 5.5909, 5.6562, 5.2639, 5.2220, 5.0516], device='cuda:1'), covar=tensor([0.0463, 0.0560, 0.0419, 0.0503, 0.0496, 0.0443, 0.1049, 0.0454], device='cuda:1'), in_proj_covar=tensor([0.0433, 0.0484, 0.0470, 0.0433, 0.0517, 0.0496, 0.0568, 0.0397], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:1') 2023-05-02 11:55:09,516 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=3.67 vs. limit=5.0 2023-05-02 11:55:21,738 INFO [optim.py:368] (1/8) 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 11:55:58,900 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 28, batch 550, loss[loss=0.1784, simple_loss=0.256, pruned_loss=0.05037, over 16711.00 frames. ], tot_loss[loss=0.1618, simple_loss=0.2487, pruned_loss=0.03744, over 3120332.96 frames. ], batch size: 134, lr: 2.42e-03, grad_scale: 1.0 2023-05-02 11:56:30,882 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=274625.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 11:57:02,382 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.4997, 2.3426, 1.8862, 2.1267, 2.6822, 2.4077, 2.5089, 2.7761], device='cuda:1'), covar=tensor([0.0277, 0.0452, 0.0617, 0.0487, 0.0262, 0.0375, 0.0209, 0.0289], device='cuda:1'), in_proj_covar=tensor([0.0230, 0.0248, 0.0236, 0.0237, 0.0248, 0.0246, 0.0243, 0.0244], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-02 11:57:07,135 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.7392, 5.0764, 4.9046, 4.8587, 4.5991, 4.6405, 4.5342, 5.1734], device='cuda:1'), covar=tensor([0.1403, 0.1021, 0.1074, 0.0943, 0.0863, 0.1078, 0.1312, 0.0887], device='cuda:1'), in_proj_covar=tensor([0.0712, 0.0864, 0.0703, 0.0663, 0.0545, 0.0540, 0.0723, 0.0671], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:1') 2023-05-02 11:57:09,862 INFO [train.py:904] (1/8) Epoch 28, batch 600, loss[loss=0.1632, simple_loss=0.2361, pruned_loss=0.04522, over 16719.00 frames. ], tot_loss[loss=0.1623, simple_loss=0.2485, pruned_loss=0.03802, over 3159273.39 frames. ], batch size: 124, lr: 2.42e-03, grad_scale: 1.0 2023-05-02 11:57:15,354 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=274657.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 11:57:17,497 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=274659.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 11:57:36,635 INFO [optim.py:368] (1/8) 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 11:58:00,245 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.86 vs. limit=2.0 2023-05-02 11:58:10,254 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.6756, 3.7603, 2.5428, 4.3883, 2.9942, 4.3025, 2.7976, 3.2100], device='cuda:1'), covar=tensor([0.0372, 0.0432, 0.1645, 0.0341, 0.0883, 0.0567, 0.1380, 0.0763], device='cuda:1'), in_proj_covar=tensor([0.0176, 0.0181, 0.0198, 0.0172, 0.0180, 0.0220, 0.0206, 0.0184], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-05-02 11:58:16,066 INFO [train.py:904] (1/8) Epoch 28, batch 650, loss[loss=0.1379, simple_loss=0.2226, pruned_loss=0.02658, over 16774.00 frames. ], tot_loss[loss=0.1609, simple_loss=0.2469, pruned_loss=0.03747, over 3199559.68 frames. ], batch size: 39, lr: 2.42e-03, grad_scale: 1.0 2023-05-02 11:58:18,639 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=274705.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 11:59:05,624 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.8772, 4.2629, 3.0147, 2.4224, 2.6399, 2.6826, 4.6227, 3.5018], device='cuda:1'), covar=tensor([0.2935, 0.0645, 0.1889, 0.3028, 0.3066, 0.2202, 0.0360, 0.1562], device='cuda:1'), in_proj_covar=tensor([0.0333, 0.0274, 0.0311, 0.0324, 0.0301, 0.0275, 0.0302, 0.0349], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:1') 2023-05-02 11:59:14,283 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.7765, 4.7549, 4.6314, 4.0784, 4.7104, 1.8583, 4.4602, 4.2861], device='cuda:1'), covar=tensor([0.0196, 0.0121, 0.0206, 0.0339, 0.0123, 0.2979, 0.0172, 0.0267], device='cuda:1'), in_proj_covar=tensor([0.0177, 0.0171, 0.0208, 0.0180, 0.0185, 0.0215, 0.0197, 0.0175], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-02 11:59:25,787 INFO [train.py:904] (1/8) Epoch 28, batch 700, loss[loss=0.1689, simple_loss=0.2456, pruned_loss=0.04613, over 16735.00 frames. ], tot_loss[loss=0.1602, simple_loss=0.2464, pruned_loss=0.03701, over 3219666.46 frames. ], batch size: 134, lr: 2.42e-03, grad_scale: 1.0 2023-05-02 11:59:48,983 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.4537, 2.3467, 1.9431, 2.0842, 2.6669, 2.3740, 2.5301, 2.7196], device='cuda:1'), covar=tensor([0.0278, 0.0449, 0.0570, 0.0523, 0.0282, 0.0406, 0.0254, 0.0353], device='cuda:1'), in_proj_covar=tensor([0.0232, 0.0249, 0.0237, 0.0238, 0.0250, 0.0247, 0.0245, 0.0245], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-02 11:59:51,869 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.6949, 2.6321, 2.3793, 2.4614, 2.9760, 2.7316, 3.2055, 3.1703], device='cuda:1'), covar=tensor([0.0175, 0.0530, 0.0642, 0.0580, 0.0372, 0.0493, 0.0319, 0.0358], device='cuda:1'), in_proj_covar=tensor([0.0232, 0.0249, 0.0237, 0.0238, 0.0250, 0.0247, 0.0245, 0.0245], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-02 11:59:54,343 INFO [optim.py:368] (1/8) 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 12:00:13,235 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=274787.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 12:00:34,709 INFO [train.py:904] (1/8) Epoch 28, batch 750, loss[loss=0.2025, simple_loss=0.2698, pruned_loss=0.0676, over 16666.00 frames. ], tot_loss[loss=0.1609, simple_loss=0.2469, pruned_loss=0.0375, over 3237732.92 frames. ], batch size: 134, lr: 2.42e-03, grad_scale: 1.0 2023-05-02 12:01:14,272 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([5.2237, 5.1945, 4.9645, 4.3971, 5.0512, 1.7867, 4.8331, 4.7918], device='cuda:1'), covar=tensor([0.0126, 0.0120, 0.0262, 0.0474, 0.0127, 0.3099, 0.0150, 0.0293], device='cuda:1'), in_proj_covar=tensor([0.0177, 0.0171, 0.0208, 0.0180, 0.0185, 0.0215, 0.0197, 0.0176], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-02 12:01:19,359 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=274835.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 12:01:45,403 INFO [train.py:904] (1/8) Epoch 28, batch 800, loss[loss=0.1812, simple_loss=0.2506, pruned_loss=0.05588, over 16847.00 frames. ], tot_loss[loss=0.1608, simple_loss=0.2475, pruned_loss=0.0371, over 3264129.49 frames. ], batch size: 102, lr: 2.42e-03, grad_scale: 2.0 2023-05-02 12:02:15,002 INFO [optim.py:368] (1/8) 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 12:02:43,377 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=274894.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 12:02:44,932 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.33 vs. limit=2.0 2023-05-02 12:02:46,201 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 28, batch 850, loss[loss=0.1579, simple_loss=0.245, pruned_loss=0.03541, over 16460.00 frames. ], tot_loss[loss=0.1604, simple_loss=0.247, pruned_loss=0.0369, over 3280824.53 frames. ], batch size: 75, lr: 2.42e-03, grad_scale: 2.0 2023-05-02 12:03:26,689 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 28, batch 900, loss[loss=0.1637, simple_loss=0.2427, pruned_loss=0.04238, over 16901.00 frames. ], tot_loss[loss=0.16, simple_loss=0.2471, pruned_loss=0.0365, over 3296942.59 frames. ], batch size: 96, lr: 2.42e-03, grad_scale: 2.0 2023-05-02 12:04:09,169 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=274955.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 12:04:13,739 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=274959.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 12:04:33,593 INFO [zipformer.py:625] (1/8) 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] (1/8) 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 12:05:14,546 INFO [train.py:904] (1/8) Epoch 28, batch 950, loss[loss=0.145, simple_loss=0.2329, pruned_loss=0.02854, over 16865.00 frames. ], tot_loss[loss=0.1604, simple_loss=0.2468, pruned_loss=0.03698, over 3301085.55 frames. ], batch size: 42, lr: 2.42e-03, grad_scale: 2.0 2023-05-02 12:05:20,444 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=275007.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 12:05:22,270 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.9242, 4.9431, 5.3167, 5.3029, 5.3290, 5.0152, 4.9468, 4.8277], device='cuda:1'), covar=tensor([0.0376, 0.0703, 0.0439, 0.0415, 0.0450, 0.0440, 0.0921, 0.0440], device='cuda:1'), in_proj_covar=tensor([0.0440, 0.0494, 0.0478, 0.0439, 0.0525, 0.0503, 0.0576, 0.0402], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:1') 2023-05-02 12:06:23,168 INFO [train.py:904] (1/8) Epoch 28, batch 1000, loss[loss=0.1562, simple_loss=0.2333, pruned_loss=0.03954, over 16452.00 frames. ], tot_loss[loss=0.1597, simple_loss=0.2455, pruned_loss=0.03695, over 3304380.91 frames. ], batch size: 146, lr: 2.42e-03, grad_scale: 2.0 2023-05-02 12:06:52,343 INFO [optim.py:368] (1/8) 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 12:07:31,781 INFO [train.py:904] (1/8) Epoch 28, batch 1050, loss[loss=0.1807, simple_loss=0.2546, pruned_loss=0.05341, over 16813.00 frames. ], tot_loss[loss=0.1595, simple_loss=0.2449, pruned_loss=0.0371, over 3313530.53 frames. ], batch size: 83, lr: 2.42e-03, grad_scale: 2.0 2023-05-02 12:08:09,131 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.5765, 3.6130, 4.1704, 2.3356, 3.2881, 2.6645, 4.0388, 3.8333], device='cuda:1'), covar=tensor([0.0259, 0.1023, 0.0476, 0.2071, 0.0820, 0.0980, 0.0543, 0.1093], device='cuda:1'), in_proj_covar=tensor([0.0161, 0.0170, 0.0170, 0.0157, 0.0148, 0.0133, 0.0147, 0.0182], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0005], device='cuda:1') 2023-05-02 12:08:39,620 INFO [train.py:904] (1/8) Epoch 28, batch 1100, loss[loss=0.1727, simple_loss=0.2594, pruned_loss=0.04301, over 17078.00 frames. ], tot_loss[loss=0.1585, simple_loss=0.2442, pruned_loss=0.03639, over 3318347.08 frames. ], batch size: 55, lr: 2.42e-03, grad_scale: 2.0 2023-05-02 12:08:47,418 INFO [zipformer.py:625] (1/8) 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] (1/8) 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 12:09:39,570 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=275196.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 12:09:49,908 INFO [train.py:904] (1/8) Epoch 28, batch 1150, loss[loss=0.1507, simple_loss=0.2316, pruned_loss=0.03487, over 17035.00 frames. ], tot_loss[loss=0.1576, simple_loss=0.2437, pruned_loss=0.03579, over 3317144.38 frames. ], batch size: 41, lr: 2.42e-03, grad_scale: 2.0 2023-05-02 12:09:52,235 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.30 vs. limit=2.0 2023-05-02 12:10:10,226 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.5704, 3.5471, 3.4912, 2.7494, 3.3311, 2.1458, 3.1585, 2.7672], device='cuda:1'), covar=tensor([0.0164, 0.0162, 0.0184, 0.0230, 0.0118, 0.2423, 0.0147, 0.0289], device='cuda:1'), in_proj_covar=tensor([0.0179, 0.0174, 0.0210, 0.0182, 0.0187, 0.0217, 0.0200, 0.0178], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-02 12:10:11,439 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=275219.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 12:10:30,508 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.6985, 2.6275, 2.2878, 2.4843, 2.9697, 2.6644, 3.2564, 3.1390], device='cuda:1'), covar=tensor([0.0194, 0.0505, 0.0626, 0.0547, 0.0359, 0.0505, 0.0276, 0.0341], device='cuda:1'), in_proj_covar=tensor([0.0234, 0.0250, 0.0238, 0.0238, 0.0251, 0.0248, 0.0247, 0.0247], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-02 12:10:44,954 INFO [zipformer.py:625] (1/8) 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] (1/8) 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] (1/8) Epoch 28, batch 1200, loss[loss=0.1595, simple_loss=0.2402, pruned_loss=0.03938, over 11785.00 frames. ], tot_loss[loss=0.1575, simple_loss=0.2438, pruned_loss=0.03564, over 3316753.47 frames. ], batch size: 246, lr: 2.42e-03, grad_scale: 4.0 2023-05-02 12:11:20,296 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.9802, 4.4236, 4.3301, 3.2284, 3.6561, 4.3711, 4.0171, 2.5647], device='cuda:1'), covar=tensor([0.0498, 0.0069, 0.0067, 0.0387, 0.0187, 0.0130, 0.0114, 0.0542], device='cuda:1'), in_proj_covar=tensor([0.0138, 0.0090, 0.0091, 0.0135, 0.0102, 0.0114, 0.0098, 0.0131], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0003, 0.0004], device='cuda:1') 2023-05-02 12:11:21,695 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.26 vs. limit=2.0 2023-05-02 12:11:26,948 INFO [optim.py:368] (1/8) 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] (1/8) Epoch 28, batch 1250, loss[loss=0.1861, simple_loss=0.2528, pruned_loss=0.05964, over 16779.00 frames. ], tot_loss[loss=0.1578, simple_loss=0.2441, pruned_loss=0.03572, over 3320599.22 frames. ], batch size: 83, lr: 2.42e-03, grad_scale: 4.0 2023-05-02 12:12:12,658 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=275307.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 12:13:08,398 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=275347.0, num_to_drop=1, layers_to_drop={0} 2023-05-02 12:13:17,199 INFO [train.py:904] (1/8) Epoch 28, batch 1300, loss[loss=0.1597, simple_loss=0.2565, pruned_loss=0.03143, over 17013.00 frames. ], tot_loss[loss=0.1573, simple_loss=0.2436, pruned_loss=0.03545, over 3320852.93 frames. ], batch size: 55, lr: 2.42e-03, grad_scale: 4.0 2023-05-02 12:13:39,180 INFO [zipformer.py:625] (1/8) 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] (1/8) 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 12:14:02,900 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=275385.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 12:14:26,884 INFO [train.py:904] (1/8) Epoch 28, batch 1350, loss[loss=0.1773, simple_loss=0.2495, pruned_loss=0.0526, over 16719.00 frames. ], tot_loss[loss=0.1575, simple_loss=0.2442, pruned_loss=0.03537, over 3324419.27 frames. ], batch size: 134, lr: 2.42e-03, grad_scale: 4.0 2023-05-02 12:14:35,325 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=275408.0, num_to_drop=1, layers_to_drop={2} 2023-05-02 12:15:28,623 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=275446.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 12:15:37,716 INFO [train.py:904] (1/8) Epoch 28, batch 1400, loss[loss=0.1541, simple_loss=0.2412, pruned_loss=0.03351, over 16831.00 frames. ], tot_loss[loss=0.1577, simple_loss=0.2446, pruned_loss=0.0354, over 3327119.54 frames. ], batch size: 42, lr: 2.42e-03, grad_scale: 4.0 2023-05-02 12:16:06,675 INFO [optim.py:368] (1/8) 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] (1/8) Epoch 28, batch 1450, loss[loss=0.1585, simple_loss=0.2503, pruned_loss=0.03339, over 17254.00 frames. ], tot_loss[loss=0.1569, simple_loss=0.2435, pruned_loss=0.03513, over 3335014.78 frames. ], batch size: 45, lr: 2.41e-03, grad_scale: 4.0 2023-05-02 12:17:03,271 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=275514.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 12:17:05,724 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.1870, 2.4133, 2.7826, 3.1054, 2.9782, 3.6210, 2.6129, 3.5970], device='cuda:1'), covar=tensor([0.0314, 0.0537, 0.0391, 0.0403, 0.0410, 0.0228, 0.0522, 0.0214], device='cuda:1'), in_proj_covar=tensor([0.0199, 0.0200, 0.0189, 0.0194, 0.0209, 0.0168, 0.0205, 0.0167], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0002], device='cuda:1') 2023-05-02 12:17:53,385 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=275550.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 12:17:57,212 INFO [train.py:904] (1/8) Epoch 28, batch 1500, loss[loss=0.1529, simple_loss=0.227, pruned_loss=0.03938, over 16356.00 frames. ], tot_loss[loss=0.157, simple_loss=0.2434, pruned_loss=0.03528, over 3333066.44 frames. ], batch size: 145, lr: 2.41e-03, grad_scale: 4.0 2023-05-02 12:18:26,333 INFO [optim.py:368] (1/8) 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 12:19:01,051 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 28, batch 1550, loss[loss=0.1846, simple_loss=0.2665, pruned_loss=0.05134, over 15450.00 frames. ], tot_loss[loss=0.159, simple_loss=0.245, pruned_loss=0.03653, over 3328914.88 frames. ], batch size: 190, lr: 2.41e-03, grad_scale: 4.0 2023-05-02 12:19:18,137 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.44 vs. limit=2.0 2023-05-02 12:19:31,473 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.5167, 3.1592, 3.5850, 1.8746, 3.6251, 3.6498, 3.0891, 2.7459], device='cuda:1'), covar=tensor([0.0798, 0.0318, 0.0221, 0.1258, 0.0132, 0.0236, 0.0432, 0.0470], device='cuda:1'), in_proj_covar=tensor([0.0151, 0.0111, 0.0103, 0.0141, 0.0087, 0.0133, 0.0131, 0.0132], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-05-02 12:20:18,406 INFO [train.py:904] (1/8) Epoch 28, batch 1600, loss[loss=0.1622, simple_loss=0.2502, pruned_loss=0.03714, over 15810.00 frames. ], tot_loss[loss=0.1611, simple_loss=0.2468, pruned_loss=0.03769, over 3318556.01 frames. ], batch size: 35, lr: 2.41e-03, grad_scale: 8.0 2023-05-02 12:20:33,167 INFO [zipformer.py:625] (1/8) 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] (1/8) 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 12:21:10,181 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.7134, 4.2435, 4.2218, 3.0303, 3.5208, 4.2073, 3.8674, 2.4028], device='cuda:1'), covar=tensor([0.0554, 0.0091, 0.0063, 0.0391, 0.0168, 0.0108, 0.0095, 0.0550], device='cuda:1'), in_proj_covar=tensor([0.0140, 0.0091, 0.0092, 0.0136, 0.0103, 0.0115, 0.0099, 0.0133], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0003, 0.0004], device='cuda:1') 2023-05-02 12:21:28,983 INFO [train.py:904] (1/8) Epoch 28, batch 1650, loss[loss=0.1746, simple_loss=0.2666, pruned_loss=0.04129, over 17064.00 frames. ], tot_loss[loss=0.1615, simple_loss=0.2475, pruned_loss=0.03772, over 3330095.24 frames. ], batch size: 53, lr: 2.41e-03, grad_scale: 8.0 2023-05-02 12:21:29,287 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=275703.0, num_to_drop=1, layers_to_drop={0} 2023-05-02 12:22:21,277 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=275741.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 12:22:28,497 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([5.3457, 5.2890, 5.2117, 4.6768, 4.8134, 5.2353, 5.1578, 4.8272], device='cuda:1'), covar=tensor([0.0626, 0.0567, 0.0328, 0.0383, 0.1160, 0.0508, 0.0385, 0.0938], device='cuda:1'), in_proj_covar=tensor([0.0319, 0.0479, 0.0373, 0.0375, 0.0370, 0.0430, 0.0256, 0.0447], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-05-02 12:22:37,918 INFO [train.py:904] (1/8) Epoch 28, batch 1700, loss[loss=0.1772, simple_loss=0.2668, pruned_loss=0.04379, over 15556.00 frames. ], tot_loss[loss=0.1636, simple_loss=0.2496, pruned_loss=0.03883, over 3319623.37 frames. ], batch size: 190, lr: 2.41e-03, grad_scale: 8.0 2023-05-02 12:23:08,712 INFO [optim.py:368] (1/8) 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 12:23:26,609 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.0678, 4.5773, 4.5354, 3.2017, 3.8025, 4.5478, 4.0792, 2.7788], device='cuda:1'), covar=tensor([0.0531, 0.0080, 0.0052, 0.0425, 0.0153, 0.0097, 0.0091, 0.0516], device='cuda:1'), in_proj_covar=tensor([0.0141, 0.0092, 0.0092, 0.0138, 0.0104, 0.0116, 0.0100, 0.0134], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0003, 0.0004], device='cuda:1') 2023-05-02 12:23:38,000 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=3.52 vs. limit=5.0 2023-05-02 12:23:49,162 INFO [train.py:904] (1/8) Epoch 28, batch 1750, loss[loss=0.1445, simple_loss=0.2305, pruned_loss=0.02932, over 16975.00 frames. ], tot_loss[loss=0.1639, simple_loss=0.2501, pruned_loss=0.03882, over 3309609.20 frames. ], batch size: 41, lr: 2.41e-03, grad_scale: 4.0 2023-05-02 12:24:05,576 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=275814.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 12:24:59,130 INFO [train.py:904] (1/8) Epoch 28, batch 1800, loss[loss=0.1855, simple_loss=0.2673, pruned_loss=0.05184, over 16909.00 frames. ], tot_loss[loss=0.1651, simple_loss=0.2519, pruned_loss=0.03912, over 3305669.99 frames. ], batch size: 116, lr: 2.41e-03, grad_scale: 4.0 2023-05-02 12:25:09,000 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=275860.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 12:25:11,182 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=275862.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 12:25:14,272 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=4.91 vs. limit=5.0 2023-05-02 12:25:30,537 INFO [optim.py:368] (1/8) 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 12:26:07,507 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.5776, 3.5692, 3.5059, 2.7636, 3.3908, 2.0929, 3.1876, 2.7701], device='cuda:1'), covar=tensor([0.0182, 0.0163, 0.0193, 0.0246, 0.0121, 0.2650, 0.0152, 0.0310], device='cuda:1'), in_proj_covar=tensor([0.0182, 0.0177, 0.0214, 0.0186, 0.0190, 0.0220, 0.0203, 0.0181], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-02 12:26:08,320 INFO [train.py:904] (1/8) Epoch 28, batch 1850, loss[loss=0.1822, simple_loss=0.2754, pruned_loss=0.04452, over 12293.00 frames. ], tot_loss[loss=0.1653, simple_loss=0.2526, pruned_loss=0.03902, over 3307858.52 frames. ], batch size: 246, lr: 2.41e-03, grad_scale: 4.0 2023-05-02 12:26:20,874 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.6598, 4.9262, 5.3044, 5.2636, 5.2723, 5.0092, 4.6894, 4.7958], device='cuda:1'), covar=tensor([0.0651, 0.0736, 0.0538, 0.0661, 0.0698, 0.0627, 0.1518, 0.0594], device='cuda:1'), in_proj_covar=tensor([0.0441, 0.0494, 0.0477, 0.0441, 0.0525, 0.0505, 0.0580, 0.0403], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:1') 2023-05-02 12:26:33,118 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=275921.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 12:26:48,363 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.5627, 3.0671, 3.5070, 1.9445, 3.5691, 3.5698, 3.1025, 2.7668], device='cuda:1'), covar=tensor([0.0766, 0.0309, 0.0196, 0.1212, 0.0119, 0.0247, 0.0407, 0.0487], device='cuda:1'), in_proj_covar=tensor([0.0149, 0.0111, 0.0102, 0.0139, 0.0087, 0.0132, 0.0130, 0.0131], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-05-02 12:27:05,249 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.7613, 5.0993, 4.8967, 4.9124, 4.6751, 4.5854, 4.5658, 5.1938], device='cuda:1'), covar=tensor([0.1263, 0.0870, 0.0999, 0.0815, 0.0816, 0.1241, 0.1263, 0.0853], device='cuda:1'), in_proj_covar=tensor([0.0724, 0.0880, 0.0716, 0.0678, 0.0556, 0.0555, 0.0740, 0.0686], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:1') 2023-05-02 12:27:16,671 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=275952.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 12:27:17,449 INFO [train.py:904] (1/8) Epoch 28, batch 1900, loss[loss=0.1615, simple_loss=0.2586, pruned_loss=0.03227, over 17124.00 frames. ], tot_loss[loss=0.1639, simple_loss=0.2516, pruned_loss=0.03811, over 3312236.20 frames. ], batch size: 48, lr: 2.41e-03, grad_scale: 4.0 2023-05-02 12:27:22,944 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.92 vs. limit=2.0 2023-05-02 12:27:31,494 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=275963.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 12:27:39,243 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.3637, 3.9511, 4.4839, 2.5509, 4.5951, 4.7149, 3.6012, 3.6653], device='cuda:1'), covar=tensor([0.0632, 0.0299, 0.0216, 0.1016, 0.0105, 0.0186, 0.0408, 0.0382], device='cuda:1'), in_proj_covar=tensor([0.0148, 0.0110, 0.0101, 0.0139, 0.0086, 0.0131, 0.0129, 0.0130], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-05-02 12:27:43,723 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.9119, 4.6722, 4.9777, 5.1488, 5.3343, 4.7021, 5.3220, 5.3393], device='cuda:1'), covar=tensor([0.2117, 0.1384, 0.1774, 0.0794, 0.0565, 0.1069, 0.0619, 0.0679], device='cuda:1'), in_proj_covar=tensor([0.0693, 0.0843, 0.0981, 0.0858, 0.0652, 0.0685, 0.0719, 0.0832], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-02 12:27:47,686 INFO [optim.py:368] (1/8) 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] (1/8) Epoch 28, batch 1950, loss[loss=0.1659, simple_loss=0.2684, pruned_loss=0.03166, over 17125.00 frames. ], tot_loss[loss=0.1634, simple_loss=0.2515, pruned_loss=0.03768, over 3315664.52 frames. ], batch size: 48, lr: 2.41e-03, grad_scale: 4.0 2023-05-02 12:28:30,811 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=276003.0, num_to_drop=1, layers_to_drop={0} 2023-05-02 12:28:40,998 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=276011.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 12:28:43,427 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=276013.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 12:29:24,686 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=276041.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 12:29:37,515 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 28, batch 2000, loss[loss=0.1407, simple_loss=0.2261, pruned_loss=0.02765, over 16988.00 frames. ], tot_loss[loss=0.1631, simple_loss=0.2516, pruned_loss=0.03727, over 3321163.19 frames. ], batch size: 41, lr: 2.41e-03, grad_scale: 8.0 2023-05-02 12:30:11,361 INFO [optim.py:368] (1/8) 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 12:30:31,278 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=276089.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 12:30:50,228 INFO [train.py:904] (1/8) Epoch 28, batch 2050, loss[loss=0.1783, simple_loss=0.2565, pruned_loss=0.05006, over 16868.00 frames. ], tot_loss[loss=0.1641, simple_loss=0.2521, pruned_loss=0.03808, over 3312848.40 frames. ], batch size: 116, lr: 2.41e-03, grad_scale: 8.0 2023-05-02 12:30:54,894 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.26 vs. limit=2.0 2023-05-02 12:31:08,979 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.88 vs. limit=2.0 2023-05-02 12:31:21,925 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.9148, 5.2504, 5.0559, 5.0445, 4.8109, 4.6915, 4.7408, 5.3788], device='cuda:1'), covar=tensor([0.1355, 0.0948, 0.1023, 0.0948, 0.0855, 0.1196, 0.1214, 0.0897], device='cuda:1'), in_proj_covar=tensor([0.0727, 0.0889, 0.0723, 0.0683, 0.0560, 0.0557, 0.0745, 0.0691], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:1') 2023-05-02 12:31:26,623 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.2989, 2.4055, 2.4188, 4.0783, 2.3185, 2.7252, 2.4730, 2.5557], device='cuda:1'), covar=tensor([0.1434, 0.3688, 0.3237, 0.0617, 0.4268, 0.2606, 0.3481, 0.3581], device='cuda:1'), in_proj_covar=tensor([0.0422, 0.0475, 0.0388, 0.0338, 0.0447, 0.0545, 0.0447, 0.0556], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-02 12:31:31,827 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.5659, 3.7273, 3.9384, 2.7533, 3.5716, 3.9807, 3.6863, 2.4771], device='cuda:1'), covar=tensor([0.0539, 0.0395, 0.0077, 0.0424, 0.0140, 0.0118, 0.0124, 0.0482], device='cuda:1'), in_proj_covar=tensor([0.0140, 0.0092, 0.0092, 0.0137, 0.0103, 0.0116, 0.0100, 0.0133], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0003, 0.0004], device='cuda:1') 2023-05-02 12:32:00,358 INFO [train.py:904] (1/8) Epoch 28, batch 2100, loss[loss=0.1737, simple_loss=0.2601, pruned_loss=0.0436, over 15571.00 frames. ], tot_loss[loss=0.1651, simple_loss=0.2526, pruned_loss=0.0388, over 3313749.67 frames. ], batch size: 191, lr: 2.41e-03, grad_scale: 8.0 2023-05-02 12:32:15,340 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.1096, 4.1078, 4.3914, 4.3890, 4.4221, 4.1894, 4.1922, 4.1513], device='cuda:1'), covar=tensor([0.0405, 0.0760, 0.0478, 0.0472, 0.0578, 0.0497, 0.0836, 0.0668], device='cuda:1'), in_proj_covar=tensor([0.0443, 0.0498, 0.0481, 0.0444, 0.0529, 0.0508, 0.0583, 0.0406], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:1') 2023-05-02 12:32:30,698 INFO [optim.py:368] (1/8) 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 12:33:09,362 INFO [train.py:904] (1/8) Epoch 28, batch 2150, loss[loss=0.1739, simple_loss=0.265, pruned_loss=0.04145, over 16704.00 frames. ], tot_loss[loss=0.1655, simple_loss=0.2533, pruned_loss=0.0389, over 3312507.17 frames. ], batch size: 57, lr: 2.41e-03, grad_scale: 8.0 2023-05-02 12:33:27,927 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=276216.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 12:34:18,066 INFO [train.py:904] (1/8) Epoch 28, batch 2200, loss[loss=0.2003, simple_loss=0.2783, pruned_loss=0.06115, over 17105.00 frames. ], tot_loss[loss=0.1667, simple_loss=0.2542, pruned_loss=0.03966, over 3306349.10 frames. ], batch size: 49, lr: 2.41e-03, grad_scale: 8.0 2023-05-02 12:34:23,901 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.5186, 4.7939, 4.6111, 4.6151, 4.3649, 4.2942, 4.3243, 4.8801], device='cuda:1'), covar=tensor([0.1206, 0.0920, 0.1019, 0.0893, 0.0867, 0.1568, 0.1220, 0.0859], device='cuda:1'), in_proj_covar=tensor([0.0728, 0.0889, 0.0721, 0.0683, 0.0559, 0.0558, 0.0745, 0.0691], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:1') 2023-05-02 12:34:50,532 INFO [optim.py:368] (1/8) 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 12:35:27,984 INFO [train.py:904] (1/8) Epoch 28, batch 2250, loss[loss=0.1584, simple_loss=0.2408, pruned_loss=0.03805, over 16243.00 frames. ], tot_loss[loss=0.1677, simple_loss=0.2552, pruned_loss=0.04009, over 3300174.28 frames. ], batch size: 165, lr: 2.41e-03, grad_scale: 4.0 2023-05-02 12:35:35,390 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=276308.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 12:35:35,449 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([5.0993, 4.8678, 5.1662, 5.3660, 5.5400, 4.8504, 5.5367, 5.5415], device='cuda:1'), covar=tensor([0.2090, 0.1392, 0.1670, 0.0746, 0.0524, 0.0965, 0.0535, 0.0627], device='cuda:1'), in_proj_covar=tensor([0.0700, 0.0852, 0.0990, 0.0867, 0.0659, 0.0691, 0.0725, 0.0840], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-02 12:35:48,046 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=276317.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 12:35:58,354 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.78 vs. limit=2.0 2023-05-02 12:36:32,844 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.4478, 4.2001, 4.2158, 4.6640, 4.7595, 4.3812, 4.6660, 4.7708], device='cuda:1'), covar=tensor([0.1875, 0.1494, 0.2316, 0.0965, 0.0928, 0.1755, 0.2235, 0.1107], device='cuda:1'), in_proj_covar=tensor([0.0699, 0.0851, 0.0988, 0.0865, 0.0658, 0.0689, 0.0724, 0.0839], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-02 12:36:37,084 INFO [train.py:904] (1/8) Epoch 28, batch 2300, loss[loss=0.1607, simple_loss=0.2448, pruned_loss=0.03829, over 16777.00 frames. ], tot_loss[loss=0.1684, simple_loss=0.2558, pruned_loss=0.04053, over 3301433.62 frames. ], batch size: 134, lr: 2.41e-03, grad_scale: 4.0 2023-05-02 12:36:52,436 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.7146, 4.6491, 4.6285, 4.3118, 4.3731, 4.6748, 4.4409, 4.4530], device='cuda:1'), covar=tensor([0.0619, 0.0913, 0.0323, 0.0335, 0.0781, 0.0560, 0.0543, 0.0680], device='cuda:1'), in_proj_covar=tensor([0.0322, 0.0482, 0.0376, 0.0378, 0.0371, 0.0433, 0.0257, 0.0450], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-05-02 12:37:08,704 INFO [optim.py:368] (1/8) 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 12:37:11,196 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.6646, 2.3882, 1.9897, 2.1473, 2.7032, 2.4464, 2.5791, 2.8005], device='cuda:1'), covar=tensor([0.0301, 0.0469, 0.0619, 0.0582, 0.0291, 0.0424, 0.0241, 0.0357], device='cuda:1'), in_proj_covar=tensor([0.0238, 0.0253, 0.0240, 0.0241, 0.0254, 0.0252, 0.0251, 0.0250], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-02 12:37:12,916 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 28, batch 2350, loss[loss=0.1641, simple_loss=0.264, pruned_loss=0.0321, over 17041.00 frames. ], tot_loss[loss=0.1693, simple_loss=0.2565, pruned_loss=0.04105, over 3294171.44 frames. ], batch size: 50, lr: 2.41e-03, grad_scale: 4.0 2023-05-02 12:38:10,927 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.8419, 2.9388, 3.1299, 2.1172, 2.7733, 2.1185, 3.3299, 3.2564], device='cuda:1'), covar=tensor([0.0267, 0.1035, 0.0668, 0.2044, 0.0934, 0.1137, 0.0580, 0.0887], device='cuda:1'), in_proj_covar=tensor([0.0162, 0.0172, 0.0170, 0.0157, 0.0149, 0.0133, 0.0147, 0.0184], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0005], device='cuda:1') 2023-05-02 12:38:54,360 INFO [train.py:904] (1/8) Epoch 28, batch 2400, loss[loss=0.1609, simple_loss=0.2478, pruned_loss=0.03698, over 16793.00 frames. ], tot_loss[loss=0.169, simple_loss=0.2565, pruned_loss=0.04073, over 3308254.95 frames. ], batch size: 42, lr: 2.41e-03, grad_scale: 8.0 2023-05-02 12:39:26,385 INFO [optim.py:368] (1/8) 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 12:39:39,847 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.0990, 4.7552, 4.5966, 3.4671, 3.8792, 4.5928, 4.0702, 3.0406], device='cuda:1'), covar=tensor([0.0488, 0.0088, 0.0048, 0.0349, 0.0178, 0.0085, 0.0111, 0.0416], device='cuda:1'), in_proj_covar=tensor([0.0139, 0.0091, 0.0092, 0.0136, 0.0103, 0.0116, 0.0100, 0.0133], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0003, 0.0004], device='cuda:1') 2023-05-02 12:40:04,313 INFO [train.py:904] (1/8) Epoch 28, batch 2450, loss[loss=0.2091, simple_loss=0.2951, pruned_loss=0.06157, over 11895.00 frames. ], tot_loss[loss=0.1681, simple_loss=0.2563, pruned_loss=0.03994, over 3311767.08 frames. ], batch size: 246, lr: 2.41e-03, grad_scale: 8.0 2023-05-02 12:40:23,636 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=276516.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 12:40:23,815 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.6988, 2.6175, 2.2550, 2.4574, 2.9553, 2.6729, 3.2497, 3.1572], device='cuda:1'), covar=tensor([0.0188, 0.0506, 0.0611, 0.0527, 0.0335, 0.0449, 0.0268, 0.0314], device='cuda:1'), in_proj_covar=tensor([0.0239, 0.0253, 0.0241, 0.0242, 0.0254, 0.0252, 0.0251, 0.0251], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-02 12:40:38,010 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.7297, 3.3527, 3.7747, 1.9739, 3.8439, 3.8924, 3.1877, 2.9753], device='cuda:1'), covar=tensor([0.0698, 0.0274, 0.0184, 0.1194, 0.0103, 0.0215, 0.0385, 0.0418], device='cuda:1'), in_proj_covar=tensor([0.0147, 0.0110, 0.0101, 0.0138, 0.0086, 0.0131, 0.0129, 0.0130], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-05-02 12:41:13,985 INFO [train.py:904] (1/8) Epoch 28, batch 2500, loss[loss=0.145, simple_loss=0.2357, pruned_loss=0.02717, over 17232.00 frames. ], tot_loss[loss=0.168, simple_loss=0.2562, pruned_loss=0.03984, over 3323931.11 frames. ], batch size: 44, lr: 2.41e-03, grad_scale: 8.0 2023-05-02 12:41:23,389 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.6028, 3.4672, 2.7305, 2.1854, 2.2860, 2.3091, 3.5900, 3.0706], device='cuda:1'), covar=tensor([0.2718, 0.0642, 0.1746, 0.3113, 0.2803, 0.2262, 0.0583, 0.1657], device='cuda:1'), in_proj_covar=tensor([0.0338, 0.0280, 0.0317, 0.0331, 0.0309, 0.0281, 0.0309, 0.0358], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:1') 2023-05-02 12:41:30,322 INFO [zipformer.py:625] (1/8) 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] (1/8) 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 12:42:23,404 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.9861, 2.1062, 2.6583, 2.9938, 2.7527, 3.3842, 2.3940, 3.4438], device='cuda:1'), covar=tensor([0.0316, 0.0598, 0.0416, 0.0393, 0.0447, 0.0252, 0.0585, 0.0213], device='cuda:1'), in_proj_covar=tensor([0.0201, 0.0201, 0.0189, 0.0196, 0.0210, 0.0169, 0.0205, 0.0168], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0002], device='cuda:1') 2023-05-02 12:42:24,131 INFO [train.py:904] (1/8) Epoch 28, batch 2550, loss[loss=0.1674, simple_loss=0.264, pruned_loss=0.03541, over 17124.00 frames. ], tot_loss[loss=0.1671, simple_loss=0.2558, pruned_loss=0.03915, over 3327811.53 frames. ], batch size: 48, lr: 2.41e-03, grad_scale: 8.0 2023-05-02 12:42:31,823 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=276608.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 12:42:35,368 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=276611.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 12:43:28,805 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.7791, 5.0035, 5.1268, 4.8756, 4.9144, 5.5345, 4.9890, 4.6898], device='cuda:1'), covar=tensor([0.1498, 0.2045, 0.2525, 0.2275, 0.2816, 0.1050, 0.1680, 0.2703], device='cuda:1'), in_proj_covar=tensor([0.0437, 0.0654, 0.0718, 0.0532, 0.0707, 0.0745, 0.0559, 0.0708], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-05-02 12:43:33,172 INFO [train.py:904] (1/8) Epoch 28, batch 2600, loss[loss=0.1542, simple_loss=0.2404, pruned_loss=0.03399, over 16873.00 frames. ], tot_loss[loss=0.1668, simple_loss=0.2559, pruned_loss=0.03887, over 3324215.16 frames. ], batch size: 90, lr: 2.41e-03, grad_scale: 8.0 2023-05-02 12:43:37,973 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=276656.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 12:43:55,732 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=276668.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 12:44:01,028 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=276672.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 12:44:01,987 INFO [zipformer.py:625] (1/8) 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] (1/8) 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 12:44:41,532 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.22 vs. limit=2.0 2023-05-02 12:44:43,682 INFO [train.py:904] (1/8) Epoch 28, batch 2650, loss[loss=0.186, simple_loss=0.2707, pruned_loss=0.05061, over 16184.00 frames. ], tot_loss[loss=0.1666, simple_loss=0.2557, pruned_loss=0.0387, over 3321637.60 frames. ], batch size: 165, lr: 2.41e-03, grad_scale: 8.0 2023-05-02 12:45:19,131 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=276729.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 12:45:50,739 INFO [train.py:904] (1/8) Epoch 28, batch 2700, loss[loss=0.1667, simple_loss=0.2639, pruned_loss=0.0347, over 16842.00 frames. ], tot_loss[loss=0.1663, simple_loss=0.2561, pruned_loss=0.03825, over 3327921.00 frames. ], batch size: 42, lr: 2.41e-03, grad_scale: 8.0 2023-05-02 12:46:23,524 INFO [optim.py:368] (1/8) 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 12:46:42,604 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=276790.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 12:46:56,322 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([1.7695, 2.7369, 2.5241, 4.2563, 3.5068, 4.1209, 1.5939, 2.9265], device='cuda:1'), covar=tensor([0.1418, 0.0724, 0.1210, 0.0181, 0.0171, 0.0376, 0.1644, 0.0855], device='cuda:1'), in_proj_covar=tensor([0.0174, 0.0181, 0.0201, 0.0205, 0.0207, 0.0220, 0.0210, 0.0200], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-05-02 12:47:00,576 INFO [train.py:904] (1/8) Epoch 28, batch 2750, loss[loss=0.2261, simple_loss=0.3036, pruned_loss=0.07429, over 11986.00 frames. ], tot_loss[loss=0.1663, simple_loss=0.2561, pruned_loss=0.03822, over 3328303.18 frames. ], batch size: 246, lr: 2.41e-03, grad_scale: 8.0 2023-05-02 12:48:07,241 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 28, batch 2800, loss[loss=0.1573, simple_loss=0.2461, pruned_loss=0.03425, over 16999.00 frames. ], tot_loss[loss=0.1658, simple_loss=0.2556, pruned_loss=0.03804, over 3334411.97 frames. ], batch size: 41, lr: 2.41e-03, grad_scale: 8.0 2023-05-02 12:48:25,007 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=276863.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 12:48:25,266 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.23 vs. limit=2.0 2023-05-02 12:48:41,263 INFO [optim.py:368] (1/8) 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] (1/8) Epoch 28, batch 2850, loss[loss=0.1536, simple_loss=0.2561, pruned_loss=0.02554, over 17250.00 frames. ], tot_loss[loss=0.1647, simple_loss=0.2545, pruned_loss=0.03749, over 3337188.39 frames. ], batch size: 52, lr: 2.41e-03, grad_scale: 8.0 2023-05-02 12:49:50,146 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=276924.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 12:49:57,668 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=276931.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 12:50:03,121 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.8747, 2.8569, 2.3720, 2.6769, 3.1162, 2.9004, 3.4640, 3.3578], device='cuda:1'), covar=tensor([0.0192, 0.0522, 0.0668, 0.0503, 0.0365, 0.0461, 0.0278, 0.0328], device='cuda:1'), in_proj_covar=tensor([0.0239, 0.0253, 0.0240, 0.0241, 0.0254, 0.0251, 0.0252, 0.0250], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-02 12:50:27,301 INFO [train.py:904] (1/8) Epoch 28, batch 2900, loss[loss=0.1744, simple_loss=0.2422, pruned_loss=0.05333, over 16886.00 frames. ], tot_loss[loss=0.1644, simple_loss=0.2534, pruned_loss=0.0377, over 3334412.71 frames. ], batch size: 116, lr: 2.41e-03, grad_scale: 8.0 2023-05-02 12:50:46,672 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=276967.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 12:50:54,974 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=276973.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 12:50:58,048 INFO [optim.py:368] (1/8) 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 12:51:22,007 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=276992.0, num_to_drop=1, layers_to_drop={0} 2023-05-02 12:51:27,122 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([5.0957, 5.4663, 5.6256, 5.3680, 5.4197, 5.9807, 5.4573, 5.1920], device='cuda:1'), covar=tensor([0.1037, 0.1993, 0.2663, 0.1991, 0.2545, 0.1000, 0.1624, 0.2523], device='cuda:1'), in_proj_covar=tensor([0.0435, 0.0655, 0.0718, 0.0532, 0.0706, 0.0742, 0.0558, 0.0708], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-05-02 12:51:36,311 INFO [train.py:904] (1/8) Epoch 28, batch 2950, loss[loss=0.1842, simple_loss=0.2552, pruned_loss=0.0566, over 16523.00 frames. ], tot_loss[loss=0.1653, simple_loss=0.2535, pruned_loss=0.03858, over 3324730.30 frames. ], batch size: 146, lr: 2.41e-03, grad_scale: 8.0 2023-05-02 12:51:48,558 INFO [zipformer.py:625] (1/8) 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] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=277021.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 12:52:05,380 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=277024.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 12:52:45,412 INFO [train.py:904] (1/8) Epoch 28, batch 3000, loss[loss=0.1566, simple_loss=0.2538, pruned_loss=0.02974, over 17050.00 frames. ], tot_loss[loss=0.1644, simple_loss=0.2528, pruned_loss=0.03804, over 3332652.11 frames. ], batch size: 53, lr: 2.41e-03, grad_scale: 8.0 2023-05-02 12:52:45,413 INFO [train.py:929] (1/8) Computing validation loss 2023-05-02 12:52:54,788 INFO [train.py:938] (1/8) Epoch 28, validation: loss=0.1335, simple_loss=0.2385, pruned_loss=0.01427, over 944034.00 frames. 2023-05-02 12:52:54,789 INFO [train.py:939] (1/8) Maximum memory allocated so far is 18145MB 2023-05-02 12:53:21,133 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=277072.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 12:53:25,663 INFO [optim.py:368] (1/8) 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 12:53:27,144 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.7406, 5.0255, 4.8372, 4.8169, 4.6255, 4.5386, 4.4613, 5.1059], device='cuda:1'), covar=tensor([0.1231, 0.0897, 0.1005, 0.0897, 0.0783, 0.1234, 0.1288, 0.0856], device='cuda:1'), in_proj_covar=tensor([0.0737, 0.0899, 0.0734, 0.0692, 0.0567, 0.0562, 0.0754, 0.0700], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:1') 2023-05-02 12:53:55,131 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.6721, 1.9257, 2.3012, 2.5262, 2.6437, 2.5978, 1.8895, 2.7891], device='cuda:1'), covar=tensor([0.0232, 0.0562, 0.0386, 0.0350, 0.0351, 0.0393, 0.0641, 0.0239], device='cuda:1'), in_proj_covar=tensor([0.0203, 0.0202, 0.0191, 0.0197, 0.0212, 0.0170, 0.0207, 0.0170], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0002], device='cuda:1') 2023-05-02 12:54:02,141 INFO [train.py:904] (1/8) Epoch 28, batch 3050, loss[loss=0.1805, simple_loss=0.2861, pruned_loss=0.03745, over 16706.00 frames. ], tot_loss[loss=0.1646, simple_loss=0.2528, pruned_loss=0.03822, over 3330671.88 frames. ], batch size: 57, lr: 2.41e-03, grad_scale: 8.0 2023-05-02 12:54:13,077 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.5763, 4.4041, 4.6233, 4.7645, 4.8839, 4.4311, 4.7681, 4.9044], device='cuda:1'), covar=tensor([0.1802, 0.1301, 0.1527, 0.0796, 0.0728, 0.1192, 0.1987, 0.0997], device='cuda:1'), in_proj_covar=tensor([0.0698, 0.0851, 0.0989, 0.0867, 0.0658, 0.0691, 0.0725, 0.0837], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-02 12:55:01,327 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 28, batch 3100, loss[loss=0.1665, simple_loss=0.2472, pruned_loss=0.04293, over 16483.00 frames. ], tot_loss[loss=0.1646, simple_loss=0.2523, pruned_loss=0.03848, over 3330279.60 frames. ], batch size: 75, lr: 2.41e-03, grad_scale: 8.0 2023-05-02 12:55:43,284 INFO [optim.py:368] (1/8) 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 12:55:45,335 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=277177.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 12:56:05,049 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([5.3075, 5.8399, 6.0361, 5.5863, 5.7922, 6.3443, 5.8034, 5.4273], device='cuda:1'), covar=tensor([0.0962, 0.1926, 0.2200, 0.2149, 0.2509, 0.0956, 0.1700, 0.2550], device='cuda:1'), in_proj_covar=tensor([0.0436, 0.0655, 0.0720, 0.0533, 0.0707, 0.0743, 0.0558, 0.0709], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-05-02 12:56:21,091 INFO [train.py:904] (1/8) Epoch 28, batch 3150, loss[loss=0.1834, simple_loss=0.26, pruned_loss=0.05338, over 16889.00 frames. ], tot_loss[loss=0.1636, simple_loss=0.2514, pruned_loss=0.03791, over 3331948.32 frames. ], batch size: 109, lr: 2.41e-03, grad_scale: 8.0 2023-05-02 12:56:44,545 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=277219.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 12:56:44,679 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=277219.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 12:56:50,337 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.8201, 4.3415, 3.0189, 2.3480, 2.6501, 2.5939, 4.7450, 3.5591], device='cuda:1'), covar=tensor([0.2993, 0.0599, 0.1925, 0.3166, 0.2858, 0.2217, 0.0332, 0.1455], device='cuda:1'), in_proj_covar=tensor([0.0337, 0.0279, 0.0315, 0.0330, 0.0308, 0.0280, 0.0308, 0.0358], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:1') 2023-05-02 12:57:10,216 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=277238.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 12:57:15,063 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.7677, 5.1028, 4.8694, 4.8582, 4.6640, 4.6297, 4.5252, 5.1763], device='cuda:1'), covar=tensor([0.1335, 0.0875, 0.1043, 0.0931, 0.0920, 0.1160, 0.1231, 0.0903], device='cuda:1'), in_proj_covar=tensor([0.0739, 0.0902, 0.0735, 0.0693, 0.0568, 0.0563, 0.0755, 0.0701], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:1') 2023-05-02 12:57:30,727 INFO [train.py:904] (1/8) Epoch 28, batch 3200, loss[loss=0.1491, simple_loss=0.2349, pruned_loss=0.03164, over 16821.00 frames. ], tot_loss[loss=0.1626, simple_loss=0.2508, pruned_loss=0.03722, over 3335974.34 frames. ], batch size: 42, lr: 2.41e-03, grad_scale: 8.0 2023-05-02 12:57:49,829 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.7939, 3.9749, 2.8985, 2.3782, 2.7067, 2.5123, 4.1231, 3.4833], device='cuda:1'), covar=tensor([0.2958, 0.0660, 0.1909, 0.2945, 0.2663, 0.2279, 0.0538, 0.1460], device='cuda:1'), in_proj_covar=tensor([0.0337, 0.0279, 0.0315, 0.0330, 0.0309, 0.0280, 0.0308, 0.0358], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:1') 2023-05-02 12:57:50,704 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=277267.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 12:58:01,812 INFO [optim.py:368] (1/8) 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 12:58:07,085 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=277280.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 12:58:17,849 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=277287.0, num_to_drop=1, layers_to_drop={3} 2023-05-02 12:58:29,264 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.8804, 2.5096, 2.0324, 2.3002, 2.8472, 2.6234, 2.8010, 2.9944], device='cuda:1'), covar=tensor([0.0285, 0.0469, 0.0639, 0.0544, 0.0285, 0.0418, 0.0237, 0.0321], device='cuda:1'), in_proj_covar=tensor([0.0240, 0.0253, 0.0240, 0.0241, 0.0255, 0.0251, 0.0252, 0.0251], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-02 12:58:39,071 INFO [train.py:904] (1/8) Epoch 28, batch 3250, loss[loss=0.1953, simple_loss=0.277, pruned_loss=0.0568, over 16959.00 frames. ], tot_loss[loss=0.163, simple_loss=0.251, pruned_loss=0.03748, over 3332364.33 frames. ], batch size: 109, lr: 2.41e-03, grad_scale: 8.0 2023-05-02 12:58:42,402 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=277305.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 12:58:55,897 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=277315.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 12:59:09,074 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=277324.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 12:59:47,703 INFO [train.py:904] (1/8) Epoch 28, batch 3300, loss[loss=0.1798, simple_loss=0.2588, pruned_loss=0.05044, over 16779.00 frames. ], tot_loss[loss=0.1642, simple_loss=0.2521, pruned_loss=0.03818, over 3323020.22 frames. ], batch size: 124, lr: 2.41e-03, grad_scale: 8.0 2023-05-02 13:00:05,309 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=277366.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 13:00:06,285 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=277367.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 13:00:12,568 INFO [zipformer.py:625] (1/8) 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] (1/8) 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 13:00:34,351 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.8633, 2.2015, 2.4298, 3.1134, 2.2272, 2.4025, 2.3550, 2.3131], device='cuda:1'), covar=tensor([0.1497, 0.3407, 0.2631, 0.0767, 0.3913, 0.2333, 0.3425, 0.3171], device='cuda:1'), in_proj_covar=tensor([0.0427, 0.0480, 0.0391, 0.0342, 0.0450, 0.0550, 0.0451, 0.0561], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-02 13:00:55,902 INFO [train.py:904] (1/8) Epoch 28, batch 3350, loss[loss=0.1741, simple_loss=0.253, pruned_loss=0.04764, over 16251.00 frames. ], tot_loss[loss=0.1656, simple_loss=0.2535, pruned_loss=0.03885, over 3320197.39 frames. ], batch size: 165, lr: 2.41e-03, grad_scale: 8.0 2023-05-02 13:01:25,864 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([5.0394, 5.1215, 5.5211, 5.4751, 5.5004, 5.1615, 5.1018, 4.9247], device='cuda:1'), covar=tensor([0.0347, 0.0493, 0.0359, 0.0414, 0.0456, 0.0397, 0.0986, 0.0476], device='cuda:1'), in_proj_covar=tensor([0.0450, 0.0509, 0.0489, 0.0452, 0.0538, 0.0516, 0.0596, 0.0414], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0003], device='cuda:1') 2023-05-02 13:01:26,034 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.4838, 3.5239, 3.9859, 2.2019, 3.1795, 2.5572, 3.8949, 3.7908], device='cuda:1'), covar=tensor([0.0292, 0.1023, 0.0499, 0.2223, 0.0885, 0.1026, 0.0641, 0.1119], device='cuda:1'), in_proj_covar=tensor([0.0163, 0.0174, 0.0171, 0.0158, 0.0150, 0.0134, 0.0148, 0.0186], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0005], device='cuda:1') 2023-05-02 13:01:55,020 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=277446.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 13:02:03,950 INFO [train.py:904] (1/8) Epoch 28, batch 3400, loss[loss=0.157, simple_loss=0.2538, pruned_loss=0.03008, over 17139.00 frames. ], tot_loss[loss=0.1656, simple_loss=0.2535, pruned_loss=0.03885, over 3329369.56 frames. ], batch size: 48, lr: 2.41e-03, grad_scale: 8.0 2023-05-02 13:02:23,262 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([5.1599, 5.5699, 5.6701, 5.3526, 5.4755, 6.0320, 5.5475, 5.2778], device='cuda:1'), covar=tensor([0.0983, 0.2046, 0.2583, 0.2392, 0.2779, 0.1019, 0.1580, 0.2592], device='cuda:1'), in_proj_covar=tensor([0.0435, 0.0654, 0.0718, 0.0533, 0.0708, 0.0742, 0.0556, 0.0707], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-05-02 13:02:34,599 INFO [optim.py:368] (1/8) Clipping_scale=2.0, grad-norm quartiles 1.382e+02 2.063e+02 2.423e+02 2.740e+02 5.592e+02, threshold=4.846e+02, percent-clipped=2.0 2023-05-02 13:03:01,406 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=277494.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 13:03:13,661 INFO [train.py:904] (1/8) Epoch 28, batch 3450, loss[loss=0.1498, simple_loss=0.2309, pruned_loss=0.03432, over 16820.00 frames. ], tot_loss[loss=0.1643, simple_loss=0.252, pruned_loss=0.03834, over 3329832.99 frames. ], batch size: 83, lr: 2.41e-03, grad_scale: 8.0 2023-05-02 13:03:14,084 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([5.3636, 5.3075, 5.1992, 4.6824, 4.8229, 5.2712, 5.1938, 4.8535], device='cuda:1'), covar=tensor([0.0618, 0.0593, 0.0333, 0.0398, 0.1158, 0.0562, 0.0363, 0.0868], device='cuda:1'), in_proj_covar=tensor([0.0326, 0.0490, 0.0382, 0.0384, 0.0378, 0.0440, 0.0260, 0.0457], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-05-02 13:03:35,940 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=277519.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 13:03:54,312 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([5.2869, 5.6426, 5.4216, 5.4787, 5.1549, 5.0796, 5.1113, 5.7259], device='cuda:1'), covar=tensor([0.1379, 0.0914, 0.0980, 0.0872, 0.0833, 0.0879, 0.1265, 0.0932], device='cuda:1'), in_proj_covar=tensor([0.0741, 0.0905, 0.0735, 0.0694, 0.0569, 0.0564, 0.0758, 0.0703], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:1') 2023-05-02 13:03:55,436 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=277533.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 13:03:57,639 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([5.3222, 5.2848, 5.1670, 4.6277, 4.7937, 5.2394, 5.1781, 4.8396], device='cuda:1'), covar=tensor([0.0598, 0.0476, 0.0339, 0.0419, 0.1170, 0.0460, 0.0338, 0.0844], device='cuda:1'), in_proj_covar=tensor([0.0326, 0.0491, 0.0382, 0.0384, 0.0379, 0.0440, 0.0260, 0.0457], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-05-02 13:04:23,507 INFO [train.py:904] (1/8) Epoch 28, batch 3500, loss[loss=0.1342, simple_loss=0.2233, pruned_loss=0.02258, over 16837.00 frames. ], tot_loss[loss=0.163, simple_loss=0.2509, pruned_loss=0.03752, over 3329871.36 frames. ], batch size: 42, lr: 2.41e-03, grad_scale: 8.0 2023-05-02 13:04:25,774 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=277554.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 13:04:27,074 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.5929, 3.5814, 2.6195, 2.2814, 2.2953, 2.1991, 3.5805, 3.0724], device='cuda:1'), covar=tensor([0.2900, 0.0716, 0.2047, 0.2968, 0.2924, 0.2558, 0.0673, 0.1743], device='cuda:1'), in_proj_covar=tensor([0.0335, 0.0277, 0.0313, 0.0328, 0.0307, 0.0278, 0.0306, 0.0356], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:1') 2023-05-02 13:04:42,186 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=277567.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 13:04:47,018 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=277570.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 13:04:54,493 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=277575.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 13:04:55,302 INFO [optim.py:368] (1/8) 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 13:05:10,621 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=277587.0, num_to_drop=1, layers_to_drop={1} 2023-05-02 13:05:29,005 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.0724, 3.8114, 4.3397, 2.1863, 4.4805, 4.6360, 3.3759, 3.4888], device='cuda:1'), covar=tensor([0.0701, 0.0292, 0.0213, 0.1179, 0.0085, 0.0169, 0.0437, 0.0427], device='cuda:1'), in_proj_covar=tensor([0.0150, 0.0112, 0.0104, 0.0141, 0.0088, 0.0134, 0.0132, 0.0132], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-05-02 13:05:32,630 INFO [train.py:904] (1/8) Epoch 28, batch 3550, loss[loss=0.144, simple_loss=0.2313, pruned_loss=0.02834, over 16875.00 frames. ], tot_loss[loss=0.1624, simple_loss=0.2505, pruned_loss=0.03719, over 3328978.73 frames. ], batch size: 42, lr: 2.41e-03, grad_scale: 8.0 2023-05-02 13:05:49,364 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=277615.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 13:06:11,019 INFO [zipformer.py:625] (1/8) 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] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=277635.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 13:06:35,599 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=4.76 vs. limit=5.0 2023-05-02 13:06:42,326 INFO [train.py:904] (1/8) Epoch 28, batch 3600, loss[loss=0.146, simple_loss=0.2339, pruned_loss=0.02902, over 16224.00 frames. ], tot_loss[loss=0.1613, simple_loss=0.249, pruned_loss=0.03678, over 3327507.34 frames. ], batch size: 36, lr: 2.41e-03, grad_scale: 8.0 2023-05-02 13:06:47,446 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.34 vs. limit=2.0 2023-05-02 13:06:52,883 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=277661.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 13:07:01,952 INFO [zipformer.py:625] (1/8) 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] (1/8) 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 13:07:36,963 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 28, batch 3650, loss[loss=0.1529, simple_loss=0.2497, pruned_loss=0.02802, over 17135.00 frames. ], tot_loss[loss=0.161, simple_loss=0.248, pruned_loss=0.03704, over 3329976.28 frames. ], batch size: 49, lr: 2.41e-03, grad_scale: 8.0 2023-05-02 13:08:12,452 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=277715.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 13:08:27,025 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.8559, 4.0509, 3.0431, 2.5038, 2.7820, 2.6789, 4.4793, 3.5224], device='cuda:1'), covar=tensor([0.2839, 0.0701, 0.1803, 0.2765, 0.2621, 0.2130, 0.0398, 0.1431], device='cuda:1'), in_proj_covar=tensor([0.0336, 0.0279, 0.0315, 0.0329, 0.0308, 0.0280, 0.0307, 0.0357], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:1') 2023-05-02 13:09:03,386 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([5.0833, 5.1849, 5.4887, 5.4678, 5.5125, 5.1987, 5.1249, 4.9307], device='cuda:1'), covar=tensor([0.0315, 0.0492, 0.0365, 0.0388, 0.0395, 0.0330, 0.0907, 0.0475], device='cuda:1'), in_proj_covar=tensor([0.0449, 0.0507, 0.0488, 0.0450, 0.0537, 0.0514, 0.0593, 0.0413], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0003], device='cuda:1') 2023-05-02 13:09:06,651 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=277752.0, num_to_drop=1, layers_to_drop={2} 2023-05-02 13:09:07,228 INFO [train.py:904] (1/8) Epoch 28, batch 3700, loss[loss=0.1587, simple_loss=0.2383, pruned_loss=0.03958, over 16800.00 frames. ], tot_loss[loss=0.1615, simple_loss=0.2466, pruned_loss=0.03817, over 3312924.84 frames. ], batch size: 102, lr: 2.41e-03, grad_scale: 8.0 2023-05-02 13:09:41,446 INFO [optim.py:368] (1/8) 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 13:10:22,497 INFO [train.py:904] (1/8) Epoch 28, batch 3750, loss[loss=0.1933, simple_loss=0.2706, pruned_loss=0.05801, over 16939.00 frames. ], tot_loss[loss=0.1634, simple_loss=0.247, pruned_loss=0.0399, over 3308452.31 frames. ], batch size: 116, lr: 2.40e-03, grad_scale: 4.0 2023-05-02 13:10:49,650 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.6303, 4.9340, 4.7437, 4.7277, 4.5266, 4.4637, 4.4211, 5.0286], device='cuda:1'), covar=tensor([0.1332, 0.0950, 0.0969, 0.0895, 0.0820, 0.1187, 0.1209, 0.0917], device='cuda:1'), in_proj_covar=tensor([0.0739, 0.0903, 0.0734, 0.0693, 0.0569, 0.0563, 0.0755, 0.0702], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:1') 2023-05-02 13:11:06,934 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=277833.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 13:11:10,975 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.34 vs. limit=2.0 2023-05-02 13:11:35,738 INFO [train.py:904] (1/8) Epoch 28, batch 3800, loss[loss=0.1927, simple_loss=0.2702, pruned_loss=0.05763, over 16687.00 frames. ], tot_loss[loss=0.1656, simple_loss=0.2484, pruned_loss=0.04146, over 3294458.27 frames. ], batch size: 35, lr: 2.40e-03, grad_scale: 4.0 2023-05-02 13:12:08,721 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=277875.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 13:12:11,081 INFO [optim.py:368] (1/8) 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] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=277881.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 13:12:48,150 INFO [train.py:904] (1/8) Epoch 28, batch 3850, loss[loss=0.1586, simple_loss=0.2384, pruned_loss=0.03944, over 16726.00 frames. ], tot_loss[loss=0.1665, simple_loss=0.2488, pruned_loss=0.04212, over 3293779.17 frames. ], batch size: 83, lr: 2.40e-03, grad_scale: 4.0 2023-05-02 13:12:59,660 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=277910.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 13:13:17,228 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.0751, 4.0770, 4.0082, 3.3605, 4.0123, 1.7374, 3.8216, 3.3690], device='cuda:1'), covar=tensor([0.0129, 0.0120, 0.0204, 0.0275, 0.0088, 0.3188, 0.0127, 0.0335], device='cuda:1'), in_proj_covar=tensor([0.0186, 0.0180, 0.0218, 0.0191, 0.0194, 0.0224, 0.0208, 0.0186], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-02 13:13:18,279 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=277923.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 13:13:23,017 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=277926.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 13:14:00,814 INFO [train.py:904] (1/8) Epoch 28, batch 3900, loss[loss=0.1538, simple_loss=0.2285, pruned_loss=0.03958, over 16811.00 frames. ], tot_loss[loss=0.1663, simple_loss=0.2477, pruned_loss=0.04249, over 3292943.31 frames. ], batch size: 96, lr: 2.40e-03, grad_scale: 4.0 2023-05-02 13:14:13,109 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=277961.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 13:14:27,646 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=277971.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 13:14:36,141 INFO [optim.py:368] (1/8) 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 13:14:55,984 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.30 vs. limit=2.0 2023-05-02 13:15:16,801 INFO [train.py:904] (1/8) Epoch 28, batch 3950, loss[loss=0.1647, simple_loss=0.2377, pruned_loss=0.04588, over 16519.00 frames. ], tot_loss[loss=0.1669, simple_loss=0.2475, pruned_loss=0.0432, over 3293239.12 frames. ], batch size: 75, lr: 2.40e-03, grad_scale: 4.0 2023-05-02 13:15:25,690 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=278009.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 13:15:58,221 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=278032.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 13:16:10,078 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.4730, 4.3311, 4.5227, 4.6523, 4.7530, 4.3341, 4.5528, 4.7585], device='cuda:1'), covar=tensor([0.1707, 0.1244, 0.1395, 0.0745, 0.0655, 0.1160, 0.2661, 0.0856], device='cuda:1'), in_proj_covar=tensor([0.0702, 0.0855, 0.0994, 0.0872, 0.0662, 0.0694, 0.0728, 0.0840], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-02 13:16:19,830 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=278047.0, num_to_drop=1, layers_to_drop={0} 2023-05-02 13:16:20,191 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.20 vs. limit=2.0 2023-05-02 13:16:26,864 INFO [train.py:904] (1/8) Epoch 28, batch 4000, loss[loss=0.1624, simple_loss=0.2477, pruned_loss=0.03853, over 16738.00 frames. ], tot_loss[loss=0.1673, simple_loss=0.2474, pruned_loss=0.04363, over 3292686.51 frames. ], batch size: 83, lr: 2.40e-03, grad_scale: 8.0 2023-05-02 13:16:27,719 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.85 vs. limit=2.0 2023-05-02 13:16:32,651 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=2.80 vs. limit=5.0 2023-05-02 13:17:01,537 INFO [optim.py:368] (1/8) 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 13:17:17,014 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.7666, 4.8339, 5.1249, 5.0745, 5.1247, 4.8264, 4.7818, 4.6613], device='cuda:1'), covar=tensor([0.0339, 0.0573, 0.0388, 0.0454, 0.0469, 0.0392, 0.0900, 0.0497], device='cuda:1'), in_proj_covar=tensor([0.0448, 0.0506, 0.0486, 0.0449, 0.0535, 0.0513, 0.0591, 0.0412], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0003], device='cuda:1') 2023-05-02 13:17:35,438 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.6320, 2.3839, 1.9747, 2.2438, 2.6816, 2.3267, 2.3489, 2.7818], device='cuda:1'), covar=tensor([0.0223, 0.0433, 0.0619, 0.0472, 0.0284, 0.0418, 0.0270, 0.0271], device='cuda:1'), in_proj_covar=tensor([0.0238, 0.0251, 0.0237, 0.0238, 0.0251, 0.0249, 0.0250, 0.0249], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-02 13:17:38,521 INFO [train.py:904] (1/8) Epoch 28, batch 4050, loss[loss=0.1771, simple_loss=0.2611, pruned_loss=0.04661, over 16543.00 frames. ], tot_loss[loss=0.168, simple_loss=0.2492, pruned_loss=0.04342, over 3287694.22 frames. ], batch size: 75, lr: 2.40e-03, grad_scale: 8.0 2023-05-02 13:17:56,698 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.5558, 4.8899, 5.1910, 5.1305, 5.1505, 4.7962, 4.5189, 4.6169], device='cuda:1'), covar=tensor([0.0606, 0.0598, 0.0480, 0.0550, 0.0769, 0.0584, 0.1393, 0.0572], device='cuda:1'), in_proj_covar=tensor([0.0447, 0.0505, 0.0485, 0.0448, 0.0535, 0.0512, 0.0589, 0.0412], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0003], device='cuda:1') 2023-05-02 13:18:00,876 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.1456, 4.1360, 4.2910, 4.3789, 4.4903, 4.0782, 4.4250, 4.5570], device='cuda:1'), covar=tensor([0.1919, 0.1084, 0.1330, 0.0668, 0.0541, 0.1274, 0.0879, 0.0581], device='cuda:1'), in_proj_covar=tensor([0.0704, 0.0858, 0.0996, 0.0873, 0.0664, 0.0695, 0.0728, 0.0842], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-02 13:18:52,733 INFO [train.py:904] (1/8) Epoch 28, batch 4100, loss[loss=0.1912, simple_loss=0.2715, pruned_loss=0.05545, over 12052.00 frames. ], tot_loss[loss=0.1687, simple_loss=0.2513, pruned_loss=0.04301, over 3270306.46 frames. ], batch size: 247, lr: 2.40e-03, grad_scale: 8.0 2023-05-02 13:18:53,564 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=2.99 vs. limit=5.0 2023-05-02 13:19:08,233 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.46 vs. limit=2.0 2023-05-02 13:19:29,061 INFO [optim.py:368] (1/8) Clipping_scale=2.0, grad-norm quartiles 1.475e+02 1.827e+02 2.078e+02 2.388e+02 3.753e+02, threshold=4.156e+02, percent-clipped=0.0 2023-05-02 13:20:09,078 INFO [train.py:904] (1/8) Epoch 28, batch 4150, loss[loss=0.2221, simple_loss=0.2981, pruned_loss=0.07307, over 11717.00 frames. ], tot_loss[loss=0.1738, simple_loss=0.2575, pruned_loss=0.04501, over 3249034.44 frames. ], batch size: 246, lr: 2.40e-03, grad_scale: 8.0 2023-05-02 13:20:19,187 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=278210.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 13:20:39,975 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.61 vs. limit=2.0 2023-05-02 13:20:42,429 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=278226.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 13:21:09,368 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.8988, 5.0796, 5.2712, 4.9155, 4.9682, 5.6186, 5.0775, 4.7960], device='cuda:1'), covar=tensor([0.0965, 0.1693, 0.1842, 0.1805, 0.2512, 0.0823, 0.1466, 0.2176], device='cuda:1'), in_proj_covar=tensor([0.0431, 0.0647, 0.0710, 0.0529, 0.0700, 0.0733, 0.0552, 0.0701], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-05-02 13:21:22,143 INFO [train.py:904] (1/8) Epoch 28, batch 4200, loss[loss=0.1989, simple_loss=0.2882, pruned_loss=0.05481, over 16943.00 frames. ], tot_loss[loss=0.1794, simple_loss=0.2652, pruned_loss=0.04686, over 3217659.34 frames. ], batch size: 109, lr: 2.40e-03, grad_scale: 8.0 2023-05-02 13:21:30,670 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=278258.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 13:21:44,118 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.0866, 2.2823, 2.1602, 3.6493, 2.1542, 2.5945, 2.3136, 2.3952], device='cuda:1'), covar=tensor([0.1565, 0.3842, 0.3415, 0.0706, 0.4553, 0.2630, 0.3850, 0.3432], device='cuda:1'), in_proj_covar=tensor([0.0423, 0.0477, 0.0386, 0.0339, 0.0444, 0.0547, 0.0448, 0.0558], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-02 13:21:54,354 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=278274.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 13:21:58,171 INFO [optim.py:368] (1/8) 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 13:22:22,670 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.13 vs. limit=2.0 2023-05-02 13:22:37,391 INFO [train.py:904] (1/8) Epoch 28, batch 4250, loss[loss=0.16, simple_loss=0.2554, pruned_loss=0.03226, over 16474.00 frames. ], tot_loss[loss=0.181, simple_loss=0.2685, pruned_loss=0.04673, over 3189110.69 frames. ], batch size: 146, lr: 2.40e-03, grad_scale: 8.0 2023-05-02 13:22:45,353 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=4.83 vs. limit=5.0 2023-05-02 13:23:07,826 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=278323.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 13:23:13,448 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=278327.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 13:23:44,105 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=278347.0, num_to_drop=1, layers_to_drop={2} 2023-05-02 13:23:52,555 INFO [train.py:904] (1/8) Epoch 28, batch 4300, loss[loss=0.1872, simple_loss=0.2764, pruned_loss=0.04898, over 16715.00 frames. ], tot_loss[loss=0.1812, simple_loss=0.2697, pruned_loss=0.04641, over 3171775.33 frames. ], batch size: 124, lr: 2.40e-03, grad_scale: 8.0 2023-05-02 13:24:25,570 INFO [zipformer.py:625] (1/8) 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] (1/8) 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 13:24:40,239 INFO [zipformer.py:625] (1/8) 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] (1/8) 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] (1/8) Epoch 28, batch 4350, loss[loss=0.1915, simple_loss=0.284, pruned_loss=0.04951, over 16903.00 frames. ], tot_loss[loss=0.1842, simple_loss=0.2733, pruned_loss=0.04758, over 3185094.34 frames. ], batch size: 109, lr: 2.40e-03, grad_scale: 8.0 2023-05-02 13:25:12,291 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.23 vs. limit=2.0 2023-05-02 13:25:56,487 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=278435.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 13:26:22,393 INFO [train.py:904] (1/8) Epoch 28, batch 4400, loss[loss=0.2055, simple_loss=0.2875, pruned_loss=0.06174, over 11925.00 frames. ], tot_loss[loss=0.1867, simple_loss=0.2756, pruned_loss=0.04891, over 3173711.06 frames. ], batch size: 248, lr: 2.40e-03, grad_scale: 8.0 2023-05-02 13:26:58,177 INFO [optim.py:368] (1/8) Clipping_scale=2.0, grad-norm quartiles 1.489e+02 2.247e+02 2.621e+02 3.049e+02 4.749e+02, threshold=5.243e+02, percent-clipped=0.0 2023-05-02 13:27:35,968 INFO [train.py:904] (1/8) Epoch 28, batch 4450, loss[loss=0.2077, simple_loss=0.2895, pruned_loss=0.06294, over 16242.00 frames. ], tot_loss[loss=0.1899, simple_loss=0.279, pruned_loss=0.05045, over 3183919.21 frames. ], batch size: 35, lr: 2.40e-03, grad_scale: 8.0 2023-05-02 13:27:43,093 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.15 vs. limit=2.0 2023-05-02 13:27:47,295 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=278511.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 13:27:58,487 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.40 vs. limit=2.0 2023-05-02 13:28:33,727 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.4692, 1.7395, 2.1010, 2.3975, 2.4545, 2.7206, 1.9396, 2.6175], device='cuda:1'), covar=tensor([0.0255, 0.0578, 0.0349, 0.0368, 0.0350, 0.0218, 0.0562, 0.0170], device='cuda:1'), in_proj_covar=tensor([0.0200, 0.0199, 0.0188, 0.0194, 0.0210, 0.0168, 0.0203, 0.0167], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-02 13:28:49,075 INFO [train.py:904] (1/8) Epoch 28, batch 4500, loss[loss=0.193, simple_loss=0.2859, pruned_loss=0.05009, over 16761.00 frames. ], tot_loss[loss=0.1909, simple_loss=0.2799, pruned_loss=0.05092, over 3199757.89 frames. ], batch size: 83, lr: 2.40e-03, grad_scale: 8.0 2023-05-02 13:29:17,520 INFO [zipformer.py:625] (1/8) 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] (1/8) 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 13:30:01,295 INFO [train.py:904] (1/8) Epoch 28, batch 4550, loss[loss=0.2038, simple_loss=0.2885, pruned_loss=0.05955, over 16660.00 frames. ], tot_loss[loss=0.192, simple_loss=0.2804, pruned_loss=0.05183, over 3203886.41 frames. ], batch size: 57, lr: 2.40e-03, grad_scale: 8.0 2023-05-02 13:30:36,964 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=278627.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 13:30:39,538 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.7208, 4.8200, 4.5866, 4.2414, 4.2860, 4.7363, 4.3812, 4.3826], device='cuda:1'), covar=tensor([0.0477, 0.0286, 0.0238, 0.0277, 0.0730, 0.0255, 0.0461, 0.0545], device='cuda:1'), in_proj_covar=tensor([0.0315, 0.0472, 0.0368, 0.0371, 0.0366, 0.0423, 0.0252, 0.0440], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002], device='cuda:1') 2023-05-02 13:31:06,125 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([5.0166, 5.1074, 5.3797, 5.3462, 5.3934, 5.0338, 5.0143, 4.7140], device='cuda:1'), covar=tensor([0.0298, 0.0399, 0.0289, 0.0349, 0.0380, 0.0329, 0.0798, 0.0502], device='cuda:1'), in_proj_covar=tensor([0.0436, 0.0493, 0.0473, 0.0438, 0.0523, 0.0499, 0.0575, 0.0404], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:1') 2023-05-02 13:31:12,169 INFO [train.py:904] (1/8) Epoch 28, batch 4600, loss[loss=0.1884, simple_loss=0.273, pruned_loss=0.05191, over 11513.00 frames. ], tot_loss[loss=0.1925, simple_loss=0.2814, pruned_loss=0.05183, over 3219065.53 frames. ], batch size: 246, lr: 2.40e-03, grad_scale: 8.0 2023-05-02 13:31:43,900 INFO [zipformer.py:625] (1/8) 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] (1/8) 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 13:31:49,267 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=278679.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 13:32:22,022 INFO [train.py:904] (1/8) Epoch 28, batch 4650, loss[loss=0.1856, simple_loss=0.2724, pruned_loss=0.0494, over 16742.00 frames. ], tot_loss[loss=0.1912, simple_loss=0.2799, pruned_loss=0.05131, over 3226944.36 frames. ], batch size: 124, lr: 2.40e-03, grad_scale: 8.0 2023-05-02 13:33:00,841 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=278730.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 13:33:33,370 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=4.86 vs. limit=5.0 2023-05-02 13:33:33,716 INFO [train.py:904] (1/8) Epoch 28, batch 4700, loss[loss=0.1834, simple_loss=0.2622, pruned_loss=0.05229, over 11324.00 frames. ], tot_loss[loss=0.189, simple_loss=0.2775, pruned_loss=0.05025, over 3213929.50 frames. ], batch size: 248, lr: 2.40e-03, grad_scale: 8.0 2023-05-02 13:33:51,560 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=4.79 vs. limit=5.0 2023-05-02 13:34:07,034 INFO [optim.py:368] (1/8) 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 13:34:45,678 INFO [train.py:904] (1/8) Epoch 28, batch 4750, loss[loss=0.1564, simple_loss=0.2517, pruned_loss=0.03061, over 16889.00 frames. ], tot_loss[loss=0.1847, simple_loss=0.2733, pruned_loss=0.04809, over 3218960.30 frames. ], batch size: 96, lr: 2.40e-03, grad_scale: 8.0 2023-05-02 13:35:57,547 INFO [train.py:904] (1/8) Epoch 28, batch 4800, loss[loss=0.1835, simple_loss=0.2786, pruned_loss=0.04421, over 16715.00 frames. ], tot_loss[loss=0.1817, simple_loss=0.2704, pruned_loss=0.04653, over 3209491.23 frames. ], batch size: 76, lr: 2.40e-03, grad_scale: 8.0 2023-05-02 13:36:18,256 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=278867.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 13:36:32,582 INFO [optim.py:368] (1/8) 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 13:37:04,417 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.5070, 3.6035, 2.7576, 2.1573, 2.3251, 2.3885, 3.7833, 3.1151], device='cuda:1'), covar=tensor([0.3136, 0.0592, 0.1909, 0.3405, 0.2693, 0.2209, 0.0565, 0.1401], device='cuda:1'), in_proj_covar=tensor([0.0334, 0.0275, 0.0312, 0.0326, 0.0306, 0.0276, 0.0304, 0.0353], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:1') 2023-05-02 13:37:13,014 INFO [train.py:904] (1/8) Epoch 28, batch 4850, loss[loss=0.1785, simple_loss=0.2773, pruned_loss=0.03988, over 16395.00 frames. ], tot_loss[loss=0.1806, simple_loss=0.2704, pruned_loss=0.04542, over 3197098.17 frames. ], batch size: 146, lr: 2.40e-03, grad_scale: 8.0 2023-05-02 13:38:26,574 INFO [train.py:904] (1/8) Epoch 28, batch 4900, loss[loss=0.1826, simple_loss=0.2788, pruned_loss=0.04317, over 16603.00 frames. ], tot_loss[loss=0.1799, simple_loss=0.2701, pruned_loss=0.04478, over 3178978.65 frames. ], batch size: 57, lr: 2.40e-03, grad_scale: 8.0 2023-05-02 13:39:00,725 INFO [optim.py:368] (1/8) 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 13:39:04,653 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=278979.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 13:39:39,801 INFO [train.py:904] (1/8) Epoch 28, batch 4950, loss[loss=0.1771, simple_loss=0.2638, pruned_loss=0.04518, over 16632.00 frames. ], tot_loss[loss=0.1793, simple_loss=0.2697, pruned_loss=0.04447, over 3180165.24 frames. ], batch size: 57, lr: 2.40e-03, grad_scale: 8.0 2023-05-02 13:39:45,053 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.1835, 2.0578, 1.7449, 1.7993, 2.3118, 1.9767, 1.8815, 2.3713], device='cuda:1'), covar=tensor([0.0229, 0.0507, 0.0670, 0.0567, 0.0314, 0.0436, 0.0284, 0.0358], device='cuda:1'), in_proj_covar=tensor([0.0233, 0.0247, 0.0235, 0.0235, 0.0247, 0.0245, 0.0245, 0.0246], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-02 13:40:14,412 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=279027.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 13:40:19,463 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=279030.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 13:40:51,499 INFO [train.py:904] (1/8) Epoch 28, batch 5000, loss[loss=0.1766, simple_loss=0.2644, pruned_loss=0.04442, over 16349.00 frames. ], tot_loss[loss=0.1796, simple_loss=0.2708, pruned_loss=0.04423, over 3182728.20 frames. ], batch size: 35, lr: 2.40e-03, grad_scale: 8.0 2023-05-02 13:41:26,569 INFO [optim.py:368] (1/8) 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 13:41:27,926 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=279078.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 13:42:04,193 INFO [train.py:904] (1/8) Epoch 28, batch 5050, loss[loss=0.1952, simple_loss=0.2814, pruned_loss=0.05449, over 12046.00 frames. ], tot_loss[loss=0.1795, simple_loss=0.271, pruned_loss=0.04398, over 3189636.67 frames. ], batch size: 247, lr: 2.40e-03, grad_scale: 8.0 2023-05-02 13:42:14,314 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.9542, 2.1595, 2.1807, 3.4531, 2.0946, 2.4306, 2.2433, 2.2860], device='cuda:1'), covar=tensor([0.1566, 0.3805, 0.3281, 0.0698, 0.4305, 0.2749, 0.3989, 0.3391], device='cuda:1'), in_proj_covar=tensor([0.0423, 0.0478, 0.0386, 0.0339, 0.0446, 0.0546, 0.0448, 0.0556], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-02 13:42:25,677 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=4.18 vs. limit=5.0 2023-05-02 13:42:45,645 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=279131.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 13:42:58,434 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.1756, 2.0827, 1.7423, 1.7609, 2.3065, 1.9444, 2.0195, 2.4070], device='cuda:1'), covar=tensor([0.0215, 0.0419, 0.0604, 0.0519, 0.0275, 0.0419, 0.0204, 0.0316], device='cuda:1'), in_proj_covar=tensor([0.0232, 0.0247, 0.0235, 0.0235, 0.0247, 0.0245, 0.0245, 0.0246], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-02 13:43:14,093 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([5.0352, 5.0579, 5.3802, 5.3792, 5.4150, 5.0985, 5.0294, 4.8386], device='cuda:1'), covar=tensor([0.0287, 0.0491, 0.0368, 0.0396, 0.0457, 0.0335, 0.0996, 0.0443], device='cuda:1'), in_proj_covar=tensor([0.0432, 0.0487, 0.0470, 0.0433, 0.0515, 0.0495, 0.0570, 0.0399], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:1') 2023-05-02 13:43:17,182 INFO [train.py:904] (1/8) Epoch 28, batch 5100, loss[loss=0.1565, simple_loss=0.2535, pruned_loss=0.02974, over 16882.00 frames. ], tot_loss[loss=0.1779, simple_loss=0.2694, pruned_loss=0.04318, over 3196477.70 frames. ], batch size: 96, lr: 2.40e-03, grad_scale: 8.0 2023-05-02 13:43:20,273 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.17 vs. limit=2.0 2023-05-02 13:43:37,533 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=279167.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 13:43:44,874 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([1.6192, 2.3313, 2.3875, 3.5133, 2.2166, 3.6873, 1.5263, 2.7853], device='cuda:1'), covar=tensor([0.1531, 0.0935, 0.1353, 0.0168, 0.0146, 0.0367, 0.1851, 0.0876], device='cuda:1'), in_proj_covar=tensor([0.0173, 0.0181, 0.0201, 0.0203, 0.0208, 0.0218, 0.0210, 0.0199], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-05-02 13:43:52,644 INFO [optim.py:368] (1/8) Clipping_scale=2.0, grad-norm quartiles 1.331e+02 1.882e+02 2.136e+02 2.468e+02 5.149e+02, threshold=4.272e+02, percent-clipped=1.0 2023-05-02 13:44:15,266 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=279192.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 13:44:30,688 INFO [train.py:904] (1/8) Epoch 28, batch 5150, loss[loss=0.1605, simple_loss=0.2551, pruned_loss=0.03296, over 16555.00 frames. ], tot_loss[loss=0.1769, simple_loss=0.269, pruned_loss=0.04237, over 3204228.92 frames. ], batch size: 76, lr: 2.40e-03, grad_scale: 8.0 2023-05-02 13:44:46,355 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.3107, 3.4241, 2.0501, 3.7583, 2.5262, 3.7847, 2.1587, 2.7388], device='cuda:1'), covar=tensor([0.0341, 0.0438, 0.1866, 0.0239, 0.0978, 0.0571, 0.1860, 0.0950], device='cuda:1'), in_proj_covar=tensor([0.0176, 0.0183, 0.0196, 0.0173, 0.0181, 0.0221, 0.0205, 0.0184], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-05-02 13:44:48,994 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=279215.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 13:45:04,206 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.9786, 4.1737, 2.5439, 4.8348, 3.1190, 4.7605, 2.6181, 3.3173], device='cuda:1'), covar=tensor([0.0273, 0.0331, 0.1783, 0.0249, 0.0829, 0.0397, 0.1711, 0.0857], device='cuda:1'), in_proj_covar=tensor([0.0176, 0.0183, 0.0196, 0.0173, 0.0182, 0.0221, 0.0205, 0.0185], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-05-02 13:45:42,233 INFO [train.py:904] (1/8) Epoch 28, batch 5200, loss[loss=0.1789, simple_loss=0.2692, pruned_loss=0.04432, over 12050.00 frames. ], tot_loss[loss=0.1749, simple_loss=0.2669, pruned_loss=0.04142, over 3214284.55 frames. ], batch size: 248, lr: 2.40e-03, grad_scale: 8.0 2023-05-02 13:46:17,348 INFO [optim.py:368] (1/8) Clipping_scale=2.0, grad-norm quartiles 1.295e+02 1.964e+02 2.228e+02 2.622e+02 8.480e+02, threshold=4.456e+02, percent-clipped=3.0 2023-05-02 13:46:53,552 INFO [train.py:904] (1/8) Epoch 28, batch 5250, loss[loss=0.1686, simple_loss=0.2615, pruned_loss=0.03785, over 16424.00 frames. ], tot_loss[loss=0.1735, simple_loss=0.2649, pruned_loss=0.04101, over 3223328.37 frames. ], batch size: 146, lr: 2.40e-03, grad_scale: 8.0 2023-05-02 13:48:06,861 INFO [train.py:904] (1/8) Epoch 28, batch 5300, loss[loss=0.1594, simple_loss=0.2404, pruned_loss=0.03922, over 16677.00 frames. ], tot_loss[loss=0.1713, simple_loss=0.2618, pruned_loss=0.04041, over 3204782.98 frames. ], batch size: 62, lr: 2.40e-03, grad_scale: 8.0 2023-05-02 13:48:07,289 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.4419, 3.3074, 3.7647, 1.8548, 3.9640, 3.9519, 3.0050, 2.9346], device='cuda:1'), covar=tensor([0.0862, 0.0300, 0.0218, 0.1240, 0.0075, 0.0132, 0.0446, 0.0500], device='cuda:1'), in_proj_covar=tensor([0.0148, 0.0110, 0.0101, 0.0137, 0.0086, 0.0129, 0.0129, 0.0130], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-05-02 13:48:41,218 INFO [optim.py:368] (1/8) 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 13:48:49,951 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.2881, 2.2467, 2.3506, 4.0037, 2.2016, 2.5682, 2.3536, 2.4667], device='cuda:1'), covar=tensor([0.1632, 0.4048, 0.3201, 0.0657, 0.4644, 0.2986, 0.3769, 0.3595], device='cuda:1'), in_proj_covar=tensor([0.0421, 0.0475, 0.0384, 0.0337, 0.0444, 0.0543, 0.0446, 0.0553], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-02 13:48:59,739 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.2302, 4.2060, 4.1150, 3.2621, 4.1507, 1.6162, 3.9351, 3.6245], device='cuda:1'), covar=tensor([0.0119, 0.0130, 0.0182, 0.0424, 0.0100, 0.3317, 0.0150, 0.0357], device='cuda:1'), in_proj_covar=tensor([0.0180, 0.0174, 0.0212, 0.0186, 0.0188, 0.0217, 0.0201, 0.0180], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-02 13:49:20,977 INFO [train.py:904] (1/8) Epoch 28, batch 5350, loss[loss=0.168, simple_loss=0.2582, pruned_loss=0.03892, over 17199.00 frames. ], tot_loss[loss=0.17, simple_loss=0.2602, pruned_loss=0.03989, over 3219695.06 frames. ], batch size: 46, lr: 2.40e-03, grad_scale: 8.0 2023-05-02 13:50:09,613 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.2146, 4.3643, 4.5201, 4.2170, 4.3070, 4.8351, 4.3567, 4.0442], device='cuda:1'), covar=tensor([0.1849, 0.1758, 0.1919, 0.2136, 0.2375, 0.0984, 0.1437, 0.2316], device='cuda:1'), in_proj_covar=tensor([0.0421, 0.0626, 0.0685, 0.0511, 0.0680, 0.0714, 0.0537, 0.0681], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-05-02 13:50:32,504 INFO [train.py:904] (1/8) Epoch 28, batch 5400, loss[loss=0.1643, simple_loss=0.2617, pruned_loss=0.03348, over 16427.00 frames. ], tot_loss[loss=0.1711, simple_loss=0.2619, pruned_loss=0.04011, over 3222339.25 frames. ], batch size: 68, lr: 2.40e-03, grad_scale: 8.0 2023-05-02 13:51:08,322 INFO [optim.py:368] (1/8) 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 13:51:24,275 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=279487.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 13:51:50,343 INFO [train.py:904] (1/8) Epoch 28, batch 5450, loss[loss=0.2, simple_loss=0.2911, pruned_loss=0.05443, over 16540.00 frames. ], tot_loss[loss=0.174, simple_loss=0.2649, pruned_loss=0.04154, over 3209416.13 frames. ], batch size: 62, lr: 2.40e-03, grad_scale: 8.0 2023-05-02 13:52:31,258 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=279529.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 13:52:52,441 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 28, batch 5500, loss[loss=0.2127, simple_loss=0.3024, pruned_loss=0.06148, over 16285.00 frames. ], tot_loss[loss=0.1819, simple_loss=0.272, pruned_loss=0.04588, over 3173395.76 frames. ], batch size: 165, lr: 2.40e-03, grad_scale: 8.0 2023-05-02 13:53:47,381 INFO [optim.py:368] (1/8) Clipping_scale=2.0, grad-norm quartiles 1.670e+02 2.730e+02 3.219e+02 3.961e+02 9.894e+02, threshold=6.439e+02, percent-clipped=12.0 2023-05-02 13:54:09,285 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=279590.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 13:54:28,771 INFO [train.py:904] (1/8) Epoch 28, batch 5550, loss[loss=0.2132, simple_loss=0.2936, pruned_loss=0.06638, over 16288.00 frames. ], tot_loss[loss=0.1895, simple_loss=0.2788, pruned_loss=0.05015, over 3143057.43 frames. ], batch size: 165, lr: 2.40e-03, grad_scale: 8.0 2023-05-02 13:54:29,304 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=279603.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 13:55:19,542 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.30 vs. limit=2.0 2023-05-02 13:55:47,876 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.3989, 3.1959, 3.5582, 1.9625, 3.7117, 3.7066, 2.9078, 2.8307], device='cuda:1'), covar=tensor([0.0816, 0.0295, 0.0249, 0.1123, 0.0101, 0.0193, 0.0450, 0.0465], device='cuda:1'), in_proj_covar=tensor([0.0147, 0.0110, 0.0101, 0.0137, 0.0086, 0.0129, 0.0128, 0.0129], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-05-02 13:55:48,648 INFO [train.py:904] (1/8) Epoch 28, batch 5600, loss[loss=0.2184, simple_loss=0.3031, pruned_loss=0.06688, over 16708.00 frames. ], tot_loss[loss=0.1963, simple_loss=0.2836, pruned_loss=0.05453, over 3096262.27 frames. ], batch size: 124, lr: 2.40e-03, grad_scale: 8.0 2023-05-02 13:56:28,890 INFO [optim.py:368] (1/8) Clipping_scale=2.0, grad-norm quartiles 2.204e+02 3.162e+02 3.859e+02 4.792e+02 1.536e+03, threshold=7.717e+02, percent-clipped=11.0 2023-05-02 13:57:11,930 INFO [train.py:904] (1/8) Epoch 28, batch 5650, loss[loss=0.2542, simple_loss=0.3135, pruned_loss=0.09749, over 11153.00 frames. ], tot_loss[loss=0.2028, simple_loss=0.2886, pruned_loss=0.05851, over 3042157.34 frames. ], batch size: 246, lr: 2.40e-03, grad_scale: 8.0 2023-05-02 13:57:33,063 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.47 vs. limit=2.0 2023-05-02 13:57:57,924 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.40 vs. limit=2.0 2023-05-02 13:58:01,650 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.5338, 3.6494, 2.7580, 2.2122, 2.4800, 2.4438, 3.9305, 3.3011], device='cuda:1'), covar=tensor([0.3115, 0.0615, 0.1929, 0.3050, 0.2748, 0.2185, 0.0465, 0.1355], device='cuda:1'), in_proj_covar=tensor([0.0337, 0.0276, 0.0314, 0.0329, 0.0307, 0.0278, 0.0307, 0.0355], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:1') 2023-05-02 13:58:17,986 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.24 vs. limit=2.0 2023-05-02 13:58:29,306 INFO [train.py:904] (1/8) Epoch 28, batch 5700, loss[loss=0.1995, simple_loss=0.2911, pruned_loss=0.05397, over 17219.00 frames. ], tot_loss[loss=0.2056, simple_loss=0.2909, pruned_loss=0.06015, over 3042053.49 frames. ], batch size: 44, lr: 2.40e-03, grad_scale: 8.0 2023-05-02 13:59:05,447 INFO [optim.py:368] (1/8) 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 13:59:22,125 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 28, batch 5750, loss[loss=0.1725, simple_loss=0.2692, pruned_loss=0.03794, over 16581.00 frames. ], tot_loss[loss=0.2093, simple_loss=0.2942, pruned_loss=0.06222, over 3004337.40 frames. ], batch size: 89, lr: 2.40e-03, grad_scale: 8.0 2023-05-02 14:00:39,194 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=279835.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 14:00:46,266 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.1441, 2.2787, 2.3769, 2.6438, 1.6753, 3.1531, 1.9988, 2.7200], device='cuda:1'), covar=tensor([0.1210, 0.0696, 0.1135, 0.0236, 0.0112, 0.0512, 0.1515, 0.0767], device='cuda:1'), in_proj_covar=tensor([0.0173, 0.0180, 0.0201, 0.0204, 0.0207, 0.0218, 0.0210, 0.0199], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-05-02 14:01:07,319 INFO [train.py:904] (1/8) Epoch 28, batch 5800, loss[loss=0.2156, simple_loss=0.3024, pruned_loss=0.06438, over 16997.00 frames. ], tot_loss[loss=0.2081, simple_loss=0.2941, pruned_loss=0.06102, over 3008779.49 frames. ], batch size: 41, lr: 2.40e-03, grad_scale: 8.0 2023-05-02 14:01:46,119 INFO [optim.py:368] (1/8) 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] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=279885.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 14:02:16,873 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=279898.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 14:02:25,207 INFO [train.py:904] (1/8) Epoch 28, batch 5850, loss[loss=0.2021, simple_loss=0.2863, pruned_loss=0.05899, over 16649.00 frames. ], tot_loss[loss=0.2048, simple_loss=0.2913, pruned_loss=0.05914, over 3028910.80 frames. ], batch size: 62, lr: 2.40e-03, grad_scale: 8.0 2023-05-02 14:02:27,441 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.2018, 4.2187, 4.5377, 4.5096, 4.5166, 4.2520, 4.2724, 4.2297], device='cuda:1'), covar=tensor([0.0409, 0.0881, 0.0538, 0.0530, 0.0615, 0.0630, 0.0866, 0.0686], device='cuda:1'), in_proj_covar=tensor([0.0433, 0.0490, 0.0473, 0.0435, 0.0518, 0.0498, 0.0574, 0.0400], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:1') 2023-05-02 14:02:31,374 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.1666, 2.0594, 1.8136, 1.8530, 2.3452, 2.0001, 1.9772, 2.3808], device='cuda:1'), covar=tensor([0.0237, 0.0387, 0.0504, 0.0435, 0.0252, 0.0351, 0.0231, 0.0270], device='cuda:1'), in_proj_covar=tensor([0.0231, 0.0244, 0.0233, 0.0233, 0.0245, 0.0243, 0.0241, 0.0243], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-02 14:02:35,919 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.44 vs. limit=2.0 2023-05-02 14:03:11,282 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.2805, 3.0884, 3.5618, 1.8339, 3.6894, 3.6975, 2.8899, 2.7211], device='cuda:1'), covar=tensor([0.0946, 0.0342, 0.0219, 0.1256, 0.0094, 0.0187, 0.0467, 0.0540], device='cuda:1'), in_proj_covar=tensor([0.0148, 0.0110, 0.0102, 0.0138, 0.0086, 0.0130, 0.0129, 0.0129], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-05-02 14:03:41,368 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.5862, 3.3966, 3.9083, 1.9261, 4.0906, 4.0888, 3.2145, 3.0287], device='cuda:1'), covar=tensor([0.0857, 0.0319, 0.0199, 0.1305, 0.0088, 0.0166, 0.0404, 0.0506], device='cuda:1'), in_proj_covar=tensor([0.0148, 0.0110, 0.0102, 0.0138, 0.0086, 0.0130, 0.0129, 0.0129], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-05-02 14:03:44,212 INFO [train.py:904] (1/8) Epoch 28, batch 5900, loss[loss=0.2027, simple_loss=0.2964, pruned_loss=0.05452, over 16786.00 frames. ], tot_loss[loss=0.2043, simple_loss=0.2907, pruned_loss=0.0589, over 3033857.01 frames. ], batch size: 124, lr: 2.40e-03, grad_scale: 8.0 2023-05-02 14:04:26,171 INFO [optim.py:368] (1/8) 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 14:05:07,642 INFO [train.py:904] (1/8) Epoch 28, batch 5950, loss[loss=0.1996, simple_loss=0.2937, pruned_loss=0.0527, over 16487.00 frames. ], tot_loss[loss=0.2032, simple_loss=0.2911, pruned_loss=0.05772, over 3048809.79 frames. ], batch size: 146, lr: 2.40e-03, grad_scale: 8.0 2023-05-02 14:06:20,839 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=280051.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 14:06:23,830 INFO [train.py:904] (1/8) Epoch 28, batch 6000, loss[loss=0.1983, simple_loss=0.2907, pruned_loss=0.05292, over 16194.00 frames. ], tot_loss[loss=0.2022, simple_loss=0.2904, pruned_loss=0.05702, over 3085375.31 frames. ], batch size: 165, lr: 2.40e-03, grad_scale: 8.0 2023-05-02 14:06:23,830 INFO [train.py:929] (1/8) Computing validation loss 2023-05-02 14:06:34,276 INFO [train.py:938] (1/8) Epoch 28, validation: loss=0.148, simple_loss=0.2602, pruned_loss=0.0179, over 944034.00 frames. 2023-05-02 14:06:34,276 INFO [train.py:939] (1/8) Maximum memory allocated so far is 18145MB 2023-05-02 14:06:47,642 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.88 vs. limit=2.0 2023-05-02 14:06:50,956 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.58 vs. limit=2.0 2023-05-02 14:07:11,167 INFO [optim.py:368] (1/8) 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 14:07:24,387 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.8591, 1.9993, 2.1778, 3.4236, 1.9228, 2.2484, 2.1504, 2.1159], device='cuda:1'), covar=tensor([0.1949, 0.4348, 0.3283, 0.0788, 0.5238, 0.3141, 0.3840, 0.4084], device='cuda:1'), in_proj_covar=tensor([0.0423, 0.0476, 0.0385, 0.0337, 0.0446, 0.0543, 0.0446, 0.0554], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-02 14:07:25,994 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 28, batch 6050, loss[loss=0.1934, simple_loss=0.2852, pruned_loss=0.05082, over 16171.00 frames. ], tot_loss[loss=0.2012, simple_loss=0.2892, pruned_loss=0.05664, over 3091332.66 frames. ], batch size: 35, lr: 2.39e-03, grad_scale: 8.0 2023-05-02 14:08:06,555 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=280112.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 14:09:01,883 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 28, batch 6100, loss[loss=0.1869, simple_loss=0.2842, pruned_loss=0.04481, over 16937.00 frames. ], tot_loss[loss=0.1997, simple_loss=0.2884, pruned_loss=0.05553, over 3086005.67 frames. ], batch size: 96, lr: 2.39e-03, grad_scale: 8.0 2023-05-02 14:09:41,683 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([1.8336, 1.4120, 1.7282, 1.7063, 1.7958, 1.8996, 1.6778, 1.8011], device='cuda:1'), covar=tensor([0.0309, 0.0438, 0.0240, 0.0352, 0.0297, 0.0228, 0.0447, 0.0150], device='cuda:1'), in_proj_covar=tensor([0.0200, 0.0200, 0.0188, 0.0194, 0.0210, 0.0167, 0.0204, 0.0167], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-02 14:09:51,397 INFO [optim.py:368] (1/8) 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 14:10:01,952 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=280185.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 14:10:22,546 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=280198.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 14:10:29,421 INFO [train.py:904] (1/8) Epoch 28, batch 6150, loss[loss=0.1786, simple_loss=0.27, pruned_loss=0.04361, over 16897.00 frames. ], tot_loss[loss=0.1974, simple_loss=0.2857, pruned_loss=0.05454, over 3091966.16 frames. ], batch size: 96, lr: 2.39e-03, grad_scale: 8.0 2023-05-02 14:10:41,389 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.1327, 3.4254, 3.5123, 2.1017, 3.0758, 2.4190, 3.5749, 3.7736], device='cuda:1'), covar=tensor([0.0230, 0.0778, 0.0708, 0.2188, 0.0849, 0.1004, 0.0614, 0.0900], device='cuda:1'), in_proj_covar=tensor([0.0162, 0.0172, 0.0172, 0.0158, 0.0149, 0.0133, 0.0147, 0.0185], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0005], device='cuda:1') 2023-05-02 14:11:16,583 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=280233.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 14:11:35,646 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 28, batch 6200, loss[loss=0.1973, simple_loss=0.2866, pruned_loss=0.05401, over 16410.00 frames. ], tot_loss[loss=0.1957, simple_loss=0.2837, pruned_loss=0.05392, over 3096044.17 frames. ], batch size: 68, lr: 2.39e-03, grad_scale: 8.0 2023-05-02 14:11:51,765 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.4120, 4.2813, 4.4760, 4.6374, 4.7766, 4.3444, 4.7388, 4.7993], device='cuda:1'), covar=tensor([0.1956, 0.1293, 0.1636, 0.0787, 0.0622, 0.1119, 0.0736, 0.0733], device='cuda:1'), in_proj_covar=tensor([0.0671, 0.0819, 0.0950, 0.0834, 0.0639, 0.0666, 0.0698, 0.0805], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-02 14:12:24,026 INFO [optim.py:368] (1/8) Clipping_scale=2.0, grad-norm quartiles 2.005e+02 2.646e+02 3.232e+02 3.925e+02 8.814e+02, threshold=6.464e+02, percent-clipped=6.0 2023-05-02 14:12:49,446 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.8842, 2.7946, 2.2688, 2.6775, 3.2304, 2.8702, 3.3213, 3.4013], device='cuda:1'), covar=tensor([0.0115, 0.0464, 0.0631, 0.0473, 0.0286, 0.0390, 0.0275, 0.0282], device='cuda:1'), in_proj_covar=tensor([0.0228, 0.0242, 0.0231, 0.0231, 0.0244, 0.0240, 0.0238, 0.0241], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-02 14:13:00,713 INFO [train.py:904] (1/8) Epoch 28, batch 6250, loss[loss=0.1806, simple_loss=0.2793, pruned_loss=0.04093, over 16837.00 frames. ], tot_loss[loss=0.1942, simple_loss=0.2827, pruned_loss=0.05282, over 3120014.17 frames. ], batch size: 116, lr: 2.39e-03, grad_scale: 8.0 2023-05-02 14:13:32,466 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.5023, 4.5162, 4.4323, 3.5732, 4.4254, 1.7694, 4.1884, 4.0118], device='cuda:1'), covar=tensor([0.0145, 0.0137, 0.0199, 0.0376, 0.0133, 0.2947, 0.0177, 0.0295], device='cuda:1'), in_proj_covar=tensor([0.0181, 0.0175, 0.0213, 0.0186, 0.0189, 0.0218, 0.0201, 0.0181], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-02 14:13:36,638 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=280327.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 14:14:07,076 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.9088, 2.7178, 2.8890, 2.1671, 2.7285, 2.1831, 2.7496, 2.9630], device='cuda:1'), covar=tensor([0.0277, 0.0782, 0.0489, 0.1812, 0.0773, 0.0882, 0.0544, 0.0721], device='cuda:1'), in_proj_covar=tensor([0.0162, 0.0172, 0.0172, 0.0158, 0.0149, 0.0133, 0.0147, 0.0185], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0005], device='cuda:1') 2023-05-02 14:14:11,261 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.6660, 1.7667, 2.3025, 2.6729, 2.5713, 3.0601, 2.0657, 3.0212], device='cuda:1'), covar=tensor([0.0287, 0.0686, 0.0405, 0.0355, 0.0397, 0.0206, 0.0614, 0.0165], device='cuda:1'), in_proj_covar=tensor([0.0199, 0.0199, 0.0187, 0.0193, 0.0209, 0.0166, 0.0203, 0.0166], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-02 14:14:14,430 INFO [train.py:904] (1/8) Epoch 28, batch 6300, loss[loss=0.1982, simple_loss=0.2866, pruned_loss=0.0549, over 16975.00 frames. ], tot_loss[loss=0.1943, simple_loss=0.2828, pruned_loss=0.05289, over 3113466.30 frames. ], batch size: 55, lr: 2.39e-03, grad_scale: 8.0 2023-05-02 14:14:53,563 INFO [optim.py:368] (1/8) 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 14:14:55,968 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.6496, 2.6174, 1.9377, 2.6885, 2.2105, 2.8008, 2.1666, 2.4309], device='cuda:1'), covar=tensor([0.0369, 0.0449, 0.1375, 0.0346, 0.0767, 0.0563, 0.1312, 0.0712], device='cuda:1'), in_proj_covar=tensor([0.0174, 0.0181, 0.0196, 0.0172, 0.0179, 0.0219, 0.0203, 0.0183], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-05-02 14:15:09,904 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 28, batch 6350, loss[loss=0.1789, simple_loss=0.2697, pruned_loss=0.04401, over 16972.00 frames. ], tot_loss[loss=0.1967, simple_loss=0.2841, pruned_loss=0.05465, over 3103728.23 frames. ], batch size: 41, lr: 2.39e-03, grad_scale: 8.0 2023-05-02 14:15:36,781 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=280407.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 14:15:56,251 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.5265, 2.3111, 1.9236, 2.1462, 2.6519, 2.3100, 2.3118, 2.7284], device='cuda:1'), covar=tensor([0.0257, 0.0428, 0.0590, 0.0501, 0.0309, 0.0393, 0.0267, 0.0309], device='cuda:1'), in_proj_covar=tensor([0.0228, 0.0241, 0.0231, 0.0231, 0.0244, 0.0240, 0.0238, 0.0241], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-02 14:16:31,362 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=280444.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 14:16:42,937 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.8811, 2.2280, 2.4737, 3.1120, 2.2526, 2.4445, 2.3730, 2.3232], device='cuda:1'), covar=tensor([0.1545, 0.3416, 0.2475, 0.0781, 0.4195, 0.2349, 0.3387, 0.3062], device='cuda:1'), in_proj_covar=tensor([0.0420, 0.0473, 0.0384, 0.0335, 0.0444, 0.0542, 0.0445, 0.0553], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-02 14:16:43,550 INFO [train.py:904] (1/8) Epoch 28, batch 6400, loss[loss=0.2497, simple_loss=0.318, pruned_loss=0.0907, over 11211.00 frames. ], tot_loss[loss=0.1957, simple_loss=0.283, pruned_loss=0.05427, over 3127422.01 frames. ], batch size: 248, lr: 2.39e-03, grad_scale: 8.0 2023-05-02 14:17:03,654 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.13 vs. limit=2.0 2023-05-02 14:17:19,317 INFO [optim.py:368] (1/8) 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 14:17:23,520 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.7124, 3.9308, 2.9547, 2.3394, 2.6859, 2.5337, 4.2732, 3.5107], device='cuda:1'), covar=tensor([0.3078, 0.0719, 0.1904, 0.3097, 0.2812, 0.2172, 0.0448, 0.1453], device='cuda:1'), in_proj_covar=tensor([0.0336, 0.0276, 0.0313, 0.0328, 0.0307, 0.0277, 0.0305, 0.0353], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:1') 2023-05-02 14:17:56,101 INFO [train.py:904] (1/8) Epoch 28, batch 6450, loss[loss=0.1879, simple_loss=0.2804, pruned_loss=0.04772, over 16711.00 frames. ], tot_loss[loss=0.1963, simple_loss=0.2837, pruned_loss=0.05449, over 3115922.86 frames. ], batch size: 76, lr: 2.39e-03, grad_scale: 8.0 2023-05-02 14:18:18,385 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=280518.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 14:18:41,524 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=280533.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 14:19:02,774 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=280547.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 14:19:13,552 INFO [train.py:904] (1/8) Epoch 28, batch 6500, loss[loss=0.2053, simple_loss=0.2736, pruned_loss=0.06854, over 11825.00 frames. ], tot_loss[loss=0.1959, simple_loss=0.2826, pruned_loss=0.05459, over 3096555.55 frames. ], batch size: 248, lr: 2.39e-03, grad_scale: 8.0 2023-05-02 14:19:25,531 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=2.93 vs. limit=5.0 2023-05-02 14:19:45,776 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.1384, 4.0277, 4.1976, 4.3213, 4.4549, 4.0851, 4.4242, 4.4861], device='cuda:1'), covar=tensor([0.1870, 0.1282, 0.1543, 0.0819, 0.0699, 0.1315, 0.0954, 0.0754], device='cuda:1'), in_proj_covar=tensor([0.0668, 0.0815, 0.0946, 0.0832, 0.0636, 0.0663, 0.0696, 0.0802], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-02 14:19:49,977 INFO [optim.py:368] (1/8) 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 14:19:52,056 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=280579.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 14:20:15,923 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=280594.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 14:20:28,521 INFO [train.py:904] (1/8) Epoch 28, batch 6550, loss[loss=0.2566, simple_loss=0.3198, pruned_loss=0.09674, over 11424.00 frames. ], tot_loss[loss=0.199, simple_loss=0.2858, pruned_loss=0.05608, over 3089442.58 frames. ], batch size: 246, lr: 2.39e-03, grad_scale: 8.0 2023-05-02 14:20:37,409 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=280608.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 14:20:38,770 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.1981, 2.8473, 3.1313, 1.7594, 3.2932, 3.2984, 2.7798, 2.5700], device='cuda:1'), covar=tensor([0.0858, 0.0324, 0.0232, 0.1272, 0.0118, 0.0280, 0.0440, 0.0513], device='cuda:1'), in_proj_covar=tensor([0.0149, 0.0111, 0.0103, 0.0140, 0.0087, 0.0131, 0.0130, 0.0131], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-05-02 14:20:45,324 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=280614.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 14:21:44,272 INFO [train.py:904] (1/8) Epoch 28, batch 6600, loss[loss=0.191, simple_loss=0.2867, pruned_loss=0.04767, over 16952.00 frames. ], tot_loss[loss=0.2, simple_loss=0.2879, pruned_loss=0.05603, over 3096252.93 frames. ], batch size: 96, lr: 2.39e-03, grad_scale: 8.0 2023-05-02 14:22:18,046 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=280675.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 14:22:21,696 INFO [optim.py:368] (1/8) 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 14:22:28,918 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=280683.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 14:22:59,936 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=280702.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 14:23:01,183 INFO [train.py:904] (1/8) Epoch 28, batch 6650, loss[loss=0.2576, simple_loss=0.3214, pruned_loss=0.09696, over 11598.00 frames. ], tot_loss[loss=0.2005, simple_loss=0.2878, pruned_loss=0.05659, over 3093147.43 frames. ], batch size: 247, lr: 2.39e-03, grad_scale: 8.0 2023-05-02 14:23:07,241 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=280707.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 14:23:27,275 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.7117, 3.4561, 4.0425, 1.9517, 4.2705, 4.2328, 3.2180, 3.1369], device='cuda:1'), covar=tensor([0.0891, 0.0363, 0.0245, 0.1395, 0.0082, 0.0206, 0.0446, 0.0526], device='cuda:1'), in_proj_covar=tensor([0.0149, 0.0111, 0.0103, 0.0139, 0.0087, 0.0131, 0.0129, 0.0130], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-05-02 14:24:02,718 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 28, batch 6700, loss[loss=0.2449, simple_loss=0.3068, pruned_loss=0.09149, over 11621.00 frames. ], tot_loss[loss=0.1997, simple_loss=0.2865, pruned_loss=0.05647, over 3107460.79 frames. ], batch size: 248, lr: 2.39e-03, grad_scale: 8.0 2023-05-02 14:24:18,762 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=280755.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 14:24:30,609 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=280763.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 14:24:51,655 INFO [optim.py:368] (1/8) 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 14:25:12,494 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=280792.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 14:25:29,257 INFO [train.py:904] (1/8) Epoch 28, batch 6750, loss[loss=0.1832, simple_loss=0.2702, pruned_loss=0.04811, over 16312.00 frames. ], tot_loss[loss=0.1983, simple_loss=0.2847, pruned_loss=0.05596, over 3091029.14 frames. ], batch size: 35, lr: 2.39e-03, grad_scale: 8.0 2023-05-02 14:25:31,661 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.8621, 2.1870, 2.3975, 3.0944, 2.1829, 2.4297, 2.3194, 2.3045], device='cuda:1'), covar=tensor([0.1688, 0.3322, 0.2712, 0.0807, 0.4405, 0.2304, 0.3684, 0.3177], device='cuda:1'), in_proj_covar=tensor([0.0421, 0.0474, 0.0385, 0.0336, 0.0446, 0.0543, 0.0446, 0.0553], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-02 14:25:49,292 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.3823, 4.4075, 4.7349, 4.6825, 4.7073, 4.4274, 4.4217, 4.3449], device='cuda:1'), covar=tensor([0.0354, 0.0579, 0.0374, 0.0414, 0.0542, 0.0402, 0.0940, 0.0528], device='cuda:1'), in_proj_covar=tensor([0.0435, 0.0490, 0.0474, 0.0439, 0.0521, 0.0500, 0.0577, 0.0401], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:1') 2023-05-02 14:26:26,046 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=4.68 vs. limit=5.0 2023-05-02 14:26:41,173 INFO [train.py:904] (1/8) Epoch 28, batch 6800, loss[loss=0.2252, simple_loss=0.2977, pruned_loss=0.07639, over 11567.00 frames. ], tot_loss[loss=0.1998, simple_loss=0.2858, pruned_loss=0.05695, over 3080162.74 frames. ], batch size: 246, lr: 2.39e-03, grad_scale: 8.0 2023-05-02 14:26:55,192 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.3588, 2.5414, 2.1587, 2.2921, 2.8706, 2.5140, 2.8429, 3.0586], device='cuda:1'), covar=tensor([0.0157, 0.0494, 0.0609, 0.0552, 0.0323, 0.0466, 0.0287, 0.0316], device='cuda:1'), in_proj_covar=tensor([0.0229, 0.0242, 0.0232, 0.0232, 0.0244, 0.0242, 0.0240, 0.0242], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-02 14:27:04,163 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.5414, 1.6689, 2.1876, 2.4245, 2.4410, 2.7841, 1.8108, 2.7066], device='cuda:1'), covar=tensor([0.0266, 0.0628, 0.0355, 0.0378, 0.0392, 0.0255, 0.0692, 0.0182], device='cuda:1'), in_proj_covar=tensor([0.0196, 0.0198, 0.0185, 0.0191, 0.0206, 0.0165, 0.0201, 0.0165], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-02 14:27:13,265 INFO [zipformer.py:625] (1/8) 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] (1/8) 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 14:27:35,918 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=280889.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 14:27:55,352 INFO [train.py:904] (1/8) Epoch 28, batch 6850, loss[loss=0.1835, simple_loss=0.2998, pruned_loss=0.03361, over 16662.00 frames. ], tot_loss[loss=0.2013, simple_loss=0.2876, pruned_loss=0.05744, over 3089137.89 frames. ], batch size: 76, lr: 2.39e-03, grad_scale: 8.0 2023-05-02 14:27:55,591 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=280903.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 14:27:59,722 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=4.65 vs. limit=5.0 2023-05-02 14:29:06,762 INFO [train.py:904] (1/8) Epoch 28, batch 6900, loss[loss=0.2118, simple_loss=0.2988, pruned_loss=0.06234, over 16512.00 frames. ], tot_loss[loss=0.2022, simple_loss=0.2899, pruned_loss=0.0573, over 3088600.24 frames. ], batch size: 62, lr: 2.39e-03, grad_scale: 8.0 2023-05-02 14:29:15,789 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.95 vs. limit=2.0 2023-05-02 14:29:34,347 INFO [zipformer.py:625] (1/8) 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] (1/8) 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 14:29:53,929 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=280983.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 14:30:23,411 INFO [train.py:904] (1/8) Epoch 28, batch 6950, loss[loss=0.2138, simple_loss=0.2892, pruned_loss=0.06918, over 16358.00 frames. ], tot_loss[loss=0.2036, simple_loss=0.2909, pruned_loss=0.05818, over 3085272.63 frames. ], batch size: 35, lr: 2.39e-03, grad_scale: 8.0 2023-05-02 14:30:32,558 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.21 vs. limit=2.0 2023-05-02 14:31:03,952 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=281031.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 14:31:30,249 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=281049.0, num_to_drop=1, layers_to_drop={0} 2023-05-02 14:31:35,801 INFO [train.py:904] (1/8) Epoch 28, batch 7000, loss[loss=0.1878, simple_loss=0.2975, pruned_loss=0.03909, over 16703.00 frames. ], tot_loss[loss=0.2027, simple_loss=0.2907, pruned_loss=0.05732, over 3089046.22 frames. ], batch size: 89, lr: 2.39e-03, grad_scale: 8.0 2023-05-02 14:31:44,104 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=281058.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 14:31:47,663 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.3652, 2.1858, 1.8618, 1.9127, 2.4995, 2.1640, 2.0994, 2.6469], device='cuda:1'), covar=tensor([0.0314, 0.0500, 0.0640, 0.0581, 0.0309, 0.0429, 0.0252, 0.0314], device='cuda:1'), in_proj_covar=tensor([0.0229, 0.0243, 0.0233, 0.0233, 0.0244, 0.0243, 0.0240, 0.0242], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-02 14:32:12,984 INFO [optim.py:368] (1/8) 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] (1/8) Epoch 28, batch 7050, loss[loss=0.1882, simple_loss=0.2814, pruned_loss=0.04756, over 17019.00 frames. ], tot_loss[loss=0.2027, simple_loss=0.2914, pruned_loss=0.05695, over 3095001.09 frames. ], batch size: 50, lr: 2.39e-03, grad_scale: 8.0 2023-05-02 14:33:01,177 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=281110.0, num_to_drop=1, layers_to_drop={3} 2023-05-02 14:33:08,363 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.4031, 3.3810, 3.4457, 3.5172, 3.5565, 3.3200, 3.5304, 3.5960], device='cuda:1'), covar=tensor([0.1355, 0.0950, 0.1006, 0.0637, 0.0713, 0.2428, 0.1161, 0.0866], device='cuda:1'), in_proj_covar=tensor([0.0664, 0.0810, 0.0938, 0.0824, 0.0631, 0.0658, 0.0692, 0.0797], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-02 14:33:38,512 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.0309, 3.0509, 1.9337, 3.2779, 2.3664, 3.3389, 2.1277, 2.5341], device='cuda:1'), covar=tensor([0.0335, 0.0442, 0.1608, 0.0224, 0.0858, 0.0588, 0.1481, 0.0820], device='cuda:1'), in_proj_covar=tensor([0.0175, 0.0181, 0.0196, 0.0172, 0.0180, 0.0220, 0.0204, 0.0183], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-05-02 14:33:43,634 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.28 vs. limit=2.0 2023-05-02 14:33:48,678 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.12 vs. limit=2.0 2023-05-02 14:34:04,177 INFO [train.py:904] (1/8) Epoch 28, batch 7100, loss[loss=0.1976, simple_loss=0.2841, pruned_loss=0.0556, over 16430.00 frames. ], tot_loss[loss=0.202, simple_loss=0.2904, pruned_loss=0.05678, over 3103777.10 frames. ], batch size: 146, lr: 2.39e-03, grad_scale: 8.0 2023-05-02 14:34:09,527 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.14 vs. limit=2.0 2023-05-02 14:34:29,302 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=281169.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 14:34:37,211 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=281174.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 14:34:43,392 INFO [optim.py:368] (1/8) 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 14:34:59,572 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=281189.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 14:35:20,727 INFO [train.py:904] (1/8) Epoch 28, batch 7150, loss[loss=0.2405, simple_loss=0.3129, pruned_loss=0.08407, over 11746.00 frames. ], tot_loss[loss=0.2017, simple_loss=0.2892, pruned_loss=0.05705, over 3085413.71 frames. ], batch size: 248, lr: 2.39e-03, grad_scale: 8.0 2023-05-02 14:35:20,998 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=281203.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 14:35:47,781 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=281222.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 14:36:00,701 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=281230.0, num_to_drop=1, layers_to_drop={2} 2023-05-02 14:36:00,947 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.70 vs. limit=2.0 2023-05-02 14:36:09,138 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=281237.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 14:36:29,668 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=281251.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 14:36:31,491 INFO [train.py:904] (1/8) Epoch 28, batch 7200, loss[loss=0.1852, simple_loss=0.2758, pruned_loss=0.04726, over 15422.00 frames. ], tot_loss[loss=0.1993, simple_loss=0.287, pruned_loss=0.05583, over 3065942.17 frames. ], batch size: 190, lr: 2.39e-03, grad_scale: 8.0 2023-05-02 14:36:32,498 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=4.83 vs. limit=5.0 2023-05-02 14:36:38,229 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.5155, 3.6006, 3.3434, 2.9749, 3.2064, 3.4769, 3.3155, 3.2616], device='cuda:1'), covar=tensor([0.0583, 0.0596, 0.0256, 0.0274, 0.0468, 0.0482, 0.1371, 0.0458], device='cuda:1'), in_proj_covar=tensor([0.0308, 0.0466, 0.0360, 0.0363, 0.0359, 0.0415, 0.0247, 0.0430], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002], device='cuda:1') 2023-05-02 14:36:43,196 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.75 vs. limit=2.0 2023-05-02 14:36:56,562 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=281270.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 14:37:07,757 INFO [optim.py:368] (1/8) 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 14:37:46,525 INFO [train.py:904] (1/8) Epoch 28, batch 7250, loss[loss=0.1834, simple_loss=0.2702, pruned_loss=0.04826, over 16543.00 frames. ], tot_loss[loss=0.1965, simple_loss=0.2845, pruned_loss=0.05431, over 3083607.57 frames. ], batch size: 62, lr: 2.39e-03, grad_scale: 8.0 2023-05-02 14:38:09,205 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=281318.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 14:38:42,288 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.15 vs. limit=2.0 2023-05-02 14:38:59,568 INFO [train.py:904] (1/8) Epoch 28, batch 7300, loss[loss=0.1951, simple_loss=0.2849, pruned_loss=0.05267, over 16694.00 frames. ], tot_loss[loss=0.1955, simple_loss=0.2834, pruned_loss=0.05378, over 3088334.37 frames. ], batch size: 134, lr: 2.39e-03, grad_scale: 2.0 2023-05-02 14:39:08,205 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=281358.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 14:39:18,635 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.5553, 2.9409, 3.2638, 2.0028, 2.8704, 2.0657, 3.1377, 3.2310], device='cuda:1'), covar=tensor([0.0293, 0.0885, 0.0602, 0.2222, 0.0870, 0.1096, 0.0734, 0.0991], device='cuda:1'), in_proj_covar=tensor([0.0161, 0.0171, 0.0172, 0.0158, 0.0148, 0.0133, 0.0147, 0.0184], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0005], device='cuda:1') 2023-05-02 14:39:39,517 INFO [optim.py:368] (1/8) 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 14:39:48,753 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.41 vs. limit=2.0 2023-05-02 14:40:13,646 INFO [train.py:904] (1/8) Epoch 28, batch 7350, loss[loss=0.2029, simple_loss=0.2893, pruned_loss=0.05822, over 17002.00 frames. ], tot_loss[loss=0.1971, simple_loss=0.2846, pruned_loss=0.05485, over 3074425.92 frames. ], batch size: 53, lr: 2.39e-03, grad_scale: 2.0 2023-05-02 14:40:16,308 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=281405.0, num_to_drop=1, layers_to_drop={0} 2023-05-02 14:40:17,489 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=281406.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 14:40:32,458 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.8731, 5.1340, 4.9032, 4.9083, 4.6654, 4.6195, 4.5458, 5.2099], device='cuda:1'), covar=tensor([0.1193, 0.0869, 0.1040, 0.0982, 0.0820, 0.1112, 0.1265, 0.0829], device='cuda:1'), in_proj_covar=tensor([0.0710, 0.0864, 0.0706, 0.0664, 0.0544, 0.0544, 0.0718, 0.0672], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:1') 2023-05-02 14:40:49,374 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=281426.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 14:41:28,055 INFO [train.py:904] (1/8) Epoch 28, batch 7400, loss[loss=0.2355, simple_loss=0.3068, pruned_loss=0.08209, over 11704.00 frames. ], tot_loss[loss=0.198, simple_loss=0.2855, pruned_loss=0.05521, over 3082202.59 frames. ], batch size: 248, lr: 2.39e-03, grad_scale: 2.0 2023-05-02 14:41:33,737 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=281456.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 14:42:08,144 INFO [optim.py:368] (1/8) 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 14:42:18,977 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=281487.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 14:42:23,302 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.15 vs. limit=2.0 2023-05-02 14:42:26,329 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.22 vs. limit=2.0 2023-05-02 14:42:33,195 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.3958, 3.3649, 3.4274, 3.5017, 3.5218, 3.2964, 3.5065, 3.5792], device='cuda:1'), covar=tensor([0.1266, 0.0940, 0.1053, 0.0648, 0.0736, 0.2179, 0.1220, 0.0913], device='cuda:1'), in_proj_covar=tensor([0.0665, 0.0812, 0.0941, 0.0825, 0.0632, 0.0660, 0.0693, 0.0799], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-02 14:42:42,678 INFO [train.py:904] (1/8) Epoch 28, batch 7450, loss[loss=0.2407, simple_loss=0.3032, pruned_loss=0.08912, over 11799.00 frames. ], tot_loss[loss=0.1993, simple_loss=0.2864, pruned_loss=0.05615, over 3070228.74 frames. ], batch size: 246, lr: 2.39e-03, grad_scale: 2.0 2023-05-02 14:43:05,947 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=281517.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 14:43:09,894 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([1.5981, 2.4537, 2.3538, 3.5322, 1.9891, 3.6937, 1.4817, 2.6631], device='cuda:1'), covar=tensor([0.1539, 0.0861, 0.1345, 0.0235, 0.0133, 0.0419, 0.1823, 0.0929], device='cuda:1'), in_proj_covar=tensor([0.0172, 0.0181, 0.0201, 0.0203, 0.0208, 0.0219, 0.0210, 0.0199], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-05-02 14:43:17,713 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=281525.0, num_to_drop=1, layers_to_drop={0} 2023-05-02 14:43:58,005 INFO [train.py:904] (1/8) Epoch 28, batch 7500, loss[loss=0.2389, simple_loss=0.3038, pruned_loss=0.08701, over 11272.00 frames. ], tot_loss[loss=0.1992, simple_loss=0.2867, pruned_loss=0.05581, over 3058287.26 frames. ], batch size: 250, lr: 2.39e-03, grad_scale: 2.0 2023-05-02 14:44:36,766 INFO [optim.py:368] (1/8) 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 14:45:11,671 INFO [train.py:904] (1/8) Epoch 28, batch 7550, loss[loss=0.1766, simple_loss=0.2679, pruned_loss=0.0426, over 16582.00 frames. ], tot_loss[loss=0.1992, simple_loss=0.2862, pruned_loss=0.0561, over 3051545.60 frames. ], batch size: 75, lr: 2.39e-03, grad_scale: 2.0 2023-05-02 14:46:25,956 INFO [train.py:904] (1/8) Epoch 28, batch 7600, loss[loss=0.1916, simple_loss=0.2909, pruned_loss=0.04614, over 16816.00 frames. ], tot_loss[loss=0.199, simple_loss=0.2856, pruned_loss=0.0562, over 3061006.90 frames. ], batch size: 83, lr: 2.39e-03, grad_scale: 4.0 2023-05-02 14:46:31,615 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.3580, 2.5323, 2.4750, 4.1746, 2.3675, 2.8527, 2.5217, 2.6272], device='cuda:1'), covar=tensor([0.1518, 0.3498, 0.2937, 0.0546, 0.4004, 0.2426, 0.3675, 0.3190], device='cuda:1'), in_proj_covar=tensor([0.0421, 0.0474, 0.0386, 0.0337, 0.0447, 0.0543, 0.0448, 0.0555], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-02 14:47:04,824 INFO [optim.py:368] (1/8) 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] (1/8) Epoch 28, batch 7650, loss[loss=0.2612, simple_loss=0.3288, pruned_loss=0.09679, over 11190.00 frames. ], tot_loss[loss=0.1991, simple_loss=0.2857, pruned_loss=0.05627, over 3070175.82 frames. ], batch size: 248, lr: 2.39e-03, grad_scale: 4.0 2023-05-02 14:47:43,622 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=281705.0, num_to_drop=1, layers_to_drop={0} 2023-05-02 14:48:20,114 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=4.78 vs. limit=5.0 2023-05-02 14:48:51,546 INFO [train.py:904] (1/8) Epoch 28, batch 7700, loss[loss=0.1808, simple_loss=0.2658, pruned_loss=0.04791, over 16368.00 frames. ], tot_loss[loss=0.2002, simple_loss=0.286, pruned_loss=0.05719, over 3056355.14 frames. ], batch size: 146, lr: 2.39e-03, grad_scale: 4.0 2023-05-02 14:48:52,554 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=281753.0, num_to_drop=1, layers_to_drop={1} 2023-05-02 14:49:31,077 INFO [optim.py:368] (1/8) 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 14:49:35,177 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 28, batch 7750, loss[loss=0.1669, simple_loss=0.258, pruned_loss=0.0379, over 17173.00 frames. ], tot_loss[loss=0.1997, simple_loss=0.2859, pruned_loss=0.05677, over 3069523.43 frames. ], batch size: 46, lr: 2.39e-03, grad_scale: 4.0 2023-05-02 14:50:21,324 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=281812.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 14:50:40,429 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=281825.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 14:50:53,684 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([1.8802, 3.1916, 3.1749, 2.0496, 3.0361, 3.2364, 3.0390, 1.7962], device='cuda:1'), covar=tensor([0.0649, 0.0090, 0.0084, 0.0492, 0.0140, 0.0145, 0.0126, 0.0578], device='cuda:1'), in_proj_covar=tensor([0.0137, 0.0089, 0.0090, 0.0134, 0.0101, 0.0115, 0.0098, 0.0131], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0003, 0.0004], device='cuda:1') 2023-05-02 14:51:03,098 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.4577, 4.4862, 4.8514, 4.8168, 4.8378, 4.5339, 4.5193, 4.4463], device='cuda:1'), covar=tensor([0.0358, 0.0624, 0.0363, 0.0383, 0.0459, 0.0415, 0.0935, 0.0477], device='cuda:1'), in_proj_covar=tensor([0.0430, 0.0484, 0.0469, 0.0433, 0.0516, 0.0494, 0.0570, 0.0397], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:1') 2023-05-02 14:51:16,454 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.7942, 3.7149, 3.8614, 3.6009, 3.8634, 4.2076, 3.8965, 3.6235], device='cuda:1'), covar=tensor([0.1898, 0.2629, 0.2957, 0.2702, 0.2611, 0.1979, 0.1983, 0.2794], device='cuda:1'), in_proj_covar=tensor([0.0426, 0.0636, 0.0700, 0.0515, 0.0690, 0.0723, 0.0546, 0.0691], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-05-02 14:51:21,105 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=281852.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 14:51:21,947 INFO [train.py:904] (1/8) Epoch 28, batch 7800, loss[loss=0.1722, simple_loss=0.2675, pruned_loss=0.0384, over 16881.00 frames. ], tot_loss[loss=0.2001, simple_loss=0.2864, pruned_loss=0.05685, over 3079558.10 frames. ], batch size: 96, lr: 2.39e-03, grad_scale: 4.0 2023-05-02 14:51:29,167 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.2396, 4.3532, 4.4918, 4.2842, 4.3784, 4.8440, 4.3745, 4.1346], device='cuda:1'), covar=tensor([0.1713, 0.1942, 0.2446, 0.2051, 0.2322, 0.1059, 0.1658, 0.2446], device='cuda:1'), in_proj_covar=tensor([0.0426, 0.0637, 0.0701, 0.0515, 0.0690, 0.0723, 0.0546, 0.0691], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-05-02 14:51:52,695 INFO [zipformer.py:625] (1/8) 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] (1/8) 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 14:52:17,525 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([1.8228, 1.3911, 1.7425, 1.6528, 1.8231, 1.8686, 1.7220, 1.8377], device='cuda:1'), covar=tensor([0.0261, 0.0396, 0.0228, 0.0324, 0.0299, 0.0216, 0.0413, 0.0164], device='cuda:1'), in_proj_covar=tensor([0.0195, 0.0197, 0.0185, 0.0190, 0.0206, 0.0164, 0.0201, 0.0164], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-02 14:52:37,335 INFO [train.py:904] (1/8) Epoch 28, batch 7850, loss[loss=0.181, simple_loss=0.2842, pruned_loss=0.03886, over 16696.00 frames. ], tot_loss[loss=0.2002, simple_loss=0.2871, pruned_loss=0.05669, over 3074949.75 frames. ], batch size: 76, lr: 2.39e-03, grad_scale: 4.0 2023-05-02 14:52:47,351 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=281910.0, num_to_drop=1, layers_to_drop={0} 2023-05-02 14:52:51,853 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=281913.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 14:53:02,773 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.7182, 4.5114, 4.4419, 3.0475, 3.9970, 4.4981, 3.8805, 2.5102], device='cuda:1'), covar=tensor([0.0540, 0.0051, 0.0054, 0.0387, 0.0095, 0.0110, 0.0104, 0.0479], device='cuda:1'), in_proj_covar=tensor([0.0136, 0.0089, 0.0090, 0.0134, 0.0101, 0.0115, 0.0097, 0.0130], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:1') 2023-05-02 14:53:07,536 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.8698, 2.1466, 2.4573, 3.0732, 2.2108, 2.3253, 2.3522, 2.2581], device='cuda:1'), covar=tensor([0.1509, 0.3484, 0.2674, 0.0819, 0.4342, 0.2409, 0.3281, 0.3519], device='cuda:1'), in_proj_covar=tensor([0.0418, 0.0473, 0.0384, 0.0335, 0.0445, 0.0541, 0.0446, 0.0552], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-02 14:53:37,331 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=281945.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 14:53:48,936 INFO [train.py:904] (1/8) Epoch 28, batch 7900, loss[loss=0.1988, simple_loss=0.2955, pruned_loss=0.05102, over 16382.00 frames. ], tot_loss[loss=0.199, simple_loss=0.2859, pruned_loss=0.05604, over 3091429.18 frames. ], batch size: 146, lr: 2.39e-03, grad_scale: 4.0 2023-05-02 14:54:16,275 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=281971.0, num_to_drop=1, layers_to_drop={2} 2023-05-02 14:54:29,909 INFO [optim.py:368] (1/8) 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 14:55:09,784 INFO [train.py:904] (1/8) Epoch 28, batch 7950, loss[loss=0.2525, simple_loss=0.3096, pruned_loss=0.09773, over 11929.00 frames. ], tot_loss[loss=0.2002, simple_loss=0.2866, pruned_loss=0.05686, over 3074689.98 frames. ], batch size: 246, lr: 2.39e-03, grad_scale: 4.0 2023-05-02 14:55:14,761 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=282006.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 14:55:49,840 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.48 vs. limit=2.0 2023-05-02 14:56:22,096 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.8594, 5.2120, 5.3948, 5.1032, 5.1943, 5.7419, 5.1588, 4.9550], device='cuda:1'), covar=tensor([0.1063, 0.1704, 0.2154, 0.1917, 0.2245, 0.0865, 0.1589, 0.2316], device='cuda:1'), in_proj_covar=tensor([0.0427, 0.0637, 0.0701, 0.0516, 0.0692, 0.0725, 0.0547, 0.0693], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-05-02 14:56:27,143 INFO [train.py:904] (1/8) Epoch 28, batch 8000, loss[loss=0.1849, simple_loss=0.2778, pruned_loss=0.04601, over 17161.00 frames. ], tot_loss[loss=0.2002, simple_loss=0.2868, pruned_loss=0.05686, over 3089393.47 frames. ], batch size: 40, lr: 2.39e-03, grad_scale: 8.0 2023-05-02 14:56:37,753 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.1296, 3.4826, 3.4907, 2.2342, 3.2802, 3.5546, 3.2764, 2.0995], device='cuda:1'), covar=tensor([0.0562, 0.0082, 0.0076, 0.0453, 0.0114, 0.0127, 0.0114, 0.0495], device='cuda:1'), in_proj_covar=tensor([0.0137, 0.0089, 0.0090, 0.0134, 0.0101, 0.0115, 0.0097, 0.0130], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:1') 2023-05-02 14:56:50,084 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.2697, 3.0102, 3.3289, 1.8174, 3.4348, 3.4752, 2.8108, 2.6097], device='cuda:1'), covar=tensor([0.0878, 0.0322, 0.0213, 0.1246, 0.0105, 0.0239, 0.0462, 0.0518], device='cuda:1'), in_proj_covar=tensor([0.0150, 0.0112, 0.0104, 0.0140, 0.0087, 0.0132, 0.0130, 0.0132], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-05-02 14:57:07,769 INFO [optim.py:368] (1/8) 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 14:57:11,190 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=282082.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 14:57:33,404 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=282096.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 14:57:42,464 INFO [train.py:904] (1/8) Epoch 28, batch 8050, loss[loss=0.2161, simple_loss=0.3092, pruned_loss=0.06155, over 15385.00 frames. ], tot_loss[loss=0.1995, simple_loss=0.2863, pruned_loss=0.05637, over 3093849.56 frames. ], batch size: 190, lr: 2.39e-03, grad_scale: 8.0 2023-05-02 14:57:56,153 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=282112.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 14:58:22,176 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=282130.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 14:58:22,644 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.30 vs. limit=2.0 2023-05-02 14:58:57,914 INFO [train.py:904] (1/8) Epoch 28, batch 8100, loss[loss=0.182, simple_loss=0.2694, pruned_loss=0.04727, over 17123.00 frames. ], tot_loss[loss=0.1983, simple_loss=0.2857, pruned_loss=0.05547, over 3110971.23 frames. ], batch size: 49, lr: 2.39e-03, grad_scale: 8.0 2023-05-02 14:59:03,794 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=282157.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 14:59:08,427 INFO [zipformer.py:625] (1/8) 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] (1/8) 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 15:00:13,682 INFO [train.py:904] (1/8) Epoch 28, batch 8150, loss[loss=0.1936, simple_loss=0.2762, pruned_loss=0.05545, over 15470.00 frames. ], tot_loss[loss=0.1969, simple_loss=0.2841, pruned_loss=0.05488, over 3121625.30 frames. ], batch size: 191, lr: 2.39e-03, grad_scale: 8.0 2023-05-02 15:00:14,136 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([1.8767, 2.7049, 2.5518, 1.9278, 2.5665, 2.7063, 2.5516, 1.9519], device='cuda:1'), covar=tensor([0.0482, 0.0104, 0.0099, 0.0395, 0.0156, 0.0157, 0.0142, 0.0432], device='cuda:1'), in_proj_covar=tensor([0.0138, 0.0089, 0.0091, 0.0134, 0.0102, 0.0115, 0.0098, 0.0131], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0003, 0.0004], device='cuda:1') 2023-05-02 15:00:21,752 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=282208.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 15:01:29,259 INFO [train.py:904] (1/8) Epoch 28, batch 8200, loss[loss=0.1767, simple_loss=0.2663, pruned_loss=0.04353, over 16632.00 frames. ], tot_loss[loss=0.1943, simple_loss=0.2809, pruned_loss=0.05385, over 3124401.15 frames. ], batch size: 134, lr: 2.39e-03, grad_scale: 8.0 2023-05-02 15:01:47,358 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.16 vs. limit=2.0 2023-05-02 15:01:50,074 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=282266.0, num_to_drop=1, layers_to_drop={0} 2023-05-02 15:02:13,183 INFO [optim.py:368] (1/8) 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 15:02:46,942 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=282301.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 15:02:50,325 INFO [train.py:904] (1/8) Epoch 28, batch 8250, loss[loss=0.1938, simple_loss=0.2914, pruned_loss=0.04809, over 16376.00 frames. ], tot_loss[loss=0.1914, simple_loss=0.2797, pruned_loss=0.05158, over 3089555.36 frames. ], batch size: 146, lr: 2.39e-03, grad_scale: 4.0 2023-05-02 15:04:07,366 INFO [train.py:904] (1/8) Epoch 28, batch 8300, loss[loss=0.1852, simple_loss=0.2767, pruned_loss=0.04686, over 16746.00 frames. ], tot_loss[loss=0.1872, simple_loss=0.2772, pruned_loss=0.04854, over 3099577.83 frames. ], batch size: 124, lr: 2.39e-03, grad_scale: 4.0 2023-05-02 15:04:19,287 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.9977, 3.2596, 3.7427, 2.1678, 3.1927, 2.2452, 3.5520, 3.4295], device='cuda:1'), covar=tensor([0.0234, 0.0929, 0.0440, 0.2183, 0.0741, 0.1070, 0.0572, 0.0952], device='cuda:1'), in_proj_covar=tensor([0.0160, 0.0170, 0.0170, 0.0157, 0.0147, 0.0132, 0.0146, 0.0182], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0005], device='cuda:1') 2023-05-02 15:04:50,785 INFO [optim.py:368] (1/8) Clipping_scale=2.0, grad-norm quartiles 1.600e+02 2.126e+02 2.473e+02 2.954e+02 5.391e+02, threshold=4.946e+02, percent-clipped=0.0 2023-05-02 15:05:26,438 INFO [train.py:904] (1/8) Epoch 28, batch 8350, loss[loss=0.1758, simple_loss=0.2789, pruned_loss=0.03639, over 16918.00 frames. ], tot_loss[loss=0.1853, simple_loss=0.277, pruned_loss=0.04685, over 3099649.63 frames. ], batch size: 96, lr: 2.39e-03, grad_scale: 4.0 2023-05-02 15:05:56,625 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.4517, 4.6953, 4.5514, 4.5623, 4.2899, 4.2090, 4.1613, 4.7515], device='cuda:1'), covar=tensor([0.1153, 0.0946, 0.0943, 0.0869, 0.0772, 0.1650, 0.1212, 0.0827], device='cuda:1'), in_proj_covar=tensor([0.0715, 0.0864, 0.0708, 0.0668, 0.0544, 0.0548, 0.0722, 0.0672], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:1') 2023-05-02 15:06:43,063 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=282452.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 15:06:43,836 INFO [train.py:904] (1/8) Epoch 28, batch 8400, loss[loss=0.1768, simple_loss=0.2703, pruned_loss=0.04164, over 16341.00 frames. ], tot_loss[loss=0.1826, simple_loss=0.2749, pruned_loss=0.04512, over 3076461.03 frames. ], batch size: 146, lr: 2.39e-03, grad_scale: 8.0 2023-05-02 15:06:59,947 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([1.7693, 1.4262, 1.7626, 1.7093, 1.8872, 1.8691, 1.7410, 1.7839], device='cuda:1'), covar=tensor([0.0285, 0.0434, 0.0230, 0.0333, 0.0328, 0.0222, 0.0467, 0.0168], device='cuda:1'), in_proj_covar=tensor([0.0194, 0.0196, 0.0184, 0.0189, 0.0206, 0.0163, 0.0200, 0.0163], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-02 15:07:26,987 INFO [optim.py:368] (1/8) Clipping_scale=2.0, grad-norm quartiles 1.421e+02 2.221e+02 2.637e+02 3.332e+02 6.864e+02, threshold=5.273e+02, percent-clipped=5.0 2023-05-02 15:07:40,041 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 28, batch 8450, loss[loss=0.1643, simple_loss=0.2621, pruned_loss=0.03322, over 16720.00 frames. ], tot_loss[loss=0.1798, simple_loss=0.2728, pruned_loss=0.04339, over 3074884.70 frames. ], batch size: 76, lr: 2.38e-03, grad_scale: 8.0 2023-05-02 15:08:11,392 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=282508.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 15:09:17,260 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 28, batch 8500, loss[loss=0.1588, simple_loss=0.252, pruned_loss=0.03281, over 16290.00 frames. ], tot_loss[loss=0.1758, simple_loss=0.269, pruned_loss=0.04128, over 3063142.18 frames. ], batch size: 165, lr: 2.38e-03, grad_scale: 8.0 2023-05-02 15:09:27,448 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=282556.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 15:09:43,299 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=282566.0, num_to_drop=1, layers_to_drop={0} 2023-05-02 15:09:54,221 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([1.8942, 2.6991, 2.5572, 1.9424, 2.4673, 2.7228, 2.6175, 1.9219], device='cuda:1'), covar=tensor([0.0416, 0.0093, 0.0087, 0.0373, 0.0153, 0.0124, 0.0112, 0.0450], device='cuda:1'), in_proj_covar=tensor([0.0136, 0.0088, 0.0089, 0.0132, 0.0100, 0.0113, 0.0096, 0.0128], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:1') 2023-05-02 15:09:54,267 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([1.7350, 1.3493, 1.7377, 1.6962, 1.8344, 1.8727, 1.7158, 1.8168], device='cuda:1'), covar=tensor([0.0273, 0.0457, 0.0259, 0.0325, 0.0345, 0.0245, 0.0464, 0.0168], device='cuda:1'), in_proj_covar=tensor([0.0195, 0.0196, 0.0184, 0.0189, 0.0206, 0.0164, 0.0200, 0.0164], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-02 15:10:07,072 INFO [optim.py:368] (1/8) 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 15:10:22,749 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.1273, 4.1274, 4.4391, 4.4204, 4.4161, 4.1738, 4.1625, 4.1833], device='cuda:1'), covar=tensor([0.0535, 0.0935, 0.0651, 0.0596, 0.0694, 0.0704, 0.1212, 0.0698], device='cuda:1'), in_proj_covar=tensor([0.0433, 0.0488, 0.0472, 0.0436, 0.0518, 0.0496, 0.0572, 0.0400], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:1') 2023-05-02 15:10:42,296 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 28, batch 8550, loss[loss=0.1785, simple_loss=0.2763, pruned_loss=0.04032, over 16771.00 frames. ], tot_loss[loss=0.1738, simple_loss=0.267, pruned_loss=0.04031, over 3059512.14 frames. ], batch size: 124, lr: 2.38e-03, grad_scale: 8.0 2023-05-02 15:11:05,481 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=282614.0, num_to_drop=1, layers_to_drop={0} 2023-05-02 15:11:45,157 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.5864, 4.8604, 4.6919, 4.6775, 4.4009, 4.3728, 4.3371, 4.9262], device='cuda:1'), covar=tensor([0.1111, 0.0930, 0.0938, 0.0866, 0.0828, 0.1457, 0.1139, 0.0890], device='cuda:1'), in_proj_covar=tensor([0.0706, 0.0853, 0.0700, 0.0660, 0.0538, 0.0542, 0.0711, 0.0664], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:1') 2023-05-02 15:12:14,466 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=282649.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 15:12:21,564 INFO [train.py:904] (1/8) Epoch 28, batch 8600, loss[loss=0.1757, simple_loss=0.2719, pruned_loss=0.0398, over 16868.00 frames. ], tot_loss[loss=0.1731, simple_loss=0.2669, pruned_loss=0.03964, over 3040479.40 frames. ], batch size: 116, lr: 2.38e-03, grad_scale: 8.0 2023-05-02 15:12:32,756 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=282658.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 15:13:11,909 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.5727, 3.8633, 3.9469, 2.7658, 3.4817, 3.9481, 3.7045, 2.1993], device='cuda:1'), covar=tensor([0.0535, 0.0056, 0.0045, 0.0396, 0.0128, 0.0088, 0.0072, 0.0531], device='cuda:1'), in_proj_covar=tensor([0.0135, 0.0087, 0.0089, 0.0132, 0.0100, 0.0112, 0.0096, 0.0128], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:1') 2023-05-02 15:13:11,925 INFO [zipformer.py:625] (1/8) 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] (1/8) Clipping_scale=2.0, grad-norm quartiles 1.534e+02 2.220e+02 2.463e+02 3.072e+02 6.042e+02, threshold=4.927e+02, percent-clipped=1.0 2023-05-02 15:13:50,091 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 28, batch 8650, loss[loss=0.1562, simple_loss=0.2573, pruned_loss=0.02761, over 15331.00 frames. ], tot_loss[loss=0.1707, simple_loss=0.2649, pruned_loss=0.03825, over 3025456.34 frames. ], batch size: 191, lr: 2.38e-03, grad_scale: 8.0 2023-05-02 15:14:20,135 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.1354, 2.3679, 2.0655, 2.1181, 2.6714, 2.3573, 2.6034, 2.8484], device='cuda:1'), covar=tensor([0.0177, 0.0514, 0.0595, 0.0599, 0.0358, 0.0502, 0.0239, 0.0338], device='cuda:1'), in_proj_covar=tensor([0.0220, 0.0235, 0.0227, 0.0227, 0.0237, 0.0235, 0.0233, 0.0235], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-02 15:14:36,227 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=282719.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 15:15:14,366 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=282738.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 15:15:20,338 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.2180, 3.3437, 2.1211, 3.6531, 2.5243, 3.6027, 2.2959, 2.7166], device='cuda:1'), covar=tensor([0.0411, 0.0432, 0.1741, 0.0302, 0.0924, 0.0660, 0.1551, 0.0861], device='cuda:1'), in_proj_covar=tensor([0.0173, 0.0177, 0.0193, 0.0168, 0.0177, 0.0216, 0.0201, 0.0180], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-05-02 15:15:40,173 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 28, batch 8700, loss[loss=0.1585, simple_loss=0.2576, pruned_loss=0.0297, over 16311.00 frames. ], tot_loss[loss=0.1685, simple_loss=0.2626, pruned_loss=0.03714, over 3054262.63 frames. ], batch size: 146, lr: 2.38e-03, grad_scale: 8.0 2023-05-02 15:15:53,207 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=282759.0, num_to_drop=1, layers_to_drop={3} 2023-05-02 15:16:29,665 INFO [optim.py:368] (1/8) 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 15:17:06,893 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=282800.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 15:17:12,010 INFO [train.py:904] (1/8) Epoch 28, batch 8750, loss[loss=0.1759, simple_loss=0.2813, pruned_loss=0.03525, over 16691.00 frames. ], tot_loss[loss=0.1683, simple_loss=0.2632, pruned_loss=0.03669, over 3068360.88 frames. ], batch size: 134, lr: 2.38e-03, grad_scale: 8.0 2023-05-02 15:17:18,923 INFO [zipformer.py:625] (1/8) 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] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=282845.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 15:19:02,845 INFO [train.py:904] (1/8) Epoch 28, batch 8800, loss[loss=0.1665, simple_loss=0.2647, pruned_loss=0.03413, over 15440.00 frames. ], tot_loss[loss=0.1669, simple_loss=0.2623, pruned_loss=0.0358, over 3070621.06 frames. ], batch size: 191, lr: 2.38e-03, grad_scale: 8.0 2023-05-02 15:19:29,129 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=282866.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 15:19:43,650 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.83 vs. limit=2.0 2023-05-02 15:20:00,725 INFO [optim.py:368] (1/8) 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 15:20:14,609 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.4741, 3.1603, 3.5026, 1.7735, 3.6447, 3.7009, 2.9064, 2.8244], device='cuda:1'), covar=tensor([0.0754, 0.0305, 0.0226, 0.1246, 0.0099, 0.0180, 0.0455, 0.0462], device='cuda:1'), in_proj_covar=tensor([0.0143, 0.0107, 0.0099, 0.0135, 0.0083, 0.0126, 0.0125, 0.0127], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0004, 0.0003], device='cuda:1') 2023-05-02 15:20:23,004 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.9637, 5.0134, 5.3375, 5.3061, 5.3487, 5.0553, 5.0088, 4.7704], device='cuda:1'), covar=tensor([0.0285, 0.0501, 0.0383, 0.0359, 0.0346, 0.0374, 0.0776, 0.0442], device='cuda:1'), in_proj_covar=tensor([0.0425, 0.0478, 0.0463, 0.0427, 0.0509, 0.0487, 0.0560, 0.0394], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:1') 2023-05-02 15:20:47,892 INFO [train.py:904] (1/8) Epoch 28, batch 8850, loss[loss=0.151, simple_loss=0.259, pruned_loss=0.02154, over 15307.00 frames. ], tot_loss[loss=0.1676, simple_loss=0.2644, pruned_loss=0.03544, over 3052144.27 frames. ], batch size: 191, lr: 2.38e-03, grad_scale: 8.0 2023-05-02 15:21:46,931 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([1.9716, 1.8522, 1.6834, 1.5067, 1.9924, 1.6313, 1.5074, 1.9597], device='cuda:1'), covar=tensor([0.0205, 0.0339, 0.0492, 0.0395, 0.0282, 0.0319, 0.0200, 0.0249], device='cuda:1'), in_proj_covar=tensor([0.0219, 0.0234, 0.0226, 0.0226, 0.0236, 0.0234, 0.0231, 0.0234], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-02 15:22:20,292 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.7992, 4.1335, 4.2150, 2.9547, 3.6645, 4.2264, 3.8893, 2.5204], device='cuda:1'), covar=tensor([0.0455, 0.0059, 0.0043, 0.0364, 0.0120, 0.0091, 0.0067, 0.0425], device='cuda:1'), in_proj_covar=tensor([0.0134, 0.0086, 0.0087, 0.0131, 0.0099, 0.0111, 0.0095, 0.0127], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-05-02 15:22:35,072 INFO [train.py:904] (1/8) Epoch 28, batch 8900, loss[loss=0.1742, simple_loss=0.2694, pruned_loss=0.03949, over 16986.00 frames. ], tot_loss[loss=0.1673, simple_loss=0.2648, pruned_loss=0.03495, over 3054915.09 frames. ], batch size: 109, lr: 2.38e-03, grad_scale: 8.0 2023-05-02 15:23:38,029 INFO [optim.py:368] (1/8) 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 15:24:19,487 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.73 vs. limit=2.0 2023-05-02 15:24:39,869 INFO [train.py:904] (1/8) Epoch 28, batch 8950, loss[loss=0.144, simple_loss=0.2448, pruned_loss=0.0216, over 16313.00 frames. ], tot_loss[loss=0.1677, simple_loss=0.2645, pruned_loss=0.03547, over 3058979.28 frames. ], batch size: 165, lr: 2.38e-03, grad_scale: 8.0 2023-05-02 15:25:04,563 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=283014.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 15:25:45,364 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=283033.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 15:26:01,323 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.27 vs. limit=2.0 2023-05-02 15:26:27,360 INFO [train.py:904] (1/8) Epoch 28, batch 9000, loss[loss=0.1661, simple_loss=0.2551, pruned_loss=0.03854, over 16941.00 frames. ], tot_loss[loss=0.1645, simple_loss=0.261, pruned_loss=0.03396, over 3077113.60 frames. ], batch size: 116, lr: 2.38e-03, grad_scale: 8.0 2023-05-02 15:26:27,360 INFO [train.py:929] (1/8) Computing validation loss 2023-05-02 15:26:38,046 INFO [train.py:938] (1/8) Epoch 28, validation: loss=0.1436, simple_loss=0.2472, pruned_loss=0.02006, over 944034.00 frames. 2023-05-02 15:26:38,047 INFO [train.py:939] (1/8) Maximum memory allocated so far is 18145MB 2023-05-02 15:26:41,524 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=283054.0, num_to_drop=1, layers_to_drop={2} 2023-05-02 15:27:36,920 INFO [optim.py:368] (1/8) 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 15:28:03,421 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.64 vs. limit=2.0 2023-05-02 15:28:21,093 INFO [train.py:904] (1/8) Epoch 28, batch 9050, loss[loss=0.1705, simple_loss=0.2599, pruned_loss=0.04056, over 12672.00 frames. ], tot_loss[loss=0.1649, simple_loss=0.2611, pruned_loss=0.03429, over 3069006.67 frames. ], batch size: 246, lr: 2.38e-03, grad_scale: 8.0 2023-05-02 15:28:48,249 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=283115.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 15:29:46,343 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=283144.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 15:29:48,388 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=283145.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 15:30:04,166 INFO [train.py:904] (1/8) Epoch 28, batch 9100, loss[loss=0.1682, simple_loss=0.2593, pruned_loss=0.03857, over 12362.00 frames. ], tot_loss[loss=0.1648, simple_loss=0.2607, pruned_loss=0.03446, over 3085159.31 frames. ], batch size: 247, lr: 2.38e-03, grad_scale: 8.0 2023-05-02 15:30:20,073 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=283161.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 15:30:57,963 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=283176.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 15:31:08,548 INFO [optim.py:368] (1/8) 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] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=283193.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 15:32:01,010 INFO [train.py:904] (1/8) Epoch 28, batch 9150, loss[loss=0.1743, simple_loss=0.2626, pruned_loss=0.04306, over 16576.00 frames. ], tot_loss[loss=0.1655, simple_loss=0.2618, pruned_loss=0.03455, over 3074126.34 frames. ], batch size: 68, lr: 2.38e-03, grad_scale: 8.0 2023-05-02 15:32:03,869 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.1459, 4.1997, 4.4469, 4.4439, 4.4471, 4.2342, 4.2161, 4.1760], device='cuda:1'), covar=tensor([0.0332, 0.0648, 0.0445, 0.0404, 0.0402, 0.0396, 0.0766, 0.0492], device='cuda:1'), in_proj_covar=tensor([0.0424, 0.0475, 0.0463, 0.0426, 0.0507, 0.0485, 0.0557, 0.0393], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:1') 2023-05-02 15:32:06,349 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=283205.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 15:32:46,233 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.5815, 4.7660, 4.9535, 4.6890, 4.7612, 5.2891, 4.7803, 4.4757], device='cuda:1'), covar=tensor([0.1244, 0.1835, 0.1866, 0.1919, 0.2245, 0.0844, 0.1605, 0.2418], device='cuda:1'), in_proj_covar=tensor([0.0413, 0.0613, 0.0676, 0.0498, 0.0666, 0.0705, 0.0529, 0.0668], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-05-02 15:33:12,067 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.0708, 4.1671, 3.9592, 3.6601, 3.7453, 4.0733, 3.7269, 3.8648], device='cuda:1'), covar=tensor([0.0597, 0.0611, 0.0295, 0.0289, 0.0552, 0.0563, 0.1061, 0.0573], device='cuda:1'), in_proj_covar=tensor([0.0302, 0.0454, 0.0353, 0.0352, 0.0348, 0.0406, 0.0242, 0.0419], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-02 15:33:16,464 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.1266, 3.3658, 3.3707, 2.2815, 3.0527, 3.4197, 3.2073, 2.0278], device='cuda:1'), covar=tensor([0.0533, 0.0068, 0.0069, 0.0431, 0.0145, 0.0106, 0.0103, 0.0531], device='cuda:1'), in_proj_covar=tensor([0.0134, 0.0086, 0.0088, 0.0131, 0.0099, 0.0111, 0.0095, 0.0127], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-05-02 15:33:44,271 INFO [train.py:904] (1/8) Epoch 28, batch 9200, loss[loss=0.1611, simple_loss=0.2611, pruned_loss=0.03057, over 16248.00 frames. ], tot_loss[loss=0.1626, simple_loss=0.2577, pruned_loss=0.03378, over 3087156.56 frames. ], batch size: 165, lr: 2.38e-03, grad_scale: 8.0 2023-05-02 15:34:10,113 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.26 vs. limit=2.0 2023-05-02 15:34:34,270 INFO [optim.py:368] (1/8) 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 15:34:39,306 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=283283.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 15:35:20,275 INFO [train.py:904] (1/8) Epoch 28, batch 9250, loss[loss=0.1451, simple_loss=0.2322, pruned_loss=0.02895, over 12436.00 frames. ], tot_loss[loss=0.1628, simple_loss=0.2574, pruned_loss=0.03412, over 3064422.55 frames. ], batch size: 250, lr: 2.38e-03, grad_scale: 8.0 2023-05-02 15:35:35,750 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.0475, 3.1135, 2.0267, 3.3278, 2.3371, 3.3144, 2.1361, 2.6215], device='cuda:1'), covar=tensor([0.0354, 0.0379, 0.1622, 0.0234, 0.0867, 0.0527, 0.1599, 0.0812], device='cuda:1'), in_proj_covar=tensor([0.0170, 0.0175, 0.0191, 0.0166, 0.0175, 0.0212, 0.0200, 0.0177], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-05-02 15:35:44,196 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=283314.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 15:36:28,224 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=283333.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 15:36:34,428 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([1.7240, 2.7517, 2.4450, 4.0134, 2.0869, 3.8871, 1.5977, 2.8989], device='cuda:1'), covar=tensor([0.1506, 0.0787, 0.1259, 0.0187, 0.0089, 0.0385, 0.1857, 0.0737], device='cuda:1'), in_proj_covar=tensor([0.0170, 0.0176, 0.0196, 0.0196, 0.0200, 0.0212, 0.0207, 0.0195], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-05-02 15:36:55,367 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=283344.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 15:37:09,905 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.9701, 2.2133, 1.8528, 2.0256, 2.5532, 2.2257, 2.3492, 2.7818], device='cuda:1'), covar=tensor([0.0223, 0.0626, 0.0843, 0.0680, 0.0445, 0.0585, 0.0244, 0.0375], device='cuda:1'), in_proj_covar=tensor([0.0219, 0.0236, 0.0228, 0.0228, 0.0238, 0.0236, 0.0231, 0.0235], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-02 15:37:14,589 INFO [train.py:904] (1/8) Epoch 28, batch 9300, loss[loss=0.1598, simple_loss=0.2487, pruned_loss=0.03547, over 16138.00 frames. ], tot_loss[loss=0.1613, simple_loss=0.2554, pruned_loss=0.03359, over 3041865.70 frames. ], batch size: 165, lr: 2.38e-03, grad_scale: 8.0 2023-05-02 15:37:16,972 INFO [zipformer.py:625] (1/8) 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] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=283362.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 15:38:16,508 INFO [optim.py:368] (1/8) 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 15:38:17,486 INFO [zipformer.py:625] (1/8) 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] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=283402.0, num_to_drop=1, layers_to_drop={0} 2023-05-02 15:38:59,593 INFO [train.py:904] (1/8) Epoch 28, batch 9350, loss[loss=0.1597, simple_loss=0.2595, pruned_loss=0.02996, over 16735.00 frames. ], tot_loss[loss=0.1615, simple_loss=0.2556, pruned_loss=0.03366, over 3050533.68 frames. ], batch size: 83, lr: 2.38e-03, grad_scale: 8.0 2023-05-02 15:39:37,653 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.4774, 3.0506, 2.7351, 2.3051, 2.2551, 2.2998, 3.0227, 2.8662], device='cuda:1'), covar=tensor([0.2575, 0.0686, 0.1690, 0.2983, 0.2767, 0.2428, 0.0465, 0.1546], device='cuda:1'), in_proj_covar=tensor([0.0330, 0.0267, 0.0307, 0.0321, 0.0296, 0.0272, 0.0297, 0.0344], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:1') 2023-05-02 15:40:00,325 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.42 vs. limit=2.0 2023-05-02 15:40:41,025 INFO [train.py:904] (1/8) Epoch 28, batch 9400, loss[loss=0.1714, simple_loss=0.2748, pruned_loss=0.03396, over 16167.00 frames. ], tot_loss[loss=0.1611, simple_loss=0.2556, pruned_loss=0.03328, over 3048920.64 frames. ], batch size: 165, lr: 2.38e-03, grad_scale: 8.0 2023-05-02 15:41:00,036 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=283461.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 15:41:18,937 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=283471.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 15:41:39,544 INFO [optim.py:368] (1/8) 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 15:42:20,611 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 28, batch 9450, loss[loss=0.1673, simple_loss=0.2635, pruned_loss=0.03561, over 16237.00 frames. ], tot_loss[loss=0.1623, simple_loss=0.2575, pruned_loss=0.03358, over 3043307.26 frames. ], batch size: 165, lr: 2.38e-03, grad_scale: 8.0 2023-05-02 15:42:36,255 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=283509.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 15:43:48,960 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=283544.0, num_to_drop=1, layers_to_drop={1} 2023-05-02 15:44:06,790 INFO [train.py:904] (1/8) Epoch 28, batch 9500, loss[loss=0.1634, simple_loss=0.2559, pruned_loss=0.03542, over 16961.00 frames. ], tot_loss[loss=0.1617, simple_loss=0.257, pruned_loss=0.03313, over 3054031.50 frames. ], batch size: 109, lr: 2.38e-03, grad_scale: 8.0 2023-05-02 15:45:03,873 INFO [optim.py:368] (1/8) 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 15:45:04,674 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.18 vs. limit=2.0 2023-05-02 15:45:53,231 INFO [train.py:904] (1/8) Epoch 28, batch 9550, loss[loss=0.1801, simple_loss=0.2812, pruned_loss=0.03955, over 15415.00 frames. ], tot_loss[loss=0.1617, simple_loss=0.2567, pruned_loss=0.03336, over 3054257.59 frames. ], batch size: 191, lr: 2.38e-03, grad_scale: 8.0 2023-05-02 15:45:59,208 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=283605.0, num_to_drop=1, layers_to_drop={2} 2023-05-02 15:46:41,465 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.25 vs. limit=2.0 2023-05-02 15:47:09,792 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 28, batch 9600, loss[loss=0.1726, simple_loss=0.2697, pruned_loss=0.03773, over 16445.00 frames. ], tot_loss[loss=0.1626, simple_loss=0.2578, pruned_loss=0.03373, over 3043383.05 frames. ], batch size: 75, lr: 2.38e-03, grad_scale: 8.0 2023-05-02 15:48:02,099 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.0984, 3.9652, 4.1452, 4.2754, 4.3786, 3.9828, 4.3581, 4.4166], device='cuda:1'), covar=tensor([0.1965, 0.1229, 0.1466, 0.0794, 0.0635, 0.1445, 0.0731, 0.0832], device='cuda:1'), in_proj_covar=tensor([0.0638, 0.0780, 0.0900, 0.0795, 0.0605, 0.0631, 0.0665, 0.0769], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-02 15:48:04,379 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=4.84 vs. limit=5.0 2023-05-02 15:48:29,440 INFO [optim.py:368] (1/8) 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 15:48:55,340 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=4.83 vs. limit=5.0 2023-05-02 15:49:23,008 INFO [train.py:904] (1/8) Epoch 28, batch 9650, loss[loss=0.1528, simple_loss=0.2508, pruned_loss=0.02746, over 16862.00 frames. ], tot_loss[loss=0.1641, simple_loss=0.26, pruned_loss=0.03412, over 3068852.83 frames. ], batch size: 116, lr: 2.38e-03, grad_scale: 8.0 2023-05-02 15:50:39,256 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([1.7775, 2.5487, 2.4595, 4.0686, 1.9763, 3.8184, 1.5488, 2.8329], device='cuda:1'), covar=tensor([0.1596, 0.1026, 0.1381, 0.0193, 0.0118, 0.0432, 0.2019, 0.0904], device='cuda:1'), in_proj_covar=tensor([0.0169, 0.0175, 0.0195, 0.0195, 0.0198, 0.0211, 0.0205, 0.0194], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-05-02 15:51:09,561 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.5214, 4.4942, 4.2909, 3.6411, 4.4113, 1.8263, 4.1979, 4.0301], device='cuda:1'), covar=tensor([0.0120, 0.0124, 0.0223, 0.0274, 0.0116, 0.2722, 0.0143, 0.0262], device='cuda:1'), in_proj_covar=tensor([0.0174, 0.0168, 0.0205, 0.0176, 0.0182, 0.0211, 0.0193, 0.0172], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-02 15:51:10,203 INFO [train.py:904] (1/8) Epoch 28, batch 9700, loss[loss=0.1687, simple_loss=0.2665, pruned_loss=0.03543, over 16719.00 frames. ], tot_loss[loss=0.1636, simple_loss=0.2591, pruned_loss=0.03406, over 3071042.70 frames. ], batch size: 134, lr: 2.38e-03, grad_scale: 8.0 2023-05-02 15:51:47,258 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=283771.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 15:52:08,987 INFO [optim.py:368] (1/8) 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 15:52:40,461 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.7784, 4.7567, 4.5819, 4.0127, 4.6400, 1.8070, 4.4440, 4.4457], device='cuda:1'), covar=tensor([0.0090, 0.0110, 0.0218, 0.0338, 0.0131, 0.2719, 0.0143, 0.0238], device='cuda:1'), in_proj_covar=tensor([0.0175, 0.0168, 0.0205, 0.0177, 0.0182, 0.0212, 0.0193, 0.0172], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0001, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-02 15:52:48,102 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 28, batch 9750, loss[loss=0.1604, simple_loss=0.2613, pruned_loss=0.02973, over 16364.00 frames. ], tot_loss[loss=0.1633, simple_loss=0.2582, pruned_loss=0.03418, over 3065730.03 frames. ], batch size: 146, lr: 2.38e-03, grad_scale: 8.0 2023-05-02 15:52:56,852 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.4561, 4.5875, 4.3731, 4.0419, 4.0855, 4.5081, 4.2265, 4.2055], device='cuda:1'), covar=tensor([0.0659, 0.0610, 0.0344, 0.0344, 0.0889, 0.0522, 0.0540, 0.0718], device='cuda:1'), in_proj_covar=tensor([0.0299, 0.0449, 0.0350, 0.0350, 0.0345, 0.0404, 0.0241, 0.0415], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-02 15:53:24,759 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=283819.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 15:54:12,833 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=283842.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 15:54:24,072 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 28, batch 9800, loss[loss=0.1517, simple_loss=0.2603, pruned_loss=0.02154, over 16940.00 frames. ], tot_loss[loss=0.1618, simple_loss=0.2574, pruned_loss=0.03315, over 3063373.82 frames. ], batch size: 102, lr: 2.38e-03, grad_scale: 8.0 2023-05-02 15:55:23,052 INFO [optim.py:368] (1/8) Clipping_scale=2.0, grad-norm quartiles 1.531e+02 2.061e+02 2.416e+02 3.000e+02 5.773e+02, threshold=4.831e+02, percent-clipped=3.0 2023-05-02 15:56:11,010 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=283900.0, num_to_drop=1, layers_to_drop={1} 2023-05-02 15:56:14,719 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.7881, 5.0484, 4.8671, 4.8925, 4.5966, 4.5844, 4.4453, 5.1388], device='cuda:1'), covar=tensor([0.1204, 0.0865, 0.0915, 0.0834, 0.0788, 0.1126, 0.1319, 0.0841], device='cuda:1'), in_proj_covar=tensor([0.0691, 0.0836, 0.0682, 0.0646, 0.0528, 0.0529, 0.0696, 0.0651], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:1') 2023-05-02 15:56:15,639 INFO [train.py:904] (1/8) Epoch 28, batch 9850, loss[loss=0.1717, simple_loss=0.2689, pruned_loss=0.0373, over 16648.00 frames. ], tot_loss[loss=0.162, simple_loss=0.2588, pruned_loss=0.03266, over 3089461.45 frames. ], batch size: 134, lr: 2.38e-03, grad_scale: 8.0 2023-05-02 15:56:17,409 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=283903.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 15:56:50,931 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.3934, 2.8118, 3.1518, 1.9848, 2.7428, 2.1511, 2.9987, 3.1046], device='cuda:1'), covar=tensor([0.0312, 0.0916, 0.0609, 0.2131, 0.0914, 0.1057, 0.0686, 0.0862], device='cuda:1'), in_proj_covar=tensor([0.0156, 0.0163, 0.0166, 0.0153, 0.0143, 0.0129, 0.0141, 0.0176], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:1') 2023-05-02 15:57:36,916 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=283939.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 15:58:06,899 INFO [train.py:904] (1/8) Epoch 28, batch 9900, loss[loss=0.1553, simple_loss=0.2498, pruned_loss=0.03036, over 12527.00 frames. ], tot_loss[loss=0.162, simple_loss=0.2587, pruned_loss=0.03266, over 3056718.16 frames. ], batch size: 248, lr: 2.38e-03, grad_scale: 8.0 2023-05-02 15:58:13,727 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.6709, 2.0976, 1.8006, 1.8595, 2.3870, 2.0650, 2.0886, 2.4940], device='cuda:1'), covar=tensor([0.0214, 0.0495, 0.0629, 0.0567, 0.0346, 0.0488, 0.0236, 0.0318], device='cuda:1'), in_proj_covar=tensor([0.0219, 0.0237, 0.0228, 0.0228, 0.0237, 0.0237, 0.0230, 0.0234], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-02 15:58:38,001 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=283966.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 15:58:44,144 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.2129, 3.7268, 3.7585, 2.5085, 3.3045, 3.7810, 3.5030, 2.1872], device='cuda:1'), covar=tensor([0.0581, 0.0061, 0.0058, 0.0442, 0.0147, 0.0088, 0.0093, 0.0507], device='cuda:1'), in_proj_covar=tensor([0.0134, 0.0086, 0.0087, 0.0131, 0.0099, 0.0110, 0.0094, 0.0126], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-05-02 15:59:13,294 INFO [optim.py:368] (1/8) 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] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=283987.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 16:00:07,758 INFO [train.py:904] (1/8) Epoch 28, batch 9950, loss[loss=0.1665, simple_loss=0.2687, pruned_loss=0.03215, over 16800.00 frames. ], tot_loss[loss=0.1635, simple_loss=0.2605, pruned_loss=0.0332, over 3036638.07 frames. ], batch size: 124, lr: 2.38e-03, grad_scale: 8.0 2023-05-02 16:01:07,745 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=284027.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 16:02:08,461 INFO [train.py:904] (1/8) Epoch 28, batch 10000, loss[loss=0.154, simple_loss=0.2489, pruned_loss=0.02956, over 16609.00 frames. ], tot_loss[loss=0.1626, simple_loss=0.2595, pruned_loss=0.03284, over 3064731.10 frames. ], batch size: 62, lr: 2.38e-03, grad_scale: 8.0 2023-05-02 16:03:03,885 INFO [optim.py:368] (1/8) 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] (1/8) Epoch 28, batch 10050, loss[loss=0.1638, simple_loss=0.261, pruned_loss=0.03327, over 16742.00 frames. ], tot_loss[loss=0.1627, simple_loss=0.2598, pruned_loss=0.03279, over 3062892.17 frames. ], batch size: 76, lr: 2.38e-03, grad_scale: 8.0 2023-05-02 16:03:54,500 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.5783, 4.5951, 4.9413, 4.9167, 4.9266, 4.6286, 4.6261, 4.5541], device='cuda:1'), covar=tensor([0.0326, 0.0628, 0.0381, 0.0382, 0.0428, 0.0371, 0.0871, 0.0407], device='cuda:1'), in_proj_covar=tensor([0.0419, 0.0469, 0.0455, 0.0419, 0.0502, 0.0479, 0.0549, 0.0386], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:1') 2023-05-02 16:05:24,423 INFO [train.py:904] (1/8) Epoch 28, batch 10100, loss[loss=0.1525, simple_loss=0.2504, pruned_loss=0.02732, over 16395.00 frames. ], tot_loss[loss=0.1626, simple_loss=0.2595, pruned_loss=0.03284, over 3054325.43 frames. ], batch size: 146, lr: 2.38e-03, grad_scale: 8.0 2023-05-02 16:05:52,254 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=4.16 vs. limit=5.0 2023-05-02 16:06:20,722 INFO [optim.py:368] (1/8) 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 16:06:39,382 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=284198.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 16:06:42,038 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=284200.0, num_to_drop=1, layers_to_drop={1} 2023-05-02 16:06:44,464 INFO [train.py:904] (1/8) Epoch 28, batch 10150, loss[loss=0.1501, simple_loss=0.2409, pruned_loss=0.0296, over 12144.00 frames. ], tot_loss[loss=0.1625, simple_loss=0.2587, pruned_loss=0.03318, over 3042267.04 frames. ], batch size: 246, lr: 2.38e-03, grad_scale: 8.0 2023-05-02 16:07:10,315 INFO [train.py:904] (1/8) Epoch 29, batch 0, loss[loss=0.1868, simple_loss=0.2735, pruned_loss=0.05009, over 17196.00 frames. ], tot_loss[loss=0.1868, simple_loss=0.2735, pruned_loss=0.05009, over 17196.00 frames. ], batch size: 46, lr: 2.34e-03, grad_scale: 8.0 2023-05-02 16:07:10,315 INFO [train.py:929] (1/8) Computing validation loss 2023-05-02 16:07:17,744 INFO [train.py:938] (1/8) Epoch 29, validation: loss=0.1427, simple_loss=0.246, pruned_loss=0.0197, over 944034.00 frames. 2023-05-02 16:07:17,745 INFO [train.py:939] (1/8) Maximum memory allocated so far is 18145MB 2023-05-02 16:08:18,231 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=284248.0, num_to_drop=1, layers_to_drop={1} 2023-05-02 16:08:26,888 INFO [train.py:904] (1/8) Epoch 29, batch 50, loss[loss=0.1603, simple_loss=0.2412, pruned_loss=0.0397, over 16802.00 frames. ], tot_loss[loss=0.1764, simple_loss=0.2651, pruned_loss=0.04388, over 750716.14 frames. ], batch size: 96, lr: 2.34e-03, grad_scale: 2.0 2023-05-02 16:09:08,278 INFO [optim.py:368] (1/8) 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] (1/8) Epoch 29, batch 100, loss[loss=0.1759, simple_loss=0.2569, pruned_loss=0.04744, over 16690.00 frames. ], tot_loss[loss=0.1745, simple_loss=0.2608, pruned_loss=0.0441, over 1308274.66 frames. ], batch size: 134, lr: 2.34e-03, grad_scale: 2.0 2023-05-02 16:09:43,746 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.5928, 3.6115, 2.3505, 3.8112, 2.8872, 3.7422, 2.3522, 2.9843], device='cuda:1'), covar=tensor([0.0319, 0.0503, 0.1564, 0.0433, 0.0811, 0.0932, 0.1542, 0.0776], device='cuda:1'), in_proj_covar=tensor([0.0173, 0.0178, 0.0195, 0.0169, 0.0178, 0.0215, 0.0203, 0.0181], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-05-02 16:09:57,857 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.86 vs. limit=2.0 2023-05-02 16:10:02,064 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=284322.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 16:10:46,097 INFO [train.py:904] (1/8) Epoch 29, batch 150, loss[loss=0.1756, simple_loss=0.2625, pruned_loss=0.04434, over 16502.00 frames. ], tot_loss[loss=0.1714, simple_loss=0.2579, pruned_loss=0.0425, over 1758912.23 frames. ], batch size: 75, lr: 2.33e-03, grad_scale: 2.0 2023-05-02 16:11:25,623 INFO [optim.py:368] (1/8) 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 16:11:55,140 INFO [train.py:904] (1/8) Epoch 29, batch 200, loss[loss=0.1776, simple_loss=0.2715, pruned_loss=0.04181, over 16733.00 frames. ], tot_loss[loss=0.1727, simple_loss=0.2591, pruned_loss=0.04315, over 2105895.65 frames. ], batch size: 57, lr: 2.33e-03, grad_scale: 2.0 2023-05-02 16:12:31,477 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.7448, 4.6190, 4.6664, 4.3543, 4.4216, 4.6999, 4.5024, 4.4838], device='cuda:1'), covar=tensor([0.0842, 0.1184, 0.0419, 0.0417, 0.0917, 0.0813, 0.0489, 0.0733], device='cuda:1'), in_proj_covar=tensor([0.0305, 0.0458, 0.0356, 0.0356, 0.0350, 0.0410, 0.0245, 0.0424], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-02 16:12:51,509 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=284443.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 16:13:03,531 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.57 vs. limit=2.0 2023-05-02 16:13:04,662 INFO [train.py:904] (1/8) Epoch 29, batch 250, loss[loss=0.1415, simple_loss=0.2428, pruned_loss=0.02016, over 17253.00 frames. ], tot_loss[loss=0.1716, simple_loss=0.2578, pruned_loss=0.04273, over 2375389.16 frames. ], batch size: 52, lr: 2.33e-03, grad_scale: 2.0 2023-05-02 16:13:36,250 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.32 vs. limit=2.0 2023-05-02 16:13:46,442 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.48 vs. limit=2.0 2023-05-02 16:13:47,576 INFO [optim.py:368] (1/8) 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 16:13:53,946 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.1547, 2.2290, 2.3548, 3.9119, 2.2373, 2.5488, 2.2918, 2.3843], device='cuda:1'), covar=tensor([0.1653, 0.3809, 0.3366, 0.0737, 0.4213, 0.2776, 0.4256, 0.3377], device='cuda:1'), in_proj_covar=tensor([0.0415, 0.0469, 0.0384, 0.0331, 0.0443, 0.0535, 0.0442, 0.0547], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-02 16:14:08,407 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 29, batch 300, loss[loss=0.1759, simple_loss=0.2483, pruned_loss=0.05171, over 16714.00 frames. ], tot_loss[loss=0.1683, simple_loss=0.2549, pruned_loss=0.04083, over 2586778.55 frames. ], batch size: 124, lr: 2.33e-03, grad_scale: 2.0 2023-05-02 16:14:17,363 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=284504.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 16:14:50,764 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([5.4846, 5.8994, 5.6368, 5.6629, 5.2582, 5.4133, 5.2799, 5.9949], device='cuda:1'), covar=tensor([0.1504, 0.1113, 0.1123, 0.0966, 0.0947, 0.0732, 0.1368, 0.0941], device='cuda:1'), in_proj_covar=tensor([0.0702, 0.0851, 0.0697, 0.0658, 0.0538, 0.0537, 0.0713, 0.0665], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:1') 2023-05-02 16:14:50,944 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([1.8391, 2.6816, 2.7604, 4.9656, 3.9040, 4.2959, 1.5998, 3.1157], device='cuda:1'), covar=tensor([0.1467, 0.0932, 0.1235, 0.0224, 0.0251, 0.0434, 0.1800, 0.0870], device='cuda:1'), in_proj_covar=tensor([0.0173, 0.0179, 0.0199, 0.0200, 0.0202, 0.0216, 0.0210, 0.0198], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-05-02 16:15:14,272 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=284546.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 16:15:23,224 INFO [train.py:904] (1/8) Epoch 29, batch 350, loss[loss=0.1697, simple_loss=0.2458, pruned_loss=0.04673, over 16452.00 frames. ], tot_loss[loss=0.1655, simple_loss=0.2529, pruned_loss=0.03903, over 2760091.37 frames. ], batch size: 146, lr: 2.33e-03, grad_scale: 2.0 2023-05-02 16:16:02,935 INFO [optim.py:368] (1/8) 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] (1/8) Epoch 29, batch 400, loss[loss=0.1743, simple_loss=0.2546, pruned_loss=0.047, over 16208.00 frames. ], tot_loss[loss=0.1645, simple_loss=0.2511, pruned_loss=0.03897, over 2885306.25 frames. ], batch size: 165, lr: 2.33e-03, grad_scale: 4.0 2023-05-02 16:16:49,957 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.7947, 5.0124, 5.1606, 4.9518, 5.0170, 5.5853, 5.0462, 4.7461], device='cuda:1'), covar=tensor([0.1341, 0.1978, 0.2480, 0.2218, 0.2418, 0.1053, 0.1851, 0.2499], device='cuda:1'), in_proj_covar=tensor([0.0420, 0.0628, 0.0698, 0.0513, 0.0685, 0.0720, 0.0541, 0.0683], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-05-02 16:16:57,001 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=284622.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 16:17:26,725 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.8446, 5.1508, 5.2755, 5.0466, 5.1078, 5.7143, 5.1805, 4.8664], device='cuda:1'), covar=tensor([0.1234, 0.2038, 0.2411, 0.2124, 0.2466, 0.1024, 0.1791, 0.2347], device='cuda:1'), in_proj_covar=tensor([0.0422, 0.0630, 0.0700, 0.0515, 0.0688, 0.0722, 0.0543, 0.0685], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-05-02 16:17:41,207 INFO [train.py:904] (1/8) Epoch 29, batch 450, loss[loss=0.1758, simple_loss=0.2604, pruned_loss=0.04558, over 16748.00 frames. ], tot_loss[loss=0.1632, simple_loss=0.2501, pruned_loss=0.03814, over 2981983.43 frames. ], batch size: 57, lr: 2.33e-03, grad_scale: 4.0 2023-05-02 16:17:45,981 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([5.0271, 5.0761, 5.4985, 5.4875, 5.4962, 5.1312, 5.0892, 4.9350], device='cuda:1'), covar=tensor([0.0394, 0.0603, 0.0431, 0.0421, 0.0547, 0.0480, 0.1039, 0.0467], device='cuda:1'), in_proj_covar=tensor([0.0435, 0.0489, 0.0471, 0.0435, 0.0521, 0.0497, 0.0569, 0.0400], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:1') 2023-05-02 16:18:02,994 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=284670.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 16:18:07,816 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.8994, 4.4131, 4.4506, 3.2640, 3.6561, 4.4001, 3.9091, 2.6314], device='cuda:1'), covar=tensor([0.0480, 0.0078, 0.0048, 0.0349, 0.0165, 0.0096, 0.0106, 0.0475], device='cuda:1'), in_proj_covar=tensor([0.0137, 0.0089, 0.0089, 0.0134, 0.0101, 0.0113, 0.0097, 0.0129], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:1') 2023-05-02 16:18:11,078 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.26 vs. limit=2.0 2023-05-02 16:18:18,521 INFO [optim.py:368] (1/8) Clipping_scale=2.0, grad-norm quartiles 1.454e+02 1.978e+02 2.312e+02 2.858e+02 4.933e+02, threshold=4.623e+02, percent-clipped=2.0 2023-05-02 16:18:47,401 INFO [train.py:904] (1/8) Epoch 29, batch 500, loss[loss=0.1833, simple_loss=0.2683, pruned_loss=0.04915, over 17002.00 frames. ], tot_loss[loss=0.1624, simple_loss=0.2492, pruned_loss=0.03782, over 3059682.95 frames. ], batch size: 55, lr: 2.33e-03, grad_scale: 4.0 2023-05-02 16:19:30,549 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.77 vs. limit=2.0 2023-05-02 16:19:56,204 INFO [train.py:904] (1/8) Epoch 29, batch 550, loss[loss=0.1771, simple_loss=0.2563, pruned_loss=0.04895, over 16807.00 frames. ], tot_loss[loss=0.162, simple_loss=0.2487, pruned_loss=0.03767, over 3119226.14 frames. ], batch size: 102, lr: 2.33e-03, grad_scale: 4.0 2023-05-02 16:20:17,687 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.82 vs. limit=2.0 2023-05-02 16:20:35,714 INFO [optim.py:368] (1/8) Clipping_scale=2.0, grad-norm quartiles 1.399e+02 1.886e+02 2.233e+02 2.634e+02 5.354e+02, threshold=4.465e+02, percent-clipped=1.0 2023-05-02 16:20:57,908 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=284799.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 16:21:04,196 INFO [train.py:904] (1/8) Epoch 29, batch 600, loss[loss=0.1503, simple_loss=0.2453, pruned_loss=0.02765, over 17055.00 frames. ], tot_loss[loss=0.1618, simple_loss=0.2478, pruned_loss=0.03789, over 3167452.73 frames. ], batch size: 50, lr: 2.33e-03, grad_scale: 4.0 2023-05-02 16:21:22,012 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=284816.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 16:21:28,924 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.21 vs. limit=2.0 2023-05-02 16:21:57,627 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.2108, 4.1738, 4.1157, 3.5305, 4.1421, 1.7818, 3.9310, 3.6412], device='cuda:1'), covar=tensor([0.0195, 0.0142, 0.0243, 0.0327, 0.0131, 0.3023, 0.0166, 0.0319], device='cuda:1'), in_proj_covar=tensor([0.0180, 0.0173, 0.0211, 0.0181, 0.0187, 0.0217, 0.0199, 0.0178], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-02 16:22:12,347 INFO [train.py:904] (1/8) Epoch 29, batch 650, loss[loss=0.159, simple_loss=0.2529, pruned_loss=0.0325, over 17237.00 frames. ], tot_loss[loss=0.1605, simple_loss=0.2467, pruned_loss=0.03717, over 3204142.18 frames. ], batch size: 45, lr: 2.33e-03, grad_scale: 4.0 2023-05-02 16:22:13,623 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.2037, 3.8555, 4.3348, 2.1844, 4.4804, 4.6532, 3.3924, 3.6888], device='cuda:1'), covar=tensor([0.0692, 0.0295, 0.0253, 0.1204, 0.0104, 0.0169, 0.0473, 0.0401], device='cuda:1'), in_proj_covar=tensor([0.0148, 0.0110, 0.0101, 0.0138, 0.0086, 0.0130, 0.0129, 0.0130], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-05-02 16:22:46,689 INFO [zipformer.py:625] (1/8) 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] (1/8) 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] (1/8) Epoch 29, batch 700, loss[loss=0.1393, simple_loss=0.2276, pruned_loss=0.02547, over 16813.00 frames. ], tot_loss[loss=0.1601, simple_loss=0.2462, pruned_loss=0.037, over 3217692.70 frames. ], batch size: 42, lr: 2.33e-03, grad_scale: 4.0 2023-05-02 16:23:31,185 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.8052, 3.5889, 4.0349, 2.0483, 4.1708, 4.1999, 3.1436, 3.2493], device='cuda:1'), covar=tensor([0.0833, 0.0305, 0.0263, 0.1286, 0.0114, 0.0229, 0.0497, 0.0489], device='cuda:1'), in_proj_covar=tensor([0.0148, 0.0111, 0.0102, 0.0139, 0.0086, 0.0131, 0.0130, 0.0130], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-05-02 16:24:12,248 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=284939.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 16:24:30,115 INFO [train.py:904] (1/8) Epoch 29, batch 750, loss[loss=0.1471, simple_loss=0.2372, pruned_loss=0.02848, over 17201.00 frames. ], tot_loss[loss=0.1606, simple_loss=0.2469, pruned_loss=0.03718, over 3248342.78 frames. ], batch size: 44, lr: 2.33e-03, grad_scale: 4.0 2023-05-02 16:24:37,708 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.0640, 4.0147, 3.9749, 3.3254, 3.9645, 1.7958, 3.7631, 3.4712], device='cuda:1'), covar=tensor([0.0197, 0.0181, 0.0229, 0.0306, 0.0124, 0.3087, 0.0169, 0.0324], device='cuda:1'), in_proj_covar=tensor([0.0182, 0.0175, 0.0214, 0.0184, 0.0189, 0.0219, 0.0201, 0.0180], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-02 16:24:49,551 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.9334, 4.6816, 4.9534, 5.1379, 5.3222, 4.7241, 5.3067, 5.3202], device='cuda:1'), covar=tensor([0.1930, 0.1340, 0.1843, 0.0762, 0.0594, 0.1083, 0.0673, 0.0694], device='cuda:1'), in_proj_covar=tensor([0.0681, 0.0830, 0.0959, 0.0845, 0.0642, 0.0667, 0.0711, 0.0820], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-02 16:25:13,140 INFO [optim.py:368] (1/8) 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 16:25:30,053 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=284996.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 16:25:37,939 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=285000.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 16:25:41,613 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.89 vs. limit=2.0 2023-05-02 16:25:42,931 INFO [train.py:904] (1/8) Epoch 29, batch 800, loss[loss=0.1653, simple_loss=0.246, pruned_loss=0.04232, over 16470.00 frames. ], tot_loss[loss=0.1608, simple_loss=0.2472, pruned_loss=0.03714, over 3260307.62 frames. ], batch size: 146, lr: 2.33e-03, grad_scale: 8.0 2023-05-02 16:26:31,545 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.15 vs. limit=2.0 2023-05-02 16:26:33,616 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=285042.0, num_to_drop=1, layers_to_drop={0} 2023-05-02 16:26:43,899 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.1280, 2.0768, 2.2002, 3.7746, 2.1329, 2.3581, 2.2182, 2.2623], device='cuda:1'), covar=tensor([0.1820, 0.4730, 0.3865, 0.0833, 0.5384, 0.3678, 0.4388, 0.4750], device='cuda:1'), in_proj_covar=tensor([0.0421, 0.0474, 0.0389, 0.0336, 0.0447, 0.0542, 0.0448, 0.0556], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-02 16:26:49,075 INFO [train.py:904] (1/8) Epoch 29, batch 850, loss[loss=0.1549, simple_loss=0.2418, pruned_loss=0.03397, over 16626.00 frames. ], tot_loss[loss=0.1601, simple_loss=0.2466, pruned_loss=0.03684, over 3269946.11 frames. ], batch size: 68, lr: 2.33e-03, grad_scale: 4.0 2023-05-02 16:26:52,994 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=285057.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 16:27:31,497 INFO [optim.py:368] (1/8) Clipping_scale=2.0, grad-norm quartiles 1.555e+02 2.142e+02 2.515e+02 3.015e+02 5.883e+02, threshold=5.030e+02, percent-clipped=4.0 2023-05-02 16:27:49,509 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=285099.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 16:27:54,223 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=285103.0, num_to_drop=1, layers_to_drop={2} 2023-05-02 16:27:55,639 INFO [train.py:904] (1/8) Epoch 29, batch 900, loss[loss=0.174, simple_loss=0.2636, pruned_loss=0.0422, over 16675.00 frames. ], tot_loss[loss=0.1592, simple_loss=0.2455, pruned_loss=0.03644, over 3280916.16 frames. ], batch size: 57, lr: 2.33e-03, grad_scale: 2.0 2023-05-02 16:28:55,569 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=285147.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 16:29:05,607 INFO [train.py:904] (1/8) Epoch 29, batch 950, loss[loss=0.1606, simple_loss=0.2483, pruned_loss=0.03642, over 16505.00 frames. ], tot_loss[loss=0.1592, simple_loss=0.2454, pruned_loss=0.03652, over 3286098.87 frames. ], batch size: 68, lr: 2.33e-03, grad_scale: 2.0 2023-05-02 16:29:30,490 INFO [zipformer.py:625] (1/8) 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] (1/8) Clipping_scale=2.0, grad-norm quartiles 1.394e+02 2.122e+02 2.524e+02 3.029e+02 7.170e+02, threshold=5.047e+02, percent-clipped=3.0 2023-05-02 16:30:14,194 INFO [train.py:904] (1/8) Epoch 29, batch 1000, loss[loss=0.1709, simple_loss=0.2621, pruned_loss=0.03984, over 16711.00 frames. ], tot_loss[loss=0.1586, simple_loss=0.2446, pruned_loss=0.03635, over 3289744.13 frames. ], batch size: 57, lr: 2.33e-03, grad_scale: 2.0 2023-05-02 16:31:02,856 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.3207, 3.4790, 3.9596, 2.0895, 3.1533, 2.4524, 3.7261, 3.7016], device='cuda:1'), covar=tensor([0.0301, 0.1005, 0.0495, 0.2200, 0.0877, 0.1021, 0.0610, 0.1095], device='cuda:1'), in_proj_covar=tensor([0.0161, 0.0171, 0.0170, 0.0158, 0.0148, 0.0133, 0.0146, 0.0183], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0005], device='cuda:1') 2023-05-02 16:31:21,163 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.4872, 3.6666, 4.0362, 2.2253, 3.1536, 2.5746, 3.8451, 3.8792], device='cuda:1'), covar=tensor([0.0285, 0.1047, 0.0484, 0.2194, 0.0929, 0.1046, 0.0643, 0.1149], device='cuda:1'), in_proj_covar=tensor([0.0161, 0.0171, 0.0170, 0.0158, 0.0148, 0.0133, 0.0146, 0.0183], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0005], device='cuda:1') 2023-05-02 16:31:24,219 INFO [train.py:904] (1/8) Epoch 29, batch 1050, loss[loss=0.1457, simple_loss=0.2419, pruned_loss=0.02474, over 17100.00 frames. ], tot_loss[loss=0.1582, simple_loss=0.2442, pruned_loss=0.03613, over 3302582.76 frames. ], batch size: 49, lr: 2.33e-03, grad_scale: 2.0 2023-05-02 16:32:05,303 INFO [optim.py:368] (1/8) 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 16:32:19,760 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=285295.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 16:32:31,248 INFO [train.py:904] (1/8) Epoch 29, batch 1100, loss[loss=0.1444, simple_loss=0.235, pruned_loss=0.02687, over 17136.00 frames. ], tot_loss[loss=0.1575, simple_loss=0.2433, pruned_loss=0.03586, over 3309930.53 frames. ], batch size: 49, lr: 2.33e-03, grad_scale: 2.0 2023-05-02 16:33:37,331 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=285352.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 16:33:37,773 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.09 vs. limit=2.0 2023-05-02 16:33:40,187 INFO [train.py:904] (1/8) Epoch 29, batch 1150, loss[loss=0.1667, simple_loss=0.2566, pruned_loss=0.03838, over 16721.00 frames. ], tot_loss[loss=0.1573, simple_loss=0.2428, pruned_loss=0.03591, over 3295424.93 frames. ], batch size: 57, lr: 2.33e-03, grad_scale: 2.0 2023-05-02 16:34:00,646 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([5.4864, 5.8736, 5.6216, 5.6982, 5.3140, 5.4253, 5.2626, 6.0239], device='cuda:1'), covar=tensor([0.1492, 0.1030, 0.1101, 0.0926, 0.0952, 0.0774, 0.1389, 0.0896], device='cuda:1'), in_proj_covar=tensor([0.0725, 0.0877, 0.0717, 0.0681, 0.0555, 0.0553, 0.0738, 0.0687], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:1') 2023-05-02 16:34:07,007 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.3018, 2.3628, 2.4457, 4.1405, 2.2980, 2.7251, 2.4405, 2.5221], device='cuda:1'), covar=tensor([0.1498, 0.3842, 0.3272, 0.0622, 0.4292, 0.2839, 0.3751, 0.3665], device='cuda:1'), in_proj_covar=tensor([0.0423, 0.0477, 0.0391, 0.0338, 0.0449, 0.0546, 0.0450, 0.0560], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-02 16:34:22,232 INFO [optim.py:368] (1/8) 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 16:34:40,264 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 29, batch 1200, loss[loss=0.153, simple_loss=0.2371, pruned_loss=0.03446, over 16904.00 frames. ], tot_loss[loss=0.1567, simple_loss=0.2427, pruned_loss=0.03538, over 3306458.18 frames. ], batch size: 109, lr: 2.33e-03, grad_scale: 4.0 2023-05-02 16:35:22,675 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.14 vs. limit=2.0 2023-05-02 16:35:56,660 INFO [train.py:904] (1/8) Epoch 29, batch 1250, loss[loss=0.1431, simple_loss=0.2287, pruned_loss=0.0288, over 15903.00 frames. ], tot_loss[loss=0.1568, simple_loss=0.2428, pruned_loss=0.03543, over 3315408.12 frames. ], batch size: 35, lr: 2.33e-03, grad_scale: 4.0 2023-05-02 16:36:21,847 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=285472.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 16:36:38,697 INFO [optim.py:368] (1/8) Clipping_scale=2.0, grad-norm quartiles 1.468e+02 2.107e+02 2.455e+02 3.038e+02 4.730e+02, threshold=4.911e+02, percent-clipped=0.0 2023-05-02 16:37:05,020 INFO [train.py:904] (1/8) Epoch 29, batch 1300, loss[loss=0.1431, simple_loss=0.2237, pruned_loss=0.03128, over 15963.00 frames. ], tot_loss[loss=0.1571, simple_loss=0.2429, pruned_loss=0.03564, over 3320208.20 frames. ], batch size: 35, lr: 2.33e-03, grad_scale: 4.0 2023-05-02 16:37:17,101 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.8835, 3.6603, 3.9973, 2.2341, 4.1556, 4.1489, 3.3466, 3.2176], device='cuda:1'), covar=tensor([0.0742, 0.0290, 0.0230, 0.1172, 0.0118, 0.0230, 0.0395, 0.0448], device='cuda:1'), in_proj_covar=tensor([0.0150, 0.0112, 0.0103, 0.0141, 0.0087, 0.0133, 0.0131, 0.0132], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-05-02 16:37:27,841 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=285520.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 16:38:13,887 INFO [train.py:904] (1/8) Epoch 29, batch 1350, loss[loss=0.1728, simple_loss=0.2681, pruned_loss=0.03881, over 16734.00 frames. ], tot_loss[loss=0.1576, simple_loss=0.2436, pruned_loss=0.03579, over 3326031.09 frames. ], batch size: 57, lr: 2.33e-03, grad_scale: 4.0 2023-05-02 16:38:30,918 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=4.66 vs. limit=5.0 2023-05-02 16:38:58,364 INFO [optim.py:368] (1/8) 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 16:39:08,994 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.4116, 4.1123, 4.5378, 2.6428, 4.7298, 4.8185, 3.5630, 3.8621], device='cuda:1'), covar=tensor([0.0631, 0.0264, 0.0221, 0.1035, 0.0084, 0.0167, 0.0435, 0.0379], device='cuda:1'), in_proj_covar=tensor([0.0150, 0.0112, 0.0104, 0.0141, 0.0088, 0.0133, 0.0131, 0.0132], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-05-02 16:39:11,790 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=285595.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 16:39:24,823 INFO [train.py:904] (1/8) Epoch 29, batch 1400, loss[loss=0.1323, simple_loss=0.2189, pruned_loss=0.02283, over 16974.00 frames. ], tot_loss[loss=0.1578, simple_loss=0.2441, pruned_loss=0.03576, over 3324889.73 frames. ], batch size: 41, lr: 2.33e-03, grad_scale: 4.0 2023-05-02 16:39:49,651 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=285622.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 16:40:19,786 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=285643.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 16:40:30,993 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=285652.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 16:40:33,553 INFO [train.py:904] (1/8) Epoch 29, batch 1450, loss[loss=0.1494, simple_loss=0.2478, pruned_loss=0.02551, over 17125.00 frames. ], tot_loss[loss=0.1573, simple_loss=0.2434, pruned_loss=0.03561, over 3324381.54 frames. ], batch size: 47, lr: 2.33e-03, grad_scale: 4.0 2023-05-02 16:41:13,916 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=285683.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 16:41:16,389 INFO [optim.py:368] (1/8) 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 16:41:34,384 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=285698.0, num_to_drop=1, layers_to_drop={2} 2023-05-02 16:41:37,510 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=285700.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 16:41:43,180 INFO [train.py:904] (1/8) Epoch 29, batch 1500, loss[loss=0.1353, simple_loss=0.2164, pruned_loss=0.02712, over 16875.00 frames. ], tot_loss[loss=0.1578, simple_loss=0.2436, pruned_loss=0.03605, over 3319361.92 frames. ], batch size: 96, lr: 2.33e-03, grad_scale: 4.0 2023-05-02 16:42:40,635 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=285746.0, num_to_drop=1, layers_to_drop={1} 2023-05-02 16:42:51,736 INFO [train.py:904] (1/8) Epoch 29, batch 1550, loss[loss=0.1865, simple_loss=0.2642, pruned_loss=0.05442, over 16855.00 frames. ], tot_loss[loss=0.1588, simple_loss=0.2442, pruned_loss=0.03673, over 3328283.82 frames. ], batch size: 109, lr: 2.33e-03, grad_scale: 4.0 2023-05-02 16:42:57,374 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.7555, 3.5897, 3.9534, 2.1960, 4.0408, 4.1098, 3.3599, 2.9760], device='cuda:1'), covar=tensor([0.0852, 0.0275, 0.0222, 0.1181, 0.0117, 0.0193, 0.0396, 0.0534], device='cuda:1'), in_proj_covar=tensor([0.0150, 0.0112, 0.0104, 0.0141, 0.0088, 0.0133, 0.0131, 0.0132], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-05-02 16:43:34,073 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.7552, 4.3499, 2.9318, 2.2971, 2.5201, 2.6034, 4.7376, 3.4081], device='cuda:1'), covar=tensor([0.3112, 0.0527, 0.1969, 0.3002, 0.3003, 0.2218, 0.0332, 0.1588], device='cuda:1'), in_proj_covar=tensor([0.0340, 0.0276, 0.0316, 0.0329, 0.0307, 0.0281, 0.0307, 0.0356], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:1') 2023-05-02 16:43:34,663 INFO [optim.py:368] (1/8) 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] (1/8) Epoch 29, batch 1600, loss[loss=0.214, simple_loss=0.2939, pruned_loss=0.06707, over 11933.00 frames. ], tot_loss[loss=0.1615, simple_loss=0.2469, pruned_loss=0.038, over 3308898.63 frames. ], batch size: 246, lr: 2.33e-03, grad_scale: 8.0 2023-05-02 16:44:14,756 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.7548, 4.8470, 5.2046, 5.1991, 5.2099, 4.8661, 4.8602, 4.6818], device='cuda:1'), covar=tensor([0.0364, 0.0597, 0.0388, 0.0393, 0.0566, 0.0491, 0.1016, 0.0506], device='cuda:1'), in_proj_covar=tensor([0.0446, 0.0503, 0.0486, 0.0447, 0.0533, 0.0509, 0.0585, 0.0409], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:1') 2023-05-02 16:44:22,270 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.23 vs. limit=2.0 2023-05-02 16:44:54,196 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([5.1426, 5.1755, 5.0159, 4.4974, 4.4632, 5.1118, 5.0488, 4.6061], device='cuda:1'), covar=tensor([0.0735, 0.0697, 0.0448, 0.0513, 0.1438, 0.0606, 0.0344, 0.0954], device='cuda:1'), in_proj_covar=tensor([0.0321, 0.0484, 0.0377, 0.0379, 0.0374, 0.0435, 0.0259, 0.0450], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-05-02 16:45:09,517 INFO [train.py:904] (1/8) Epoch 29, batch 1650, loss[loss=0.1749, simple_loss=0.2588, pruned_loss=0.04551, over 16364.00 frames. ], tot_loss[loss=0.1629, simple_loss=0.2484, pruned_loss=0.03866, over 3310471.24 frames. ], batch size: 68, lr: 2.33e-03, grad_scale: 8.0 2023-05-02 16:45:31,433 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.4443, 4.4130, 4.3707, 4.0468, 4.1100, 4.4337, 4.1988, 4.1635], device='cuda:1'), covar=tensor([0.0706, 0.0761, 0.0348, 0.0391, 0.0873, 0.0547, 0.0713, 0.0679], device='cuda:1'), in_proj_covar=tensor([0.0321, 0.0484, 0.0377, 0.0379, 0.0374, 0.0436, 0.0259, 0.0451], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-05-02 16:45:50,989 INFO [optim.py:368] (1/8) 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] (1/8) Epoch 29, batch 1700, loss[loss=0.1581, simple_loss=0.2588, pruned_loss=0.02869, over 17238.00 frames. ], tot_loss[loss=0.165, simple_loss=0.2507, pruned_loss=0.03965, over 3310588.04 frames. ], batch size: 52, lr: 2.33e-03, grad_scale: 8.0 2023-05-02 16:46:25,011 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.1224, 2.3108, 2.8481, 3.1009, 2.9528, 3.5917, 2.6972, 3.6149], device='cuda:1'), covar=tensor([0.0316, 0.0580, 0.0366, 0.0399, 0.0419, 0.0228, 0.0502, 0.0189], device='cuda:1'), in_proj_covar=tensor([0.0201, 0.0202, 0.0189, 0.0195, 0.0212, 0.0169, 0.0205, 0.0168], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0002], device='cuda:1') 2023-05-02 16:46:32,872 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.5486, 3.4723, 4.1104, 2.2688, 3.2348, 2.6326, 3.9315, 3.7202], device='cuda:1'), covar=tensor([0.0287, 0.1109, 0.0484, 0.2238, 0.0873, 0.1073, 0.0600, 0.1218], device='cuda:1'), in_proj_covar=tensor([0.0161, 0.0171, 0.0171, 0.0158, 0.0148, 0.0133, 0.0146, 0.0184], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0005], device='cuda:1') 2023-05-02 16:47:24,308 INFO [train.py:904] (1/8) Epoch 29, batch 1750, loss[loss=0.1687, simple_loss=0.2494, pruned_loss=0.04399, over 16431.00 frames. ], tot_loss[loss=0.1641, simple_loss=0.2506, pruned_loss=0.03877, over 3320224.08 frames. ], batch size: 75, lr: 2.33e-03, grad_scale: 8.0 2023-05-02 16:47:58,301 INFO [zipformer.py:625] (1/8) 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] (1/8) 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 16:48:36,744 INFO [train.py:904] (1/8) Epoch 29, batch 1800, loss[loss=0.1655, simple_loss=0.2496, pruned_loss=0.04073, over 16780.00 frames. ], tot_loss[loss=0.1638, simple_loss=0.2509, pruned_loss=0.03832, over 3321432.86 frames. ], batch size: 102, lr: 2.33e-03, grad_scale: 8.0 2023-05-02 16:49:43,956 INFO [train.py:904] (1/8) Epoch 29, batch 1850, loss[loss=0.1499, simple_loss=0.2389, pruned_loss=0.03043, over 16825.00 frames. ], tot_loss[loss=0.1643, simple_loss=0.2521, pruned_loss=0.03828, over 3326866.43 frames. ], batch size: 42, lr: 2.33e-03, grad_scale: 4.0 2023-05-02 16:50:17,266 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.4411, 4.4012, 4.3204, 3.7480, 4.3827, 1.7425, 4.1191, 3.8109], device='cuda:1'), covar=tensor([0.0178, 0.0173, 0.0214, 0.0341, 0.0119, 0.3165, 0.0161, 0.0295], device='cuda:1'), in_proj_covar=tensor([0.0185, 0.0178, 0.0216, 0.0187, 0.0193, 0.0221, 0.0204, 0.0183], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-02 16:50:28,094 INFO [optim.py:368] (1/8) 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 16:50:53,100 INFO [train.py:904] (1/8) Epoch 29, batch 1900, loss[loss=0.1458, simple_loss=0.2404, pruned_loss=0.02561, over 17143.00 frames. ], tot_loss[loss=0.1636, simple_loss=0.2513, pruned_loss=0.03793, over 3323822.26 frames. ], batch size: 47, lr: 2.33e-03, grad_scale: 4.0 2023-05-02 16:51:20,253 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=286123.0, num_to_drop=1, layers_to_drop={1} 2023-05-02 16:51:29,250 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=286129.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 16:52:01,736 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.2884, 3.3093, 2.2305, 3.4571, 2.6727, 3.4313, 2.2983, 2.7996], device='cuda:1'), covar=tensor([0.0311, 0.0499, 0.1505, 0.0362, 0.0770, 0.0911, 0.1413, 0.0743], device='cuda:1'), in_proj_covar=tensor([0.0179, 0.0184, 0.0200, 0.0178, 0.0183, 0.0225, 0.0209, 0.0186], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-05-02 16:52:04,245 INFO [train.py:904] (1/8) Epoch 29, batch 1950, loss[loss=0.1723, simple_loss=0.2559, pruned_loss=0.04437, over 16697.00 frames. ], tot_loss[loss=0.1632, simple_loss=0.2514, pruned_loss=0.03749, over 3319146.54 frames. ], batch size: 89, lr: 2.33e-03, grad_scale: 4.0 2023-05-02 16:52:25,399 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.5811, 3.9962, 4.1184, 2.8618, 3.5774, 4.0732, 3.6836, 2.4732], device='cuda:1'), covar=tensor([0.0536, 0.0285, 0.0068, 0.0422, 0.0167, 0.0141, 0.0136, 0.0496], device='cuda:1'), in_proj_covar=tensor([0.0139, 0.0091, 0.0092, 0.0137, 0.0104, 0.0117, 0.0099, 0.0132], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0003, 0.0004], device='cuda:1') 2023-05-02 16:52:46,895 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=286184.0, num_to_drop=1, layers_to_drop={1} 2023-05-02 16:52:48,710 INFO [optim.py:368] (1/8) Clipping_scale=2.0, grad-norm quartiles 1.778e+02 2.194e+02 2.537e+02 3.082e+02 2.053e+03, threshold=5.073e+02, percent-clipped=1.0 2023-05-02 16:52:55,329 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=286190.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 16:53:13,092 INFO [train.py:904] (1/8) Epoch 29, batch 2000, loss[loss=0.161, simple_loss=0.2364, pruned_loss=0.04274, over 16887.00 frames. ], tot_loss[loss=0.1619, simple_loss=0.2503, pruned_loss=0.03679, over 3322573.64 frames. ], batch size: 96, lr: 2.33e-03, grad_scale: 8.0 2023-05-02 16:54:21,795 INFO [train.py:904] (1/8) Epoch 29, batch 2050, loss[loss=0.1862, simple_loss=0.2786, pruned_loss=0.04689, over 15473.00 frames. ], tot_loss[loss=0.1631, simple_loss=0.2513, pruned_loss=0.03744, over 3320509.90 frames. ], batch size: 190, lr: 2.33e-03, grad_scale: 8.0 2023-05-02 16:54:23,373 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([5.5498, 5.4786, 5.4276, 4.9534, 5.1223, 5.4545, 5.4402, 5.1332], device='cuda:1'), covar=tensor([0.0560, 0.0526, 0.0301, 0.0344, 0.0937, 0.0467, 0.0261, 0.0673], device='cuda:1'), in_proj_covar=tensor([0.0323, 0.0488, 0.0379, 0.0381, 0.0376, 0.0438, 0.0260, 0.0454], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-05-02 16:54:54,932 INFO [zipformer.py:625] (1/8) 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] (1/8) 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 16:55:14,414 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.7422, 3.8500, 2.4998, 4.4004, 3.0066, 4.3128, 2.6858, 3.1874], device='cuda:1'), covar=tensor([0.0333, 0.0416, 0.1688, 0.0358, 0.0871, 0.0600, 0.1472, 0.0803], device='cuda:1'), in_proj_covar=tensor([0.0179, 0.0185, 0.0201, 0.0179, 0.0183, 0.0225, 0.0209, 0.0186], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-05-02 16:55:18,549 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=286295.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 16:55:30,407 INFO [train.py:904] (1/8) Epoch 29, batch 2100, loss[loss=0.1851, simple_loss=0.2763, pruned_loss=0.04698, over 16663.00 frames. ], tot_loss[loss=0.1644, simple_loss=0.2525, pruned_loss=0.03815, over 3307489.67 frames. ], batch size: 62, lr: 2.33e-03, grad_scale: 8.0 2023-05-02 16:56:00,344 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=286326.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 16:56:26,275 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=286344.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 16:56:40,463 INFO [train.py:904] (1/8) Epoch 29, batch 2150, loss[loss=0.1599, simple_loss=0.2413, pruned_loss=0.03923, over 16699.00 frames. ], tot_loss[loss=0.1646, simple_loss=0.2526, pruned_loss=0.03834, over 3315374.83 frames. ], batch size: 89, lr: 2.33e-03, grad_scale: 8.0 2023-05-02 16:56:44,736 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=286356.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 16:57:24,981 INFO [optim.py:368] (1/8) 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] (1/8) Epoch 29, batch 2200, loss[loss=0.1619, simple_loss=0.2581, pruned_loss=0.0328, over 17118.00 frames. ], tot_loss[loss=0.1655, simple_loss=0.2531, pruned_loss=0.03893, over 3317939.81 frames. ], batch size: 49, lr: 2.33e-03, grad_scale: 4.0 2023-05-02 16:57:51,142 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.0836, 2.8453, 2.5869, 4.9868, 3.7318, 4.1766, 1.7642, 3.0860], device='cuda:1'), covar=tensor([0.1327, 0.0915, 0.1413, 0.0176, 0.0206, 0.0504, 0.1715, 0.0904], device='cuda:1'), in_proj_covar=tensor([0.0174, 0.0181, 0.0200, 0.0206, 0.0206, 0.0219, 0.0210, 0.0199], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-05-02 16:57:52,983 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=286405.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 16:58:00,476 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.4415, 3.3162, 2.6913, 2.1957, 2.2467, 2.3256, 3.4101, 2.9248], device='cuda:1'), covar=tensor([0.2905, 0.0701, 0.1878, 0.2934, 0.2749, 0.2336, 0.0602, 0.1585], device='cuda:1'), in_proj_covar=tensor([0.0340, 0.0278, 0.0317, 0.0330, 0.0308, 0.0282, 0.0308, 0.0358], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:1') 2023-05-02 16:58:07,674 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.7812, 2.3844, 1.9499, 2.2085, 2.7659, 2.5456, 2.7050, 2.8770], device='cuda:1'), covar=tensor([0.0265, 0.0474, 0.0628, 0.0521, 0.0265, 0.0372, 0.0235, 0.0331], device='cuda:1'), in_proj_covar=tensor([0.0239, 0.0252, 0.0240, 0.0241, 0.0253, 0.0252, 0.0249, 0.0251], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-02 16:58:49,133 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.9831, 2.9720, 2.6919, 2.9735, 3.2431, 3.0790, 3.5165, 3.4544], device='cuda:1'), covar=tensor([0.0144, 0.0463, 0.0513, 0.0437, 0.0324, 0.0425, 0.0286, 0.0301], device='cuda:1'), in_proj_covar=tensor([0.0239, 0.0252, 0.0240, 0.0241, 0.0252, 0.0251, 0.0249, 0.0251], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-02 16:58:59,516 INFO [train.py:904] (1/8) Epoch 29, batch 2250, loss[loss=0.1387, simple_loss=0.2273, pruned_loss=0.0251, over 16842.00 frames. ], tot_loss[loss=0.1649, simple_loss=0.2526, pruned_loss=0.03856, over 3326445.86 frames. ], batch size: 42, lr: 2.33e-03, grad_scale: 4.0 2023-05-02 16:59:33,905 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=286479.0, num_to_drop=1, layers_to_drop={2} 2023-05-02 16:59:43,648 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=286485.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 16:59:45,654 INFO [optim.py:368] (1/8) Clipping_scale=2.0, grad-norm quartiles 1.691e+02 2.269e+02 2.585e+02 3.162e+02 6.397e+02, threshold=5.169e+02, percent-clipped=1.0 2023-05-02 17:00:08,671 INFO [train.py:904] (1/8) Epoch 29, batch 2300, loss[loss=0.1549, simple_loss=0.2371, pruned_loss=0.03637, over 15815.00 frames. ], tot_loss[loss=0.1655, simple_loss=0.2532, pruned_loss=0.03889, over 3323981.78 frames. ], batch size: 35, lr: 2.33e-03, grad_scale: 4.0 2023-05-02 17:00:17,311 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.30 vs. limit=2.0 2023-05-02 17:01:17,619 INFO [train.py:904] (1/8) Epoch 29, batch 2350, loss[loss=0.1436, simple_loss=0.2313, pruned_loss=0.02791, over 17215.00 frames. ], tot_loss[loss=0.1659, simple_loss=0.2538, pruned_loss=0.03898, over 3334093.65 frames. ], batch size: 45, lr: 2.33e-03, grad_scale: 4.0 2023-05-02 17:01:27,592 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.1061, 3.2402, 3.3091, 2.3259, 3.0064, 3.4192, 3.1218, 2.0454], device='cuda:1'), covar=tensor([0.0572, 0.0138, 0.0087, 0.0441, 0.0165, 0.0121, 0.0133, 0.0529], device='cuda:1'), in_proj_covar=tensor([0.0138, 0.0090, 0.0092, 0.0136, 0.0103, 0.0116, 0.0099, 0.0131], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0003, 0.0004], device='cuda:1') 2023-05-02 17:02:03,068 INFO [optim.py:368] (1/8) 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 17:02:04,841 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 29, batch 2400, loss[loss=0.1514, simple_loss=0.2398, pruned_loss=0.03148, over 16571.00 frames. ], tot_loss[loss=0.166, simple_loss=0.2544, pruned_loss=0.03881, over 3329072.17 frames. ], batch size: 75, lr: 2.33e-03, grad_scale: 8.0 2023-05-02 17:03:08,351 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.9601, 4.2091, 4.0604, 4.1192, 3.7682, 3.7959, 3.8355, 4.2239], device='cuda:1'), covar=tensor([0.1232, 0.1022, 0.1078, 0.0845, 0.0835, 0.1895, 0.1046, 0.1045], device='cuda:1'), in_proj_covar=tensor([0.0735, 0.0887, 0.0727, 0.0688, 0.0560, 0.0559, 0.0746, 0.0697], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:1') 2023-05-02 17:03:30,952 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=286649.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 17:03:34,192 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 29, batch 2450, loss[loss=0.2071, simple_loss=0.2878, pruned_loss=0.0632, over 11779.00 frames. ], tot_loss[loss=0.1675, simple_loss=0.2563, pruned_loss=0.03935, over 3308659.31 frames. ], batch size: 246, lr: 2.33e-03, grad_scale: 4.0 2023-05-02 17:03:59,974 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=286671.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 17:04:23,750 INFO [optim.py:368] (1/8) 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 17:04:41,144 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=286700.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 17:04:46,152 INFO [train.py:904] (1/8) Epoch 29, batch 2500, loss[loss=0.1403, simple_loss=0.2266, pruned_loss=0.02702, over 16738.00 frames. ], tot_loss[loss=0.1665, simple_loss=0.2554, pruned_loss=0.03883, over 3310542.75 frames. ], batch size: 39, lr: 2.33e-03, grad_scale: 4.0 2023-05-02 17:05:24,952 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=286732.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 17:05:55,670 INFO [train.py:904] (1/8) Epoch 29, batch 2550, loss[loss=0.1476, simple_loss=0.2374, pruned_loss=0.02892, over 17229.00 frames. ], tot_loss[loss=0.1671, simple_loss=0.2562, pruned_loss=0.03905, over 3317377.99 frames. ], batch size: 43, lr: 2.33e-03, grad_scale: 4.0 2023-05-02 17:06:18,986 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=286770.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 17:06:28,979 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.80 vs. limit=2.0 2023-05-02 17:06:32,970 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=286779.0, num_to_drop=1, layers_to_drop={1} 2023-05-02 17:06:40,930 INFO [zipformer.py:625] (1/8) 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] (1/8) 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 17:06:49,068 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=4.86 vs. limit=5.0 2023-05-02 17:07:07,742 INFO [train.py:904] (1/8) Epoch 29, batch 2600, loss[loss=0.1653, simple_loss=0.2585, pruned_loss=0.03605, over 17231.00 frames. ], tot_loss[loss=0.1667, simple_loss=0.2558, pruned_loss=0.03886, over 3321985.65 frames. ], batch size: 44, lr: 2.32e-03, grad_scale: 4.0 2023-05-02 17:07:21,051 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.0131, 3.0908, 3.1171, 2.1197, 2.9629, 3.2618, 2.9796, 1.9547], device='cuda:1'), covar=tensor([0.0589, 0.0156, 0.0099, 0.0470, 0.0162, 0.0131, 0.0146, 0.0525], device='cuda:1'), in_proj_covar=tensor([0.0141, 0.0092, 0.0093, 0.0137, 0.0105, 0.0118, 0.0100, 0.0133], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0003, 0.0004], device='cuda:1') 2023-05-02 17:07:39,058 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=286827.0, num_to_drop=1, layers_to_drop={0} 2023-05-02 17:07:44,493 INFO [zipformer.py:625] (1/8) 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] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=286833.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 17:07:55,615 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.66 vs. limit=2.0 2023-05-02 17:08:15,898 INFO [train.py:904] (1/8) Epoch 29, batch 2650, loss[loss=0.1443, simple_loss=0.2405, pruned_loss=0.02406, over 17253.00 frames. ], tot_loss[loss=0.1672, simple_loss=0.2566, pruned_loss=0.03887, over 3329641.28 frames. ], batch size: 45, lr: 2.32e-03, grad_scale: 4.0 2023-05-02 17:08:18,588 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.7156, 2.6414, 2.3763, 2.4072, 2.9646, 2.7287, 3.2510, 3.1898], device='cuda:1'), covar=tensor([0.0198, 0.0504, 0.0586, 0.0562, 0.0332, 0.0433, 0.0258, 0.0327], device='cuda:1'), in_proj_covar=tensor([0.0240, 0.0252, 0.0240, 0.0241, 0.0253, 0.0252, 0.0249, 0.0251], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-02 17:08:28,889 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.90 vs. limit=2.0 2023-05-02 17:08:41,289 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.14 vs. limit=2.0 2023-05-02 17:09:00,291 INFO [optim.py:368] (1/8) 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] (1/8) Epoch 29, batch 2700, loss[loss=0.1713, simple_loss=0.255, pruned_loss=0.04381, over 16711.00 frames. ], tot_loss[loss=0.1671, simple_loss=0.2568, pruned_loss=0.03868, over 3330705.71 frames. ], batch size: 124, lr: 2.32e-03, grad_scale: 4.0 2023-05-02 17:09:40,859 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.8395, 3.9698, 4.1327, 4.1290, 4.1715, 3.9448, 3.9914, 3.9131], device='cuda:1'), covar=tensor([0.0440, 0.0671, 0.0472, 0.0437, 0.0515, 0.0512, 0.0722, 0.0596], device='cuda:1'), in_proj_covar=tensor([0.0451, 0.0510, 0.0493, 0.0451, 0.0539, 0.0516, 0.0594, 0.0416], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0003], device='cuda:1') 2023-05-02 17:10:17,468 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=286944.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 17:10:26,213 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 29, batch 2750, loss[loss=0.1661, simple_loss=0.2519, pruned_loss=0.04022, over 16201.00 frames. ], tot_loss[loss=0.1663, simple_loss=0.2563, pruned_loss=0.03815, over 3322342.22 frames. ], batch size: 164, lr: 2.32e-03, grad_scale: 4.0 2023-05-02 17:11:17,993 INFO [optim.py:368] (1/8) 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] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=286999.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 17:11:35,317 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=287000.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 17:11:39,449 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.7721, 4.0172, 2.6073, 4.6686, 3.2351, 4.5660, 2.7427, 3.3630], device='cuda:1'), covar=tensor([0.0401, 0.0431, 0.1709, 0.0320, 0.0837, 0.0541, 0.1560, 0.0769], device='cuda:1'), in_proj_covar=tensor([0.0181, 0.0186, 0.0201, 0.0180, 0.0184, 0.0227, 0.0209, 0.0187], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-05-02 17:11:40,751 INFO [train.py:904] (1/8) Epoch 29, batch 2800, loss[loss=0.1529, simple_loss=0.2432, pruned_loss=0.0313, over 17199.00 frames. ], tot_loss[loss=0.1653, simple_loss=0.2553, pruned_loss=0.03763, over 3326080.70 frames. ], batch size: 44, lr: 2.32e-03, grad_scale: 4.0 2023-05-02 17:12:12,974 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=287027.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 17:12:19,191 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([5.5769, 5.5053, 5.4486, 4.9354, 5.1298, 5.4457, 5.3686, 5.0993], device='cuda:1'), covar=tensor([0.0516, 0.0489, 0.0273, 0.0337, 0.0901, 0.0436, 0.0301, 0.0740], device='cuda:1'), in_proj_covar=tensor([0.0324, 0.0488, 0.0380, 0.0381, 0.0375, 0.0438, 0.0260, 0.0454], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-05-02 17:12:42,702 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=287048.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 17:12:46,988 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.7938, 5.1039, 4.9048, 4.8959, 4.6991, 4.6463, 4.5490, 5.1879], device='cuda:1'), covar=tensor([0.1318, 0.0922, 0.1030, 0.0837, 0.0793, 0.1090, 0.1312, 0.0895], device='cuda:1'), in_proj_covar=tensor([0.0733, 0.0888, 0.0727, 0.0688, 0.0561, 0.0559, 0.0745, 0.0698], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:1') 2023-05-02 17:12:50,053 INFO [train.py:904] (1/8) Epoch 29, batch 2850, loss[loss=0.152, simple_loss=0.2394, pruned_loss=0.03224, over 17219.00 frames. ], tot_loss[loss=0.1646, simple_loss=0.2544, pruned_loss=0.03745, over 3330983.12 frames. ], batch size: 44, lr: 2.32e-03, grad_scale: 4.0 2023-05-02 17:13:39,986 INFO [optim.py:368] (1/8) 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] (1/8) Epoch 29, batch 2900, loss[loss=0.1544, simple_loss=0.2557, pruned_loss=0.02656, over 17129.00 frames. ], tot_loss[loss=0.1652, simple_loss=0.2541, pruned_loss=0.03814, over 3321320.73 frames. ], batch size: 49, lr: 2.32e-03, grad_scale: 4.0 2023-05-02 17:14:23,287 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([5.2502, 5.2429, 4.9760, 4.4215, 5.0670, 1.8540, 4.8236, 4.7485], device='cuda:1'), covar=tensor([0.0106, 0.0104, 0.0251, 0.0446, 0.0121, 0.3271, 0.0166, 0.0285], device='cuda:1'), in_proj_covar=tensor([0.0186, 0.0179, 0.0217, 0.0189, 0.0195, 0.0222, 0.0205, 0.0184], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-02 17:14:31,456 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=287126.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 17:15:10,987 INFO [train.py:904] (1/8) Epoch 29, batch 2950, loss[loss=0.2048, simple_loss=0.2915, pruned_loss=0.05907, over 15489.00 frames. ], tot_loss[loss=0.1649, simple_loss=0.2532, pruned_loss=0.03828, over 3324372.35 frames. ], batch size: 190, lr: 2.32e-03, grad_scale: 4.0 2023-05-02 17:15:59,392 INFO [optim.py:368] (1/8) Clipping_scale=2.0, grad-norm quartiles 1.489e+02 2.092e+02 2.416e+02 3.171e+02 5.496e+02, threshold=4.831e+02, percent-clipped=1.0 2023-05-02 17:16:12,074 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.13 vs. limit=2.0 2023-05-02 17:16:20,171 INFO [train.py:904] (1/8) Epoch 29, batch 3000, loss[loss=0.1624, simple_loss=0.2454, pruned_loss=0.03971, over 16779.00 frames. ], tot_loss[loss=0.1653, simple_loss=0.253, pruned_loss=0.03874, over 3322166.56 frames. ], batch size: 83, lr: 2.32e-03, grad_scale: 4.0 2023-05-02 17:16:20,172 INFO [train.py:929] (1/8) Computing validation loss 2023-05-02 17:16:28,746 INFO [train.py:938] (1/8) Epoch 29, validation: loss=0.1336, simple_loss=0.2385, pruned_loss=0.01438, over 944034.00 frames. 2023-05-02 17:16:28,746 INFO [train.py:939] (1/8) Maximum memory allocated so far is 18145MB 2023-05-02 17:17:25,959 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=287244.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 17:17:40,003 INFO [train.py:904] (1/8) Epoch 29, batch 3050, loss[loss=0.1845, simple_loss=0.2605, pruned_loss=0.05424, over 16896.00 frames. ], tot_loss[loss=0.165, simple_loss=0.2527, pruned_loss=0.03861, over 3326835.98 frames. ], batch size: 116, lr: 2.32e-03, grad_scale: 4.0 2023-05-02 17:17:58,036 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.5243, 4.5626, 4.8728, 4.8555, 4.9171, 4.5891, 4.5774, 4.4356], device='cuda:1'), covar=tensor([0.0411, 0.0636, 0.0448, 0.0450, 0.0564, 0.0483, 0.0919, 0.0705], device='cuda:1'), in_proj_covar=tensor([0.0452, 0.0512, 0.0495, 0.0453, 0.0541, 0.0518, 0.0597, 0.0417], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0003], device='cuda:1') 2023-05-02 17:18:29,494 INFO [optim.py:368] (1/8) 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 17:18:34,329 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=287292.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 17:18:49,727 INFO [train.py:904] (1/8) Epoch 29, batch 3100, loss[loss=0.1692, simple_loss=0.2592, pruned_loss=0.03962, over 16728.00 frames. ], tot_loss[loss=0.1658, simple_loss=0.2533, pruned_loss=0.03917, over 3323403.30 frames. ], batch size: 57, lr: 2.32e-03, grad_scale: 4.0 2023-05-02 17:19:15,275 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.9300, 2.0784, 2.6235, 2.8779, 2.8138, 3.3362, 2.4708, 3.3257], device='cuda:1'), covar=tensor([0.0295, 0.0573, 0.0391, 0.0381, 0.0385, 0.0236, 0.0553, 0.0193], device='cuda:1'), in_proj_covar=tensor([0.0204, 0.0204, 0.0192, 0.0198, 0.0216, 0.0172, 0.0208, 0.0172], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0002], device='cuda:1') 2023-05-02 17:19:22,407 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=287327.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 17:20:00,322 INFO [train.py:904] (1/8) Epoch 29, batch 3150, loss[loss=0.1506, simple_loss=0.2424, pruned_loss=0.02935, over 17122.00 frames. ], tot_loss[loss=0.1651, simple_loss=0.2524, pruned_loss=0.03891, over 3325281.05 frames. ], batch size: 49, lr: 2.32e-03, grad_scale: 4.0 2023-05-02 17:20:30,294 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=287375.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 17:20:49,322 INFO [optim.py:368] (1/8) Clipping_scale=2.0, grad-norm quartiles 1.513e+02 2.202e+02 2.547e+02 3.057e+02 5.331e+02, threshold=5.094e+02, percent-clipped=1.0 2023-05-02 17:21:10,099 INFO [train.py:904] (1/8) Epoch 29, batch 3200, loss[loss=0.1431, simple_loss=0.2263, pruned_loss=0.02992, over 16496.00 frames. ], tot_loss[loss=0.1644, simple_loss=0.2516, pruned_loss=0.03862, over 3326402.60 frames. ], batch size: 75, lr: 2.32e-03, grad_scale: 8.0 2023-05-02 17:21:20,087 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=287411.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 17:21:40,333 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=287426.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 17:22:18,321 INFO [train.py:904] (1/8) Epoch 29, batch 3250, loss[loss=0.1345, simple_loss=0.2237, pruned_loss=0.0227, over 16812.00 frames. ], tot_loss[loss=0.1637, simple_loss=0.2505, pruned_loss=0.03842, over 3322505.76 frames. ], batch size: 39, lr: 2.32e-03, grad_scale: 8.0 2023-05-02 17:22:43,507 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=287472.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 17:22:45,675 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=287474.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 17:22:58,091 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.6911, 4.6393, 4.5612, 3.9678, 4.6249, 1.8454, 4.3796, 4.2127], device='cuda:1'), covar=tensor([0.0168, 0.0131, 0.0213, 0.0357, 0.0123, 0.2910, 0.0166, 0.0256], device='cuda:1'), in_proj_covar=tensor([0.0187, 0.0181, 0.0220, 0.0190, 0.0197, 0.0224, 0.0208, 0.0186], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-02 17:23:00,330 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.1482, 4.0296, 4.2072, 4.3184, 4.3834, 4.0062, 4.2247, 4.4089], device='cuda:1'), covar=tensor([0.1619, 0.1132, 0.1241, 0.0673, 0.0623, 0.1514, 0.2631, 0.0729], device='cuda:1'), in_proj_covar=tensor([0.0712, 0.0872, 0.1005, 0.0885, 0.0674, 0.0700, 0.0738, 0.0856], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-02 17:23:05,921 INFO [optim.py:368] (1/8) 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 17:23:26,856 INFO [train.py:904] (1/8) Epoch 29, batch 3300, loss[loss=0.1627, simple_loss=0.2489, pruned_loss=0.03824, over 16800.00 frames. ], tot_loss[loss=0.1652, simple_loss=0.252, pruned_loss=0.03919, over 3324874.22 frames. ], batch size: 83, lr: 2.32e-03, grad_scale: 8.0 2023-05-02 17:23:36,893 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.26 vs. limit=2.0 2023-05-02 17:24:11,402 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 29, batch 3350, loss[loss=0.1932, simple_loss=0.2846, pruned_loss=0.05089, over 12009.00 frames. ], tot_loss[loss=0.1667, simple_loss=0.2539, pruned_loss=0.0398, over 3318259.90 frames. ], batch size: 247, lr: 2.32e-03, grad_scale: 8.0 2023-05-02 17:24:46,577 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=287562.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 17:24:50,080 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.9126, 3.6287, 4.0835, 2.2834, 4.2317, 4.2473, 3.2484, 3.1565], device='cuda:1'), covar=tensor([0.0755, 0.0272, 0.0228, 0.1105, 0.0103, 0.0231, 0.0436, 0.0485], device='cuda:1'), in_proj_covar=tensor([0.0149, 0.0112, 0.0103, 0.0140, 0.0088, 0.0133, 0.0131, 0.0132], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-05-02 17:25:13,862 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=4.75 vs. limit=5.0 2023-05-02 17:25:14,557 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=287583.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 17:25:21,671 INFO [optim.py:368] (1/8) 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 17:25:33,252 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=287597.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 17:25:43,306 INFO [train.py:904] (1/8) Epoch 29, batch 3400, loss[loss=0.1709, simple_loss=0.2561, pruned_loss=0.04285, over 16471.00 frames. ], tot_loss[loss=0.1658, simple_loss=0.2535, pruned_loss=0.03904, over 3311825.18 frames. ], batch size: 146, lr: 2.32e-03, grad_scale: 8.0 2023-05-02 17:26:10,408 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=287623.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 17:26:40,115 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=287644.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 17:26:53,548 INFO [train.py:904] (1/8) Epoch 29, batch 3450, loss[loss=0.1401, simple_loss=0.2273, pruned_loss=0.02641, over 17222.00 frames. ], tot_loss[loss=0.1637, simple_loss=0.2512, pruned_loss=0.03804, over 3315898.37 frames. ], batch size: 44, lr: 2.32e-03, grad_scale: 8.0 2023-05-02 17:27:16,553 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([5.2064, 5.2642, 5.0590, 4.5837, 4.5714, 5.1875, 5.1111, 4.6533], device='cuda:1'), covar=tensor([0.0702, 0.0519, 0.0397, 0.0492, 0.1406, 0.0498, 0.0354, 0.0899], device='cuda:1'), in_proj_covar=tensor([0.0328, 0.0496, 0.0384, 0.0386, 0.0380, 0.0444, 0.0263, 0.0459], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-05-02 17:27:43,291 INFO [optim.py:368] (1/8) Clipping_scale=2.0, grad-norm quartiles 1.405e+02 1.986e+02 2.391e+02 2.898e+02 4.066e+02, threshold=4.783e+02, percent-clipped=0.0 2023-05-02 17:28:04,223 INFO [train.py:904] (1/8) Epoch 29, batch 3500, loss[loss=0.1887, simple_loss=0.2676, pruned_loss=0.05492, over 16775.00 frames. ], tot_loss[loss=0.1627, simple_loss=0.2496, pruned_loss=0.03789, over 3318855.03 frames. ], batch size: 134, lr: 2.32e-03, grad_scale: 8.0 2023-05-02 17:28:26,806 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=287720.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 17:28:40,622 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.3589, 3.4275, 3.6262, 2.6398, 3.2799, 3.7214, 3.4463, 2.1264], device='cuda:1'), covar=tensor([0.0542, 0.0227, 0.0080, 0.0406, 0.0132, 0.0114, 0.0123, 0.0556], device='cuda:1'), in_proj_covar=tensor([0.0142, 0.0093, 0.0094, 0.0140, 0.0106, 0.0120, 0.0102, 0.0135], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0003, 0.0004], device='cuda:1') 2023-05-02 17:28:59,788 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.3099, 3.7260, 4.0937, 2.3484, 3.2679, 2.6661, 3.7491, 3.8357], device='cuda:1'), covar=tensor([0.0440, 0.1064, 0.0479, 0.2109, 0.0878, 0.1031, 0.0791, 0.1222], device='cuda:1'), in_proj_covar=tensor([0.0163, 0.0173, 0.0171, 0.0158, 0.0148, 0.0134, 0.0147, 0.0186], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0005], device='cuda:1') 2023-05-02 17:29:14,560 INFO [train.py:904] (1/8) Epoch 29, batch 3550, loss[loss=0.1609, simple_loss=0.2474, pruned_loss=0.03718, over 16574.00 frames. ], tot_loss[loss=0.1621, simple_loss=0.2486, pruned_loss=0.03781, over 3318583.23 frames. ], batch size: 68, lr: 2.32e-03, grad_scale: 8.0 2023-05-02 17:29:33,112 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=287767.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 17:29:42,518 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=287773.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 17:29:54,006 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=287781.0, num_to_drop=1, layers_to_drop={1} 2023-05-02 17:30:07,186 INFO [optim.py:368] (1/8) Clipping_scale=2.0, grad-norm quartiles 1.331e+02 2.118e+02 2.537e+02 3.010e+02 5.172e+02, threshold=5.074e+02, percent-clipped=1.0 2023-05-02 17:30:26,883 INFO [train.py:904] (1/8) Epoch 29, batch 3600, loss[loss=0.1672, simple_loss=0.2434, pruned_loss=0.04549, over 16770.00 frames. ], tot_loss[loss=0.162, simple_loss=0.2483, pruned_loss=0.03781, over 3304475.72 frames. ], batch size: 134, lr: 2.32e-03, grad_scale: 8.0 2023-05-02 17:31:09,632 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.3498, 4.1948, 4.4168, 4.5522, 4.6286, 4.2272, 4.4232, 4.6417], device='cuda:1'), covar=tensor([0.1777, 0.1363, 0.1406, 0.0737, 0.0680, 0.1288, 0.3523, 0.1006], device='cuda:1'), in_proj_covar=tensor([0.0715, 0.0874, 0.1008, 0.0886, 0.0675, 0.0701, 0.0740, 0.0857], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-02 17:31:10,906 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=287834.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 17:31:40,143 INFO [train.py:904] (1/8) Epoch 29, batch 3650, loss[loss=0.1502, simple_loss=0.2217, pruned_loss=0.03931, over 16820.00 frames. ], tot_loss[loss=0.1614, simple_loss=0.2473, pruned_loss=0.03773, over 3288805.95 frames. ], batch size: 83, lr: 2.32e-03, grad_scale: 8.0 2023-05-02 17:32:35,499 INFO [optim.py:368] (1/8) 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] (1/8) 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] (1/8) Epoch 29, batch 3700, loss[loss=0.1579, simple_loss=0.2348, pruned_loss=0.04052, over 16837.00 frames. ], tot_loss[loss=0.1627, simple_loss=0.2464, pruned_loss=0.03951, over 3267471.37 frames. ], batch size: 116, lr: 2.32e-03, grad_scale: 8.0 2023-05-02 17:33:18,786 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=287918.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 17:33:48,730 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=287939.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 17:34:02,443 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.8166, 2.8056, 2.3570, 2.6816, 3.1689, 2.8280, 3.2644, 3.3933], device='cuda:1'), covar=tensor([0.0108, 0.0461, 0.0614, 0.0490, 0.0289, 0.0463, 0.0292, 0.0285], device='cuda:1'), in_proj_covar=tensor([0.0241, 0.0251, 0.0240, 0.0241, 0.0253, 0.0251, 0.0249, 0.0252], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-02 17:34:10,008 INFO [train.py:904] (1/8) Epoch 29, batch 3750, loss[loss=0.1574, simple_loss=0.2403, pruned_loss=0.03729, over 16798.00 frames. ], tot_loss[loss=0.1639, simple_loss=0.247, pruned_loss=0.04039, over 3260027.78 frames. ], batch size: 102, lr: 2.32e-03, grad_scale: 8.0 2023-05-02 17:34:49,963 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.5806, 4.3910, 4.5051, 4.7829, 4.8642, 4.4789, 4.8783, 4.9343], device='cuda:1'), covar=tensor([0.1816, 0.1567, 0.1987, 0.0959, 0.0903, 0.1197, 0.1918, 0.1070], device='cuda:1'), in_proj_covar=tensor([0.0710, 0.0866, 0.0998, 0.0879, 0.0670, 0.0695, 0.0733, 0.0850], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-02 17:35:05,542 INFO [optim.py:368] (1/8) 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 17:35:29,221 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=288003.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 17:35:30,027 INFO [train.py:904] (1/8) Epoch 29, batch 3800, loss[loss=0.1641, simple_loss=0.2416, pruned_loss=0.04333, over 16861.00 frames. ], tot_loss[loss=0.1659, simple_loss=0.2483, pruned_loss=0.0417, over 3245028.94 frames. ], batch size: 90, lr: 2.32e-03, grad_scale: 8.0 2023-05-02 17:36:15,831 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.2625, 3.2714, 2.1644, 3.4547, 2.6517, 3.4833, 2.2603, 2.7743], device='cuda:1'), covar=tensor([0.0346, 0.0496, 0.1577, 0.0366, 0.0834, 0.0849, 0.1461, 0.0757], device='cuda:1'), in_proj_covar=tensor([0.0181, 0.0185, 0.0200, 0.0180, 0.0183, 0.0227, 0.0208, 0.0186], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-05-02 17:36:40,374 INFO [train.py:904] (1/8) Epoch 29, batch 3850, loss[loss=0.173, simple_loss=0.2552, pruned_loss=0.04535, over 15616.00 frames. ], tot_loss[loss=0.1673, simple_loss=0.2492, pruned_loss=0.04268, over 3247304.41 frames. ], batch size: 191, lr: 2.32e-03, grad_scale: 8.0 2023-05-02 17:36:55,417 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=288064.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 17:36:58,986 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=288067.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 17:37:12,438 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=288076.0, num_to_drop=1, layers_to_drop={1} 2023-05-02 17:37:30,811 INFO [optim.py:368] (1/8) 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] (1/8) Epoch 29, batch 3900, loss[loss=0.1579, simple_loss=0.2393, pruned_loss=0.03823, over 16405.00 frames. ], tot_loss[loss=0.1673, simple_loss=0.2485, pruned_loss=0.04306, over 3265266.67 frames. ], batch size: 68, lr: 2.32e-03, grad_scale: 8.0 2023-05-02 17:37:53,588 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=288106.0, num_to_drop=1, layers_to_drop={0} 2023-05-02 17:38:05,837 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=288115.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 17:38:24,046 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.45 vs. limit=2.0 2023-05-02 17:38:27,298 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=288129.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 17:38:35,901 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.7851, 3.8572, 2.5247, 4.3524, 3.0943, 4.3452, 2.6421, 3.2373], device='cuda:1'), covar=tensor([0.0293, 0.0370, 0.1493, 0.0194, 0.0711, 0.0421, 0.1346, 0.0669], device='cuda:1'), in_proj_covar=tensor([0.0180, 0.0184, 0.0199, 0.0179, 0.0183, 0.0226, 0.0207, 0.0185], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-05-02 17:39:00,609 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=288152.0, num_to_drop=1, layers_to_drop={0} 2023-05-02 17:39:03,101 INFO [train.py:904] (1/8) Epoch 29, batch 3950, loss[loss=0.1712, simple_loss=0.2407, pruned_loss=0.05087, over 16735.00 frames. ], tot_loss[loss=0.1674, simple_loss=0.2479, pruned_loss=0.04345, over 3254654.27 frames. ], batch size: 83, lr: 2.32e-03, grad_scale: 8.0 2023-05-02 17:39:22,307 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=288167.0, num_to_drop=1, layers_to_drop={3} 2023-05-02 17:39:55,368 INFO [optim.py:368] (1/8) 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 17:39:58,706 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=288192.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 17:40:16,238 INFO [train.py:904] (1/8) Epoch 29, batch 4000, loss[loss=0.1829, simple_loss=0.2678, pruned_loss=0.04898, over 16704.00 frames. ], tot_loss[loss=0.169, simple_loss=0.2489, pruned_loss=0.04459, over 3263977.73 frames. ], batch size: 134, lr: 2.32e-03, grad_scale: 8.0 2023-05-02 17:40:30,275 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=288213.0, num_to_drop=1, layers_to_drop={3} 2023-05-02 17:40:36,696 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=288218.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 17:41:07,469 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=288239.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 17:41:08,416 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=288240.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 17:41:29,240 INFO [train.py:904] (1/8) Epoch 29, batch 4050, loss[loss=0.1583, simple_loss=0.2459, pruned_loss=0.0353, over 17100.00 frames. ], tot_loss[loss=0.1691, simple_loss=0.2504, pruned_loss=0.04387, over 3265898.08 frames. ], batch size: 48, lr: 2.32e-03, grad_scale: 8.0 2023-05-02 17:41:47,213 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=288266.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 17:42:17,218 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=288287.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 17:42:21,078 INFO [optim.py:368] (1/8) 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 17:42:34,372 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.4404, 3.1378, 3.5305, 1.8375, 3.6790, 3.6570, 2.9620, 2.7305], device='cuda:1'), covar=tensor([0.0855, 0.0347, 0.0193, 0.1309, 0.0098, 0.0153, 0.0421, 0.0539], device='cuda:1'), in_proj_covar=tensor([0.0150, 0.0113, 0.0103, 0.0140, 0.0088, 0.0134, 0.0131, 0.0132], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-05-02 17:42:42,070 INFO [train.py:904] (1/8) Epoch 29, batch 4100, loss[loss=0.1892, simple_loss=0.2837, pruned_loss=0.04732, over 16224.00 frames. ], tot_loss[loss=0.1693, simple_loss=0.252, pruned_loss=0.04329, over 3252557.53 frames. ], batch size: 165, lr: 2.32e-03, grad_scale: 8.0 2023-05-02 17:43:17,908 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=288327.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 17:43:45,756 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 29, batch 4150, loss[loss=0.2003, simple_loss=0.2902, pruned_loss=0.05524, over 15342.00 frames. ], tot_loss[loss=0.1744, simple_loss=0.2585, pruned_loss=0.04508, over 3227977.08 frames. ], batch size: 190, lr: 2.32e-03, grad_scale: 8.0 2023-05-02 17:44:07,903 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=288359.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 17:44:14,203 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([5.0433, 5.0211, 4.7598, 4.1326, 4.9713, 1.8416, 4.7051, 4.3145], device='cuda:1'), covar=tensor([0.0084, 0.0083, 0.0194, 0.0331, 0.0079, 0.2992, 0.0120, 0.0294], device='cuda:1'), in_proj_covar=tensor([0.0188, 0.0181, 0.0221, 0.0192, 0.0198, 0.0224, 0.0209, 0.0187], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-02 17:44:33,292 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=288376.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 17:44:51,098 INFO [zipformer.py:625] (1/8) 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] (1/8) 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 17:45:02,578 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=288396.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 17:45:13,928 INFO [train.py:904] (1/8) Epoch 29, batch 4200, loss[loss=0.2065, simple_loss=0.3078, pruned_loss=0.05265, over 16922.00 frames. ], tot_loss[loss=0.179, simple_loss=0.2654, pruned_loss=0.04636, over 3211570.65 frames. ], batch size: 109, lr: 2.32e-03, grad_scale: 8.0 2023-05-02 17:45:18,021 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=288406.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 17:45:19,799 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.6843, 4.6867, 4.4258, 3.6949, 4.5956, 1.6562, 4.3496, 3.9271], device='cuda:1'), covar=tensor([0.0087, 0.0082, 0.0220, 0.0360, 0.0081, 0.3379, 0.0114, 0.0343], device='cuda:1'), in_proj_covar=tensor([0.0188, 0.0181, 0.0221, 0.0192, 0.0198, 0.0224, 0.0209, 0.0187], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-02 17:45:45,767 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=288424.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 17:45:53,468 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=288429.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 17:46:30,280 INFO [train.py:904] (1/8) Epoch 29, batch 4250, loss[loss=0.2012, simple_loss=0.2922, pruned_loss=0.05515, over 17027.00 frames. ], tot_loss[loss=0.1802, simple_loss=0.2687, pruned_loss=0.04586, over 3209683.32 frames. ], batch size: 55, lr: 2.32e-03, grad_scale: 8.0 2023-05-02 17:46:36,668 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=288457.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 17:46:42,513 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=288462.0, num_to_drop=1, layers_to_drop={0} 2023-05-02 17:46:45,728 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.8082, 3.1270, 3.4135, 2.1546, 2.9522, 2.2044, 3.2924, 3.3636], device='cuda:1'), covar=tensor([0.0306, 0.0920, 0.0641, 0.2121, 0.0867, 0.1075, 0.0713, 0.0941], device='cuda:1'), in_proj_covar=tensor([0.0162, 0.0173, 0.0171, 0.0158, 0.0148, 0.0133, 0.0147, 0.0185], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0005], device='cuda:1') 2023-05-02 17:47:05,112 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=288477.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 17:47:24,426 INFO [optim.py:368] (1/8) 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] (1/8) Epoch 29, batch 4300, loss[loss=0.1677, simple_loss=0.2634, pruned_loss=0.036, over 16319.00 frames. ], tot_loss[loss=0.1797, simple_loss=0.2697, pruned_loss=0.04487, over 3217021.89 frames. ], batch size: 146, lr: 2.32e-03, grad_scale: 8.0 2023-05-02 17:47:51,117 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=288508.0, num_to_drop=1, layers_to_drop={0} 2023-05-02 17:48:59,343 INFO [train.py:904] (1/8) Epoch 29, batch 4350, loss[loss=0.2072, simple_loss=0.3032, pruned_loss=0.05557, over 17029.00 frames. ], tot_loss[loss=0.1824, simple_loss=0.2729, pruned_loss=0.04598, over 3211336.13 frames. ], batch size: 55, lr: 2.32e-03, grad_scale: 8.0 2023-05-02 17:49:53,411 INFO [optim.py:368] (1/8) 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 17:50:13,840 INFO [train.py:904] (1/8) Epoch 29, batch 4400, loss[loss=0.1987, simple_loss=0.2879, pruned_loss=0.05474, over 16912.00 frames. ], tot_loss[loss=0.185, simple_loss=0.2751, pruned_loss=0.04748, over 3197888.38 frames. ], batch size: 109, lr: 2.32e-03, grad_scale: 8.0 2023-05-02 17:50:22,313 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=288609.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 17:51:26,299 INFO [train.py:904] (1/8) Epoch 29, batch 4450, loss[loss=0.195, simple_loss=0.2869, pruned_loss=0.05153, over 16409.00 frames. ], tot_loss[loss=0.1887, simple_loss=0.2787, pruned_loss=0.04929, over 3212805.90 frames. ], batch size: 68, lr: 2.32e-03, grad_scale: 8.0 2023-05-02 17:51:33,491 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=288659.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 17:51:49,633 INFO [zipformer.py:625] (1/8) 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] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=288683.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 17:52:18,787 INFO [optim.py:368] (1/8) Clipping_scale=2.0, grad-norm quartiles 1.398e+02 1.892e+02 2.239e+02 2.604e+02 5.228e+02, threshold=4.478e+02, percent-clipped=1.0 2023-05-02 17:52:34,489 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=288701.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 17:52:37,593 INFO [train.py:904] (1/8) Epoch 29, batch 4500, loss[loss=0.2045, simple_loss=0.2819, pruned_loss=0.06353, over 15381.00 frames. ], tot_loss[loss=0.1905, simple_loss=0.2801, pruned_loss=0.05044, over 3219182.15 frames. ], batch size: 190, lr: 2.32e-03, grad_scale: 8.0 2023-05-02 17:52:42,937 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=288707.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 17:53:01,795 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.2647, 3.4567, 3.7030, 2.1901, 3.2683, 2.3567, 3.6032, 3.7682], device='cuda:1'), covar=tensor([0.0209, 0.0812, 0.0553, 0.2270, 0.0812, 0.1066, 0.0541, 0.0911], device='cuda:1'), in_proj_covar=tensor([0.0163, 0.0173, 0.0172, 0.0158, 0.0149, 0.0134, 0.0147, 0.0186], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0005], device='cuda:1') 2023-05-02 17:53:49,131 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=288752.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 17:53:50,941 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=4.91 vs. limit=5.0 2023-05-02 17:53:51,862 INFO [train.py:904] (1/8) Epoch 29, batch 4550, loss[loss=0.186, simple_loss=0.2763, pruned_loss=0.0479, over 17016.00 frames. ], tot_loss[loss=0.1915, simple_loss=0.2807, pruned_loss=0.05113, over 3242915.98 frames. ], batch size: 50, lr: 2.32e-03, grad_scale: 8.0 2023-05-02 17:54:03,012 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=288762.0, num_to_drop=1, layers_to_drop={1} 2023-05-02 17:54:13,431 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=4.89 vs. limit=5.0 2023-05-02 17:54:40,328 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.58 vs. limit=2.0 2023-05-02 17:54:44,107 INFO [optim.py:368] (1/8) 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 17:54:53,746 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=288797.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 17:54:58,825 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.7467, 4.9998, 5.1939, 4.8496, 4.9499, 5.5626, 4.9833, 4.6473], device='cuda:1'), covar=tensor([0.1115, 0.1844, 0.2115, 0.1923, 0.2391, 0.0864, 0.1566, 0.2360], device='cuda:1'), in_proj_covar=tensor([0.0431, 0.0640, 0.0706, 0.0519, 0.0696, 0.0728, 0.0548, 0.0690], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-05-02 17:55:04,695 INFO [train.py:904] (1/8) Epoch 29, batch 4600, loss[loss=0.2077, simple_loss=0.2858, pruned_loss=0.06483, over 12120.00 frames. ], tot_loss[loss=0.192, simple_loss=0.2811, pruned_loss=0.05142, over 3227850.95 frames. ], batch size: 248, lr: 2.32e-03, grad_scale: 8.0 2023-05-02 17:55:11,573 INFO [zipformer.py:625] (1/8) 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] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=288810.0, num_to_drop=1, layers_to_drop={1} 2023-05-02 17:55:14,682 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.1471, 2.5214, 2.4661, 3.9033, 2.2549, 2.7867, 2.5427, 2.5866], device='cuda:1'), covar=tensor([0.1510, 0.3088, 0.2795, 0.0606, 0.4040, 0.2180, 0.2975, 0.3352], device='cuda:1'), in_proj_covar=tensor([0.0424, 0.0478, 0.0388, 0.0339, 0.0447, 0.0550, 0.0450, 0.0560], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-02 17:56:18,390 INFO [train.py:904] (1/8) Epoch 29, batch 4650, loss[loss=0.1859, simple_loss=0.2648, pruned_loss=0.05348, over 16653.00 frames. ], tot_loss[loss=0.1911, simple_loss=0.2799, pruned_loss=0.05116, over 3235282.36 frames. ], batch size: 57, lr: 2.32e-03, grad_scale: 8.0 2023-05-02 17:56:19,944 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.9222, 4.9535, 4.7764, 4.4172, 4.5117, 4.8508, 4.5691, 4.5671], device='cuda:1'), covar=tensor([0.0434, 0.0327, 0.0240, 0.0256, 0.0704, 0.0329, 0.0459, 0.0501], device='cuda:1'), in_proj_covar=tensor([0.0314, 0.0476, 0.0369, 0.0371, 0.0366, 0.0426, 0.0254, 0.0442], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002], device='cuda:1') 2023-05-02 17:56:21,002 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=288856.0, num_to_drop=1, layers_to_drop={0} 2023-05-02 17:56:24,232 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=288858.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 17:57:10,667 INFO [optim.py:368] (1/8) Clipping_scale=2.0, grad-norm quartiles 1.194e+02 1.769e+02 2.025e+02 2.417e+02 4.191e+02, threshold=4.050e+02, percent-clipped=1.0 2023-05-02 17:57:29,733 INFO [train.py:904] (1/8) Epoch 29, batch 4700, loss[loss=0.1688, simple_loss=0.262, pruned_loss=0.03777, over 16873.00 frames. ], tot_loss[loss=0.1886, simple_loss=0.2771, pruned_loss=0.05002, over 3225075.35 frames. ], batch size: 102, lr: 2.32e-03, grad_scale: 4.0 2023-05-02 17:58:41,594 INFO [train.py:904] (1/8) Epoch 29, batch 4750, loss[loss=0.1856, simple_loss=0.2702, pruned_loss=0.05053, over 12109.00 frames. ], tot_loss[loss=0.1843, simple_loss=0.2729, pruned_loss=0.04788, over 3232032.03 frames. ], batch size: 247, lr: 2.32e-03, grad_scale: 4.0 2023-05-02 17:58:57,734 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=288965.0, num_to_drop=1, layers_to_drop={0} 2023-05-02 17:59:23,637 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=288983.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 17:59:35,273 INFO [optim.py:368] (1/8) 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 17:59:45,543 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.7413, 4.9752, 4.8160, 4.8408, 4.5627, 4.5016, 4.4436, 5.0569], device='cuda:1'), covar=tensor([0.1186, 0.0795, 0.0835, 0.0764, 0.0784, 0.1237, 0.1077, 0.0890], device='cuda:1'), in_proj_covar=tensor([0.0726, 0.0872, 0.0714, 0.0678, 0.0557, 0.0552, 0.0734, 0.0687], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:1') 2023-05-02 17:59:49,187 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([5.3273, 5.2996, 5.1613, 4.7148, 4.7444, 5.1925, 5.1171, 4.8860], device='cuda:1'), covar=tensor([0.0626, 0.0596, 0.0316, 0.0329, 0.1199, 0.0685, 0.0251, 0.0654], device='cuda:1'), in_proj_covar=tensor([0.0316, 0.0479, 0.0371, 0.0373, 0.0368, 0.0428, 0.0255, 0.0443], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002], device='cuda:1') 2023-05-02 17:59:50,950 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 29, batch 4800, loss[loss=0.1928, simple_loss=0.2734, pruned_loss=0.05603, over 12098.00 frames. ], tot_loss[loss=0.1808, simple_loss=0.2693, pruned_loss=0.04621, over 3210966.85 frames. ], batch size: 246, lr: 2.32e-03, grad_scale: 8.0 2023-05-02 18:00:03,227 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=4.90 vs. limit=5.0 2023-05-02 18:00:37,054 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=289031.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 18:00:59,788 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=289046.0, num_to_drop=1, layers_to_drop={1} 2023-05-02 18:01:05,206 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=289049.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 18:01:08,912 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=289052.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 18:01:12,128 INFO [train.py:904] (1/8) Epoch 29, batch 4850, loss[loss=0.1827, simple_loss=0.2836, pruned_loss=0.04089, over 16699.00 frames. ], tot_loss[loss=0.18, simple_loss=0.2695, pruned_loss=0.0453, over 3185559.05 frames. ], batch size: 134, lr: 2.32e-03, grad_scale: 8.0 2023-05-02 18:02:08,678 INFO [optim.py:368] (1/8) 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 18:02:22,358 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 29, batch 4900, loss[loss=0.1755, simple_loss=0.2702, pruned_loss=0.04038, over 16432.00 frames. ], tot_loss[loss=0.1784, simple_loss=0.2686, pruned_loss=0.04406, over 3153661.01 frames. ], batch size: 146, lr: 2.32e-03, grad_scale: 8.0 2023-05-02 18:02:33,618 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=289107.0, num_to_drop=1, layers_to_drop={3} 2023-05-02 18:03:06,135 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=4.83 vs. limit=5.0 2023-05-02 18:03:16,463 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=289136.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 18:03:40,271 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.8549, 4.9639, 5.2534, 5.2025, 5.2263, 4.9442, 4.8583, 4.7386], device='cuda:1'), covar=tensor([0.0312, 0.0439, 0.0320, 0.0364, 0.0455, 0.0317, 0.0901, 0.0428], device='cuda:1'), in_proj_covar=tensor([0.0434, 0.0491, 0.0476, 0.0435, 0.0521, 0.0498, 0.0577, 0.0402], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:1') 2023-05-02 18:03:42,153 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 29, batch 4950, loss[loss=0.1625, simple_loss=0.268, pruned_loss=0.02852, over 16750.00 frames. ], tot_loss[loss=0.1772, simple_loss=0.2681, pruned_loss=0.04317, over 3169064.16 frames. ], batch size: 89, lr: 2.32e-03, grad_scale: 8.0 2023-05-02 18:04:38,082 INFO [optim.py:368] (1/8) 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 18:04:46,879 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=289197.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 18:04:56,975 INFO [train.py:904] (1/8) Epoch 29, batch 5000, loss[loss=0.1705, simple_loss=0.2627, pruned_loss=0.03914, over 16605.00 frames. ], tot_loss[loss=0.1776, simple_loss=0.2692, pruned_loss=0.04305, over 3184492.53 frames. ], batch size: 68, lr: 2.32e-03, grad_scale: 8.0 2023-05-02 18:05:05,554 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=2.68 vs. limit=5.0 2023-05-02 18:05:27,934 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([5.1035, 5.0691, 4.9363, 4.1558, 5.0461, 1.7188, 4.7345, 4.6840], device='cuda:1'), covar=tensor([0.0122, 0.0112, 0.0187, 0.0564, 0.0129, 0.3122, 0.0157, 0.0262], device='cuda:1'), in_proj_covar=tensor([0.0184, 0.0179, 0.0217, 0.0188, 0.0194, 0.0220, 0.0205, 0.0184], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-02 18:06:09,979 INFO [train.py:904] (1/8) Epoch 29, batch 5050, loss[loss=0.1943, simple_loss=0.2773, pruned_loss=0.05562, over 16591.00 frames. ], tot_loss[loss=0.1778, simple_loss=0.2699, pruned_loss=0.04281, over 3204173.45 frames. ], batch size: 62, lr: 2.32e-03, grad_scale: 8.0 2023-05-02 18:06:26,202 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=289265.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 18:07:03,092 INFO [optim.py:368] (1/8) 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 18:07:22,111 INFO [train.py:904] (1/8) Epoch 29, batch 5100, loss[loss=0.1699, simple_loss=0.2552, pruned_loss=0.04228, over 16630.00 frames. ], tot_loss[loss=0.1767, simple_loss=0.2683, pruned_loss=0.0425, over 3208778.33 frames. ], batch size: 57, lr: 2.31e-03, grad_scale: 8.0 2023-05-02 18:07:34,856 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=289313.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 18:08:36,361 INFO [train.py:904] (1/8) Epoch 29, batch 5150, loss[loss=0.1688, simple_loss=0.2603, pruned_loss=0.03862, over 11846.00 frames. ], tot_loss[loss=0.1761, simple_loss=0.2681, pruned_loss=0.042, over 3199832.08 frames. ], batch size: 248, lr: 2.31e-03, grad_scale: 8.0 2023-05-02 18:08:51,207 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.8130, 3.7166, 3.8715, 3.9914, 4.0857, 3.7326, 4.0599, 4.1377], device='cuda:1'), covar=tensor([0.1594, 0.1142, 0.1390, 0.0729, 0.0572, 0.1702, 0.0763, 0.0676], device='cuda:1'), in_proj_covar=tensor([0.0685, 0.0835, 0.0963, 0.0850, 0.0647, 0.0668, 0.0705, 0.0821], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-02 18:09:29,039 INFO [optim.py:368] (1/8) 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] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=289402.0, num_to_drop=1, layers_to_drop={0} 2023-05-02 18:09:47,652 INFO [train.py:904] (1/8) Epoch 29, batch 5200, loss[loss=0.1588, simple_loss=0.2466, pruned_loss=0.03549, over 16455.00 frames. ], tot_loss[loss=0.1741, simple_loss=0.2663, pruned_loss=0.04093, over 3209512.89 frames. ], batch size: 68, lr: 2.31e-03, grad_scale: 8.0 2023-05-02 18:10:59,622 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=289453.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 18:11:00,512 INFO [train.py:904] (1/8) Epoch 29, batch 5250, loss[loss=0.1641, simple_loss=0.2481, pruned_loss=0.04005, over 16258.00 frames. ], tot_loss[loss=0.1719, simple_loss=0.2632, pruned_loss=0.04025, over 3220724.69 frames. ], batch size: 35, lr: 2.31e-03, grad_scale: 8.0 2023-05-02 18:11:55,409 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=2.92 vs. limit=5.0 2023-05-02 18:11:55,944 INFO [optim.py:368] (1/8) 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 18:11:58,179 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=289492.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 18:12:02,368 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.5934, 3.5579, 3.4985, 2.6709, 3.3791, 1.9928, 3.2426, 2.8573], device='cuda:1'), covar=tensor([0.0185, 0.0189, 0.0215, 0.0341, 0.0136, 0.2739, 0.0174, 0.0316], device='cuda:1'), in_proj_covar=tensor([0.0183, 0.0178, 0.0216, 0.0187, 0.0193, 0.0220, 0.0204, 0.0183], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-02 18:12:10,716 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=289501.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 18:12:14,313 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=289503.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 18:12:15,674 INFO [train.py:904] (1/8) Epoch 29, batch 5300, loss[loss=0.1481, simple_loss=0.2427, pruned_loss=0.02677, over 16493.00 frames. ], tot_loss[loss=0.1699, simple_loss=0.2605, pruned_loss=0.03962, over 3209532.38 frames. ], batch size: 75, lr: 2.31e-03, grad_scale: 8.0 2023-05-02 18:12:46,729 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.4638, 3.0498, 2.6467, 2.2870, 2.3195, 2.3129, 3.0797, 2.8463], device='cuda:1'), covar=tensor([0.2727, 0.0720, 0.1853, 0.2604, 0.2630, 0.2279, 0.0603, 0.1419], device='cuda:1'), in_proj_covar=tensor([0.0334, 0.0274, 0.0312, 0.0326, 0.0306, 0.0277, 0.0305, 0.0352], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:1') 2023-05-02 18:13:26,897 INFO [train.py:904] (1/8) Epoch 29, batch 5350, loss[loss=0.1848, simple_loss=0.2791, pruned_loss=0.04519, over 16720.00 frames. ], tot_loss[loss=0.1691, simple_loss=0.2595, pruned_loss=0.03931, over 3197899.62 frames. ], batch size: 89, lr: 2.31e-03, grad_scale: 8.0 2023-05-02 18:13:42,029 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=289564.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 18:14:21,403 INFO [optim.py:368] (1/8) 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 18:14:41,092 INFO [train.py:904] (1/8) Epoch 29, batch 5400, loss[loss=0.1826, simple_loss=0.2809, pruned_loss=0.04218, over 16702.00 frames. ], tot_loss[loss=0.1707, simple_loss=0.2621, pruned_loss=0.03968, over 3224649.39 frames. ], batch size: 76, lr: 2.31e-03, grad_scale: 8.0 2023-05-02 18:15:06,611 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.2975, 3.3597, 2.0918, 3.7475, 2.5282, 3.7230, 2.2603, 2.7012], device='cuda:1'), covar=tensor([0.0316, 0.0407, 0.1742, 0.0162, 0.0854, 0.0558, 0.1537, 0.0827], device='cuda:1'), in_proj_covar=tensor([0.0177, 0.0183, 0.0197, 0.0174, 0.0181, 0.0222, 0.0206, 0.0184], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-05-02 18:15:41,625 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.75 vs. limit=2.0 2023-05-02 18:15:57,432 INFO [train.py:904] (1/8) Epoch 29, batch 5450, loss[loss=0.1786, simple_loss=0.2689, pruned_loss=0.0441, over 17211.00 frames. ], tot_loss[loss=0.1732, simple_loss=0.2644, pruned_loss=0.04097, over 3214738.66 frames. ], batch size: 44, lr: 2.31e-03, grad_scale: 8.0 2023-05-02 18:16:53,636 INFO [optim.py:368] (1/8) 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 18:17:11,498 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=289702.0, num_to_drop=1, layers_to_drop={0} 2023-05-02 18:17:14,644 INFO [train.py:904] (1/8) Epoch 29, batch 5500, loss[loss=0.2461, simple_loss=0.3267, pruned_loss=0.08275, over 15383.00 frames. ], tot_loss[loss=0.1808, simple_loss=0.2719, pruned_loss=0.04484, over 3200768.73 frames. ], batch size: 190, lr: 2.31e-03, grad_scale: 8.0 2023-05-02 18:18:26,454 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=289750.0, num_to_drop=1, layers_to_drop={1} 2023-05-02 18:18:31,706 INFO [train.py:904] (1/8) Epoch 29, batch 5550, loss[loss=0.2003, simple_loss=0.291, pruned_loss=0.05481, over 16475.00 frames. ], tot_loss[loss=0.1885, simple_loss=0.2785, pruned_loss=0.04925, over 3173543.83 frames. ], batch size: 68, lr: 2.31e-03, grad_scale: 8.0 2023-05-02 18:18:36,186 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.3852, 4.0461, 4.0496, 2.6711, 3.5840, 4.1019, 3.6625, 2.3262], device='cuda:1'), covar=tensor([0.0651, 0.0083, 0.0077, 0.0494, 0.0149, 0.0131, 0.0118, 0.0536], device='cuda:1'), in_proj_covar=tensor([0.0138, 0.0090, 0.0092, 0.0136, 0.0103, 0.0116, 0.0099, 0.0131], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0003, 0.0004], device='cuda:1') 2023-05-02 18:18:51,079 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([5.7778, 6.0844, 5.8104, 5.9435, 5.5265, 5.4525, 5.5431, 6.2485], device='cuda:1'), covar=tensor([0.1220, 0.0797, 0.0978, 0.0919, 0.0843, 0.0583, 0.1171, 0.0806], device='cuda:1'), in_proj_covar=tensor([0.0719, 0.0864, 0.0712, 0.0670, 0.0552, 0.0549, 0.0724, 0.0680], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:1') 2023-05-02 18:18:59,423 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.8245, 2.7687, 2.9035, 2.2187, 2.7558, 2.2093, 2.7138, 2.9558], device='cuda:1'), covar=tensor([0.0273, 0.0782, 0.0482, 0.1702, 0.0763, 0.0874, 0.0521, 0.0685], device='cuda:1'), in_proj_covar=tensor([0.0161, 0.0171, 0.0170, 0.0157, 0.0148, 0.0133, 0.0146, 0.0184], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0005], device='cuda:1') 2023-05-02 18:19:30,745 INFO [optim.py:368] (1/8) 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 18:19:33,696 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=289792.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 18:19:53,194 INFO [train.py:904] (1/8) Epoch 29, batch 5600, loss[loss=0.2061, simple_loss=0.2912, pruned_loss=0.06051, over 17061.00 frames. ], tot_loss[loss=0.1947, simple_loss=0.2832, pruned_loss=0.05309, over 3139004.32 frames. ], batch size: 53, lr: 2.31e-03, grad_scale: 8.0 2023-05-02 18:20:53,921 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=289840.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 18:21:16,879 INFO [train.py:904] (1/8) Epoch 29, batch 5650, loss[loss=0.218, simple_loss=0.2972, pruned_loss=0.06945, over 15387.00 frames. ], tot_loss[loss=0.2022, simple_loss=0.2888, pruned_loss=0.05778, over 3090841.54 frames. ], batch size: 190, lr: 2.31e-03, grad_scale: 8.0 2023-05-02 18:21:25,654 INFO [zipformer.py:625] (1/8) 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] (1/8) Clipping_scale=2.0, grad-norm quartiles 1.920e+02 3.088e+02 3.696e+02 4.617e+02 9.164e+02, threshold=7.392e+02, percent-clipped=2.0 2023-05-02 18:22:35,917 INFO [train.py:904] (1/8) Epoch 29, batch 5700, loss[loss=0.233, simple_loss=0.2983, pruned_loss=0.08387, over 11103.00 frames. ], tot_loss[loss=0.2038, simple_loss=0.2897, pruned_loss=0.05897, over 3073960.39 frames. ], batch size: 248, lr: 2.31e-03, grad_scale: 8.0 2023-05-02 18:23:01,286 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=289920.0, num_to_drop=1, layers_to_drop={1} 2023-05-02 18:23:19,998 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([5.0785, 5.0949, 4.8483, 4.1626, 5.0335, 1.9015, 4.7501, 4.5758], device='cuda:1'), covar=tensor([0.0095, 0.0091, 0.0203, 0.0423, 0.0093, 0.3058, 0.0130, 0.0255], device='cuda:1'), in_proj_covar=tensor([0.0185, 0.0179, 0.0218, 0.0189, 0.0194, 0.0221, 0.0205, 0.0184], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-02 18:23:55,707 INFO [train.py:904] (1/8) Epoch 29, batch 5750, loss[loss=0.1976, simple_loss=0.2861, pruned_loss=0.05452, over 16423.00 frames. ], tot_loss[loss=0.2062, simple_loss=0.292, pruned_loss=0.06022, over 3056033.27 frames. ], batch size: 75, lr: 2.31e-03, grad_scale: 8.0 2023-05-02 18:24:32,111 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.72 vs. limit=2.0 2023-05-02 18:24:40,786 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=289981.0, num_to_drop=1, layers_to_drop={3} 2023-05-02 18:24:56,848 INFO [optim.py:368] (1/8) 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 18:25:08,669 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=289998.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 18:25:20,969 INFO [train.py:904] (1/8) Epoch 29, batch 5800, loss[loss=0.169, simple_loss=0.2687, pruned_loss=0.03467, over 16907.00 frames. ], tot_loss[loss=0.2049, simple_loss=0.2916, pruned_loss=0.0591, over 3045479.76 frames. ], batch size: 96, lr: 2.31e-03, grad_scale: 8.0 2023-05-02 18:26:31,196 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.3842, 4.5208, 4.6751, 4.4608, 4.5147, 5.0338, 4.5965, 4.3164], device='cuda:1'), covar=tensor([0.1574, 0.1841, 0.2235, 0.1990, 0.2387, 0.1010, 0.1580, 0.2363], device='cuda:1'), in_proj_covar=tensor([0.0431, 0.0638, 0.0704, 0.0517, 0.0692, 0.0727, 0.0548, 0.0691], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-05-02 18:26:34,248 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 29, batch 5850, loss[loss=0.183, simple_loss=0.2748, pruned_loss=0.04564, over 16906.00 frames. ], tot_loss[loss=0.2019, simple_loss=0.2891, pruned_loss=0.0573, over 3043675.22 frames. ], batch size: 116, lr: 2.31e-03, grad_scale: 8.0 2023-05-02 18:26:48,246 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=290059.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 18:26:51,340 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([1.9255, 3.1166, 3.1728, 1.9987, 2.9935, 3.2014, 3.0773, 1.8278], device='cuda:1'), covar=tensor([0.0658, 0.0103, 0.0104, 0.0552, 0.0134, 0.0162, 0.0131, 0.0605], device='cuda:1'), in_proj_covar=tensor([0.0139, 0.0091, 0.0092, 0.0137, 0.0103, 0.0117, 0.0099, 0.0132], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0003, 0.0004], device='cuda:1') 2023-05-02 18:27:16,660 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.3251, 3.9333, 3.8676, 2.4774, 3.5874, 3.9804, 3.6580, 1.9263], device='cuda:1'), covar=tensor([0.0673, 0.0097, 0.0108, 0.0567, 0.0143, 0.0188, 0.0133, 0.0707], device='cuda:1'), in_proj_covar=tensor([0.0139, 0.0091, 0.0092, 0.0137, 0.0104, 0.0117, 0.0099, 0.0132], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0003, 0.0004], device='cuda:1') 2023-05-02 18:27:20,472 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.8149, 4.1015, 3.1433, 2.4606, 2.9061, 2.7085, 4.6270, 3.6661], device='cuda:1'), covar=tensor([0.2969, 0.0729, 0.1798, 0.2730, 0.2584, 0.2084, 0.0452, 0.1353], device='cuda:1'), in_proj_covar=tensor([0.0336, 0.0275, 0.0314, 0.0329, 0.0307, 0.0278, 0.0307, 0.0353], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:1') 2023-05-02 18:27:39,839 INFO [optim.py:368] (1/8) 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 18:27:45,154 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.0101, 3.3129, 3.3858, 2.1607, 3.1503, 3.3981, 3.2285, 1.9817], device='cuda:1'), covar=tensor([0.0648, 0.0084, 0.0083, 0.0520, 0.0129, 0.0143, 0.0112, 0.0552], device='cuda:1'), in_proj_covar=tensor([0.0139, 0.0091, 0.0092, 0.0137, 0.0104, 0.0117, 0.0100, 0.0132], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0003, 0.0004], device='cuda:1') 2023-05-02 18:28:01,356 INFO [train.py:904] (1/8) Epoch 29, batch 5900, loss[loss=0.2243, simple_loss=0.2981, pruned_loss=0.0753, over 11754.00 frames. ], tot_loss[loss=0.202, simple_loss=0.289, pruned_loss=0.05747, over 3035065.17 frames. ], batch size: 248, lr: 2.31e-03, grad_scale: 8.0 2023-05-02 18:28:15,100 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=290111.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 18:29:20,949 INFO [train.py:904] (1/8) Epoch 29, batch 5950, loss[loss=0.2096, simple_loss=0.2847, pruned_loss=0.06721, over 11561.00 frames. ], tot_loss[loss=0.2012, simple_loss=0.2899, pruned_loss=0.05629, over 3043605.22 frames. ], batch size: 246, lr: 2.31e-03, grad_scale: 8.0 2023-05-02 18:29:30,144 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=290159.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 18:30:18,751 INFO [optim.py:368] (1/8) Clipping_scale=2.0, grad-norm quartiles 1.883e+02 2.594e+02 3.044e+02 3.971e+02 8.061e+02, threshold=6.088e+02, percent-clipped=1.0 2023-05-02 18:30:38,257 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.5996, 3.9005, 2.9218, 2.3587, 2.5821, 2.5252, 4.2198, 3.3281], device='cuda:1'), covar=tensor([0.3123, 0.0621, 0.1910, 0.2814, 0.2803, 0.2260, 0.0433, 0.1483], device='cuda:1'), in_proj_covar=tensor([0.0337, 0.0276, 0.0315, 0.0330, 0.0308, 0.0279, 0.0308, 0.0354], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:1') 2023-05-02 18:30:40,234 INFO [train.py:904] (1/8) Epoch 29, batch 6000, loss[loss=0.1979, simple_loss=0.2851, pruned_loss=0.05532, over 16742.00 frames. ], tot_loss[loss=0.2001, simple_loss=0.2888, pruned_loss=0.0557, over 3061834.90 frames. ], batch size: 83, lr: 2.31e-03, grad_scale: 8.0 2023-05-02 18:30:40,234 INFO [train.py:929] (1/8) Computing validation loss 2023-05-02 18:30:50,250 INFO [train.py:938] (1/8) Epoch 29, validation: loss=0.1475, simple_loss=0.2594, pruned_loss=0.01778, over 944034.00 frames. 2023-05-02 18:30:50,251 INFO [train.py:939] (1/8) Maximum memory allocated so far is 18145MB 2023-05-02 18:30:55,860 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=290207.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 18:30:59,089 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.3970, 3.3599, 3.4078, 3.4883, 3.5357, 3.3177, 3.5193, 3.5861], device='cuda:1'), covar=tensor([0.1243, 0.0965, 0.1031, 0.0623, 0.0701, 0.2403, 0.1114, 0.0886], device='cuda:1'), in_proj_covar=tensor([0.0684, 0.0834, 0.0961, 0.0847, 0.0646, 0.0667, 0.0706, 0.0822], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-02 18:31:52,405 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.0495, 2.3438, 2.3403, 2.7254, 1.9824, 3.1947, 1.8704, 2.7372], device='cuda:1'), covar=tensor([0.1144, 0.0679, 0.1080, 0.0200, 0.0108, 0.0357, 0.1433, 0.0710], device='cuda:1'), in_proj_covar=tensor([0.0173, 0.0182, 0.0201, 0.0206, 0.0208, 0.0219, 0.0211, 0.0200], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-05-02 18:32:11,209 INFO [train.py:904] (1/8) Epoch 29, batch 6050, loss[loss=0.1854, simple_loss=0.2837, pruned_loss=0.04354, over 16772.00 frames. ], tot_loss[loss=0.1986, simple_loss=0.2871, pruned_loss=0.05506, over 3068400.49 frames. ], batch size: 83, lr: 2.31e-03, grad_scale: 8.0 2023-05-02 18:32:16,139 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.9270, 4.7456, 4.9277, 5.1545, 5.3898, 4.8421, 5.3214, 5.4067], device='cuda:1'), covar=tensor([0.2485, 0.1612, 0.2263, 0.1052, 0.0812, 0.1189, 0.1002, 0.0744], device='cuda:1'), in_proj_covar=tensor([0.0684, 0.0835, 0.0961, 0.0848, 0.0646, 0.0667, 0.0707, 0.0823], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-02 18:32:42,688 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=290276.0, num_to_drop=1, layers_to_drop={2} 2023-05-02 18:33:04,446 INFO [optim.py:368] (1/8) 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 18:33:17,284 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.28 vs. limit=2.0 2023-05-02 18:33:29,590 INFO [train.py:904] (1/8) Epoch 29, batch 6100, loss[loss=0.2106, simple_loss=0.2955, pruned_loss=0.06289, over 15386.00 frames. ], tot_loss[loss=0.1974, simple_loss=0.2868, pruned_loss=0.05398, over 3092220.01 frames. ], batch size: 190, lr: 2.31e-03, grad_scale: 8.0 2023-05-02 18:33:53,064 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=290318.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 18:34:29,026 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.22 vs. limit=2.0 2023-05-02 18:34:31,293 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.5527, 4.6258, 4.9627, 4.9144, 4.9505, 4.6253, 4.6030, 4.5000], device='cuda:1'), covar=tensor([0.0353, 0.0613, 0.0394, 0.0413, 0.0450, 0.0417, 0.1024, 0.0535], device='cuda:1'), in_proj_covar=tensor([0.0440, 0.0499, 0.0483, 0.0442, 0.0528, 0.0507, 0.0585, 0.0408], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:1') 2023-05-02 18:34:46,605 INFO [train.py:904] (1/8) Epoch 29, batch 6150, loss[loss=0.2048, simple_loss=0.2894, pruned_loss=0.06005, over 15344.00 frames. ], tot_loss[loss=0.1964, simple_loss=0.2852, pruned_loss=0.05383, over 3084011.58 frames. ], batch size: 190, lr: 2.31e-03, grad_scale: 8.0 2023-05-02 18:34:47,115 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=290354.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 18:35:11,370 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([5.3448, 5.6363, 5.2446, 5.4566, 5.0929, 4.9188, 5.1917, 5.6885], device='cuda:1'), covar=tensor([0.1948, 0.1098, 0.2038, 0.1409, 0.1365, 0.1321, 0.2224, 0.1489], device='cuda:1'), in_proj_covar=tensor([0.0717, 0.0860, 0.0708, 0.0670, 0.0549, 0.0549, 0.0722, 0.0677], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:1') 2023-05-02 18:35:27,411 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=290379.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 18:35:44,982 INFO [optim.py:368] (1/8) 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 18:36:04,350 INFO [train.py:904] (1/8) Epoch 29, batch 6200, loss[loss=0.1734, simple_loss=0.2643, pruned_loss=0.04124, over 16195.00 frames. ], tot_loss[loss=0.195, simple_loss=0.2831, pruned_loss=0.05342, over 3081732.55 frames. ], batch size: 165, lr: 2.31e-03, grad_scale: 8.0 2023-05-02 18:36:08,271 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=290406.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 18:37:21,097 INFO [train.py:904] (1/8) Epoch 29, batch 6250, loss[loss=0.187, simple_loss=0.278, pruned_loss=0.04795, over 16241.00 frames. ], tot_loss[loss=0.1949, simple_loss=0.2829, pruned_loss=0.05346, over 3085526.15 frames. ], batch size: 165, lr: 2.31e-03, grad_scale: 8.0 2023-05-02 18:37:41,324 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.4382, 2.5171, 2.5317, 4.4284, 2.3967, 2.9070, 2.5548, 2.6981], device='cuda:1'), covar=tensor([0.1444, 0.3616, 0.2981, 0.0468, 0.3977, 0.2499, 0.3797, 0.3192], device='cuda:1'), in_proj_covar=tensor([0.0423, 0.0475, 0.0387, 0.0337, 0.0446, 0.0547, 0.0449, 0.0558], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-02 18:38:15,680 INFO [optim.py:368] (1/8) Clipping_scale=2.0, grad-norm quartiles 1.611e+02 2.626e+02 3.170e+02 3.673e+02 5.508e+02, threshold=6.341e+02, percent-clipped=0.0 2023-05-02 18:38:34,374 INFO [train.py:904] (1/8) Epoch 29, batch 6300, loss[loss=0.1932, simple_loss=0.282, pruned_loss=0.05218, over 16555.00 frames. ], tot_loss[loss=0.1937, simple_loss=0.2827, pruned_loss=0.05236, over 3115527.21 frames. ], batch size: 68, lr: 2.31e-03, grad_scale: 8.0 2023-05-02 18:38:38,246 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.7344, 2.4266, 1.9548, 2.0870, 2.7254, 2.3230, 2.5519, 2.8451], device='cuda:1'), covar=tensor([0.0259, 0.0464, 0.0644, 0.0573, 0.0292, 0.0430, 0.0270, 0.0292], device='cuda:1'), in_proj_covar=tensor([0.0232, 0.0244, 0.0234, 0.0235, 0.0245, 0.0244, 0.0240, 0.0244], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-02 18:39:03,933 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.1273, 2.4055, 2.5289, 1.9910, 2.6872, 2.7788, 2.4484, 2.3670], device='cuda:1'), covar=tensor([0.0747, 0.0300, 0.0265, 0.0951, 0.0139, 0.0309, 0.0499, 0.0493], device='cuda:1'), in_proj_covar=tensor([0.0150, 0.0111, 0.0103, 0.0139, 0.0088, 0.0133, 0.0130, 0.0131], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-05-02 18:39:52,799 INFO [train.py:904] (1/8) Epoch 29, batch 6350, loss[loss=0.1865, simple_loss=0.2752, pruned_loss=0.04893, over 16755.00 frames. ], tot_loss[loss=0.1947, simple_loss=0.2828, pruned_loss=0.0533, over 3103762.45 frames. ], batch size: 89, lr: 2.31e-03, grad_scale: 8.0 2023-05-02 18:40:28,259 INFO [zipformer.py:625] (1/8) 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] (1/8) Clipping_scale=2.0, grad-norm quartiles 1.932e+02 2.799e+02 3.423e+02 4.274e+02 8.913e+02, threshold=6.847e+02, percent-clipped=6.0 2023-05-02 18:41:09,297 INFO [train.py:904] (1/8) Epoch 29, batch 6400, loss[loss=0.1834, simple_loss=0.2632, pruned_loss=0.05174, over 16230.00 frames. ], tot_loss[loss=0.1961, simple_loss=0.2833, pruned_loss=0.05443, over 3104975.79 frames. ], batch size: 35, lr: 2.31e-03, grad_scale: 8.0 2023-05-02 18:41:17,204 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.4385, 3.5325, 3.9026, 2.0791, 3.2544, 2.4024, 3.8089, 3.9492], device='cuda:1'), covar=tensor([0.0242, 0.0895, 0.0555, 0.2329, 0.0851, 0.1076, 0.0607, 0.0920], device='cuda:1'), in_proj_covar=tensor([0.0161, 0.0171, 0.0171, 0.0157, 0.0148, 0.0133, 0.0146, 0.0184], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0005], device='cuda:1') 2023-05-02 18:41:39,256 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=290624.0, num_to_drop=1, layers_to_drop={0} 2023-05-02 18:42:23,292 INFO [train.py:904] (1/8) Epoch 29, batch 6450, loss[loss=0.2119, simple_loss=0.3016, pruned_loss=0.0611, over 17058.00 frames. ], tot_loss[loss=0.1968, simple_loss=0.2841, pruned_loss=0.05472, over 3086042.05 frames. ], batch size: 53, lr: 2.31e-03, grad_scale: 8.0 2023-05-02 18:42:23,721 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=290654.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 18:42:38,820 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([5.5207, 5.5264, 5.3208, 4.4985, 5.4640, 1.9337, 5.1658, 4.9852], device='cuda:1'), covar=tensor([0.0149, 0.0142, 0.0210, 0.0476, 0.0120, 0.3018, 0.0184, 0.0243], device='cuda:1'), in_proj_covar=tensor([0.0184, 0.0179, 0.0217, 0.0189, 0.0194, 0.0221, 0.0205, 0.0183], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-02 18:42:53,362 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=290674.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 18:43:19,913 INFO [optim.py:368] (1/8) 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 18:43:37,901 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=290702.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 18:43:40,038 INFO [train.py:904] (1/8) Epoch 29, batch 6500, loss[loss=0.2103, simple_loss=0.2913, pruned_loss=0.06461, over 15349.00 frames. ], tot_loss[loss=0.1961, simple_loss=0.2835, pruned_loss=0.05432, over 3099897.23 frames. ], batch size: 190, lr: 2.31e-03, grad_scale: 8.0 2023-05-02 18:43:44,248 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=290706.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 18:44:05,224 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.5863, 4.8780, 4.6568, 4.6672, 4.3818, 4.3809, 4.3230, 4.9404], device='cuda:1'), covar=tensor([0.1185, 0.0821, 0.0976, 0.0961, 0.0780, 0.1307, 0.1183, 0.0805], device='cuda:1'), in_proj_covar=tensor([0.0722, 0.0865, 0.0714, 0.0674, 0.0552, 0.0553, 0.0727, 0.0681], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:1') 2023-05-02 18:44:31,049 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.22 vs. limit=2.0 2023-05-02 18:44:59,497 INFO [train.py:904] (1/8) Epoch 29, batch 6550, loss[loss=0.2032, simple_loss=0.3073, pruned_loss=0.04958, over 16911.00 frames. ], tot_loss[loss=0.1976, simple_loss=0.2858, pruned_loss=0.05476, over 3089789.48 frames. ], batch size: 116, lr: 2.31e-03, grad_scale: 8.0 2023-05-02 18:44:59,830 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=290754.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 18:45:58,154 INFO [optim.py:368] (1/8) 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 18:46:19,475 INFO [train.py:904] (1/8) Epoch 29, batch 6600, loss[loss=0.2083, simple_loss=0.2907, pruned_loss=0.06295, over 16645.00 frames. ], tot_loss[loss=0.1988, simple_loss=0.2874, pruned_loss=0.05508, over 3094914.64 frames. ], batch size: 57, lr: 2.31e-03, grad_scale: 8.0 2023-05-02 18:47:38,810 INFO [train.py:904] (1/8) Epoch 29, batch 6650, loss[loss=0.199, simple_loss=0.2759, pruned_loss=0.06108, over 16661.00 frames. ], tot_loss[loss=0.1996, simple_loss=0.2874, pruned_loss=0.05586, over 3099622.88 frames. ], batch size: 57, lr: 2.31e-03, grad_scale: 8.0 2023-05-02 18:48:35,967 INFO [optim.py:368] (1/8) 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] (1/8) Epoch 29, batch 6700, loss[loss=0.2156, simple_loss=0.295, pruned_loss=0.06812, over 16875.00 frames. ], tot_loss[loss=0.1991, simple_loss=0.2862, pruned_loss=0.05595, over 3102347.67 frames. ], batch size: 116, lr: 2.31e-03, grad_scale: 8.0 2023-05-02 18:50:12,971 INFO [train.py:904] (1/8) Epoch 29, batch 6750, loss[loss=0.1884, simple_loss=0.2706, pruned_loss=0.05304, over 17131.00 frames. ], tot_loss[loss=0.1988, simple_loss=0.2856, pruned_loss=0.05605, over 3104033.79 frames. ], batch size: 48, lr: 2.31e-03, grad_scale: 8.0 2023-05-02 18:50:43,272 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=290974.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 18:51:10,927 INFO [optim.py:368] (1/8) Clipping_scale=2.0, grad-norm quartiles 1.985e+02 2.623e+02 3.095e+02 3.882e+02 8.048e+02, threshold=6.190e+02, percent-clipped=2.0 2023-05-02 18:51:28,425 INFO [train.py:904] (1/8) Epoch 29, batch 6800, loss[loss=0.2323, simple_loss=0.2996, pruned_loss=0.08248, over 11491.00 frames. ], tot_loss[loss=0.2001, simple_loss=0.2861, pruned_loss=0.05699, over 3085617.51 frames. ], batch size: 248, lr: 2.31e-03, grad_scale: 8.0 2023-05-02 18:51:57,163 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=291022.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 18:52:03,252 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.61 vs. limit=2.0 2023-05-02 18:52:46,127 INFO [train.py:904] (1/8) Epoch 29, batch 6850, loss[loss=0.2011, simple_loss=0.303, pruned_loss=0.04961, over 16340.00 frames. ], tot_loss[loss=0.2008, simple_loss=0.2871, pruned_loss=0.05721, over 3087923.07 frames. ], batch size: 35, lr: 2.31e-03, grad_scale: 8.0 2023-05-02 18:53:09,415 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=291069.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 18:53:28,407 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.3019, 3.3855, 2.1001, 3.6058, 2.6222, 3.5976, 2.1093, 2.6332], device='cuda:1'), covar=tensor([0.0309, 0.0384, 0.1742, 0.0299, 0.0885, 0.0690, 0.1720, 0.0871], device='cuda:1'), in_proj_covar=tensor([0.0177, 0.0181, 0.0196, 0.0174, 0.0180, 0.0221, 0.0204, 0.0183], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-05-02 18:53:42,652 INFO [optim.py:368] (1/8) 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 18:54:02,728 INFO [train.py:904] (1/8) Epoch 29, batch 6900, loss[loss=0.1887, simple_loss=0.2814, pruned_loss=0.04802, over 16520.00 frames. ], tot_loss[loss=0.2005, simple_loss=0.2887, pruned_loss=0.05612, over 3103443.75 frames. ], batch size: 75, lr: 2.31e-03, grad_scale: 8.0 2023-05-02 18:54:44,537 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 29, batch 6950, loss[loss=0.178, simple_loss=0.2704, pruned_loss=0.04281, over 16734.00 frames. ], tot_loss[loss=0.2019, simple_loss=0.2894, pruned_loss=0.05716, over 3109825.38 frames. ], batch size: 83, lr: 2.31e-03, grad_scale: 8.0 2023-05-02 18:56:18,587 INFO [optim.py:368] (1/8) Clipping_scale=2.0, grad-norm quartiles 1.816e+02 3.069e+02 3.631e+02 4.410e+02 8.633e+02, threshold=7.261e+02, percent-clipped=4.0 2023-05-02 18:56:36,625 INFO [train.py:904] (1/8) Epoch 29, batch 7000, loss[loss=0.2014, simple_loss=0.2789, pruned_loss=0.06193, over 11493.00 frames. ], tot_loss[loss=0.2012, simple_loss=0.2896, pruned_loss=0.05638, over 3116943.72 frames. ], batch size: 248, lr: 2.31e-03, grad_scale: 8.0 2023-05-02 18:56:41,501 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.44 vs. limit=2.0 2023-05-02 18:57:04,035 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([1.9341, 3.3529, 3.3921, 2.0448, 3.1529, 3.4207, 3.2284, 1.9419], device='cuda:1'), covar=tensor([0.0672, 0.0081, 0.0080, 0.0532, 0.0124, 0.0129, 0.0116, 0.0546], device='cuda:1'), in_proj_covar=tensor([0.0140, 0.0091, 0.0092, 0.0137, 0.0104, 0.0117, 0.0100, 0.0132], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0003, 0.0004], device='cuda:1') 2023-05-02 18:57:50,283 INFO [train.py:904] (1/8) Epoch 29, batch 7050, loss[loss=0.215, simple_loss=0.3028, pruned_loss=0.06357, over 16738.00 frames. ], tot_loss[loss=0.2017, simple_loss=0.2907, pruned_loss=0.05632, over 3120124.98 frames. ], batch size: 124, lr: 2.31e-03, grad_scale: 8.0 2023-05-02 18:57:55,834 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.3483, 3.4235, 2.1470, 3.7979, 2.6382, 3.8050, 2.1931, 2.7107], device='cuda:1'), covar=tensor([0.0320, 0.0356, 0.1616, 0.0204, 0.0883, 0.0541, 0.1598, 0.0836], device='cuda:1'), in_proj_covar=tensor([0.0176, 0.0181, 0.0196, 0.0174, 0.0180, 0.0221, 0.0204, 0.0183], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-05-02 18:58:49,444 INFO [optim.py:368] (1/8) 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 18:59:07,327 INFO [train.py:904] (1/8) Epoch 29, batch 7100, loss[loss=0.1796, simple_loss=0.2724, pruned_loss=0.04345, over 17201.00 frames. ], tot_loss[loss=0.2018, simple_loss=0.2899, pruned_loss=0.05689, over 3089441.02 frames. ], batch size: 44, lr: 2.31e-03, grad_scale: 8.0 2023-05-02 18:59:19,273 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.3349, 3.3172, 2.4270, 2.1370, 2.2298, 2.1668, 3.4028, 2.8660], device='cuda:1'), covar=tensor([0.3691, 0.1079, 0.2520, 0.3136, 0.3024, 0.2712, 0.0800, 0.1777], device='cuda:1'), in_proj_covar=tensor([0.0337, 0.0276, 0.0315, 0.0329, 0.0307, 0.0279, 0.0306, 0.0353], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:1') 2023-05-02 19:00:25,886 INFO [train.py:904] (1/8) Epoch 29, batch 7150, loss[loss=0.1921, simple_loss=0.2843, pruned_loss=0.04996, over 16751.00 frames. ], tot_loss[loss=0.2006, simple_loss=0.288, pruned_loss=0.05664, over 3097745.09 frames. ], batch size: 83, lr: 2.31e-03, grad_scale: 8.0 2023-05-02 19:01:23,566 INFO [optim.py:368] (1/8) 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 19:01:41,836 INFO [train.py:904] (1/8) Epoch 29, batch 7200, loss[loss=0.2071, simple_loss=0.2904, pruned_loss=0.06196, over 11884.00 frames. ], tot_loss[loss=0.1986, simple_loss=0.2862, pruned_loss=0.05552, over 3072130.82 frames. ], batch size: 246, lr: 2.31e-03, grad_scale: 8.0 2023-05-02 19:01:46,003 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([5.0216, 5.0382, 4.7444, 3.8135, 4.9122, 1.8906, 4.6068, 4.4558], device='cuda:1'), covar=tensor([0.0111, 0.0109, 0.0268, 0.0604, 0.0123, 0.3185, 0.0173, 0.0332], device='cuda:1'), in_proj_covar=tensor([0.0182, 0.0176, 0.0215, 0.0186, 0.0191, 0.0219, 0.0202, 0.0180], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-02 19:02:14,495 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=291425.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 19:02:40,001 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.5369, 4.6205, 4.9017, 4.8521, 4.9024, 4.6054, 4.6067, 4.5122], device='cuda:1'), covar=tensor([0.0336, 0.0551, 0.0357, 0.0414, 0.0447, 0.0376, 0.0917, 0.0469], device='cuda:1'), in_proj_covar=tensor([0.0438, 0.0498, 0.0478, 0.0440, 0.0524, 0.0505, 0.0583, 0.0405], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:1') 2023-05-02 19:02:40,146 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([1.7165, 2.5950, 2.4704, 4.3304, 2.9973, 3.8841, 1.4917, 2.8349], device='cuda:1'), covar=tensor([0.1398, 0.0857, 0.1315, 0.0184, 0.0259, 0.0441, 0.1769, 0.0880], device='cuda:1'), in_proj_covar=tensor([0.0174, 0.0182, 0.0203, 0.0206, 0.0209, 0.0220, 0.0213, 0.0202], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-05-02 19:03:00,077 INFO [train.py:904] (1/8) Epoch 29, batch 7250, loss[loss=0.1789, simple_loss=0.2619, pruned_loss=0.04794, over 17015.00 frames. ], tot_loss[loss=0.1949, simple_loss=0.2829, pruned_loss=0.05344, over 3093494.45 frames. ], batch size: 55, lr: 2.31e-03, grad_scale: 4.0 2023-05-02 19:03:06,858 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.7406, 3.0087, 3.2402, 2.0504, 2.8868, 2.1683, 3.2842, 3.2697], device='cuda:1'), covar=tensor([0.0291, 0.0928, 0.0603, 0.2105, 0.0896, 0.1050, 0.0693, 0.1085], device='cuda:1'), in_proj_covar=tensor([0.0161, 0.0173, 0.0172, 0.0158, 0.0149, 0.0134, 0.0147, 0.0185], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0005], device='cuda:1') 2023-05-02 19:03:58,887 INFO [optim.py:368] (1/8) 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 19:04:16,116 INFO [train.py:904] (1/8) Epoch 29, batch 7300, loss[loss=0.272, simple_loss=0.3256, pruned_loss=0.1092, over 11341.00 frames. ], tot_loss[loss=0.1946, simple_loss=0.2826, pruned_loss=0.05328, over 3089797.96 frames. ], batch size: 248, lr: 2.31e-03, grad_scale: 4.0 2023-05-02 19:05:25,685 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.99 vs. limit=2.0 2023-05-02 19:05:32,025 INFO [train.py:904] (1/8) Epoch 29, batch 7350, loss[loss=0.1961, simple_loss=0.2834, pruned_loss=0.05441, over 16504.00 frames. ], tot_loss[loss=0.1954, simple_loss=0.2835, pruned_loss=0.05369, over 3095725.27 frames. ], batch size: 75, lr: 2.31e-03, grad_scale: 4.0 2023-05-02 19:06:32,138 INFO [optim.py:368] (1/8) 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 19:06:49,563 INFO [train.py:904] (1/8) Epoch 29, batch 7400, loss[loss=0.2058, simple_loss=0.2935, pruned_loss=0.0591, over 15242.00 frames. ], tot_loss[loss=0.1965, simple_loss=0.2847, pruned_loss=0.05413, over 3107051.46 frames. ], batch size: 190, lr: 2.31e-03, grad_scale: 4.0 2023-05-02 19:06:53,627 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.22 vs. limit=2.0 2023-05-02 19:08:07,280 INFO [train.py:904] (1/8) Epoch 29, batch 7450, loss[loss=0.2077, simple_loss=0.3036, pruned_loss=0.05589, over 16163.00 frames. ], tot_loss[loss=0.1977, simple_loss=0.2854, pruned_loss=0.05503, over 3117678.04 frames. ], batch size: 165, lr: 2.31e-03, grad_scale: 4.0 2023-05-02 19:09:04,504 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.5501, 3.5389, 3.5127, 2.7426, 3.4214, 2.0417, 3.2453, 2.9777], device='cuda:1'), covar=tensor([0.0191, 0.0175, 0.0214, 0.0275, 0.0134, 0.2602, 0.0163, 0.0293], device='cuda:1'), in_proj_covar=tensor([0.0182, 0.0176, 0.0215, 0.0186, 0.0191, 0.0218, 0.0202, 0.0179], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-02 19:09:10,902 INFO [optim.py:368] (1/8) Clipping_scale=2.0, grad-norm quartiles 1.991e+02 2.789e+02 3.286e+02 3.823e+02 6.627e+02, threshold=6.571e+02, percent-clipped=0.0 2023-05-02 19:09:24,855 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.66 vs. limit=2.0 2023-05-02 19:09:25,607 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([1.7052, 1.7899, 1.5552, 1.4257, 1.9184, 1.5324, 1.4753, 1.8631], device='cuda:1'), covar=tensor([0.0261, 0.0338, 0.0485, 0.0447, 0.0252, 0.0349, 0.0200, 0.0254], device='cuda:1'), in_proj_covar=tensor([0.0230, 0.0244, 0.0234, 0.0234, 0.0244, 0.0241, 0.0239, 0.0242], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-02 19:09:28,109 INFO [train.py:904] (1/8) Epoch 29, batch 7500, loss[loss=0.1802, simple_loss=0.2596, pruned_loss=0.05037, over 17024.00 frames. ], tot_loss[loss=0.1975, simple_loss=0.2859, pruned_loss=0.05453, over 3098166.82 frames. ], batch size: 55, lr: 2.31e-03, grad_scale: 4.0 2023-05-02 19:10:01,610 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=291725.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 19:10:25,140 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.91 vs. limit=2.0 2023-05-02 19:10:42,354 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.74 vs. limit=2.0 2023-05-02 19:10:45,759 INFO [train.py:904] (1/8) Epoch 29, batch 7550, loss[loss=0.1638, simple_loss=0.2565, pruned_loss=0.03549, over 16716.00 frames. ], tot_loss[loss=0.1975, simple_loss=0.2853, pruned_loss=0.05488, over 3095645.00 frames. ], batch size: 89, lr: 2.31e-03, grad_scale: 4.0 2023-05-02 19:11:15,343 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=291773.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 19:11:19,691 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=291776.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 19:11:28,336 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.2237, 2.3497, 2.4528, 3.9029, 2.2874, 2.7198, 2.4321, 2.4900], device='cuda:1'), covar=tensor([0.1445, 0.3480, 0.2953, 0.0606, 0.4279, 0.2456, 0.3480, 0.3378], device='cuda:1'), in_proj_covar=tensor([0.0423, 0.0476, 0.0386, 0.0337, 0.0446, 0.0546, 0.0449, 0.0558], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-02 19:11:44,714 INFO [optim.py:368] (1/8) 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 19:12:01,405 INFO [train.py:904] (1/8) Epoch 29, batch 7600, loss[loss=0.1895, simple_loss=0.2743, pruned_loss=0.05238, over 16736.00 frames. ], tot_loss[loss=0.1975, simple_loss=0.2846, pruned_loss=0.05525, over 3087082.64 frames. ], batch size: 124, lr: 2.30e-03, grad_scale: 8.0 2023-05-02 19:12:52,107 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=291837.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 19:13:12,956 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.5690, 2.6398, 2.2247, 2.4385, 3.0348, 2.6731, 3.1079, 3.2609], device='cuda:1'), covar=tensor([0.0140, 0.0463, 0.0592, 0.0485, 0.0267, 0.0404, 0.0258, 0.0267], device='cuda:1'), in_proj_covar=tensor([0.0229, 0.0243, 0.0233, 0.0233, 0.0243, 0.0240, 0.0238, 0.0241], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-02 19:13:18,970 INFO [train.py:904] (1/8) Epoch 29, batch 7650, loss[loss=0.1863, simple_loss=0.2756, pruned_loss=0.04855, over 16814.00 frames. ], tot_loss[loss=0.1983, simple_loss=0.2848, pruned_loss=0.05593, over 3069751.56 frames. ], batch size: 116, lr: 2.30e-03, grad_scale: 4.0 2023-05-02 19:14:00,001 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.61 vs. limit=2.0 2023-05-02 19:14:20,910 INFO [optim.py:368] (1/8) Clipping_scale=2.0, grad-norm quartiles 1.976e+02 2.915e+02 3.313e+02 4.275e+02 8.487e+02, threshold=6.627e+02, percent-clipped=5.0 2023-05-02 19:14:36,023 INFO [train.py:904] (1/8) Epoch 29, batch 7700, loss[loss=0.1934, simple_loss=0.2807, pruned_loss=0.05308, over 16415.00 frames. ], tot_loss[loss=0.1986, simple_loss=0.2849, pruned_loss=0.05615, over 3081007.26 frames. ], batch size: 35, lr: 2.30e-03, grad_scale: 4.0 2023-05-02 19:15:53,493 INFO [train.py:904] (1/8) Epoch 29, batch 7750, loss[loss=0.1723, simple_loss=0.2725, pruned_loss=0.03608, over 16884.00 frames. ], tot_loss[loss=0.1983, simple_loss=0.2851, pruned_loss=0.05576, over 3079635.81 frames. ], batch size: 96, lr: 2.30e-03, grad_scale: 4.0 2023-05-02 19:16:37,866 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.0302, 2.1972, 2.2222, 3.5710, 2.1294, 2.5121, 2.2894, 2.3436], device='cuda:1'), covar=tensor([0.1593, 0.3678, 0.3279, 0.0683, 0.4442, 0.2616, 0.3844, 0.3310], device='cuda:1'), in_proj_covar=tensor([0.0423, 0.0476, 0.0387, 0.0337, 0.0447, 0.0547, 0.0449, 0.0558], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-02 19:16:42,164 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.4993, 3.5745, 3.3503, 2.9537, 3.2333, 3.4610, 3.3570, 3.3164], device='cuda:1'), covar=tensor([0.0623, 0.0713, 0.0309, 0.0305, 0.0484, 0.0510, 0.1130, 0.0497], device='cuda:1'), in_proj_covar=tensor([0.0309, 0.0468, 0.0361, 0.0362, 0.0358, 0.0417, 0.0249, 0.0434], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002], device='cuda:1') 2023-05-02 19:16:55,756 INFO [optim.py:368] (1/8) 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 19:17:14,569 INFO [train.py:904] (1/8) Epoch 29, batch 7800, loss[loss=0.2177, simple_loss=0.2918, pruned_loss=0.07182, over 11423.00 frames. ], tot_loss[loss=0.1997, simple_loss=0.2861, pruned_loss=0.05667, over 3077785.10 frames. ], batch size: 248, lr: 2.30e-03, grad_scale: 4.0 2023-05-02 19:18:09,645 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.4205, 4.7126, 4.5135, 4.5100, 4.2559, 4.1974, 4.2367, 4.7450], device='cuda:1'), covar=tensor([0.1265, 0.0912, 0.1030, 0.0915, 0.0844, 0.1503, 0.1131, 0.0929], device='cuda:1'), in_proj_covar=tensor([0.0722, 0.0866, 0.0715, 0.0675, 0.0553, 0.0556, 0.0726, 0.0683], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:1') 2023-05-02 19:18:30,015 INFO [train.py:904] (1/8) Epoch 29, batch 7850, loss[loss=0.1756, simple_loss=0.2769, pruned_loss=0.03719, over 16857.00 frames. ], tot_loss[loss=0.1992, simple_loss=0.2862, pruned_loss=0.05611, over 3075377.70 frames. ], batch size: 102, lr: 2.30e-03, grad_scale: 2.0 2023-05-02 19:18:30,451 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=292054.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 19:18:32,228 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.6039, 3.6443, 1.8241, 4.1717, 2.7026, 4.0437, 2.0510, 2.8612], device='cuda:1'), covar=tensor([0.0328, 0.0402, 0.2274, 0.0260, 0.0911, 0.0630, 0.2135, 0.0893], device='cuda:1'), in_proj_covar=tensor([0.0177, 0.0182, 0.0198, 0.0175, 0.0182, 0.0222, 0.0206, 0.0185], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-05-02 19:19:30,095 INFO [optim.py:368] (1/8) 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 19:19:43,448 INFO [train.py:904] (1/8) Epoch 29, batch 7900, loss[loss=0.2331, simple_loss=0.3167, pruned_loss=0.07474, over 15426.00 frames. ], tot_loss[loss=0.1982, simple_loss=0.2855, pruned_loss=0.05543, over 3089117.51 frames. ], batch size: 191, lr: 2.30e-03, grad_scale: 2.0 2023-05-02 19:19:57,927 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.47 vs. limit=2.0 2023-05-02 19:19:59,935 INFO [zipformer.py:625] (1/8) 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] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=292132.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 19:21:01,246 INFO [train.py:904] (1/8) Epoch 29, batch 7950, loss[loss=0.2173, simple_loss=0.294, pruned_loss=0.07027, over 17008.00 frames. ], tot_loss[loss=0.1986, simple_loss=0.2854, pruned_loss=0.05584, over 3089639.32 frames. ], batch size: 55, lr: 2.30e-03, grad_scale: 2.0 2023-05-02 19:21:12,965 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.23 vs. limit=2.0 2023-05-02 19:21:13,943 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.8654, 4.0278, 2.5681, 4.8059, 3.1587, 4.6980, 2.7554, 3.2889], device='cuda:1'), covar=tensor([0.0327, 0.0375, 0.1624, 0.0224, 0.0824, 0.0545, 0.1408, 0.0774], device='cuda:1'), in_proj_covar=tensor([0.0176, 0.0181, 0.0197, 0.0174, 0.0181, 0.0221, 0.0205, 0.0184], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-05-02 19:22:01,848 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.4789, 3.5593, 2.7012, 2.1491, 2.2884, 2.2900, 3.7715, 3.1506], device='cuda:1'), covar=tensor([0.3304, 0.0724, 0.2055, 0.3368, 0.3143, 0.2481, 0.0554, 0.1454], device='cuda:1'), in_proj_covar=tensor([0.0338, 0.0276, 0.0316, 0.0329, 0.0308, 0.0279, 0.0306, 0.0353], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:1') 2023-05-02 19:22:03,657 INFO [optim.py:368] (1/8) 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 19:22:18,472 INFO [train.py:904] (1/8) Epoch 29, batch 8000, loss[loss=0.1823, simple_loss=0.27, pruned_loss=0.04727, over 17087.00 frames. ], tot_loss[loss=0.1994, simple_loss=0.2863, pruned_loss=0.0562, over 3089946.66 frames. ], batch size: 49, lr: 2.30e-03, grad_scale: 4.0 2023-05-02 19:23:31,107 INFO [train.py:904] (1/8) Epoch 29, batch 8050, loss[loss=0.1787, simple_loss=0.2727, pruned_loss=0.04233, over 17111.00 frames. ], tot_loss[loss=0.1986, simple_loss=0.2857, pruned_loss=0.05573, over 3097504.36 frames. ], batch size: 47, lr: 2.30e-03, grad_scale: 4.0 2023-05-02 19:23:59,585 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.8349, 2.1468, 2.4453, 3.1201, 2.2142, 2.3180, 2.3390, 2.2745], device='cuda:1'), covar=tensor([0.1527, 0.3631, 0.2726, 0.0764, 0.4191, 0.2765, 0.3359, 0.3553], device='cuda:1'), in_proj_covar=tensor([0.0423, 0.0477, 0.0387, 0.0337, 0.0448, 0.0549, 0.0449, 0.0558], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-02 19:24:32,479 INFO [optim.py:368] (1/8) 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] (1/8) Epoch 29, batch 8100, loss[loss=0.2086, simple_loss=0.2951, pruned_loss=0.06104, over 12043.00 frames. ], tot_loss[loss=0.1982, simple_loss=0.2855, pruned_loss=0.05546, over 3087722.99 frames. ], batch size: 247, lr: 2.30e-03, grad_scale: 4.0 2023-05-02 19:26:01,419 INFO [train.py:904] (1/8) Epoch 29, batch 8150, loss[loss=0.2282, simple_loss=0.2977, pruned_loss=0.07939, over 11775.00 frames. ], tot_loss[loss=0.1972, simple_loss=0.2839, pruned_loss=0.05524, over 3070241.65 frames. ], batch size: 247, lr: 2.30e-03, grad_scale: 4.0 2023-05-02 19:26:40,458 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.50 vs. limit=2.0 2023-05-02 19:27:01,239 INFO [optim.py:368] (1/8) 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 19:27:15,046 INFO [train.py:904] (1/8) Epoch 29, batch 8200, loss[loss=0.1552, simple_loss=0.2563, pruned_loss=0.02709, over 16831.00 frames. ], tot_loss[loss=0.195, simple_loss=0.281, pruned_loss=0.05448, over 3073657.06 frames. ], batch size: 102, lr: 2.30e-03, grad_scale: 4.0 2023-05-02 19:27:25,890 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=292410.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 19:27:51,043 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.9284, 5.2807, 5.4499, 5.2053, 5.2995, 5.8209, 5.3623, 5.0727], device='cuda:1'), covar=tensor([0.0992, 0.1991, 0.2370, 0.2052, 0.2417, 0.0952, 0.1594, 0.2456], device='cuda:1'), in_proj_covar=tensor([0.0430, 0.0640, 0.0709, 0.0517, 0.0692, 0.0730, 0.0550, 0.0694], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-05-02 19:28:00,216 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=292432.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 19:28:17,711 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.0517, 2.3571, 2.4639, 1.9630, 2.6349, 2.6659, 2.5039, 2.4799], device='cuda:1'), covar=tensor([0.0697, 0.0283, 0.0363, 0.1029, 0.0196, 0.0366, 0.0438, 0.0436], device='cuda:1'), in_proj_covar=tensor([0.0149, 0.0111, 0.0103, 0.0139, 0.0088, 0.0132, 0.0130, 0.0130], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-05-02 19:28:34,807 INFO [train.py:904] (1/8) Epoch 29, batch 8250, loss[loss=0.1741, simple_loss=0.2643, pruned_loss=0.04193, over 12236.00 frames. ], tot_loss[loss=0.1918, simple_loss=0.2796, pruned_loss=0.05199, over 3054786.04 frames. ], batch size: 247, lr: 2.30e-03, grad_scale: 4.0 2023-05-02 19:29:17,989 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=292480.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 19:29:41,434 INFO [optim.py:368] (1/8) Clipping_scale=2.0, grad-norm quartiles 1.407e+02 2.184e+02 2.675e+02 3.477e+02 6.200e+02, threshold=5.351e+02, percent-clipped=0.0 2023-05-02 19:29:45,229 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.3366, 3.3908, 3.4129, 2.4723, 3.1646, 3.4665, 3.2872, 2.0875], device='cuda:1'), covar=tensor([0.0501, 0.0095, 0.0080, 0.0378, 0.0142, 0.0137, 0.0102, 0.0552], device='cuda:1'), in_proj_covar=tensor([0.0138, 0.0090, 0.0091, 0.0135, 0.0103, 0.0116, 0.0099, 0.0131], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0003, 0.0004], device='cuda:1') 2023-05-02 19:29:46,740 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.84 vs. limit=2.0 2023-05-02 19:29:55,807 INFO [train.py:904] (1/8) Epoch 29, batch 8300, loss[loss=0.1473, simple_loss=0.2386, pruned_loss=0.02799, over 12367.00 frames. ], tot_loss[loss=0.1875, simple_loss=0.277, pruned_loss=0.04903, over 3055331.03 frames. ], batch size: 250, lr: 2.30e-03, grad_scale: 4.0 2023-05-02 19:31:15,559 INFO [train.py:904] (1/8) Epoch 29, batch 8350, loss[loss=0.1723, simple_loss=0.2763, pruned_loss=0.03412, over 16400.00 frames. ], tot_loss[loss=0.1855, simple_loss=0.2766, pruned_loss=0.0472, over 3053471.93 frames. ], batch size: 146, lr: 2.30e-03, grad_scale: 4.0 2023-05-02 19:32:20,504 INFO [optim.py:368] (1/8) 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 19:32:34,779 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.4211, 4.3824, 4.2184, 3.3264, 4.2585, 1.7155, 4.0256, 3.9352], device='cuda:1'), covar=tensor([0.0135, 0.0131, 0.0249, 0.0453, 0.0149, 0.3191, 0.0188, 0.0345], device='cuda:1'), in_proj_covar=tensor([0.0182, 0.0175, 0.0214, 0.0185, 0.0190, 0.0219, 0.0202, 0.0179], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-02 19:32:36,081 INFO [train.py:904] (1/8) Epoch 29, batch 8400, loss[loss=0.1733, simple_loss=0.274, pruned_loss=0.03633, over 16678.00 frames. ], tot_loss[loss=0.182, simple_loss=0.2738, pruned_loss=0.04512, over 3049024.59 frames. ], batch size: 134, lr: 2.30e-03, grad_scale: 8.0 2023-05-02 19:32:52,470 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.5547, 2.8999, 3.2706, 1.8959, 2.8529, 2.1354, 3.1386, 3.1220], device='cuda:1'), covar=tensor([0.0294, 0.0977, 0.0605, 0.2358, 0.0870, 0.1136, 0.0684, 0.0995], device='cuda:1'), in_proj_covar=tensor([0.0159, 0.0169, 0.0169, 0.0156, 0.0147, 0.0132, 0.0144, 0.0181], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-05-02 19:33:07,595 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.73 vs. limit=2.0 2023-05-02 19:33:53,569 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.3396, 4.4292, 4.2261, 3.9260, 3.9701, 4.3437, 4.0316, 4.0744], device='cuda:1'), covar=tensor([0.0630, 0.0601, 0.0346, 0.0352, 0.0765, 0.0506, 0.0727, 0.0716], device='cuda:1'), in_proj_covar=tensor([0.0308, 0.0468, 0.0361, 0.0361, 0.0356, 0.0416, 0.0248, 0.0432], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-02 19:33:56,325 INFO [train.py:904] (1/8) Epoch 29, batch 8450, loss[loss=0.1901, simple_loss=0.2839, pruned_loss=0.04812, over 16627.00 frames. ], tot_loss[loss=0.1793, simple_loss=0.2722, pruned_loss=0.0432, over 3065989.64 frames. ], batch size: 134, lr: 2.30e-03, grad_scale: 8.0 2023-05-02 19:34:24,685 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.5439, 3.5409, 2.8413, 2.1784, 2.2288, 2.3851, 3.6787, 3.0864], device='cuda:1'), covar=tensor([0.3070, 0.0594, 0.1816, 0.3386, 0.3149, 0.2436, 0.0455, 0.1412], device='cuda:1'), in_proj_covar=tensor([0.0333, 0.0272, 0.0310, 0.0325, 0.0303, 0.0276, 0.0301, 0.0347], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:1') 2023-05-02 19:35:03,479 INFO [optim.py:368] (1/8) 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 19:35:19,250 INFO [train.py:904] (1/8) Epoch 29, batch 8500, loss[loss=0.1517, simple_loss=0.252, pruned_loss=0.02568, over 16446.00 frames. ], tot_loss[loss=0.1756, simple_loss=0.2688, pruned_loss=0.04116, over 3065272.77 frames. ], batch size: 68, lr: 2.30e-03, grad_scale: 8.0 2023-05-02 19:35:28,835 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=292710.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 19:36:43,005 INFO [train.py:904] (1/8) Epoch 29, batch 8550, loss[loss=0.2007, simple_loss=0.3015, pruned_loss=0.04993, over 16702.00 frames. ], tot_loss[loss=0.174, simple_loss=0.2671, pruned_loss=0.04045, over 3052050.24 frames. ], batch size: 134, lr: 2.30e-03, grad_scale: 8.0 2023-05-02 19:36:52,466 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=292758.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 19:36:54,484 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=4.96 vs. limit=5.0 2023-05-02 19:38:01,155 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.0294, 2.7106, 2.9712, 2.1359, 2.6878, 2.1883, 2.7193, 2.8947], device='cuda:1'), covar=tensor([0.0303, 0.0970, 0.0522, 0.2040, 0.0859, 0.1002, 0.0681, 0.0816], device='cuda:1'), in_proj_covar=tensor([0.0159, 0.0168, 0.0168, 0.0155, 0.0146, 0.0132, 0.0144, 0.0180], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:1') 2023-05-02 19:38:04,313 INFO [optim.py:368] (1/8) 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 19:38:21,504 INFO [train.py:904] (1/8) Epoch 29, batch 8600, loss[loss=0.1695, simple_loss=0.2533, pruned_loss=0.04283, over 12737.00 frames. ], tot_loss[loss=0.1727, simple_loss=0.2666, pruned_loss=0.03942, over 3039926.16 frames. ], batch size: 248, lr: 2.30e-03, grad_scale: 8.0 2023-05-02 19:40:01,867 INFO [train.py:904] (1/8) Epoch 29, batch 8650, loss[loss=0.1651, simple_loss=0.2648, pruned_loss=0.03268, over 15256.00 frames. ], tot_loss[loss=0.171, simple_loss=0.2653, pruned_loss=0.03836, over 3046807.77 frames. ], batch size: 190, lr: 2.30e-03, grad_scale: 8.0 2023-05-02 19:41:31,353 INFO [optim.py:368] (1/8) 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 19:41:48,478 INFO [train.py:904] (1/8) Epoch 29, batch 8700, loss[loss=0.1516, simple_loss=0.2513, pruned_loss=0.02594, over 16899.00 frames. ], tot_loss[loss=0.1687, simple_loss=0.263, pruned_loss=0.03721, over 3065647.90 frames. ], batch size: 102, lr: 2.30e-03, grad_scale: 8.0 2023-05-02 19:43:26,847 INFO [train.py:904] (1/8) Epoch 29, batch 8750, loss[loss=0.161, simple_loss=0.266, pruned_loss=0.02802, over 16719.00 frames. ], tot_loss[loss=0.1682, simple_loss=0.2628, pruned_loss=0.03683, over 3060907.68 frames. ], batch size: 76, lr: 2.30e-03, grad_scale: 8.0 2023-05-02 19:44:01,880 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=292967.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 19:45:00,945 INFO [optim.py:368] (1/8) Clipping_scale=2.0, grad-norm quartiles 1.419e+02 2.165e+02 2.592e+02 3.174e+02 6.684e+02, threshold=5.184e+02, percent-clipped=4.0 2023-05-02 19:45:17,889 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.0010, 2.6843, 2.9471, 2.0700, 2.7055, 2.1482, 2.6873, 2.8731], device='cuda:1'), covar=tensor([0.0315, 0.1029, 0.0522, 0.2032, 0.0852, 0.1025, 0.0636, 0.0847], device='cuda:1'), in_proj_covar=tensor([0.0158, 0.0167, 0.0168, 0.0155, 0.0146, 0.0131, 0.0143, 0.0179], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:1') 2023-05-02 19:45:20,127 INFO [train.py:904] (1/8) Epoch 29, batch 8800, loss[loss=0.1791, simple_loss=0.2729, pruned_loss=0.0427, over 16939.00 frames. ], tot_loss[loss=0.1668, simple_loss=0.2617, pruned_loss=0.03597, over 3067348.08 frames. ], batch size: 109, lr: 2.30e-03, grad_scale: 8.0 2023-05-02 19:46:11,651 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=293028.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 19:46:15,915 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.4429, 4.5933, 4.7416, 4.5167, 4.5708, 5.0668, 4.5980, 4.3894], device='cuda:1'), covar=tensor([0.1470, 0.1612, 0.1639, 0.1982, 0.2407, 0.0954, 0.1551, 0.2479], device='cuda:1'), in_proj_covar=tensor([0.0417, 0.0623, 0.0693, 0.0506, 0.0676, 0.0712, 0.0538, 0.0676], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-05-02 19:46:28,938 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.4303, 4.5044, 4.7612, 4.7480, 4.7685, 4.5433, 4.5183, 4.4067], device='cuda:1'), covar=tensor([0.0298, 0.0525, 0.0396, 0.0370, 0.0383, 0.0365, 0.0789, 0.0461], device='cuda:1'), in_proj_covar=tensor([0.0430, 0.0488, 0.0469, 0.0434, 0.0513, 0.0494, 0.0569, 0.0397], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:1') 2023-05-02 19:47:05,852 INFO [train.py:904] (1/8) Epoch 29, batch 8850, loss[loss=0.1578, simple_loss=0.2622, pruned_loss=0.02675, over 15382.00 frames. ], tot_loss[loss=0.1673, simple_loss=0.264, pruned_loss=0.03533, over 3057956.86 frames. ], batch size: 191, lr: 2.30e-03, grad_scale: 8.0 2023-05-02 19:48:05,752 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=293081.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 19:48:35,604 INFO [optim.py:368] (1/8) 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 19:48:55,719 INFO [train.py:904] (1/8) Epoch 29, batch 8900, loss[loss=0.1828, simple_loss=0.2735, pruned_loss=0.04604, over 16970.00 frames. ], tot_loss[loss=0.1673, simple_loss=0.2644, pruned_loss=0.03508, over 3058827.56 frames. ], batch size: 109, lr: 2.30e-03, grad_scale: 8.0 2023-05-02 19:49:23,409 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=293117.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 19:50:34,637 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=293142.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 19:50:57,008 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.91 vs. limit=2.0 2023-05-02 19:51:01,970 INFO [train.py:904] (1/8) Epoch 29, batch 8950, loss[loss=0.151, simple_loss=0.2483, pruned_loss=0.02681, over 16825.00 frames. ], tot_loss[loss=0.167, simple_loss=0.2637, pruned_loss=0.03514, over 3067474.72 frames. ], batch size: 124, lr: 2.30e-03, grad_scale: 8.0 2023-05-02 19:51:52,457 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=293178.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 19:52:32,341 INFO [optim.py:368] (1/8) 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] (1/8) Epoch 29, batch 9000, loss[loss=0.1439, simple_loss=0.2399, pruned_loss=0.02397, over 16340.00 frames. ], tot_loss[loss=0.1641, simple_loss=0.2603, pruned_loss=0.03389, over 3068586.72 frames. ], batch size: 146, lr: 2.30e-03, grad_scale: 8.0 2023-05-02 19:52:53,145 INFO [train.py:929] (1/8) Computing validation loss 2023-05-02 19:53:02,751 INFO [train.py:938] (1/8) Epoch 29, validation: loss=0.1431, simple_loss=0.2465, pruned_loss=0.01987, over 944034.00 frames. 2023-05-02 19:53:02,751 INFO [train.py:939] (1/8) Maximum memory allocated so far is 18145MB 2023-05-02 19:53:47,419 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.3226, 3.9187, 3.9060, 2.5514, 3.4822, 3.9749, 3.6867, 2.0515], device='cuda:1'), covar=tensor([0.0608, 0.0055, 0.0052, 0.0455, 0.0110, 0.0078, 0.0077, 0.0607], device='cuda:1'), in_proj_covar=tensor([0.0136, 0.0088, 0.0090, 0.0133, 0.0101, 0.0114, 0.0097, 0.0129], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:1') 2023-05-02 19:54:23,836 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.32 vs. limit=2.0 2023-05-02 19:54:47,961 INFO [train.py:904] (1/8) Epoch 29, batch 9050, loss[loss=0.1561, simple_loss=0.2546, pruned_loss=0.02885, over 15489.00 frames. ], tot_loss[loss=0.1642, simple_loss=0.2605, pruned_loss=0.03392, over 3067377.20 frames. ], batch size: 191, lr: 2.30e-03, grad_scale: 8.0 2023-05-02 19:55:12,463 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.0589, 1.8853, 1.7202, 1.5370, 2.0202, 1.6856, 1.5554, 1.9794], device='cuda:1'), covar=tensor([0.0199, 0.0348, 0.0464, 0.0416, 0.0265, 0.0320, 0.0171, 0.0229], device='cuda:1'), in_proj_covar=tensor([0.0224, 0.0240, 0.0228, 0.0229, 0.0239, 0.0237, 0.0233, 0.0236], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-02 19:55:35,400 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.5463, 4.8781, 4.6894, 4.7216, 4.3960, 4.3589, 4.3696, 4.9384], device='cuda:1'), covar=tensor([0.1302, 0.0909, 0.1019, 0.0796, 0.0817, 0.1552, 0.1227, 0.0858], device='cuda:1'), in_proj_covar=tensor([0.0707, 0.0852, 0.0699, 0.0660, 0.0542, 0.0544, 0.0713, 0.0670], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:1') 2023-05-02 19:55:55,612 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.69 vs. limit=2.0 2023-05-02 19:56:15,025 INFO [optim.py:368] (1/8) 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 19:56:29,436 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.8405, 3.8928, 4.1366, 4.1288, 4.1401, 3.9428, 3.9294, 3.9727], device='cuda:1'), covar=tensor([0.0364, 0.0663, 0.0459, 0.0435, 0.0468, 0.0451, 0.0823, 0.0461], device='cuda:1'), in_proj_covar=tensor([0.0430, 0.0488, 0.0471, 0.0434, 0.0515, 0.0495, 0.0569, 0.0398], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:1') 2023-05-02 19:56:34,919 INFO [train.py:904] (1/8) Epoch 29, batch 9100, loss[loss=0.1602, simple_loss=0.249, pruned_loss=0.03565, over 12310.00 frames. ], tot_loss[loss=0.1644, simple_loss=0.2602, pruned_loss=0.03434, over 3075833.48 frames. ], batch size: 247, lr: 2.30e-03, grad_scale: 8.0 2023-05-02 19:57:17,187 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=293323.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 19:58:34,114 INFO [train.py:904] (1/8) Epoch 29, batch 9150, loss[loss=0.1614, simple_loss=0.2533, pruned_loss=0.03472, over 11846.00 frames. ], tot_loss[loss=0.1649, simple_loss=0.2609, pruned_loss=0.03447, over 3059265.05 frames. ], batch size: 248, lr: 2.30e-03, grad_scale: 8.0 2023-05-02 19:58:44,071 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.4515, 3.4105, 2.5140, 2.1505, 2.1315, 2.1358, 3.4297, 2.8398], device='cuda:1'), covar=tensor([0.3428, 0.0775, 0.2339, 0.3279, 0.3083, 0.2791, 0.0634, 0.1718], device='cuda:1'), in_proj_covar=tensor([0.0331, 0.0270, 0.0308, 0.0323, 0.0299, 0.0274, 0.0300, 0.0344], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:1') 2023-05-02 19:58:51,958 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=293361.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 19:58:56,741 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=293364.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 19:58:59,348 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.6670, 2.5930, 1.8921, 2.7548, 2.1401, 2.7807, 2.1032, 2.3757], device='cuda:1'), covar=tensor([0.0371, 0.0378, 0.1424, 0.0297, 0.0804, 0.0469, 0.1406, 0.0677], device='cuda:1'), in_proj_covar=tensor([0.0172, 0.0174, 0.0191, 0.0166, 0.0175, 0.0213, 0.0199, 0.0177], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-05-02 19:59:02,894 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.51 vs. limit=2.0 2023-05-02 20:00:04,471 INFO [optim.py:368] (1/8) 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] (1/8) Epoch 29, batch 9200, loss[loss=0.1726, simple_loss=0.2594, pruned_loss=0.04293, over 11849.00 frames. ], tot_loss[loss=0.1623, simple_loss=0.2571, pruned_loss=0.03376, over 3058906.20 frames. ], batch size: 246, lr: 2.30e-03, grad_scale: 8.0 2023-05-02 20:00:54,528 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=293422.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 20:01:00,076 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=293425.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 20:01:23,096 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=293437.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 20:01:36,498 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.28 vs. limit=2.0 2023-05-02 20:01:58,342 INFO [train.py:904] (1/8) Epoch 29, batch 9250, loss[loss=0.1563, simple_loss=0.2501, pruned_loss=0.03123, over 16352.00 frames. ], tot_loss[loss=0.1621, simple_loss=0.2565, pruned_loss=0.03379, over 3050760.06 frames. ], batch size: 165, lr: 2.30e-03, grad_scale: 8.0 2023-05-02 20:02:39,017 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=293473.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 20:03:04,233 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.90 vs. limit=2.0 2023-05-02 20:03:27,234 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.2523, 4.1270, 4.3372, 4.4451, 4.5771, 4.2276, 4.5605, 4.6124], device='cuda:1'), covar=tensor([0.1965, 0.1410, 0.1584, 0.0846, 0.0653, 0.1017, 0.0771, 0.0849], device='cuda:1'), in_proj_covar=tensor([0.0658, 0.0801, 0.0921, 0.0822, 0.0625, 0.0643, 0.0682, 0.0797], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-02 20:03:28,486 INFO [optim.py:368] (1/8) 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 20:03:49,264 INFO [train.py:904] (1/8) Epoch 29, batch 9300, loss[loss=0.1646, simple_loss=0.2478, pruned_loss=0.04065, over 12293.00 frames. ], tot_loss[loss=0.1605, simple_loss=0.2544, pruned_loss=0.03334, over 3024676.39 frames. ], batch size: 248, lr: 2.30e-03, grad_scale: 8.0 2023-05-02 20:05:33,313 INFO [train.py:904] (1/8) Epoch 29, batch 9350, loss[loss=0.1753, simple_loss=0.268, pruned_loss=0.04127, over 16960.00 frames. ], tot_loss[loss=0.16, simple_loss=0.2541, pruned_loss=0.03297, over 3029300.64 frames. ], batch size: 109, lr: 2.30e-03, grad_scale: 8.0 2023-05-02 20:06:21,757 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.33 vs. limit=2.0 2023-05-02 20:06:35,200 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.5623, 4.6977, 4.8714, 4.6161, 4.7162, 5.1993, 4.6818, 4.3668], device='cuda:1'), covar=tensor([0.1235, 0.1705, 0.1873, 0.1881, 0.1990, 0.0815, 0.1441, 0.2611], device='cuda:1'), in_proj_covar=tensor([0.0413, 0.0618, 0.0687, 0.0501, 0.0670, 0.0709, 0.0533, 0.0670], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-05-02 20:06:56,970 INFO [optim.py:368] (1/8) 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 20:07:15,889 INFO [train.py:904] (1/8) Epoch 29, batch 9400, loss[loss=0.1643, simple_loss=0.2692, pruned_loss=0.02968, over 16228.00 frames. ], tot_loss[loss=0.1606, simple_loss=0.2551, pruned_loss=0.03302, over 3028718.66 frames. ], batch size: 165, lr: 2.30e-03, grad_scale: 8.0 2023-05-02 20:07:53,837 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=293623.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 20:08:55,065 INFO [train.py:904] (1/8) Epoch 29, batch 9450, loss[loss=0.149, simple_loss=0.2512, pruned_loss=0.02339, over 16920.00 frames. ], tot_loss[loss=0.1615, simple_loss=0.2568, pruned_loss=0.03311, over 3030305.45 frames. ], batch size: 116, lr: 2.30e-03, grad_scale: 8.0 2023-05-02 20:09:29,418 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=293671.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 20:09:32,021 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.0408, 2.7080, 3.0160, 2.1939, 2.7874, 2.1983, 2.8257, 2.9036], device='cuda:1'), covar=tensor([0.0289, 0.0967, 0.0480, 0.1833, 0.0762, 0.0996, 0.0564, 0.0851], device='cuda:1'), in_proj_covar=tensor([0.0156, 0.0165, 0.0166, 0.0153, 0.0144, 0.0130, 0.0142, 0.0177], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:1') 2023-05-02 20:09:43,325 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.9700, 2.0769, 2.2131, 3.4826, 2.0052, 2.3032, 2.2298, 2.2080], device='cuda:1'), covar=tensor([0.1551, 0.4024, 0.3389, 0.0726, 0.4877, 0.3056, 0.4225, 0.3911], device='cuda:1'), in_proj_covar=tensor([0.0411, 0.0464, 0.0379, 0.0327, 0.0436, 0.0531, 0.0438, 0.0544], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-02 20:10:18,284 INFO [optim.py:368] (1/8) 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] (1/8) Epoch 29, batch 9500, loss[loss=0.1508, simple_loss=0.2472, pruned_loss=0.02724, over 15488.00 frames. ], tot_loss[loss=0.1607, simple_loss=0.2561, pruned_loss=0.03262, over 3053660.33 frames. ], batch size: 191, lr: 2.30e-03, grad_scale: 8.0 2023-05-02 20:10:40,715 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.2876, 4.3966, 4.4973, 4.3070, 4.3834, 4.8593, 4.3632, 4.0663], device='cuda:1'), covar=tensor([0.1674, 0.1842, 0.2173, 0.1943, 0.2330, 0.0950, 0.1661, 0.2745], device='cuda:1'), in_proj_covar=tensor([0.0413, 0.0618, 0.0687, 0.0501, 0.0670, 0.0709, 0.0533, 0.0669], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-05-02 20:11:02,445 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=293717.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 20:11:08,938 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=293720.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 20:11:41,951 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=293737.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 20:12:10,973 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.7726, 4.7209, 4.5141, 3.6125, 4.6436, 1.6018, 4.3435, 4.1536], device='cuda:1'), covar=tensor([0.0120, 0.0131, 0.0256, 0.0448, 0.0120, 0.3347, 0.0170, 0.0352], device='cuda:1'), in_proj_covar=tensor([0.0177, 0.0170, 0.0207, 0.0179, 0.0185, 0.0214, 0.0196, 0.0174], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-02 20:12:19,484 INFO [train.py:904] (1/8) Epoch 29, batch 9550, loss[loss=0.1735, simple_loss=0.2742, pruned_loss=0.03644, over 15312.00 frames. ], tot_loss[loss=0.1608, simple_loss=0.2561, pruned_loss=0.0328, over 3057780.18 frames. ], batch size: 191, lr: 2.30e-03, grad_scale: 8.0 2023-05-02 20:12:59,190 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=293773.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 20:13:24,951 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=293785.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 20:13:35,660 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.4568, 2.0488, 1.8113, 1.6994, 2.3094, 1.9396, 1.8180, 2.3293], device='cuda:1'), covar=tensor([0.0215, 0.0483, 0.0564, 0.0604, 0.0330, 0.0448, 0.0223, 0.0306], device='cuda:1'), in_proj_covar=tensor([0.0221, 0.0238, 0.0227, 0.0227, 0.0236, 0.0235, 0.0231, 0.0234], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-02 20:13:43,637 INFO [optim.py:368] (1/8) 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 20:13:55,622 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.4404, 4.5081, 4.3228, 4.0000, 4.0745, 4.4236, 4.1771, 4.1653], device='cuda:1'), covar=tensor([0.0570, 0.0550, 0.0345, 0.0317, 0.0785, 0.0500, 0.0631, 0.0649], device='cuda:1'), in_proj_covar=tensor([0.0303, 0.0456, 0.0354, 0.0353, 0.0348, 0.0406, 0.0244, 0.0421], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-02 20:13:59,135 INFO [train.py:904] (1/8) Epoch 29, batch 9600, loss[loss=0.1784, simple_loss=0.2718, pruned_loss=0.04245, over 16604.00 frames. ], tot_loss[loss=0.1632, simple_loss=0.2583, pruned_loss=0.03403, over 3057377.95 frames. ], batch size: 57, lr: 2.30e-03, grad_scale: 8.0 2023-05-02 20:14:32,484 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=293821.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 20:14:59,746 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.24 vs. limit=2.0 2023-05-02 20:15:45,739 INFO [train.py:904] (1/8) Epoch 29, batch 9650, loss[loss=0.1705, simple_loss=0.2672, pruned_loss=0.03692, over 16870.00 frames. ], tot_loss[loss=0.1641, simple_loss=0.26, pruned_loss=0.03412, over 3061274.58 frames. ], batch size: 124, lr: 2.30e-03, grad_scale: 8.0 2023-05-02 20:17:16,475 INFO [optim.py:368] (1/8) 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] (1/8) Epoch 29, batch 9700, loss[loss=0.1534, simple_loss=0.2526, pruned_loss=0.02715, over 16600.00 frames. ], tot_loss[loss=0.164, simple_loss=0.2593, pruned_loss=0.03431, over 3060201.34 frames. ], batch size: 89, lr: 2.30e-03, grad_scale: 8.0 2023-05-02 20:19:17,209 INFO [train.py:904] (1/8) Epoch 29, batch 9750, loss[loss=0.1673, simple_loss=0.2647, pruned_loss=0.03499, over 16786.00 frames. ], tot_loss[loss=0.1636, simple_loss=0.2585, pruned_loss=0.03429, over 3070411.98 frames. ], batch size: 124, lr: 2.30e-03, grad_scale: 8.0 2023-05-02 20:20:08,082 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=4.57 vs. limit=5.0 2023-05-02 20:20:24,580 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=293987.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 20:20:38,043 INFO [optim.py:368] (1/8) Clipping_scale=2.0, grad-norm quartiles 1.578e+02 2.282e+02 2.658e+02 3.259e+02 1.207e+03, threshold=5.315e+02, percent-clipped=3.0 2023-05-02 20:20:56,835 INFO [train.py:904] (1/8) Epoch 29, batch 9800, loss[loss=0.1574, simple_loss=0.2631, pruned_loss=0.02587, over 16199.00 frames. ], tot_loss[loss=0.1627, simple_loss=0.2589, pruned_loss=0.03322, over 3099083.71 frames. ], batch size: 165, lr: 2.30e-03, grad_scale: 8.0 2023-05-02 20:20:59,711 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.5864, 3.5333, 3.5012, 2.7588, 3.4211, 2.1237, 3.2479, 2.8488], device='cuda:1'), covar=tensor([0.0127, 0.0122, 0.0175, 0.0181, 0.0100, 0.2597, 0.0119, 0.0270], device='cuda:1'), in_proj_covar=tensor([0.0177, 0.0170, 0.0208, 0.0179, 0.0185, 0.0214, 0.0196, 0.0174], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-02 20:21:22,358 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=294017.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 20:21:27,432 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=294020.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 20:21:40,400 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.25 vs. limit=2.0 2023-05-02 20:21:45,314 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.5778, 3.3480, 3.7607, 1.9814, 3.8861, 3.9737, 3.0642, 3.0512], device='cuda:1'), covar=tensor([0.0766, 0.0300, 0.0195, 0.1199, 0.0086, 0.0161, 0.0432, 0.0473], device='cuda:1'), in_proj_covar=tensor([0.0144, 0.0108, 0.0098, 0.0135, 0.0084, 0.0127, 0.0126, 0.0126], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0004, 0.0003], device='cuda:1') 2023-05-02 20:22:29,329 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=294048.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 20:22:39,768 INFO [train.py:904] (1/8) Epoch 29, batch 9850, loss[loss=0.159, simple_loss=0.2438, pruned_loss=0.03707, over 12539.00 frames. ], tot_loss[loss=0.1631, simple_loss=0.2597, pruned_loss=0.03323, over 3077852.17 frames. ], batch size: 249, lr: 2.30e-03, grad_scale: 16.0 2023-05-02 20:23:02,393 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=294065.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 20:23:09,661 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=294068.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 20:24:14,911 INFO [optim.py:368] (1/8) 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] (1/8) Epoch 29, batch 9900, loss[loss=0.1678, simple_loss=0.2746, pruned_loss=0.03054, over 16676.00 frames. ], tot_loss[loss=0.1627, simple_loss=0.2596, pruned_loss=0.03291, over 3060965.28 frames. ], batch size: 83, lr: 2.30e-03, grad_scale: 8.0 2023-05-02 20:26:11,499 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.0469, 4.1185, 3.9661, 3.6750, 3.7162, 4.0329, 3.7284, 3.8067], device='cuda:1'), covar=tensor([0.0557, 0.0659, 0.0358, 0.0316, 0.0772, 0.0521, 0.1077, 0.0594], device='cuda:1'), in_proj_covar=tensor([0.0301, 0.0454, 0.0352, 0.0352, 0.0347, 0.0404, 0.0243, 0.0420], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-02 20:26:27,774 INFO [train.py:904] (1/8) Epoch 29, batch 9950, loss[loss=0.1604, simple_loss=0.2686, pruned_loss=0.02611, over 16362.00 frames. ], tot_loss[loss=0.1642, simple_loss=0.2615, pruned_loss=0.03345, over 3061211.68 frames. ], batch size: 146, lr: 2.30e-03, grad_scale: 8.0 2023-05-02 20:27:06,747 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=294170.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 20:28:08,140 INFO [optim.py:368] (1/8) 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 20:28:14,612 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.3249, 3.4194, 2.0921, 3.6304, 2.5899, 3.6739, 2.1583, 2.7187], device='cuda:1'), covar=tensor([0.0316, 0.0337, 0.1588, 0.0291, 0.0817, 0.0445, 0.1668, 0.0750], device='cuda:1'), in_proj_covar=tensor([0.0171, 0.0174, 0.0189, 0.0165, 0.0174, 0.0211, 0.0199, 0.0177], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0003, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-05-02 20:28:27,677 INFO [train.py:904] (1/8) Epoch 29, batch 10000, loss[loss=0.1546, simple_loss=0.2545, pruned_loss=0.02728, over 15289.00 frames. ], tot_loss[loss=0.1636, simple_loss=0.2607, pruned_loss=0.03327, over 3073818.49 frames. ], batch size: 190, lr: 2.30e-03, grad_scale: 8.0 2023-05-02 20:29:21,507 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=294231.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 20:30:08,528 INFO [train.py:904] (1/8) Epoch 29, batch 10050, loss[loss=0.1735, simple_loss=0.2775, pruned_loss=0.03468, over 16198.00 frames. ], tot_loss[loss=0.1637, simple_loss=0.2607, pruned_loss=0.03331, over 3057728.65 frames. ], batch size: 165, lr: 2.30e-03, grad_scale: 8.0 2023-05-02 20:31:27,889 INFO [optim.py:368] (1/8) Clipping_scale=2.0, grad-norm quartiles 1.622e+02 2.082e+02 2.498e+02 3.055e+02 5.920e+02, threshold=4.997e+02, percent-clipped=2.0 2023-05-02 20:31:35,485 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.7304, 2.6927, 2.3137, 2.5215, 2.9799, 2.6233, 3.1073, 3.2081], device='cuda:1'), covar=tensor([0.0164, 0.0518, 0.0646, 0.0573, 0.0368, 0.0529, 0.0279, 0.0329], device='cuda:1'), in_proj_covar=tensor([0.0219, 0.0237, 0.0226, 0.0227, 0.0235, 0.0234, 0.0229, 0.0233], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-02 20:31:41,187 INFO [train.py:904] (1/8) Epoch 29, batch 10100, loss[loss=0.1608, simple_loss=0.2474, pruned_loss=0.03716, over 12574.00 frames. ], tot_loss[loss=0.1633, simple_loss=0.2606, pruned_loss=0.033, over 3075087.40 frames. ], batch size: 248, lr: 2.30e-03, grad_scale: 8.0 2023-05-02 20:32:42,414 INFO [zipformer.py:625] (1/8) 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] (1/8) 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] (1/8) Epoch 30, batch 0, loss[loss=0.2097, simple_loss=0.2944, pruned_loss=0.06249, over 16258.00 frames. ], tot_loss[loss=0.2097, simple_loss=0.2944, pruned_loss=0.06249, over 16258.00 frames. ], batch size: 165, lr: 2.26e-03, grad_scale: 8.0 2023-05-02 20:33:20,747 INFO [train.py:929] (1/8) Computing validation loss 2023-05-02 20:33:28,210 INFO [train.py:938] (1/8) Epoch 30, validation: loss=0.1426, simple_loss=0.2458, pruned_loss=0.01974, over 944034.00 frames. 2023-05-02 20:33:28,211 INFO [train.py:939] (1/8) Maximum memory allocated so far is 18145MB 2023-05-02 20:34:28,806 INFO [optim.py:368] (1/8) 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] (1/8) 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] (1/8) Epoch 30, batch 50, loss[loss=0.1635, simple_loss=0.2376, pruned_loss=0.04466, over 16833.00 frames. ], tot_loss[loss=0.1771, simple_loss=0.2637, pruned_loss=0.04522, over 753001.98 frames. ], batch size: 102, lr: 2.26e-03, grad_scale: 1.0 2023-05-02 20:35:01,531 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=294423.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 20:35:43,886 INFO [train.py:904] (1/8) Epoch 30, batch 100, loss[loss=0.1701, simple_loss=0.2574, pruned_loss=0.04142, over 16397.00 frames. ], tot_loss[loss=0.1714, simple_loss=0.2598, pruned_loss=0.04153, over 1327186.49 frames. ], batch size: 75, lr: 2.26e-03, grad_scale: 1.0 2023-05-02 20:36:24,493 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=294484.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 20:36:32,168 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=294488.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 20:36:46,207 INFO [optim.py:368] (1/8) 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] (1/8) Epoch 30, batch 150, loss[loss=0.2002, simple_loss=0.2853, pruned_loss=0.05761, over 16272.00 frames. ], tot_loss[loss=0.17, simple_loss=0.2577, pruned_loss=0.04111, over 1772035.74 frames. ], batch size: 165, lr: 2.25e-03, grad_scale: 1.0 2023-05-02 20:37:20,779 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=294526.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 20:37:22,150 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.9154, 2.9750, 3.2739, 2.0681, 2.8531, 2.1950, 3.3508, 3.3789], device='cuda:1'), covar=tensor([0.0274, 0.1057, 0.0647, 0.2133, 0.0930, 0.1149, 0.0623, 0.0991], device='cuda:1'), in_proj_covar=tensor([0.0158, 0.0166, 0.0168, 0.0155, 0.0145, 0.0130, 0.0143, 0.0179], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:1') 2023-05-02 20:37:53,453 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 30, batch 200, loss[loss=0.1798, simple_loss=0.2754, pruned_loss=0.04205, over 17096.00 frames. ], tot_loss[loss=0.1689, simple_loss=0.257, pruned_loss=0.04043, over 2117476.80 frames. ], batch size: 53, lr: 2.25e-03, grad_scale: 1.0 2023-05-02 20:38:23,435 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=294571.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 20:38:25,662 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.0269, 3.8739, 4.0690, 4.1857, 4.2604, 3.8534, 4.0813, 4.2717], device='cuda:1'), covar=tensor([0.1686, 0.1190, 0.1252, 0.0726, 0.0633, 0.1543, 0.2587, 0.0727], device='cuda:1'), in_proj_covar=tensor([0.0664, 0.0812, 0.0933, 0.0829, 0.0631, 0.0650, 0.0690, 0.0802], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-02 20:39:01,687 INFO [optim.py:368] (1/8) 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 20:39:06,655 INFO [train.py:904] (1/8) Epoch 30, batch 250, loss[loss=0.1456, simple_loss=0.2437, pruned_loss=0.02373, over 17127.00 frames. ], tot_loss[loss=0.1694, simple_loss=0.2562, pruned_loss=0.04129, over 2382877.70 frames. ], batch size: 48, lr: 2.25e-03, grad_scale: 1.0 2023-05-02 20:39:12,461 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([5.0468, 4.7382, 5.0332, 5.2048, 5.4511, 4.7385, 5.3940, 5.4154], device='cuda:1'), covar=tensor([0.2050, 0.1554, 0.1938, 0.0901, 0.0596, 0.0968, 0.0611, 0.0619], device='cuda:1'), in_proj_covar=tensor([0.0666, 0.0814, 0.0936, 0.0830, 0.0632, 0.0651, 0.0692, 0.0804], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-02 20:39:45,370 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=294632.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 20:40:01,638 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=294643.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 20:40:15,313 INFO [train.py:904] (1/8) Epoch 30, batch 300, loss[loss=0.1709, simple_loss=0.2514, pruned_loss=0.04517, over 16668.00 frames. ], tot_loss[loss=0.1676, simple_loss=0.2545, pruned_loss=0.04038, over 2585971.13 frames. ], batch size: 89, lr: 2.25e-03, grad_scale: 1.0 2023-05-02 20:40:44,144 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.91 vs. limit=2.0 2023-05-02 20:41:05,815 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.6737, 3.8634, 2.7956, 2.2472, 2.4474, 2.2964, 3.8983, 3.2007], device='cuda:1'), covar=tensor([0.3116, 0.0570, 0.2084, 0.3422, 0.3042, 0.2598, 0.0540, 0.1658], device='cuda:1'), in_proj_covar=tensor([0.0335, 0.0273, 0.0313, 0.0327, 0.0303, 0.0278, 0.0302, 0.0350], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:1') 2023-05-02 20:41:05,950 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.18 vs. limit=2.0 2023-05-02 20:41:06,778 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=294691.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 20:41:12,872 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=294695.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 20:41:17,999 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([5.0232, 4.7587, 5.0321, 5.2112, 5.4376, 4.7850, 5.3972, 5.4208], device='cuda:1'), covar=tensor([0.2112, 0.1560, 0.1951, 0.0901, 0.0627, 0.0947, 0.0589, 0.0722], device='cuda:1'), in_proj_covar=tensor([0.0670, 0.0819, 0.0941, 0.0834, 0.0635, 0.0656, 0.0695, 0.0808], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-02 20:41:19,993 INFO [optim.py:368] (1/8) 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] (1/8) Epoch 30, batch 350, loss[loss=0.1729, simple_loss=0.2661, pruned_loss=0.03982, over 17078.00 frames. ], tot_loss[loss=0.1657, simple_loss=0.2522, pruned_loss=0.03957, over 2741964.90 frames. ], batch size: 53, lr: 2.25e-03, grad_scale: 1.0 2023-05-02 20:41:30,025 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.58 vs. limit=2.0 2023-05-02 20:41:43,990 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.1791, 2.3629, 2.8394, 3.1642, 2.9894, 3.6489, 2.7974, 3.5981], device='cuda:1'), covar=tensor([0.0287, 0.0569, 0.0362, 0.0357, 0.0381, 0.0220, 0.0479, 0.0214], device='cuda:1'), in_proj_covar=tensor([0.0196, 0.0197, 0.0186, 0.0190, 0.0209, 0.0165, 0.0203, 0.0166], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-02 20:42:33,013 INFO [train.py:904] (1/8) Epoch 30, batch 400, loss[loss=0.1458, simple_loss=0.238, pruned_loss=0.02683, over 16839.00 frames. ], tot_loss[loss=0.1643, simple_loss=0.2503, pruned_loss=0.03914, over 2874432.00 frames. ], batch size: 42, lr: 2.25e-03, grad_scale: 2.0 2023-05-02 20:43:06,905 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=294779.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 20:43:34,529 INFO [optim.py:368] (1/8) Clipping_scale=2.0, grad-norm quartiles 1.286e+02 2.158e+02 2.521e+02 2.985e+02 1.602e+03, threshold=5.041e+02, percent-clipped=2.0 2023-05-02 20:43:41,856 INFO [train.py:904] (1/8) Epoch 30, batch 450, loss[loss=0.1514, simple_loss=0.2314, pruned_loss=0.03572, over 16839.00 frames. ], tot_loss[loss=0.1633, simple_loss=0.2494, pruned_loss=0.03864, over 2973974.15 frames. ], batch size: 83, lr: 2.25e-03, grad_scale: 2.0 2023-05-02 20:43:49,959 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([5.7087, 6.1100, 5.8251, 5.8830, 5.4281, 5.5246, 5.4088, 6.2419], device='cuda:1'), covar=tensor([0.1526, 0.1090, 0.1166, 0.1008, 0.1073, 0.0639, 0.1439, 0.1001], device='cuda:1'), in_proj_covar=tensor([0.0719, 0.0868, 0.0714, 0.0676, 0.0554, 0.0552, 0.0728, 0.0682], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:1') 2023-05-02 20:44:12,080 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=294826.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 20:44:36,120 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 30, batch 500, loss[loss=0.1757, simple_loss=0.2485, pruned_loss=0.05145, over 16876.00 frames. ], tot_loss[loss=0.1615, simple_loss=0.2475, pruned_loss=0.03777, over 3050543.88 frames. ], batch size: 109, lr: 2.25e-03, grad_scale: 2.0 2023-05-02 20:45:17,741 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=294874.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 20:45:40,667 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.7811, 4.7889, 5.1218, 5.1263, 5.1536, 4.8629, 4.8345, 4.6931], device='cuda:1'), covar=tensor([0.0372, 0.0852, 0.0488, 0.0442, 0.0493, 0.0510, 0.0894, 0.0596], device='cuda:1'), in_proj_covar=tensor([0.0435, 0.0495, 0.0476, 0.0439, 0.0521, 0.0504, 0.0576, 0.0403], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:1') 2023-05-02 20:45:52,585 INFO [optim.py:368] (1/8) 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] (1/8) Epoch 30, batch 550, loss[loss=0.1883, simple_loss=0.2603, pruned_loss=0.05812, over 16657.00 frames. ], tot_loss[loss=0.1614, simple_loss=0.2476, pruned_loss=0.03759, over 3102357.01 frames. ], batch size: 134, lr: 2.25e-03, grad_scale: 2.0 2023-05-02 20:46:28,988 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=294927.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 20:47:05,460 INFO [train.py:904] (1/8) Epoch 30, batch 600, loss[loss=0.1587, simple_loss=0.2336, pruned_loss=0.04187, over 16674.00 frames. ], tot_loss[loss=0.161, simple_loss=0.2467, pruned_loss=0.03767, over 3151144.69 frames. ], batch size: 134, lr: 2.25e-03, grad_scale: 2.0 2023-05-02 20:47:39,897 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.25 vs. limit=2.0 2023-05-02 20:47:52,937 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=294989.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 20:48:01,953 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=294995.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 20:48:02,354 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=4.92 vs. limit=5.0 2023-05-02 20:48:07,356 INFO [optim.py:368] (1/8) 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 20:48:08,171 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.2854, 4.1276, 4.3404, 4.4593, 4.5637, 4.1461, 4.4232, 4.5471], device='cuda:1'), covar=tensor([0.1751, 0.1331, 0.1463, 0.0807, 0.0680, 0.1391, 0.2028, 0.0946], device='cuda:1'), in_proj_covar=tensor([0.0682, 0.0831, 0.0959, 0.0846, 0.0644, 0.0665, 0.0706, 0.0820], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-02 20:48:13,113 INFO [train.py:904] (1/8) Epoch 30, batch 650, loss[loss=0.1596, simple_loss=0.2351, pruned_loss=0.04209, over 16927.00 frames. ], tot_loss[loss=0.1599, simple_loss=0.2458, pruned_loss=0.03703, over 3192677.79 frames. ], batch size: 109, lr: 2.25e-03, grad_scale: 2.0 2023-05-02 20:49:04,907 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=295043.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 20:49:15,067 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=295050.0, num_to_drop=1, layers_to_drop={1} 2023-05-02 20:49:20,780 INFO [train.py:904] (1/8) Epoch 30, batch 700, loss[loss=0.156, simple_loss=0.2479, pruned_loss=0.03208, over 16644.00 frames. ], tot_loss[loss=0.1595, simple_loss=0.2455, pruned_loss=0.03673, over 3212695.80 frames. ], batch size: 62, lr: 2.25e-03, grad_scale: 2.0 2023-05-02 20:49:39,348 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=295068.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 20:49:40,426 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([5.3722, 5.3306, 5.0764, 4.5415, 5.1451, 1.9859, 4.9062, 4.8606], device='cuda:1'), covar=tensor([0.0096, 0.0099, 0.0222, 0.0411, 0.0119, 0.2913, 0.0137, 0.0272], device='cuda:1'), in_proj_covar=tensor([0.0181, 0.0174, 0.0211, 0.0182, 0.0190, 0.0218, 0.0200, 0.0179], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-02 20:49:53,840 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.50 vs. limit=2.0 2023-05-02 20:49:54,482 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=295079.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 20:50:12,101 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.8502, 4.1948, 4.2226, 3.0750, 3.6283, 4.1884, 3.7741, 2.5504], device='cuda:1'), covar=tensor([0.0480, 0.0095, 0.0065, 0.0378, 0.0159, 0.0128, 0.0121, 0.0497], device='cuda:1'), in_proj_covar=tensor([0.0140, 0.0092, 0.0093, 0.0137, 0.0104, 0.0117, 0.0100, 0.0133], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0003, 0.0004], device='cuda:1') 2023-05-02 20:50:23,317 INFO [optim.py:368] (1/8) 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] (1/8) Epoch 30, batch 750, loss[loss=0.1514, simple_loss=0.2447, pruned_loss=0.02899, over 17200.00 frames. ], tot_loss[loss=0.1602, simple_loss=0.246, pruned_loss=0.03722, over 3235964.54 frames. ], batch size: 46, lr: 2.25e-03, grad_scale: 2.0 2023-05-02 20:50:55,380 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.8193, 3.5771, 3.8405, 2.0880, 3.9362, 3.9729, 3.2150, 2.9411], device='cuda:1'), covar=tensor([0.0741, 0.0256, 0.0221, 0.1205, 0.0138, 0.0238, 0.0449, 0.0512], device='cuda:1'), in_proj_covar=tensor([0.0149, 0.0111, 0.0102, 0.0139, 0.0087, 0.0132, 0.0130, 0.0131], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-05-02 20:51:00,105 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=295127.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 20:51:03,312 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=295129.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 20:51:24,629 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.1824, 4.1690, 4.1104, 3.8618, 3.8116, 4.2115, 3.8596, 3.9230], device='cuda:1'), covar=tensor([0.0752, 0.1007, 0.0442, 0.0388, 0.0811, 0.0551, 0.1034, 0.0767], device='cuda:1'), in_proj_covar=tensor([0.0317, 0.0479, 0.0370, 0.0372, 0.0366, 0.0427, 0.0255, 0.0443], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002], device='cuda:1') 2023-05-02 20:51:24,638 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 30, batch 800, loss[loss=0.1537, simple_loss=0.2365, pruned_loss=0.03539, over 15957.00 frames. ], tot_loss[loss=0.1601, simple_loss=0.246, pruned_loss=0.03708, over 3240055.06 frames. ], batch size: 35, lr: 2.25e-03, grad_scale: 4.0 2023-05-02 20:52:29,857 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=295192.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 20:52:42,510 INFO [optim.py:368] (1/8) 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 20:52:48,746 INFO [train.py:904] (1/8) Epoch 30, batch 850, loss[loss=0.157, simple_loss=0.2574, pruned_loss=0.02826, over 17286.00 frames. ], tot_loss[loss=0.16, simple_loss=0.2457, pruned_loss=0.03712, over 3256352.55 frames. ], batch size: 52, lr: 2.25e-03, grad_scale: 4.0 2023-05-02 20:53:02,166 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.5428, 4.3310, 4.5910, 4.7242, 4.8540, 4.3809, 4.7362, 4.8207], device='cuda:1'), covar=tensor([0.1710, 0.1314, 0.1448, 0.0715, 0.0590, 0.1168, 0.2360, 0.0832], device='cuda:1'), in_proj_covar=tensor([0.0693, 0.0843, 0.0973, 0.0860, 0.0653, 0.0675, 0.0717, 0.0833], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-02 20:53:18,704 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=295225.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 20:53:21,211 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=295227.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 20:53:23,883 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=3.90 vs. limit=5.0 2023-05-02 20:53:35,978 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.3771, 4.4450, 4.6043, 4.4144, 4.4667, 5.0288, 4.5269, 4.2128], device='cuda:1'), covar=tensor([0.1978, 0.2258, 0.2675, 0.2294, 0.2843, 0.1201, 0.1964, 0.2851], device='cuda:1'), in_proj_covar=tensor([0.0433, 0.0650, 0.0725, 0.0526, 0.0703, 0.0743, 0.0557, 0.0698], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-05-02 20:53:56,732 INFO [train.py:904] (1/8) Epoch 30, batch 900, loss[loss=0.1472, simple_loss=0.2367, pruned_loss=0.02888, over 17160.00 frames. ], tot_loss[loss=0.1593, simple_loss=0.2451, pruned_loss=0.03676, over 3266465.11 frames. ], batch size: 46, lr: 2.25e-03, grad_scale: 4.0 2023-05-02 20:54:27,485 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=295275.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 20:54:33,509 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=295279.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 20:54:42,990 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=295286.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 20:55:02,819 INFO [optim.py:368] (1/8) 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 20:55:03,224 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.5606, 3.7996, 4.0084, 2.7406, 3.6067, 4.0428, 3.7001, 2.3863], device='cuda:1'), covar=tensor([0.0550, 0.0259, 0.0071, 0.0411, 0.0137, 0.0141, 0.0114, 0.0502], device='cuda:1'), in_proj_covar=tensor([0.0140, 0.0091, 0.0092, 0.0136, 0.0104, 0.0117, 0.0099, 0.0132], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0003, 0.0004], device='cuda:1') 2023-05-02 20:55:08,080 INFO [train.py:904] (1/8) Epoch 30, batch 950, loss[loss=0.1539, simple_loss=0.23, pruned_loss=0.03891, over 16756.00 frames. ], tot_loss[loss=0.1591, simple_loss=0.2453, pruned_loss=0.03647, over 3266287.38 frames. ], batch size: 124, lr: 2.25e-03, grad_scale: 4.0 2023-05-02 20:55:42,486 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([5.4777, 5.4723, 5.3807, 4.8685, 4.9915, 5.4076, 5.2959, 4.9785], device='cuda:1'), covar=tensor([0.0610, 0.0475, 0.0306, 0.0352, 0.1115, 0.0515, 0.0300, 0.0851], device='cuda:1'), in_proj_covar=tensor([0.0320, 0.0483, 0.0374, 0.0376, 0.0371, 0.0432, 0.0258, 0.0449], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-05-02 20:55:49,168 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=295333.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 20:55:59,179 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=295340.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 20:56:06,345 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=295345.0, num_to_drop=1, layers_to_drop={1} 2023-05-02 20:56:09,303 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.8088, 4.0472, 2.7910, 4.6597, 3.2930, 4.5686, 2.9296, 3.3354], device='cuda:1'), covar=tensor([0.0382, 0.0409, 0.1509, 0.0284, 0.0813, 0.0620, 0.1357, 0.0823], device='cuda:1'), in_proj_covar=tensor([0.0178, 0.0183, 0.0197, 0.0176, 0.0181, 0.0223, 0.0207, 0.0185], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-05-02 20:56:17,470 INFO [train.py:904] (1/8) Epoch 30, batch 1000, loss[loss=0.1575, simple_loss=0.2385, pruned_loss=0.03823, over 16267.00 frames. ], tot_loss[loss=0.1584, simple_loss=0.2444, pruned_loss=0.03625, over 3278698.66 frames. ], batch size: 165, lr: 2.25e-03, grad_scale: 4.0 2023-05-02 20:56:24,294 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=295359.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 20:57:12,872 INFO [zipformer.py:625] (1/8) 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] (1/8) 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 20:57:26,453 INFO [train.py:904] (1/8) Epoch 30, batch 1050, loss[loss=0.1403, simple_loss=0.2349, pruned_loss=0.02282, over 17204.00 frames. ], tot_loss[loss=0.158, simple_loss=0.2442, pruned_loss=0.03594, over 3290581.38 frames. ], batch size: 44, lr: 2.25e-03, grad_scale: 4.0 2023-05-02 20:57:41,215 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=3.47 vs. limit=5.0 2023-05-02 20:57:48,319 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=295420.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 20:57:53,562 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=295424.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 20:58:36,207 INFO [train.py:904] (1/8) Epoch 30, batch 1100, loss[loss=0.1548, simple_loss=0.2332, pruned_loss=0.03815, over 16481.00 frames. ], tot_loss[loss=0.1573, simple_loss=0.2433, pruned_loss=0.03568, over 3291808.83 frames. ], batch size: 146, lr: 2.25e-03, grad_scale: 4.0 2023-05-02 20:59:38,543 INFO [optim.py:368] (1/8) 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 20:59:41,306 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.9038, 1.8619, 2.4419, 2.7496, 2.8021, 2.8877, 1.9504, 2.9985], device='cuda:1'), covar=tensor([0.0203, 0.0661, 0.0377, 0.0310, 0.0312, 0.0307, 0.0752, 0.0210], device='cuda:1'), in_proj_covar=tensor([0.0200, 0.0199, 0.0187, 0.0193, 0.0211, 0.0168, 0.0205, 0.0169], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-02 20:59:43,304 INFO [train.py:904] (1/8) Epoch 30, batch 1150, loss[loss=0.137, simple_loss=0.2254, pruned_loss=0.02425, over 16965.00 frames. ], tot_loss[loss=0.1568, simple_loss=0.2432, pruned_loss=0.03516, over 3295839.97 frames. ], batch size: 41, lr: 2.25e-03, grad_scale: 4.0 2023-05-02 21:00:52,295 INFO [train.py:904] (1/8) Epoch 30, batch 1200, loss[loss=0.159, simple_loss=0.2562, pruned_loss=0.0309, over 17047.00 frames. ], tot_loss[loss=0.1562, simple_loss=0.2426, pruned_loss=0.03488, over 3299919.22 frames. ], batch size: 50, lr: 2.25e-03, grad_scale: 8.0 2023-05-02 21:00:55,038 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=295556.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 21:00:57,940 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=295558.0, num_to_drop=1, layers_to_drop={1} 2023-05-02 21:01:29,883 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=295581.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 21:01:46,995 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.6551, 2.5816, 1.9343, 2.6553, 2.1910, 2.8111, 2.1562, 2.3747], device='cuda:1'), covar=tensor([0.0368, 0.0410, 0.1355, 0.0317, 0.0683, 0.0475, 0.1265, 0.0678], device='cuda:1'), in_proj_covar=tensor([0.0177, 0.0182, 0.0196, 0.0175, 0.0180, 0.0222, 0.0206, 0.0184], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-05-02 21:01:55,306 INFO [optim.py:368] (1/8) 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 21:02:00,804 INFO [train.py:904] (1/8) Epoch 30, batch 1250, loss[loss=0.1613, simple_loss=0.254, pruned_loss=0.03427, over 17277.00 frames. ], tot_loss[loss=0.1561, simple_loss=0.2421, pruned_loss=0.03503, over 3295677.70 frames. ], batch size: 52, lr: 2.25e-03, grad_scale: 8.0 2023-05-02 21:02:12,096 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.7931, 1.9950, 2.4107, 2.6396, 2.7074, 2.7337, 2.1386, 2.9160], device='cuda:1'), covar=tensor([0.0229, 0.0562, 0.0371, 0.0350, 0.0384, 0.0388, 0.0578, 0.0214], device='cuda:1'), in_proj_covar=tensor([0.0201, 0.0201, 0.0189, 0.0195, 0.0212, 0.0169, 0.0206, 0.0170], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0002], device='cuda:1') 2023-05-02 21:02:18,626 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=295617.0, num_to_drop=1, layers_to_drop={3} 2023-05-02 21:02:21,342 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=295619.0, num_to_drop=1, layers_to_drop={3} 2023-05-02 21:02:34,287 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=295629.0, num_to_drop=1, layers_to_drop={0} 2023-05-02 21:02:42,475 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=295635.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 21:02:58,116 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=295645.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 21:03:07,729 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.9599, 4.4270, 4.4608, 3.2707, 3.6724, 4.3509, 3.9636, 2.6109], device='cuda:1'), covar=tensor([0.0496, 0.0075, 0.0050, 0.0357, 0.0162, 0.0111, 0.0101, 0.0500], device='cuda:1'), in_proj_covar=tensor([0.0140, 0.0092, 0.0093, 0.0136, 0.0104, 0.0117, 0.0100, 0.0132], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0003, 0.0004], device='cuda:1') 2023-05-02 21:03:08,458 INFO [train.py:904] (1/8) Epoch 30, batch 1300, loss[loss=0.1459, simple_loss=0.2443, pruned_loss=0.02371, over 17049.00 frames. ], tot_loss[loss=0.1559, simple_loss=0.2423, pruned_loss=0.03474, over 3298672.92 frames. ], batch size: 50, lr: 2.25e-03, grad_scale: 8.0 2023-05-02 21:03:13,164 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.39 vs. limit=2.0 2023-05-02 21:03:52,629 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=295685.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 21:03:57,726 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=295689.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 21:03:59,032 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=295690.0, num_to_drop=1, layers_to_drop={2} 2023-05-02 21:04:02,886 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=295693.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 21:04:11,309 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.0865, 2.2589, 2.7627, 3.0594, 2.9107, 3.5508, 2.5739, 3.5582], device='cuda:1'), covar=tensor([0.0333, 0.0600, 0.0402, 0.0386, 0.0420, 0.0230, 0.0539, 0.0190], device='cuda:1'), in_proj_covar=tensor([0.0202, 0.0202, 0.0189, 0.0196, 0.0213, 0.0170, 0.0207, 0.0171], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0002], device='cuda:1') 2023-05-02 21:04:11,933 INFO [optim.py:368] (1/8) 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 21:04:14,718 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.21 vs. limit=2.0 2023-05-02 21:04:18,058 INFO [train.py:904] (1/8) Epoch 30, batch 1350, loss[loss=0.1393, simple_loss=0.2332, pruned_loss=0.02267, over 17223.00 frames. ], tot_loss[loss=0.1559, simple_loss=0.2423, pruned_loss=0.03474, over 3311610.69 frames. ], batch size: 44, lr: 2.25e-03, grad_scale: 8.0 2023-05-02 21:04:33,096 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=295715.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 21:04:46,288 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=295724.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 21:05:17,200 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=295746.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 21:05:26,449 INFO [train.py:904] (1/8) Epoch 30, batch 1400, loss[loss=0.1741, simple_loss=0.2665, pruned_loss=0.04085, over 16797.00 frames. ], tot_loss[loss=0.1564, simple_loss=0.2425, pruned_loss=0.03509, over 3317818.71 frames. ], batch size: 62, lr: 2.25e-03, grad_scale: 8.0 2023-05-02 21:05:32,883 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.7433, 2.7557, 2.3767, 2.5684, 3.0419, 2.7168, 3.3000, 3.2516], device='cuda:1'), covar=tensor([0.0187, 0.0501, 0.0617, 0.0542, 0.0354, 0.0488, 0.0294, 0.0363], device='cuda:1'), in_proj_covar=tensor([0.0238, 0.0251, 0.0238, 0.0240, 0.0249, 0.0248, 0.0246, 0.0249], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-02 21:05:48,230 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([5.2027, 5.1821, 5.0759, 4.5180, 4.6792, 5.1089, 5.1034, 4.7260], device='cuda:1'), covar=tensor([0.0705, 0.0622, 0.0432, 0.0442, 0.1360, 0.0604, 0.0401, 0.0914], device='cuda:1'), in_proj_covar=tensor([0.0324, 0.0490, 0.0380, 0.0381, 0.0376, 0.0438, 0.0261, 0.0454], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-05-02 21:05:52,092 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=295772.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 21:05:52,616 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.23 vs. limit=2.0 2023-05-02 21:05:53,495 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.0951, 2.2197, 2.3814, 3.7183, 2.1907, 2.4853, 2.2809, 2.3606], device='cuda:1'), covar=tensor([0.1733, 0.4105, 0.3444, 0.0813, 0.4315, 0.2825, 0.4246, 0.3389], device='cuda:1'), in_proj_covar=tensor([0.0428, 0.0481, 0.0393, 0.0343, 0.0450, 0.0551, 0.0454, 0.0565], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-02 21:06:30,027 INFO [optim.py:368] (1/8) 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] (1/8) Epoch 30, batch 1450, loss[loss=0.1475, simple_loss=0.2287, pruned_loss=0.03315, over 15420.00 frames. ], tot_loss[loss=0.1555, simple_loss=0.2418, pruned_loss=0.03461, over 3325864.54 frames. ], batch size: 190, lr: 2.25e-03, grad_scale: 8.0 2023-05-02 21:06:57,472 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=295820.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 21:07:20,868 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([5.2103, 5.1775, 5.0813, 4.5682, 4.7684, 5.1102, 5.0856, 4.7640], device='cuda:1'), covar=tensor([0.0630, 0.0614, 0.0360, 0.0391, 0.1141, 0.0572, 0.0361, 0.0860], device='cuda:1'), in_proj_covar=tensor([0.0323, 0.0489, 0.0379, 0.0380, 0.0375, 0.0437, 0.0261, 0.0453], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-05-02 21:07:29,907 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.25 vs. limit=2.0 2023-05-02 21:07:38,887 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=4.60 vs. limit=5.0 2023-05-02 21:07:44,322 INFO [train.py:904] (1/8) Epoch 30, batch 1500, loss[loss=0.1462, simple_loss=0.2263, pruned_loss=0.03306, over 16241.00 frames. ], tot_loss[loss=0.156, simple_loss=0.2419, pruned_loss=0.03499, over 3320571.76 frames. ], batch size: 165, lr: 2.25e-03, grad_scale: 8.0 2023-05-02 21:08:11,310 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=295873.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 21:08:22,055 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=295881.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 21:08:22,106 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=295881.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 21:08:43,889 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.89 vs. limit=2.0 2023-05-02 21:08:47,908 INFO [optim.py:368] (1/8) 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] (1/8) Epoch 30, batch 1550, loss[loss=0.1681, simple_loss=0.2601, pruned_loss=0.03803, over 16752.00 frames. ], tot_loss[loss=0.1573, simple_loss=0.2435, pruned_loss=0.03559, over 3329838.70 frames. ], batch size: 57, lr: 2.25e-03, grad_scale: 8.0 2023-05-02 21:09:05,571 INFO [zipformer.py:625] (1/8) 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] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=295914.0, num_to_drop=1, layers_to_drop={3} 2023-05-02 21:09:29,567 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=295929.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 21:09:36,329 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=295934.0, num_to_drop=1, layers_to_drop={3} 2023-05-02 21:09:37,293 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=295935.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 21:09:46,041 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.93 vs. limit=2.0 2023-05-02 21:09:58,498 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.69 vs. limit=2.0 2023-05-02 21:10:02,943 INFO [train.py:904] (1/8) Epoch 30, batch 1600, loss[loss=0.1464, simple_loss=0.2303, pruned_loss=0.03123, over 17052.00 frames. ], tot_loss[loss=0.1588, simple_loss=0.2452, pruned_loss=0.03619, over 3328815.11 frames. ], batch size: 41, lr: 2.25e-03, grad_scale: 8.0 2023-05-02 21:10:42,972 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=295983.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 21:10:45,977 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=295985.0, num_to_drop=1, layers_to_drop={2} 2023-05-02 21:10:50,677 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=295989.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 21:11:00,521 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.4199, 4.4727, 4.8080, 4.7696, 4.8320, 4.5121, 4.5062, 4.4434], device='cuda:1'), covar=tensor([0.0440, 0.0758, 0.0488, 0.0504, 0.0621, 0.0525, 0.0969, 0.0700], device='cuda:1'), in_proj_covar=tensor([0.0447, 0.0509, 0.0487, 0.0451, 0.0535, 0.0518, 0.0593, 0.0415], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0003], device='cuda:1') 2023-05-02 21:11:05,442 INFO [optim.py:368] (1/8) Clipping_scale=2.0, grad-norm quartiles 1.556e+02 2.204e+02 2.497e+02 2.999e+02 5.396e+02, threshold=4.994e+02, percent-clipped=1.0 2023-05-02 21:11:14,063 INFO [train.py:904] (1/8) Epoch 30, batch 1650, loss[loss=0.1627, simple_loss=0.2629, pruned_loss=0.03123, over 17123.00 frames. ], tot_loss[loss=0.1608, simple_loss=0.2468, pruned_loss=0.03742, over 3331003.33 frames. ], batch size: 49, lr: 2.25e-03, grad_scale: 8.0 2023-05-02 21:11:30,017 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=296015.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 21:12:00,954 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=296037.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 21:12:05,672 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.6591, 4.6989, 4.9214, 4.6824, 4.7809, 5.3387, 4.8481, 4.5515], device='cuda:1'), covar=tensor([0.1638, 0.2326, 0.2529, 0.2436, 0.2757, 0.1186, 0.1985, 0.2801], device='cuda:1'), in_proj_covar=tensor([0.0439, 0.0657, 0.0732, 0.0532, 0.0710, 0.0749, 0.0561, 0.0707], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-05-02 21:12:05,682 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=296041.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 21:12:07,564 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.5020, 3.5472, 2.2255, 3.7498, 2.9098, 3.7285, 2.3908, 2.9274], device='cuda:1'), covar=tensor([0.0313, 0.0433, 0.1603, 0.0420, 0.0765, 0.0805, 0.1396, 0.0742], device='cuda:1'), in_proj_covar=tensor([0.0180, 0.0185, 0.0198, 0.0178, 0.0183, 0.0225, 0.0208, 0.0185], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-05-02 21:12:16,907 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=296049.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 21:12:24,439 INFO [train.py:904] (1/8) Epoch 30, batch 1700, loss[loss=0.1774, simple_loss=0.2714, pruned_loss=0.04167, over 17061.00 frames. ], tot_loss[loss=0.1634, simple_loss=0.2499, pruned_loss=0.03844, over 3312329.73 frames. ], batch size: 55, lr: 2.25e-03, grad_scale: 8.0 2023-05-02 21:12:36,066 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=296063.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 21:13:26,988 INFO [optim.py:368] (1/8) Clipping_scale=2.0, grad-norm quartiles 1.510e+02 2.214e+02 2.660e+02 3.287e+02 5.248e+02, threshold=5.320e+02, percent-clipped=2.0 2023-05-02 21:13:32,339 INFO [train.py:904] (1/8) Epoch 30, batch 1750, loss[loss=0.1465, simple_loss=0.2524, pruned_loss=0.02029, over 17267.00 frames. ], tot_loss[loss=0.1633, simple_loss=0.2503, pruned_loss=0.03815, over 3317200.45 frames. ], batch size: 52, lr: 2.25e-03, grad_scale: 8.0 2023-05-02 21:13:40,842 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=296110.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 21:14:18,155 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.5473, 3.6515, 3.7813, 2.6310, 3.5217, 3.8785, 3.6107, 1.9670], device='cuda:1'), covar=tensor([0.0604, 0.0366, 0.0111, 0.0516, 0.0186, 0.0175, 0.0145, 0.0741], device='cuda:1'), in_proj_covar=tensor([0.0140, 0.0092, 0.0093, 0.0137, 0.0105, 0.0117, 0.0100, 0.0133], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0003, 0.0004], device='cuda:1') 2023-05-02 21:14:29,295 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([5.4811, 5.9051, 5.6800, 5.7347, 5.3618, 5.3885, 5.3277, 6.0722], device='cuda:1'), covar=tensor([0.1654, 0.1081, 0.1100, 0.0942, 0.0931, 0.0736, 0.1374, 0.0968], device='cuda:1'), in_proj_covar=tensor([0.0741, 0.0891, 0.0736, 0.0696, 0.0571, 0.0566, 0.0751, 0.0702], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:1') 2023-05-02 21:14:41,095 INFO [train.py:904] (1/8) Epoch 30, batch 1800, loss[loss=0.1767, simple_loss=0.2694, pruned_loss=0.04203, over 16711.00 frames. ], tot_loss[loss=0.1645, simple_loss=0.2515, pruned_loss=0.03875, over 3324075.62 frames. ], batch size: 57, lr: 2.25e-03, grad_scale: 8.0 2023-05-02 21:14:49,268 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.89 vs. limit=2.0 2023-05-02 21:15:12,948 INFO [zipformer.py:625] (1/8) 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] (1/8) Clipping_scale=2.0, grad-norm quartiles 1.503e+02 2.165e+02 2.480e+02 2.953e+02 5.408e+02, threshold=4.959e+02, percent-clipped=1.0 2023-05-02 21:15:51,606 INFO [train.py:904] (1/8) Epoch 30, batch 1850, loss[loss=0.169, simple_loss=0.255, pruned_loss=0.04151, over 16852.00 frames. ], tot_loss[loss=0.1654, simple_loss=0.2524, pruned_loss=0.03916, over 3313316.25 frames. ], batch size: 109, lr: 2.25e-03, grad_scale: 4.0 2023-05-02 21:16:03,694 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=296212.0, num_to_drop=1, layers_to_drop={1} 2023-05-02 21:16:06,120 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=296214.0, num_to_drop=1, layers_to_drop={0} 2023-05-02 21:16:22,460 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.31 vs. limit=2.0 2023-05-02 21:16:27,209 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=296229.0, num_to_drop=1, layers_to_drop={0} 2023-05-02 21:17:02,019 INFO [train.py:904] (1/8) Epoch 30, batch 1900, loss[loss=0.1422, simple_loss=0.2283, pruned_loss=0.02799, over 16866.00 frames. ], tot_loss[loss=0.1645, simple_loss=0.2518, pruned_loss=0.03854, over 3306698.83 frames. ], batch size: 96, lr: 2.25e-03, grad_scale: 4.0 2023-05-02 21:17:09,977 INFO [zipformer.py:625] (1/8) 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] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=296262.0, num_to_drop=1, layers_to_drop={1} 2023-05-02 21:17:45,546 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([5.1652, 4.9201, 5.2112, 5.3840, 5.5795, 4.8468, 5.5126, 5.5385], device='cuda:1'), covar=tensor([0.1940, 0.1453, 0.1888, 0.0889, 0.0633, 0.1132, 0.0653, 0.0708], device='cuda:1'), in_proj_covar=tensor([0.0713, 0.0867, 0.1001, 0.0883, 0.0671, 0.0694, 0.0733, 0.0856], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-02 21:17:45,563 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=296285.0, num_to_drop=1, layers_to_drop={1} 2023-05-02 21:18:06,422 INFO [optim.py:368] (1/8) 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 21:18:10,610 INFO [train.py:904] (1/8) Epoch 30, batch 1950, loss[loss=0.1678, simple_loss=0.2506, pruned_loss=0.04249, over 16711.00 frames. ], tot_loss[loss=0.1646, simple_loss=0.2523, pruned_loss=0.03845, over 3295786.69 frames. ], batch size: 89, lr: 2.25e-03, grad_scale: 4.0 2023-05-02 21:18:12,843 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.6430, 4.7267, 4.9142, 4.6730, 4.7419, 5.3491, 4.7941, 4.4898], device='cuda:1'), covar=tensor([0.1715, 0.2318, 0.2443, 0.2369, 0.2839, 0.1080, 0.1879, 0.2624], device='cuda:1'), in_proj_covar=tensor([0.0436, 0.0651, 0.0725, 0.0529, 0.0705, 0.0745, 0.0558, 0.0703], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-05-02 21:18:26,936 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=296315.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 21:18:52,569 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=296333.0, num_to_drop=1, layers_to_drop={1} 2023-05-02 21:19:02,402 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=296341.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 21:19:16,801 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=4.77 vs. limit=5.0 2023-05-02 21:19:21,053 INFO [train.py:904] (1/8) Epoch 30, batch 2000, loss[loss=0.1561, simple_loss=0.2534, pruned_loss=0.02942, over 17051.00 frames. ], tot_loss[loss=0.1636, simple_loss=0.2513, pruned_loss=0.03791, over 3301875.45 frames. ], batch size: 50, lr: 2.25e-03, grad_scale: 8.0 2023-05-02 21:19:38,777 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.09 vs. limit=2.0 2023-05-02 21:19:51,055 INFO [zipformer.py:625] (1/8) 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] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=296389.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 21:20:25,431 INFO [optim.py:368] (1/8) 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] (1/8) Epoch 30, batch 2050, loss[loss=0.1809, simple_loss=0.2581, pruned_loss=0.05187, over 16307.00 frames. ], tot_loss[loss=0.1631, simple_loss=0.2506, pruned_loss=0.0378, over 3314861.12 frames. ], batch size: 165, lr: 2.25e-03, grad_scale: 8.0 2023-05-02 21:20:32,249 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=296405.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 21:21:12,449 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=296434.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 21:21:39,724 INFO [train.py:904] (1/8) Epoch 30, batch 2100, loss[loss=0.1456, simple_loss=0.236, pruned_loss=0.02766, over 15899.00 frames. ], tot_loss[loss=0.1644, simple_loss=0.2514, pruned_loss=0.03865, over 3305896.40 frames. ], batch size: 35, lr: 2.25e-03, grad_scale: 8.0 2023-05-02 21:22:09,550 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=296476.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 21:22:33,135 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=296492.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 21:22:36,753 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=296495.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 21:22:39,050 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=296497.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 21:22:46,671 INFO [optim.py:368] (1/8) 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 21:22:48,892 INFO [train.py:904] (1/8) Epoch 30, batch 2150, loss[loss=0.2165, simple_loss=0.295, pruned_loss=0.06896, over 12168.00 frames. ], tot_loss[loss=0.1651, simple_loss=0.2524, pruned_loss=0.03895, over 3311571.00 frames. ], batch size: 246, lr: 2.25e-03, grad_scale: 4.0 2023-05-02 21:23:03,759 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=296514.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 21:23:16,223 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=296524.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 21:23:23,453 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=296529.0, num_to_drop=1, layers_to_drop={2} 2023-05-02 21:23:56,838 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=296553.0, num_to_drop=1, layers_to_drop={3} 2023-05-02 21:23:57,527 INFO [train.py:904] (1/8) Epoch 30, batch 2200, loss[loss=0.1913, simple_loss=0.2922, pruned_loss=0.04519, over 16733.00 frames. ], tot_loss[loss=0.1654, simple_loss=0.2529, pruned_loss=0.03901, over 3315827.83 frames. ], batch size: 57, lr: 2.25e-03, grad_scale: 4.0 2023-05-02 21:24:04,211 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=296558.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 21:24:27,033 INFO [zipformer.py:625] (1/8) 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] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=296577.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 21:24:33,632 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=4.80 vs. limit=5.0 2023-05-02 21:25:04,056 INFO [optim.py:368] (1/8) 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 21:25:06,333 INFO [train.py:904] (1/8) Epoch 30, batch 2250, loss[loss=0.1542, simple_loss=0.2394, pruned_loss=0.0345, over 17216.00 frames. ], tot_loss[loss=0.166, simple_loss=0.2535, pruned_loss=0.0393, over 3312806.78 frames. ], batch size: 45, lr: 2.25e-03, grad_scale: 4.0 2023-05-02 21:25:12,331 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.9399, 2.0754, 2.4913, 2.8256, 2.7728, 2.9589, 2.0526, 3.1231], device='cuda:1'), covar=tensor([0.0235, 0.0537, 0.0400, 0.0314, 0.0374, 0.0326, 0.0651, 0.0176], device='cuda:1'), in_proj_covar=tensor([0.0203, 0.0202, 0.0191, 0.0196, 0.0215, 0.0171, 0.0207, 0.0171], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0002], device='cuda:1') 2023-05-02 21:25:20,119 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=296613.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 21:25:37,257 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.8424, 4.0452, 2.5663, 4.6309, 3.2486, 4.5781, 2.6637, 3.3279], device='cuda:1'), covar=tensor([0.0351, 0.0460, 0.1700, 0.0344, 0.0867, 0.0622, 0.1614, 0.0778], device='cuda:1'), in_proj_covar=tensor([0.0180, 0.0186, 0.0199, 0.0179, 0.0184, 0.0226, 0.0209, 0.0186], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-05-02 21:26:17,731 INFO [train.py:904] (1/8) Epoch 30, batch 2300, loss[loss=0.1406, simple_loss=0.2321, pruned_loss=0.02457, over 17211.00 frames. ], tot_loss[loss=0.1657, simple_loss=0.2534, pruned_loss=0.03903, over 3317112.32 frames. ], batch size: 44, lr: 2.25e-03, grad_scale: 4.0 2023-05-02 21:26:41,379 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=296671.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 21:26:45,082 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=296674.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 21:27:04,990 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=296688.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 21:27:24,093 INFO [optim.py:368] (1/8) 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 21:27:27,334 INFO [train.py:904] (1/8) Epoch 30, batch 2350, loss[loss=0.1899, simple_loss=0.2766, pruned_loss=0.05159, over 15505.00 frames. ], tot_loss[loss=0.1657, simple_loss=0.2534, pruned_loss=0.03899, over 3314379.74 frames. ], batch size: 191, lr: 2.25e-03, grad_scale: 4.0 2023-05-02 21:27:28,740 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=296705.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 21:28:30,519 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=296749.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 21:28:33,631 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.0059, 2.1808, 2.2492, 3.6876, 2.2330, 2.4848, 2.2556, 2.3346], device='cuda:1'), covar=tensor([0.1797, 0.4067, 0.3508, 0.0816, 0.4098, 0.2878, 0.4278, 0.3397], device='cuda:1'), in_proj_covar=tensor([0.0432, 0.0485, 0.0396, 0.0346, 0.0453, 0.0557, 0.0458, 0.0569], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-02 21:28:36,424 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 30, batch 2400, loss[loss=0.1647, simple_loss=0.2596, pruned_loss=0.03496, over 17184.00 frames. ], tot_loss[loss=0.166, simple_loss=0.254, pruned_loss=0.03895, over 3315302.43 frames. ], batch size: 46, lr: 2.25e-03, grad_scale: 8.0 2023-05-02 21:29:27,115 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=296790.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 21:29:43,967 INFO [optim.py:368] (1/8) Clipping_scale=2.0, grad-norm quartiles 1.575e+02 2.201e+02 2.707e+02 3.099e+02 6.549e+02, threshold=5.413e+02, percent-clipped=3.0 2023-05-02 21:29:46,349 INFO [train.py:904] (1/8) Epoch 30, batch 2450, loss[loss=0.1602, simple_loss=0.2552, pruned_loss=0.03264, over 17133.00 frames. ], tot_loss[loss=0.166, simple_loss=0.2542, pruned_loss=0.03893, over 3325747.63 frames. ], batch size: 48, lr: 2.25e-03, grad_scale: 8.0 2023-05-02 21:30:39,238 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.65 vs. limit=2.0 2023-05-02 21:30:48,290 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=296848.0, num_to_drop=1, layers_to_drop={1} 2023-05-02 21:30:52,450 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([5.1112, 4.8516, 5.1113, 5.2916, 5.5188, 4.8633, 5.4823, 5.5080], device='cuda:1'), covar=tensor([0.2042, 0.1401, 0.1951, 0.0899, 0.0612, 0.0994, 0.0682, 0.0679], device='cuda:1'), in_proj_covar=tensor([0.0715, 0.0867, 0.1002, 0.0886, 0.0672, 0.0699, 0.0737, 0.0859], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-02 21:30:55,429 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=296853.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 21:30:56,409 INFO [train.py:904] (1/8) Epoch 30, batch 2500, loss[loss=0.196, simple_loss=0.2806, pruned_loss=0.05571, over 12135.00 frames. ], tot_loss[loss=0.1653, simple_loss=0.2538, pruned_loss=0.03845, over 3319222.92 frames. ], batch size: 246, lr: 2.25e-03, grad_scale: 8.0 2023-05-02 21:31:19,657 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=296870.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 21:31:23,971 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=296873.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 21:32:04,914 INFO [optim.py:368] (1/8) 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] (1/8) Epoch 30, batch 2550, loss[loss=0.1489, simple_loss=0.2429, pruned_loss=0.02748, over 17233.00 frames. ], tot_loss[loss=0.1656, simple_loss=0.254, pruned_loss=0.03864, over 3321243.02 frames. ], batch size: 45, lr: 2.25e-03, grad_scale: 8.0 2023-05-02 21:32:34,477 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.6642, 2.4243, 1.8782, 2.1705, 2.7256, 2.4993, 2.5717, 2.8310], device='cuda:1'), covar=tensor([0.0314, 0.0541, 0.0708, 0.0641, 0.0301, 0.0458, 0.0274, 0.0364], device='cuda:1'), in_proj_covar=tensor([0.0243, 0.0253, 0.0241, 0.0243, 0.0254, 0.0252, 0.0251, 0.0254], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-02 21:32:49,082 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=296934.0, num_to_drop=1, layers_to_drop={3} 2023-05-02 21:32:54,158 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 30, batch 2600, loss[loss=0.1518, simple_loss=0.2458, pruned_loss=0.02895, over 17220.00 frames. ], tot_loss[loss=0.1655, simple_loss=0.254, pruned_loss=0.03845, over 3323593.02 frames. ], batch size: 44, lr: 2.25e-03, grad_scale: 8.0 2023-05-02 21:33:31,352 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.0958, 4.5030, 4.5665, 3.3360, 3.8082, 4.5087, 3.9585, 2.7927], device='cuda:1'), covar=tensor([0.0464, 0.0089, 0.0047, 0.0344, 0.0158, 0.0095, 0.0121, 0.0477], device='cuda:1'), in_proj_covar=tensor([0.0140, 0.0092, 0.0093, 0.0136, 0.0105, 0.0117, 0.0100, 0.0132], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0003, 0.0004], device='cuda:1') 2023-05-02 21:33:36,660 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=296969.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 21:33:38,329 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.84 vs. limit=2.0 2023-05-02 21:33:39,689 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=296971.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 21:34:17,867 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=296999.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 21:34:22,260 INFO [optim.py:368] (1/8) Clipping_scale=2.0, grad-norm quartiles 1.341e+02 2.007e+02 2.417e+02 2.925e+02 5.072e+02, threshold=4.833e+02, percent-clipped=0.0 2023-05-02 21:34:24,415 INFO [train.py:904] (1/8) Epoch 30, batch 2650, loss[loss=0.1887, simple_loss=0.2969, pruned_loss=0.04027, over 16619.00 frames. ], tot_loss[loss=0.1654, simple_loss=0.2545, pruned_loss=0.0381, over 3321584.04 frames. ], batch size: 57, lr: 2.25e-03, grad_scale: 8.0 2023-05-02 21:34:46,004 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=297019.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 21:35:11,950 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.6906, 2.3927, 1.8989, 2.2004, 2.7384, 2.5136, 2.6602, 2.8535], device='cuda:1'), covar=tensor([0.0310, 0.0503, 0.0641, 0.0504, 0.0286, 0.0398, 0.0208, 0.0325], device='cuda:1'), in_proj_covar=tensor([0.0242, 0.0252, 0.0240, 0.0242, 0.0252, 0.0250, 0.0250, 0.0252], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-02 21:35:20,544 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=297044.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 21:35:34,414 INFO [train.py:904] (1/8) Epoch 30, batch 2700, loss[loss=0.1824, simple_loss=0.2774, pruned_loss=0.04367, over 17225.00 frames. ], tot_loss[loss=0.1657, simple_loss=0.2553, pruned_loss=0.0381, over 3317980.54 frames. ], batch size: 52, lr: 2.25e-03, grad_scale: 8.0 2023-05-02 21:35:51,482 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.3615, 2.7012, 2.7737, 4.3711, 2.5682, 2.9285, 2.6892, 2.7561], device='cuda:1'), covar=tensor([0.1471, 0.3366, 0.2937, 0.0622, 0.3933, 0.2555, 0.3453, 0.3545], device='cuda:1'), in_proj_covar=tensor([0.0430, 0.0483, 0.0394, 0.0345, 0.0451, 0.0555, 0.0456, 0.0566], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-02 21:36:23,933 INFO [zipformer.py:625] (1/8) 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] (1/8) 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 21:36:43,153 INFO [train.py:904] (1/8) Epoch 30, batch 2750, loss[loss=0.1475, simple_loss=0.244, pruned_loss=0.02554, over 17172.00 frames. ], tot_loss[loss=0.1659, simple_loss=0.2554, pruned_loss=0.03821, over 3316113.50 frames. ], batch size: 46, lr: 2.25e-03, grad_scale: 4.0 2023-05-02 21:37:03,127 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([1.7488, 2.5132, 2.4681, 3.8659, 3.0167, 3.9105, 1.6057, 2.7742], device='cuda:1'), covar=tensor([0.1445, 0.0778, 0.1291, 0.0188, 0.0127, 0.0364, 0.1720, 0.0914], device='cuda:1'), in_proj_covar=tensor([0.0175, 0.0185, 0.0204, 0.0210, 0.0210, 0.0222, 0.0214, 0.0202], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-05-02 21:37:17,907 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.9170, 4.0519, 2.6286, 4.7018, 3.2579, 4.6112, 2.7775, 3.4689], device='cuda:1'), covar=tensor([0.0318, 0.0386, 0.1545, 0.0314, 0.0803, 0.0520, 0.1439, 0.0713], device='cuda:1'), in_proj_covar=tensor([0.0181, 0.0186, 0.0199, 0.0180, 0.0184, 0.0226, 0.0209, 0.0187], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-05-02 21:37:30,683 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=297138.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 21:37:44,462 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=297148.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 21:37:50,744 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=297153.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 21:37:51,152 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.22 vs. limit=2.0 2023-05-02 21:37:51,587 INFO [train.py:904] (1/8) Epoch 30, batch 2800, loss[loss=0.1647, simple_loss=0.2616, pruned_loss=0.03388, over 16638.00 frames. ], tot_loss[loss=0.1657, simple_loss=0.2549, pruned_loss=0.03831, over 3310808.56 frames. ], batch size: 62, lr: 2.24e-03, grad_scale: 8.0 2023-05-02 21:38:14,206 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=297170.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 21:38:14,667 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.27 vs. limit=2.0 2023-05-02 21:38:49,782 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=297196.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 21:38:57,321 INFO [zipformer.py:625] (1/8) 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] (1/8) 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] (1/8) Epoch 30, batch 2850, loss[loss=0.1888, simple_loss=0.2686, pruned_loss=0.05449, over 16194.00 frames. ], tot_loss[loss=0.1655, simple_loss=0.2545, pruned_loss=0.03822, over 3303019.97 frames. ], batch size: 165, lr: 2.24e-03, grad_scale: 8.0 2023-05-02 21:39:13,252 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([1.9876, 3.2005, 2.9923, 5.2751, 4.3549, 4.4721, 1.8176, 3.3856], device='cuda:1'), covar=tensor([0.1296, 0.0757, 0.1101, 0.0174, 0.0204, 0.0407, 0.1600, 0.0706], device='cuda:1'), in_proj_covar=tensor([0.0175, 0.0184, 0.0203, 0.0210, 0.0209, 0.0221, 0.0214, 0.0201], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-05-02 21:39:21,040 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=297218.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 21:39:26,927 INFO [zipformer.py:625] (1/8) 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] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=297229.0, num_to_drop=1, layers_to_drop={1} 2023-05-02 21:39:45,666 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.8049, 3.9938, 2.2751, 4.6635, 3.1342, 4.4918, 2.5762, 3.3141], device='cuda:1'), covar=tensor([0.0375, 0.0456, 0.2024, 0.0283, 0.0891, 0.0620, 0.1698, 0.0804], device='cuda:1'), in_proj_covar=tensor([0.0182, 0.0187, 0.0200, 0.0180, 0.0185, 0.0228, 0.0210, 0.0188], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-05-02 21:40:10,536 INFO [train.py:904] (1/8) Epoch 30, batch 2900, loss[loss=0.1418, simple_loss=0.2257, pruned_loss=0.029, over 16887.00 frames. ], tot_loss[loss=0.1651, simple_loss=0.2534, pruned_loss=0.0384, over 3311861.30 frames. ], batch size: 90, lr: 2.24e-03, grad_scale: 8.0 2023-05-02 21:40:19,415 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.41 vs. limit=2.0 2023-05-02 21:40:24,158 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.8149, 2.7747, 2.4465, 2.7021, 3.0873, 2.7790, 3.3636, 3.2730], device='cuda:1'), covar=tensor([0.0195, 0.0536, 0.0607, 0.0510, 0.0333, 0.0468, 0.0283, 0.0339], device='cuda:1'), in_proj_covar=tensor([0.0242, 0.0251, 0.0239, 0.0241, 0.0252, 0.0250, 0.0249, 0.0252], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-02 21:40:33,158 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=297269.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 21:40:48,987 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.9231, 4.1149, 3.1842, 2.5995, 2.8203, 2.8133, 4.4068, 3.6604], device='cuda:1'), covar=tensor([0.2802, 0.0586, 0.1738, 0.2958, 0.2706, 0.2009, 0.0510, 0.1420], device='cuda:1'), in_proj_covar=tensor([0.0342, 0.0280, 0.0319, 0.0333, 0.0312, 0.0285, 0.0310, 0.0361], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:1') 2023-05-02 21:40:51,657 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=297283.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 21:40:59,767 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([1.7702, 2.9229, 2.6875, 4.9686, 3.9473, 4.2991, 1.7336, 3.0733], device='cuda:1'), covar=tensor([0.1458, 0.0852, 0.1276, 0.0224, 0.0223, 0.0424, 0.1743, 0.0837], device='cuda:1'), in_proj_covar=tensor([0.0175, 0.0184, 0.0204, 0.0210, 0.0209, 0.0222, 0.0214, 0.0202], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-05-02 21:41:08,333 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=297294.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 21:41:08,656 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.5784, 2.5379, 2.5257, 4.4070, 2.5302, 2.9251, 2.6060, 2.7114], device='cuda:1'), covar=tensor([0.1359, 0.3606, 0.3333, 0.0595, 0.4147, 0.2536, 0.3693, 0.3490], device='cuda:1'), in_proj_covar=tensor([0.0431, 0.0484, 0.0395, 0.0346, 0.0451, 0.0556, 0.0457, 0.0567], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-02 21:41:11,855 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([5.0865, 5.5753, 5.7340, 5.3451, 5.5059, 6.0859, 5.5454, 5.2680], device='cuda:1'), covar=tensor([0.1059, 0.2151, 0.2776, 0.2186, 0.2533, 0.0953, 0.1607, 0.2422], device='cuda:1'), in_proj_covar=tensor([0.0441, 0.0657, 0.0732, 0.0532, 0.0712, 0.0750, 0.0563, 0.0710], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-05-02 21:41:22,345 INFO [optim.py:368] (1/8) 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 21:41:22,360 INFO [train.py:904] (1/8) Epoch 30, batch 2950, loss[loss=0.1341, simple_loss=0.2181, pruned_loss=0.02505, over 16775.00 frames. ], tot_loss[loss=0.1657, simple_loss=0.2536, pruned_loss=0.03896, over 3303023.85 frames. ], batch size: 39, lr: 2.24e-03, grad_scale: 4.0 2023-05-02 21:41:41,510 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=297317.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 21:42:04,284 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.4041, 4.4121, 4.3657, 4.0851, 4.1341, 4.4362, 4.1415, 4.1888], device='cuda:1'), covar=tensor([0.0670, 0.0692, 0.0305, 0.0291, 0.0754, 0.0439, 0.0705, 0.0626], device='cuda:1'), in_proj_covar=tensor([0.0329, 0.0499, 0.0385, 0.0389, 0.0382, 0.0446, 0.0265, 0.0462], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-05-02 21:42:18,983 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=297344.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 21:42:33,297 INFO [train.py:904] (1/8) Epoch 30, batch 3000, loss[loss=0.1587, simple_loss=0.2452, pruned_loss=0.03615, over 16875.00 frames. ], tot_loss[loss=0.1668, simple_loss=0.2547, pruned_loss=0.03949, over 3299378.24 frames. ], batch size: 90, lr: 2.24e-03, grad_scale: 4.0 2023-05-02 21:42:33,297 INFO [train.py:929] (1/8) Computing validation loss 2023-05-02 21:42:40,910 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.9573, 4.1005, 4.2775, 3.1725, 3.7220, 4.3022, 3.9164, 2.8162], device='cuda:1'), covar=tensor([0.0482, 0.0117, 0.0071, 0.0327, 0.0124, 0.0114, 0.0096, 0.0448], device='cuda:1'), in_proj_covar=tensor([0.0140, 0.0092, 0.0093, 0.0136, 0.0104, 0.0117, 0.0100, 0.0132], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0003, 0.0004], device='cuda:1') 2023-05-02 21:42:42,092 INFO [train.py:938] (1/8) Epoch 30, validation: loss=0.1331, simple_loss=0.238, pruned_loss=0.01415, over 944034.00 frames. 2023-05-02 21:42:42,092 INFO [train.py:939] (1/8) Maximum memory allocated so far is 18145MB 2023-05-02 21:43:05,015 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=3.37 vs. limit=5.0 2023-05-02 21:43:35,186 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=297392.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 21:43:41,020 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.2802, 2.4778, 3.0714, 3.2050, 3.1749, 3.7453, 2.8558, 3.7514], device='cuda:1'), covar=tensor([0.0299, 0.0548, 0.0334, 0.0395, 0.0356, 0.0223, 0.0516, 0.0193], device='cuda:1'), in_proj_covar=tensor([0.0205, 0.0204, 0.0193, 0.0199, 0.0217, 0.0174, 0.0210, 0.0173], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0002], device='cuda:1') 2023-05-02 21:43:51,743 INFO [optim.py:368] (1/8) 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] (1/8) Epoch 30, batch 3050, loss[loss=0.1739, simple_loss=0.2663, pruned_loss=0.04074, over 16685.00 frames. ], tot_loss[loss=0.1659, simple_loss=0.2535, pruned_loss=0.03911, over 3310855.16 frames. ], batch size: 62, lr: 2.24e-03, grad_scale: 4.0 2023-05-02 21:44:09,603 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([5.1823, 5.1949, 5.5667, 5.5358, 5.6041, 5.2494, 5.1740, 5.0047], device='cuda:1'), covar=tensor([0.0342, 0.0582, 0.0422, 0.0479, 0.0556, 0.0426, 0.1036, 0.0476], device='cuda:1'), in_proj_covar=tensor([0.0452, 0.0515, 0.0491, 0.0454, 0.0537, 0.0519, 0.0598, 0.0417], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0003], device='cuda:1') 2023-05-02 21:45:01,428 INFO [train.py:904] (1/8) Epoch 30, batch 3100, loss[loss=0.1667, simple_loss=0.242, pruned_loss=0.04571, over 16863.00 frames. ], tot_loss[loss=0.1646, simple_loss=0.2523, pruned_loss=0.03846, over 3323115.89 frames. ], batch size: 102, lr: 2.24e-03, grad_scale: 4.0 2023-05-02 21:45:25,623 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.43 vs. limit=2.0 2023-05-02 21:46:07,372 INFO [optim.py:368] (1/8) 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 21:46:07,387 INFO [train.py:904] (1/8) Epoch 30, batch 3150, loss[loss=0.1787, simple_loss=0.2521, pruned_loss=0.05265, over 16776.00 frames. ], tot_loss[loss=0.1643, simple_loss=0.2515, pruned_loss=0.03857, over 3327545.31 frames. ], batch size: 124, lr: 2.24e-03, grad_scale: 4.0 2023-05-02 21:46:14,556 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.3119, 2.5498, 2.5996, 3.9994, 2.3490, 2.8835, 2.5824, 2.6924], device='cuda:1'), covar=tensor([0.1550, 0.3466, 0.3032, 0.0714, 0.4086, 0.2440, 0.3656, 0.3374], device='cuda:1'), in_proj_covar=tensor([0.0433, 0.0487, 0.0397, 0.0348, 0.0454, 0.0559, 0.0459, 0.0570], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:1') 2023-05-02 21:46:15,598 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.5580, 3.5058, 3.4827, 2.7598, 3.2595, 2.1485, 3.1914, 2.7075], device='cuda:1'), covar=tensor([0.0159, 0.0153, 0.0194, 0.0220, 0.0108, 0.2355, 0.0142, 0.0272], device='cuda:1'), in_proj_covar=tensor([0.0188, 0.0180, 0.0219, 0.0190, 0.0197, 0.0223, 0.0209, 0.0186], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-02 21:46:21,454 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([5.3232, 5.9029, 6.0038, 5.6202, 5.8623, 6.3874, 5.8794, 5.4507], device='cuda:1'), covar=tensor([0.0859, 0.2114, 0.2742, 0.2094, 0.2382, 0.0920, 0.1623, 0.2464], device='cuda:1'), in_proj_covar=tensor([0.0442, 0.0660, 0.0734, 0.0533, 0.0715, 0.0752, 0.0563, 0.0712], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-05-02 21:46:41,939 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=297529.0, num_to_drop=1, layers_to_drop={0} 2023-05-02 21:47:17,203 INFO [train.py:904] (1/8) Epoch 30, batch 3200, loss[loss=0.1494, simple_loss=0.2316, pruned_loss=0.03366, over 16735.00 frames. ], tot_loss[loss=0.1636, simple_loss=0.2511, pruned_loss=0.03803, over 3327173.41 frames. ], batch size: 124, lr: 2.24e-03, grad_scale: 8.0 2023-05-02 21:47:49,841 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=297577.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 21:47:51,695 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=297578.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 21:48:13,115 INFO [zipformer.py:625] (1/8) 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] (1/8) 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] (1/8) Epoch 30, batch 3250, loss[loss=0.1679, simple_loss=0.2519, pruned_loss=0.04196, over 16873.00 frames. ], tot_loss[loss=0.1638, simple_loss=0.2512, pruned_loss=0.03821, over 3320437.42 frames. ], batch size: 90, lr: 2.24e-03, grad_scale: 8.0 2023-05-02 21:48:53,873 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.8049, 1.7553, 2.4119, 2.6410, 2.7050, 2.7255, 1.8095, 2.9216], device='cuda:1'), covar=tensor([0.0210, 0.0704, 0.0390, 0.0346, 0.0363, 0.0365, 0.0826, 0.0213], device='cuda:1'), in_proj_covar=tensor([0.0205, 0.0204, 0.0194, 0.0199, 0.0218, 0.0174, 0.0210, 0.0174], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0002], device='cuda:1') 2023-05-02 21:49:20,233 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 30, batch 3300, loss[loss=0.1645, simple_loss=0.2565, pruned_loss=0.03622, over 16728.00 frames. ], tot_loss[loss=0.1643, simple_loss=0.2518, pruned_loss=0.03841, over 3327842.32 frames. ], batch size: 57, lr: 2.24e-03, grad_scale: 8.0 2023-05-02 21:50:13,022 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([5.6714, 6.0820, 5.8394, 5.9114, 5.5090, 5.5312, 5.4316, 6.2544], device='cuda:1'), covar=tensor([0.1780, 0.1177, 0.1260, 0.1011, 0.1153, 0.0775, 0.1500, 0.1052], device='cuda:1'), in_proj_covar=tensor([0.0755, 0.0908, 0.0748, 0.0707, 0.0583, 0.0574, 0.0764, 0.0714], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:1') 2023-05-02 21:50:31,676 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.3202, 3.9000, 4.4409, 2.3558, 4.5993, 4.7374, 3.4983, 3.7834], device='cuda:1'), covar=tensor([0.0622, 0.0281, 0.0253, 0.1125, 0.0091, 0.0163, 0.0424, 0.0360], device='cuda:1'), in_proj_covar=tensor([0.0150, 0.0112, 0.0104, 0.0139, 0.0089, 0.0135, 0.0132, 0.0132], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-05-02 21:50:46,297 INFO [optim.py:368] (1/8) 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] (1/8) Epoch 30, batch 3350, loss[loss=0.1896, simple_loss=0.2752, pruned_loss=0.05199, over 15579.00 frames. ], tot_loss[loss=0.163, simple_loss=0.2511, pruned_loss=0.03742, over 3338130.47 frames. ], batch size: 190, lr: 2.24e-03, grad_scale: 8.0 2023-05-02 21:51:56,066 INFO [train.py:904] (1/8) Epoch 30, batch 3400, loss[loss=0.2109, simple_loss=0.3035, pruned_loss=0.0591, over 12064.00 frames. ], tot_loss[loss=0.1642, simple_loss=0.2522, pruned_loss=0.03809, over 3320750.32 frames. ], batch size: 246, lr: 2.24e-03, grad_scale: 8.0 2023-05-02 21:52:16,577 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.2721, 2.3715, 2.4167, 4.0061, 2.3493, 2.7199, 2.4318, 2.5390], device='cuda:1'), covar=tensor([0.1558, 0.3783, 0.3267, 0.0705, 0.4233, 0.2676, 0.3853, 0.3303], device='cuda:1'), in_proj_covar=tensor([0.0435, 0.0487, 0.0396, 0.0348, 0.0455, 0.0560, 0.0460, 0.0571], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:1') 2023-05-02 21:52:36,819 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.0310, 4.2362, 4.5365, 2.3181, 4.7599, 4.9480, 3.7508, 3.7520], device='cuda:1'), covar=tensor([0.0955, 0.0183, 0.0222, 0.1247, 0.0082, 0.0141, 0.0349, 0.0471], device='cuda:1'), in_proj_covar=tensor([0.0150, 0.0112, 0.0104, 0.0140, 0.0089, 0.0135, 0.0132, 0.0132], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-05-02 21:52:51,582 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([5.4566, 5.4332, 5.1642, 4.5782, 5.2328, 2.0870, 4.9687, 5.0160], device='cuda:1'), covar=tensor([0.0088, 0.0089, 0.0231, 0.0435, 0.0103, 0.2811, 0.0150, 0.0249], device='cuda:1'), in_proj_covar=tensor([0.0188, 0.0180, 0.0220, 0.0190, 0.0197, 0.0224, 0.0209, 0.0186], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-02 21:53:07,130 INFO [optim.py:368] (1/8) 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] (1/8) Epoch 30, batch 3450, loss[loss=0.1755, simple_loss=0.2576, pruned_loss=0.04677, over 15343.00 frames. ], tot_loss[loss=0.1629, simple_loss=0.2505, pruned_loss=0.03765, over 3318930.72 frames. ], batch size: 190, lr: 2.24e-03, grad_scale: 8.0 2023-05-02 21:54:17,197 INFO [train.py:904] (1/8) Epoch 30, batch 3500, loss[loss=0.1515, simple_loss=0.2393, pruned_loss=0.0318, over 16742.00 frames. ], tot_loss[loss=0.162, simple_loss=0.2495, pruned_loss=0.03721, over 3316620.21 frames. ], batch size: 83, lr: 2.24e-03, grad_scale: 4.0 2023-05-02 21:54:51,992 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 30, batch 3550, loss[loss=0.1692, simple_loss=0.2573, pruned_loss=0.04054, over 16461.00 frames. ], tot_loss[loss=0.1611, simple_loss=0.2486, pruned_loss=0.03686, over 3322579.90 frames. ], batch size: 68, lr: 2.24e-03, grad_scale: 4.0 2023-05-02 21:55:29,305 INFO [optim.py:368] (1/8) 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 21:55:58,197 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.8322, 1.9982, 2.5434, 2.8613, 2.6939, 3.2921, 2.3181, 3.3544], device='cuda:1'), covar=tensor([0.0363, 0.0667, 0.0432, 0.0382, 0.0441, 0.0287, 0.0618, 0.0210], device='cuda:1'), in_proj_covar=tensor([0.0206, 0.0204, 0.0194, 0.0199, 0.0218, 0.0174, 0.0210, 0.0174], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0002], device='cuda:1') 2023-05-02 21:55:59,066 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 30, batch 3600, loss[loss=0.1847, simple_loss=0.2621, pruned_loss=0.05364, over 11297.00 frames. ], tot_loss[loss=0.1604, simple_loss=0.2478, pruned_loss=0.03652, over 3308349.09 frames. ], batch size: 247, lr: 2.24e-03, grad_scale: 8.0 2023-05-02 21:56:42,443 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=297957.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 21:57:07,234 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([5.1971, 5.5345, 5.2803, 5.3204, 5.0788, 5.0052, 4.9524, 5.6656], device='cuda:1'), covar=tensor([0.1379, 0.0907, 0.1081, 0.0963, 0.0809, 0.0942, 0.1397, 0.0818], device='cuda:1'), in_proj_covar=tensor([0.0750, 0.0904, 0.0743, 0.0703, 0.0580, 0.0569, 0.0760, 0.0711], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:1') 2023-05-02 21:57:07,285 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.7123, 4.5690, 4.7180, 4.8989, 5.0138, 4.5281, 4.9501, 4.9880], device='cuda:1'), covar=tensor([0.1846, 0.1435, 0.1691, 0.0865, 0.0804, 0.1163, 0.1588, 0.1466], device='cuda:1'), in_proj_covar=tensor([0.0726, 0.0882, 0.1022, 0.0902, 0.0682, 0.0713, 0.0750, 0.0873], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-02 21:57:48,778 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.9645, 2.5292, 2.0334, 2.3839, 2.9021, 2.6635, 2.9016, 3.0077], device='cuda:1'), covar=tensor([0.0262, 0.0530, 0.0768, 0.0572, 0.0327, 0.0445, 0.0293, 0.0359], device='cuda:1'), in_proj_covar=tensor([0.0243, 0.0252, 0.0241, 0.0243, 0.0252, 0.0251, 0.0252, 0.0254], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-02 21:57:53,331 INFO [train.py:904] (1/8) Epoch 30, batch 3650, loss[loss=0.1518, simple_loss=0.2388, pruned_loss=0.0324, over 16809.00 frames. ], tot_loss[loss=0.1609, simple_loss=0.2473, pruned_loss=0.03727, over 3290265.16 frames. ], batch size: 83, lr: 2.24e-03, grad_scale: 8.0 2023-05-02 21:57:55,115 INFO [optim.py:368] (1/8) 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 21:58:14,214 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=298018.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 21:59:06,232 INFO [train.py:904] (1/8) Epoch 30, batch 3700, loss[loss=0.1839, simple_loss=0.2692, pruned_loss=0.04924, over 16681.00 frames. ], tot_loss[loss=0.1622, simple_loss=0.2465, pruned_loss=0.03901, over 3288244.02 frames. ], batch size: 57, lr: 2.24e-03, grad_scale: 8.0 2023-05-02 22:00:17,692 INFO [train.py:904] (1/8) Epoch 30, batch 3750, loss[loss=0.1696, simple_loss=0.2497, pruned_loss=0.04481, over 16221.00 frames. ], tot_loss[loss=0.164, simple_loss=0.2474, pruned_loss=0.04027, over 3285026.99 frames. ], batch size: 165, lr: 2.24e-03, grad_scale: 8.0 2023-05-02 22:00:19,704 INFO [optim.py:368] (1/8) 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 22:00:51,110 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([1.8583, 2.4456, 2.5201, 3.8769, 3.0579, 3.8886, 1.6357, 2.8923], device='cuda:1'), covar=tensor([0.1393, 0.0806, 0.1225, 0.0203, 0.0176, 0.0359, 0.1678, 0.0867], device='cuda:1'), in_proj_covar=tensor([0.0175, 0.0184, 0.0203, 0.0210, 0.0209, 0.0221, 0.0213, 0.0202], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-05-02 22:01:30,455 INFO [train.py:904] (1/8) Epoch 30, batch 3800, loss[loss=0.1979, simple_loss=0.2642, pruned_loss=0.06581, over 16720.00 frames. ], tot_loss[loss=0.1657, simple_loss=0.2483, pruned_loss=0.04157, over 3291442.30 frames. ], batch size: 124, lr: 2.24e-03, grad_scale: 8.0 2023-05-02 22:02:05,822 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.1803, 2.6866, 2.6340, 4.5585, 3.3919, 4.0637, 1.9239, 3.0961], device='cuda:1'), covar=tensor([0.1209, 0.0901, 0.1351, 0.0173, 0.0308, 0.0442, 0.1571, 0.0905], device='cuda:1'), in_proj_covar=tensor([0.0175, 0.0184, 0.0203, 0.0210, 0.0209, 0.0221, 0.0213, 0.0202], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-05-02 22:02:43,875 INFO [train.py:904] (1/8) Epoch 30, batch 3850, loss[loss=0.1918, simple_loss=0.2749, pruned_loss=0.05436, over 12549.00 frames. ], tot_loss[loss=0.1666, simple_loss=0.2489, pruned_loss=0.04212, over 3282558.37 frames. ], batch size: 246, lr: 2.24e-03, grad_scale: 8.0 2023-05-02 22:02:44,948 INFO [optim.py:368] (1/8) 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 22:03:18,677 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.4405, 3.4208, 3.4580, 3.5531, 3.6041, 3.3297, 3.5687, 3.6711], device='cuda:1'), covar=tensor([0.1256, 0.0933, 0.1094, 0.0642, 0.0648, 0.2464, 0.1212, 0.0753], device='cuda:1'), in_proj_covar=tensor([0.0721, 0.0873, 0.1012, 0.0893, 0.0678, 0.0706, 0.0743, 0.0866], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-02 22:03:41,190 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.9541, 4.7136, 4.6041, 3.3164, 3.9372, 4.5834, 3.8878, 2.5993], device='cuda:1'), covar=tensor([0.0498, 0.0036, 0.0045, 0.0347, 0.0113, 0.0113, 0.0122, 0.0466], device='cuda:1'), in_proj_covar=tensor([0.0141, 0.0092, 0.0094, 0.0137, 0.0105, 0.0118, 0.0101, 0.0133], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0003, 0.0004], device='cuda:1') 2023-05-02 22:03:53,264 INFO [train.py:904] (1/8) Epoch 30, batch 3900, loss[loss=0.1923, simple_loss=0.2677, pruned_loss=0.05846, over 16240.00 frames. ], tot_loss[loss=0.1666, simple_loss=0.2485, pruned_loss=0.04239, over 3275532.07 frames. ], batch size: 35, lr: 2.24e-03, grad_scale: 8.0 2023-05-02 22:04:05,154 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.44 vs. limit=2.0 2023-05-02 22:05:04,297 INFO [train.py:904] (1/8) Epoch 30, batch 3950, loss[loss=0.1813, simple_loss=0.2652, pruned_loss=0.04875, over 15500.00 frames. ], tot_loss[loss=0.167, simple_loss=0.2482, pruned_loss=0.04292, over 3280318.93 frames. ], batch size: 190, lr: 2.24e-03, grad_scale: 8.0 2023-05-02 22:05:05,533 INFO [optim.py:368] (1/8) 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 22:05:17,510 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=298313.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 22:05:50,668 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.42 vs. limit=2.0 2023-05-02 22:06:15,411 INFO [train.py:904] (1/8) Epoch 30, batch 4000, loss[loss=0.1847, simple_loss=0.2574, pruned_loss=0.05606, over 16711.00 frames. ], tot_loss[loss=0.1676, simple_loss=0.2484, pruned_loss=0.04337, over 3279157.85 frames. ], batch size: 134, lr: 2.24e-03, grad_scale: 8.0 2023-05-02 22:07:25,647 INFO [train.py:904] (1/8) Epoch 30, batch 4050, loss[loss=0.1574, simple_loss=0.247, pruned_loss=0.0339, over 16743.00 frames. ], tot_loss[loss=0.1668, simple_loss=0.2488, pruned_loss=0.04241, over 3282894.54 frames. ], batch size: 76, lr: 2.24e-03, grad_scale: 8.0 2023-05-02 22:07:27,597 INFO [optim.py:368] (1/8) 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 22:08:16,070 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.4125, 3.2750, 3.6919, 1.8948, 3.8170, 3.8227, 2.9657, 2.8674], device='cuda:1'), covar=tensor([0.0846, 0.0292, 0.0205, 0.1176, 0.0090, 0.0148, 0.0436, 0.0492], device='cuda:1'), in_proj_covar=tensor([0.0150, 0.0112, 0.0104, 0.0140, 0.0089, 0.0135, 0.0132, 0.0132], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-05-02 22:08:37,048 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.8628, 3.0213, 3.4597, 2.1049, 2.9626, 2.2419, 3.3711, 3.3321], device='cuda:1'), covar=tensor([0.0236, 0.0978, 0.0547, 0.2239, 0.0872, 0.1046, 0.0598, 0.1109], device='cuda:1'), in_proj_covar=tensor([0.0165, 0.0175, 0.0173, 0.0160, 0.0151, 0.0135, 0.0148, 0.0189], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0005], device='cuda:1') 2023-05-02 22:08:37,711 INFO [train.py:904] (1/8) Epoch 30, batch 4100, loss[loss=0.164, simple_loss=0.2535, pruned_loss=0.03725, over 16925.00 frames. ], tot_loss[loss=0.1676, simple_loss=0.2507, pruned_loss=0.04224, over 3274280.55 frames. ], batch size: 109, lr: 2.24e-03, grad_scale: 8.0 2023-05-02 22:09:19,616 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=298482.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 22:09:48,888 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.4982, 2.7084, 2.2689, 2.4148, 3.0386, 2.6078, 3.0188, 3.1723], device='cuda:1'), covar=tensor([0.0136, 0.0437, 0.0608, 0.0544, 0.0270, 0.0431, 0.0232, 0.0288], device='cuda:1'), in_proj_covar=tensor([0.0240, 0.0249, 0.0239, 0.0240, 0.0250, 0.0247, 0.0249, 0.0251], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-02 22:09:53,963 INFO [train.py:904] (1/8) Epoch 30, batch 4150, loss[loss=0.182, simple_loss=0.2759, pruned_loss=0.04406, over 15314.00 frames. ], tot_loss[loss=0.1725, simple_loss=0.2569, pruned_loss=0.04403, over 3235740.46 frames. ], batch size: 190, lr: 2.24e-03, grad_scale: 8.0 2023-05-02 22:09:56,053 INFO [optim.py:368] (1/8) 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 22:10:35,425 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=298530.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 22:10:55,722 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=298543.0, num_to_drop=1, layers_to_drop={0} 2023-05-02 22:11:11,635 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.8343, 2.2570, 1.8124, 2.0421, 2.5699, 2.2187, 2.4596, 2.7454], device='cuda:1'), covar=tensor([0.0243, 0.0523, 0.0717, 0.0642, 0.0354, 0.0487, 0.0279, 0.0328], device='cuda:1'), in_proj_covar=tensor([0.0240, 0.0249, 0.0239, 0.0240, 0.0250, 0.0247, 0.0249, 0.0251], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-02 22:11:12,252 INFO [train.py:904] (1/8) Epoch 30, batch 4200, loss[loss=0.2056, simple_loss=0.3029, pruned_loss=0.05418, over 16891.00 frames. ], tot_loss[loss=0.1779, simple_loss=0.2638, pruned_loss=0.04603, over 3181888.41 frames. ], batch size: 116, lr: 2.24e-03, grad_scale: 8.0 2023-05-02 22:11:16,052 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.4642, 3.5701, 2.1858, 4.1514, 2.8056, 4.0608, 2.3948, 2.8774], device='cuda:1'), covar=tensor([0.0373, 0.0469, 0.1884, 0.0228, 0.0890, 0.0629, 0.1614, 0.0899], device='cuda:1'), in_proj_covar=tensor([0.0180, 0.0185, 0.0198, 0.0178, 0.0183, 0.0225, 0.0206, 0.0185], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-05-02 22:11:42,419 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.5058, 3.5961, 3.3003, 2.8103, 3.1547, 3.4401, 3.3055, 3.2594], device='cuda:1'), covar=tensor([0.0628, 0.0700, 0.0338, 0.0324, 0.0520, 0.0470, 0.1438, 0.0515], device='cuda:1'), in_proj_covar=tensor([0.0327, 0.0495, 0.0385, 0.0386, 0.0380, 0.0444, 0.0263, 0.0457], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-05-02 22:11:51,669 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.9228, 5.1755, 5.3739, 5.0556, 5.1912, 5.7366, 5.1746, 4.9043], device='cuda:1'), covar=tensor([0.0935, 0.1812, 0.1484, 0.1860, 0.2026, 0.0698, 0.1331, 0.2403], device='cuda:1'), in_proj_covar=tensor([0.0437, 0.0648, 0.0719, 0.0527, 0.0702, 0.0740, 0.0555, 0.0700], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-05-02 22:11:58,520 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([5.3239, 5.3995, 5.7128, 5.6648, 5.7900, 5.4228, 5.2895, 5.1274], device='cuda:1'), covar=tensor([0.0307, 0.0436, 0.0421, 0.0439, 0.0381, 0.0340, 0.1202, 0.0479], device='cuda:1'), in_proj_covar=tensor([0.0450, 0.0514, 0.0489, 0.0453, 0.0535, 0.0517, 0.0597, 0.0417], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:1') 2023-05-02 22:12:08,272 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=298591.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 22:12:26,985 INFO [train.py:904] (1/8) Epoch 30, batch 4250, loss[loss=0.1778, simple_loss=0.2705, pruned_loss=0.04256, over 16970.00 frames. ], tot_loss[loss=0.1789, simple_loss=0.2669, pruned_loss=0.04547, over 3170605.84 frames. ], batch size: 41, lr: 2.24e-03, grad_scale: 8.0 2023-05-02 22:12:28,280 INFO [optim.py:368] (1/8) 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 22:12:40,711 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=298613.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 22:13:04,087 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=298629.0, num_to_drop=1, layers_to_drop={1} 2023-05-02 22:13:41,923 INFO [train.py:904] (1/8) Epoch 30, batch 4300, loss[loss=0.1703, simple_loss=0.2635, pruned_loss=0.03854, over 16730.00 frames. ], tot_loss[loss=0.1792, simple_loss=0.2685, pruned_loss=0.04496, over 3174221.94 frames. ], batch size: 83, lr: 2.24e-03, grad_scale: 8.0 2023-05-02 22:13:52,426 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=298661.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 22:14:15,361 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.2285, 3.9890, 3.8402, 2.4359, 3.4651, 3.9559, 3.4989, 2.2391], device='cuda:1'), covar=tensor([0.0644, 0.0044, 0.0076, 0.0498, 0.0128, 0.0097, 0.0117, 0.0492], device='cuda:1'), in_proj_covar=tensor([0.0141, 0.0092, 0.0094, 0.0137, 0.0106, 0.0118, 0.0101, 0.0133], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0003, 0.0004], device='cuda:1') 2023-05-02 22:14:15,701 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.33 vs. limit=2.0 2023-05-02 22:14:35,122 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=298689.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 22:14:36,373 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=298690.0, num_to_drop=1, layers_to_drop={0} 2023-05-02 22:14:44,029 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.24 vs. limit=2.0 2023-05-02 22:14:56,148 INFO [train.py:904] (1/8) Epoch 30, batch 4350, loss[loss=0.2009, simple_loss=0.2906, pruned_loss=0.0556, over 16233.00 frames. ], tot_loss[loss=0.1814, simple_loss=0.2714, pruned_loss=0.04575, over 3179853.91 frames. ], batch size: 165, lr: 2.24e-03, grad_scale: 8.0 2023-05-02 22:14:57,381 INFO [optim.py:368] (1/8) 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 22:15:34,870 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=4.28 vs. limit=5.0 2023-05-02 22:16:03,942 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=298750.0, num_to_drop=1, layers_to_drop={2} 2023-05-02 22:16:08,893 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.1727, 1.6243, 2.0639, 2.1433, 2.2672, 2.4375, 1.7998, 2.3739], device='cuda:1'), covar=tensor([0.0290, 0.0572, 0.0324, 0.0372, 0.0353, 0.0254, 0.0595, 0.0171], device='cuda:1'), in_proj_covar=tensor([0.0203, 0.0202, 0.0191, 0.0197, 0.0215, 0.0172, 0.0207, 0.0172], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0002], device='cuda:1') 2023-05-02 22:16:09,613 INFO [train.py:904] (1/8) Epoch 30, batch 4400, loss[loss=0.1991, simple_loss=0.2822, pruned_loss=0.05798, over 15518.00 frames. ], tot_loss[loss=0.1837, simple_loss=0.2737, pruned_loss=0.04689, over 3191197.61 frames. ], batch size: 191, lr: 2.24e-03, grad_scale: 8.0 2023-05-02 22:16:22,398 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.25 vs. limit=2.0 2023-05-02 22:17:21,872 INFO [train.py:904] (1/8) Epoch 30, batch 4450, loss[loss=0.209, simple_loss=0.303, pruned_loss=0.05755, over 17092.00 frames. ], tot_loss[loss=0.1872, simple_loss=0.2776, pruned_loss=0.04835, over 3208143.54 frames. ], batch size: 47, lr: 2.24e-03, grad_scale: 8.0 2023-05-02 22:17:23,559 INFO [optim.py:368] (1/8) 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 22:18:09,786 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=298838.0, num_to_drop=1, layers_to_drop={0} 2023-05-02 22:18:32,505 INFO [train.py:904] (1/8) Epoch 30, batch 4500, loss[loss=0.1845, simple_loss=0.276, pruned_loss=0.04644, over 16449.00 frames. ], tot_loss[loss=0.1879, simple_loss=0.278, pruned_loss=0.04894, over 3211337.10 frames. ], batch size: 68, lr: 2.24e-03, grad_scale: 8.0 2023-05-02 22:19:15,948 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.69 vs. limit=2.0 2023-05-02 22:19:19,727 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=298886.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 22:19:29,767 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.3024, 2.3859, 3.0375, 3.2386, 3.2858, 3.8287, 2.5485, 3.8685], device='cuda:1'), covar=tensor([0.0222, 0.0523, 0.0318, 0.0282, 0.0284, 0.0161, 0.0579, 0.0134], device='cuda:1'), in_proj_covar=tensor([0.0203, 0.0202, 0.0191, 0.0198, 0.0216, 0.0172, 0.0208, 0.0172], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0002], device='cuda:1') 2023-05-02 22:19:32,975 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.5724, 1.7903, 2.2843, 2.5325, 2.5935, 2.8707, 1.9478, 2.8553], device='cuda:1'), covar=tensor([0.0265, 0.0630, 0.0371, 0.0381, 0.0361, 0.0240, 0.0636, 0.0173], device='cuda:1'), in_proj_covar=tensor([0.0203, 0.0202, 0.0191, 0.0198, 0.0216, 0.0172, 0.0208, 0.0173], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0002], device='cuda:1') 2023-05-02 22:19:44,878 INFO [train.py:904] (1/8) Epoch 30, batch 4550, loss[loss=0.1899, simple_loss=0.2864, pruned_loss=0.04673, over 16773.00 frames. ], tot_loss[loss=0.1896, simple_loss=0.2791, pruned_loss=0.05007, over 3222919.14 frames. ], batch size: 83, lr: 2.24e-03, grad_scale: 8.0 2023-05-02 22:19:46,102 INFO [optim.py:368] (1/8) 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 22:20:27,785 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.0041, 2.4528, 2.0322, 2.1832, 2.7612, 2.3748, 2.6091, 2.8740], device='cuda:1'), covar=tensor([0.0224, 0.0447, 0.0636, 0.0528, 0.0306, 0.0424, 0.0245, 0.0298], device='cuda:1'), in_proj_covar=tensor([0.0237, 0.0248, 0.0237, 0.0238, 0.0248, 0.0246, 0.0247, 0.0248], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-02 22:20:57,288 INFO [train.py:904] (1/8) Epoch 30, batch 4600, loss[loss=0.1813, simple_loss=0.2754, pruned_loss=0.04359, over 16705.00 frames. ], tot_loss[loss=0.1904, simple_loss=0.2799, pruned_loss=0.05044, over 3228607.78 frames. ], batch size: 89, lr: 2.24e-03, grad_scale: 8.0 2023-05-02 22:21:09,681 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.44 vs. limit=2.0 2023-05-02 22:21:42,888 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=298985.0, num_to_drop=1, layers_to_drop={1} 2023-05-02 22:22:09,592 INFO [train.py:904] (1/8) Epoch 30, batch 4650, loss[loss=0.212, simple_loss=0.2872, pruned_loss=0.06843, over 12054.00 frames. ], tot_loss[loss=0.1908, simple_loss=0.2797, pruned_loss=0.05098, over 3228230.44 frames. ], batch size: 248, lr: 2.24e-03, grad_scale: 8.0 2023-05-02 22:22:10,879 INFO [optim.py:368] (1/8) 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 22:22:41,785 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.8068, 3.8306, 3.9329, 3.6166, 3.8423, 4.2531, 3.8507, 3.5572], device='cuda:1'), covar=tensor([0.2086, 0.2230, 0.2405, 0.2800, 0.2741, 0.1765, 0.1623, 0.2676], device='cuda:1'), in_proj_covar=tensor([0.0433, 0.0638, 0.0710, 0.0520, 0.0692, 0.0731, 0.0548, 0.0691], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-05-02 22:23:09,774 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=299045.0, num_to_drop=1, layers_to_drop={2} 2023-05-02 22:23:11,055 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([1.5619, 2.3457, 2.2585, 3.5746, 2.2516, 3.6224, 1.5292, 2.6358], device='cuda:1'), covar=tensor([0.1604, 0.1030, 0.1494, 0.0223, 0.0217, 0.0425, 0.1982, 0.0990], device='cuda:1'), in_proj_covar=tensor([0.0175, 0.0184, 0.0203, 0.0209, 0.0209, 0.0220, 0.0213, 0.0202], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-05-02 22:23:22,504 INFO [train.py:904] (1/8) Epoch 30, batch 4700, loss[loss=0.1887, simple_loss=0.2647, pruned_loss=0.05634, over 11694.00 frames. ], tot_loss[loss=0.1891, simple_loss=0.2775, pruned_loss=0.05034, over 3198460.90 frames. ], batch size: 248, lr: 2.24e-03, grad_scale: 8.0 2023-05-02 22:23:25,653 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.14 vs. limit=2.0 2023-05-02 22:23:39,700 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.5926, 2.4906, 1.8783, 2.6853, 2.0699, 2.7497, 2.0953, 2.3008], device='cuda:1'), covar=tensor([0.0365, 0.0426, 0.1398, 0.0239, 0.0764, 0.0490, 0.1362, 0.0778], device='cuda:1'), in_proj_covar=tensor([0.0179, 0.0184, 0.0197, 0.0176, 0.0182, 0.0224, 0.0206, 0.0185], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-05-02 22:24:36,507 INFO [train.py:904] (1/8) Epoch 30, batch 4750, loss[loss=0.1434, simple_loss=0.2385, pruned_loss=0.02413, over 16792.00 frames. ], tot_loss[loss=0.1851, simple_loss=0.2736, pruned_loss=0.04829, over 3183975.79 frames. ], batch size: 83, lr: 2.24e-03, grad_scale: 8.0 2023-05-02 22:24:37,716 INFO [optim.py:368] (1/8) 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 22:25:05,265 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=4.18 vs. limit=5.0 2023-05-02 22:25:07,699 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=299126.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 22:25:25,734 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=299138.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 22:25:25,888 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([1.8316, 1.3807, 1.7438, 1.6497, 1.8485, 1.9417, 1.7015, 1.8852], device='cuda:1'), covar=tensor([0.0285, 0.0538, 0.0269, 0.0422, 0.0368, 0.0279, 0.0551, 0.0221], device='cuda:1'), in_proj_covar=tensor([0.0202, 0.0202, 0.0190, 0.0197, 0.0215, 0.0171, 0.0207, 0.0172], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0002], device='cuda:1') 2023-05-02 22:25:49,328 INFO [train.py:904] (1/8) Epoch 30, batch 4800, loss[loss=0.1916, simple_loss=0.2662, pruned_loss=0.05849, over 11823.00 frames. ], tot_loss[loss=0.1808, simple_loss=0.2696, pruned_loss=0.04597, over 3189807.21 frames. ], batch size: 246, lr: 2.24e-03, grad_scale: 8.0 2023-05-02 22:26:37,231 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=299186.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 22:26:37,307 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=299186.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 22:26:38,487 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=299187.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 22:26:45,068 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=299191.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 22:26:49,640 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.2622, 3.9436, 3.9779, 2.6241, 3.5091, 3.9868, 3.5873, 2.2776], device='cuda:1'), covar=tensor([0.0645, 0.0069, 0.0062, 0.0456, 0.0118, 0.0141, 0.0128, 0.0519], device='cuda:1'), in_proj_covar=tensor([0.0140, 0.0092, 0.0094, 0.0137, 0.0105, 0.0118, 0.0101, 0.0132], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0003, 0.0004], device='cuda:1') 2023-05-02 22:27:04,357 INFO [train.py:904] (1/8) Epoch 30, batch 4850, loss[loss=0.1911, simple_loss=0.2928, pruned_loss=0.04471, over 16747.00 frames. ], tot_loss[loss=0.1804, simple_loss=0.2703, pruned_loss=0.04532, over 3174123.38 frames. ], batch size: 124, lr: 2.24e-03, grad_scale: 8.0 2023-05-02 22:27:06,457 INFO [optim.py:368] (1/8) 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] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=299234.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 22:28:02,961 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=299243.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 22:28:16,423 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=299252.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 22:28:18,992 INFO [train.py:904] (1/8) Epoch 30, batch 4900, loss[loss=0.1535, simple_loss=0.2483, pruned_loss=0.0293, over 16781.00 frames. ], tot_loss[loss=0.1783, simple_loss=0.269, pruned_loss=0.04382, over 3170957.80 frames. ], batch size: 83, lr: 2.24e-03, grad_scale: 8.0 2023-05-02 22:28:20,374 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.70 vs. limit=2.0 2023-05-02 22:28:21,425 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.16 vs. limit=2.0 2023-05-02 22:29:00,248 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=4.81 vs. limit=5.0 2023-05-02 22:29:05,979 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=299285.0, num_to_drop=1, layers_to_drop={0} 2023-05-02 22:29:10,824 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([1.7386, 2.5234, 2.6366, 4.2733, 2.8913, 4.0292, 1.5083, 2.9763], device='cuda:1'), covar=tensor([0.1328, 0.0866, 0.1202, 0.0152, 0.0204, 0.0318, 0.1718, 0.0797], device='cuda:1'), in_proj_covar=tensor([0.0174, 0.0183, 0.0202, 0.0207, 0.0207, 0.0218, 0.0211, 0.0201], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-05-02 22:29:33,775 INFO [train.py:904] (1/8) Epoch 30, batch 4950, loss[loss=0.1767, simple_loss=0.2688, pruned_loss=0.04235, over 16502.00 frames. ], tot_loss[loss=0.177, simple_loss=0.2679, pruned_loss=0.04309, over 3192028.26 frames. ], batch size: 68, lr: 2.24e-03, grad_scale: 8.0 2023-05-02 22:29:34,206 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=299304.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 22:29:34,846 INFO [optim.py:368] (1/8) 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 22:30:02,251 INFO [zipformer.py:625] (1/8) 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] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=299333.0, num_to_drop=1, layers_to_drop={0} 2023-05-02 22:30:36,117 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 30, batch 5000, loss[loss=0.1736, simple_loss=0.2694, pruned_loss=0.03887, over 16869.00 frames. ], tot_loss[loss=0.1781, simple_loss=0.2695, pruned_loss=0.04338, over 3180113.84 frames. ], batch size: 109, lr: 2.24e-03, grad_scale: 8.0 2023-05-02 22:30:51,175 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.25 vs. limit=2.0 2023-05-02 22:31:33,628 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=299384.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 22:31:45,863 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 30, batch 5050, loss[loss=0.1642, simple_loss=0.2592, pruned_loss=0.03459, over 16901.00 frames. ], tot_loss[loss=0.178, simple_loss=0.2701, pruned_loss=0.04294, over 3204169.03 frames. ], batch size: 109, lr: 2.24e-03, grad_scale: 8.0 2023-05-02 22:32:02,875 INFO [optim.py:368] (1/8) 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 22:32:06,434 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.8110, 1.9343, 2.5334, 2.8015, 2.8082, 3.2402, 2.2608, 3.2171], device='cuda:1'), covar=tensor([0.0271, 0.0625, 0.0400, 0.0410, 0.0359, 0.0239, 0.0619, 0.0194], device='cuda:1'), in_proj_covar=tensor([0.0201, 0.0200, 0.0189, 0.0196, 0.0214, 0.0170, 0.0207, 0.0171], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0002], device='cuda:1') 2023-05-02 22:33:13,772 INFO [train.py:904] (1/8) Epoch 30, batch 5100, loss[loss=0.1549, simple_loss=0.2443, pruned_loss=0.03276, over 16525.00 frames. ], tot_loss[loss=0.1768, simple_loss=0.2687, pruned_loss=0.04246, over 3204832.67 frames. ], batch size: 75, lr: 2.24e-03, grad_scale: 8.0 2023-05-02 22:33:17,270 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.4599, 4.2705, 4.1834, 2.7676, 3.6662, 4.1813, 3.6715, 2.3948], device='cuda:1'), covar=tensor([0.0606, 0.0044, 0.0055, 0.0448, 0.0111, 0.0123, 0.0130, 0.0489], device='cuda:1'), in_proj_covar=tensor([0.0141, 0.0092, 0.0094, 0.0137, 0.0105, 0.0118, 0.0101, 0.0132], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0003, 0.0004], device='cuda:1') 2023-05-02 22:33:58,043 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 30, batch 5150, loss[loss=0.1787, simple_loss=0.2757, pruned_loss=0.0408, over 16883.00 frames. ], tot_loss[loss=0.176, simple_loss=0.2683, pruned_loss=0.04183, over 3198534.00 frames. ], batch size: 116, lr: 2.24e-03, grad_scale: 8.0 2023-05-02 22:34:31,338 INFO [optim.py:368] (1/8) 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 22:34:34,024 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.22 vs. limit=2.0 2023-05-02 22:34:57,243 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.3836, 4.3344, 4.2659, 2.7471, 3.7682, 4.2535, 3.6958, 2.2939], device='cuda:1'), covar=tensor([0.0712, 0.0053, 0.0050, 0.0484, 0.0112, 0.0130, 0.0133, 0.0539], device='cuda:1'), in_proj_covar=tensor([0.0142, 0.0093, 0.0095, 0.0138, 0.0106, 0.0119, 0.0102, 0.0133], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0003, 0.0004], device='cuda:1') 2023-05-02 22:35:33,227 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=299547.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 22:35:43,064 INFO [train.py:904] (1/8) Epoch 30, batch 5200, loss[loss=0.1729, simple_loss=0.261, pruned_loss=0.04245, over 16597.00 frames. ], tot_loss[loss=0.174, simple_loss=0.2666, pruned_loss=0.04069, over 3204706.83 frames. ], batch size: 57, lr: 2.24e-03, grad_scale: 8.0 2023-05-02 22:35:58,601 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.0429, 3.9880, 3.9466, 3.1497, 3.9377, 1.8215, 3.7494, 3.4659], device='cuda:1'), covar=tensor([0.0143, 0.0156, 0.0209, 0.0399, 0.0122, 0.3053, 0.0172, 0.0339], device='cuda:1'), in_proj_covar=tensor([0.0187, 0.0178, 0.0217, 0.0188, 0.0195, 0.0221, 0.0205, 0.0183], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-02 22:36:48,483 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=299599.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 22:36:55,730 INFO [train.py:904] (1/8) Epoch 30, batch 5250, loss[loss=0.1561, simple_loss=0.2566, pruned_loss=0.02778, over 16930.00 frames. ], tot_loss[loss=0.1721, simple_loss=0.264, pruned_loss=0.04008, over 3210473.66 frames. ], batch size: 90, lr: 2.24e-03, grad_scale: 8.0 2023-05-02 22:36:56,956 INFO [optim.py:368] (1/8) 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 22:37:18,250 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=4.04 vs. limit=5.0 2023-05-02 22:38:06,895 INFO [train.py:904] (1/8) Epoch 30, batch 5300, loss[loss=0.1525, simple_loss=0.2477, pruned_loss=0.02867, over 16190.00 frames. ], tot_loss[loss=0.1691, simple_loss=0.2605, pruned_loss=0.03891, over 3213152.67 frames. ], batch size: 165, lr: 2.24e-03, grad_scale: 8.0 2023-05-02 22:38:31,751 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.19 vs. limit=2.0 2023-05-02 22:38:42,815 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=299679.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 22:39:18,177 INFO [train.py:904] (1/8) Epoch 30, batch 5350, loss[loss=0.1588, simple_loss=0.2476, pruned_loss=0.03501, over 17168.00 frames. ], tot_loss[loss=0.1681, simple_loss=0.2591, pruned_loss=0.03856, over 3200718.25 frames. ], batch size: 46, lr: 2.24e-03, grad_scale: 8.0 2023-05-02 22:39:19,343 INFO [optim.py:368] (1/8) 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 22:39:30,407 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.0623, 2.4828, 2.5892, 1.9230, 2.7698, 2.8193, 2.4926, 2.3962], device='cuda:1'), covar=tensor([0.0760, 0.0296, 0.0237, 0.1008, 0.0136, 0.0249, 0.0459, 0.0509], device='cuda:1'), in_proj_covar=tensor([0.0150, 0.0112, 0.0103, 0.0140, 0.0088, 0.0134, 0.0132, 0.0132], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-05-02 22:40:29,714 INFO [train.py:904] (1/8) Epoch 30, batch 5400, loss[loss=0.1722, simple_loss=0.2618, pruned_loss=0.04125, over 17131.00 frames. ], tot_loss[loss=0.1707, simple_loss=0.2622, pruned_loss=0.03959, over 3196322.86 frames. ], batch size: 47, lr: 2.24e-03, grad_scale: 8.0 2023-05-02 22:40:51,533 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=299769.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 22:41:02,313 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.77 vs. limit=2.0 2023-05-02 22:41:09,381 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=299782.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 22:41:34,618 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=299798.0, num_to_drop=1, layers_to_drop={0} 2023-05-02 22:41:46,068 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([1.9840, 2.2791, 2.3344, 2.8943, 1.8643, 3.1870, 1.8435, 2.7443], device='cuda:1'), covar=tensor([0.1227, 0.0770, 0.1199, 0.0190, 0.0115, 0.0319, 0.1658, 0.0739], device='cuda:1'), in_proj_covar=tensor([0.0175, 0.0184, 0.0203, 0.0208, 0.0209, 0.0219, 0.0213, 0.0202], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-05-02 22:41:46,646 INFO [train.py:904] (1/8) Epoch 30, batch 5450, loss[loss=0.21, simple_loss=0.2877, pruned_loss=0.0661, over 12005.00 frames. ], tot_loss[loss=0.173, simple_loss=0.2649, pruned_loss=0.04055, over 3187230.45 frames. ], batch size: 248, lr: 2.23e-03, grad_scale: 8.0 2023-05-02 22:41:47,797 INFO [optim.py:368] (1/8) 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] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=299830.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 22:42:27,221 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=299830.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 22:42:45,605 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.3706, 3.4816, 3.6081, 3.5888, 3.6190, 3.4554, 3.4709, 3.4941], device='cuda:1'), covar=tensor([0.0444, 0.0737, 0.0517, 0.0495, 0.0550, 0.0593, 0.0823, 0.0577], device='cuda:1'), in_proj_covar=tensor([0.0439, 0.0502, 0.0480, 0.0445, 0.0525, 0.0508, 0.0585, 0.0410], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:1') 2023-05-02 22:42:51,976 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 30, batch 5500, loss[loss=0.2276, simple_loss=0.3013, pruned_loss=0.07697, over 11559.00 frames. ], tot_loss[loss=0.1803, simple_loss=0.2716, pruned_loss=0.04454, over 3151569.02 frames. ], batch size: 247, lr: 2.23e-03, grad_scale: 16.0 2023-05-02 22:43:11,020 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=299859.0, num_to_drop=1, layers_to_drop={0} 2023-05-02 22:44:07,185 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.4239, 3.0408, 2.6012, 2.2623, 2.3437, 2.2410, 3.0858, 2.9229], device='cuda:1'), covar=tensor([0.2772, 0.0679, 0.1804, 0.2569, 0.2534, 0.2342, 0.0592, 0.1305], device='cuda:1'), in_proj_covar=tensor([0.0337, 0.0276, 0.0315, 0.0330, 0.0307, 0.0281, 0.0305, 0.0353], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:1') 2023-05-02 22:44:09,163 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=299895.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 22:44:14,130 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=299899.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 22:44:21,843 INFO [train.py:904] (1/8) Epoch 30, batch 5550, loss[loss=0.2212, simple_loss=0.3051, pruned_loss=0.06868, over 16264.00 frames. ], tot_loss[loss=0.1886, simple_loss=0.2786, pruned_loss=0.04929, over 3128740.58 frames. ], batch size: 165, lr: 2.23e-03, grad_scale: 16.0 2023-05-02 22:44:23,798 INFO [optim.py:368] (1/8) 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 22:44:31,938 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.78 vs. limit=2.0 2023-05-02 22:45:31,925 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 30, batch 5600, loss[loss=0.2117, simple_loss=0.2929, pruned_loss=0.06523, over 16690.00 frames. ], tot_loss[loss=0.195, simple_loss=0.2832, pruned_loss=0.05339, over 3089119.75 frames. ], batch size: 134, lr: 2.23e-03, grad_scale: 16.0 2023-05-02 22:46:24,522 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.3532, 3.3497, 3.3788, 3.4663, 3.4856, 3.2663, 3.4665, 3.5296], device='cuda:1'), covar=tensor([0.1211, 0.0938, 0.1092, 0.0720, 0.0724, 0.2150, 0.1187, 0.0966], device='cuda:1'), in_proj_covar=tensor([0.0679, 0.0827, 0.0957, 0.0846, 0.0642, 0.0668, 0.0701, 0.0817], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-02 22:46:27,194 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=299979.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 22:46:47,539 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=4.45 vs. limit=5.0 2023-05-02 22:47:11,381 INFO [train.py:904] (1/8) Epoch 30, batch 5650, loss[loss=0.2023, simple_loss=0.2876, pruned_loss=0.05845, over 16717.00 frames. ], tot_loss[loss=0.2013, simple_loss=0.2881, pruned_loss=0.05726, over 3062888.42 frames. ], batch size: 57, lr: 2.23e-03, grad_scale: 16.0 2023-05-02 22:47:13,274 INFO [optim.py:368] (1/8) 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] (1/8) 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] (1/8) Epoch 30, batch 5700, loss[loss=0.1963, simple_loss=0.2838, pruned_loss=0.05438, over 16684.00 frames. ], tot_loss[loss=0.2027, simple_loss=0.2889, pruned_loss=0.05826, over 3071124.92 frames. ], batch size: 57, lr: 2.23e-03, grad_scale: 16.0 2023-05-02 22:48:57,315 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=300073.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 22:49:45,350 INFO [train.py:904] (1/8) Epoch 30, batch 5750, loss[loss=0.2003, simple_loss=0.2905, pruned_loss=0.05504, over 16468.00 frames. ], tot_loss[loss=0.2053, simple_loss=0.2915, pruned_loss=0.05957, over 3037487.14 frames. ], batch size: 75, lr: 2.23e-03, grad_scale: 8.0 2023-05-02 22:49:49,226 INFO [optim.py:368] (1/8) 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 22:50:18,718 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=300125.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 22:50:34,999 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=300134.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 22:51:06,732 INFO [train.py:904] (1/8) Epoch 30, batch 5800, loss[loss=0.1562, simple_loss=0.2554, pruned_loss=0.0285, over 16883.00 frames. ], tot_loss[loss=0.2037, simple_loss=0.2912, pruned_loss=0.05806, over 3036923.84 frames. ], batch size: 96, lr: 2.23e-03, grad_scale: 8.0 2023-05-02 22:51:07,247 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=300154.0, num_to_drop=1, layers_to_drop={0} 2023-05-02 22:52:25,619 INFO [train.py:904] (1/8) Epoch 30, batch 5850, loss[loss=0.2392, simple_loss=0.3029, pruned_loss=0.08771, over 11812.00 frames. ], tot_loss[loss=0.2013, simple_loss=0.2892, pruned_loss=0.05674, over 3044871.57 frames. ], batch size: 248, lr: 2.23e-03, grad_scale: 8.0 2023-05-02 22:52:28,949 INFO [optim.py:368] (1/8) Clipping_scale=2.0, grad-norm quartiles 1.843e+02 2.638e+02 2.992e+02 3.617e+02 7.830e+02, threshold=5.985e+02, percent-clipped=1.0 2023-05-02 22:53:48,213 INFO [train.py:904] (1/8) Epoch 30, batch 5900, loss[loss=0.2002, simple_loss=0.2741, pruned_loss=0.0632, over 11472.00 frames. ], tot_loss[loss=0.2013, simple_loss=0.2886, pruned_loss=0.05702, over 3030203.23 frames. ], batch size: 246, lr: 2.23e-03, grad_scale: 8.0 2023-05-02 22:54:19,126 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.26 vs. limit=2.0 2023-05-02 22:54:38,423 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.5680, 2.6242, 2.3157, 2.4259, 3.0231, 2.6905, 3.0656, 3.2609], device='cuda:1'), covar=tensor([0.0177, 0.0515, 0.0608, 0.0537, 0.0328, 0.0450, 0.0287, 0.0297], device='cuda:1'), in_proj_covar=tensor([0.0234, 0.0244, 0.0234, 0.0235, 0.0245, 0.0242, 0.0242, 0.0244], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-02 22:55:09,236 INFO [train.py:904] (1/8) Epoch 30, batch 5950, loss[loss=0.1803, simple_loss=0.2785, pruned_loss=0.04101, over 16666.00 frames. ], tot_loss[loss=0.2003, simple_loss=0.2889, pruned_loss=0.05584, over 3052931.41 frames. ], batch size: 62, lr: 2.23e-03, grad_scale: 8.0 2023-05-02 22:55:12,894 INFO [optim.py:368] (1/8) Clipping_scale=2.0, grad-norm quartiles 1.875e+02 2.782e+02 3.249e+02 4.004e+02 6.648e+02, threshold=6.499e+02, percent-clipped=2.0 2023-05-02 22:56:29,061 INFO [train.py:904] (1/8) Epoch 30, batch 6000, loss[loss=0.1725, simple_loss=0.2623, pruned_loss=0.04134, over 16778.00 frames. ], tot_loss[loss=0.1993, simple_loss=0.2879, pruned_loss=0.05536, over 3077187.78 frames. ], batch size: 83, lr: 2.23e-03, grad_scale: 8.0 2023-05-02 22:56:29,061 INFO [train.py:929] (1/8) Computing validation loss 2023-05-02 22:56:39,790 INFO [train.py:938] (1/8) Epoch 30, validation: loss=0.1471, simple_loss=0.2591, pruned_loss=0.01755, over 944034.00 frames. 2023-05-02 22:56:39,790 INFO [train.py:939] (1/8) Maximum memory allocated so far is 18145MB 2023-05-02 22:57:57,799 INFO [train.py:904] (1/8) Epoch 30, batch 6050, loss[loss=0.1814, simple_loss=0.2803, pruned_loss=0.04132, over 17002.00 frames. ], tot_loss[loss=0.198, simple_loss=0.2867, pruned_loss=0.05469, over 3105242.93 frames. ], batch size: 55, lr: 2.23e-03, grad_scale: 8.0 2023-05-02 22:58:01,295 INFO [optim.py:368] (1/8) 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 22:58:29,386 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=300425.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 22:58:35,424 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 30, batch 6100, loss[loss=0.2043, simple_loss=0.2941, pruned_loss=0.05725, over 16644.00 frames. ], tot_loss[loss=0.1971, simple_loss=0.2862, pruned_loss=0.05396, over 3117986.36 frames. ], batch size: 62, lr: 2.23e-03, grad_scale: 8.0 2023-05-02 22:59:16,323 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=300454.0, num_to_drop=1, layers_to_drop={1} 2023-05-02 22:59:42,192 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.3214, 3.0786, 3.3852, 1.8159, 3.5429, 3.5431, 2.8484, 2.6995], device='cuda:1'), covar=tensor([0.0870, 0.0334, 0.0237, 0.1313, 0.0106, 0.0247, 0.0485, 0.0546], device='cuda:1'), in_proj_covar=tensor([0.0150, 0.0111, 0.0103, 0.0139, 0.0088, 0.0134, 0.0132, 0.0132], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-05-02 22:59:47,466 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=300473.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 22:59:51,324 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.2892, 4.3457, 4.4996, 4.3113, 4.3951, 4.8667, 4.4074, 4.1402], device='cuda:1'), covar=tensor([0.1614, 0.1878, 0.2538, 0.1871, 0.2228, 0.1031, 0.1741, 0.2453], device='cuda:1'), in_proj_covar=tensor([0.0433, 0.0642, 0.0717, 0.0520, 0.0693, 0.0732, 0.0551, 0.0692], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-05-02 23:00:32,802 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=300502.0, num_to_drop=1, layers_to_drop={1} 2023-05-02 23:00:34,965 INFO [train.py:904] (1/8) Epoch 30, batch 6150, loss[loss=0.1836, simple_loss=0.2757, pruned_loss=0.04575, over 16732.00 frames. ], tot_loss[loss=0.1958, simple_loss=0.2847, pruned_loss=0.05347, over 3122276.37 frames. ], batch size: 83, lr: 2.23e-03, grad_scale: 8.0 2023-05-02 23:00:38,226 INFO [optim.py:368] (1/8) 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 23:01:53,680 INFO [train.py:904] (1/8) Epoch 30, batch 6200, loss[loss=0.1927, simple_loss=0.27, pruned_loss=0.05768, over 11422.00 frames. ], tot_loss[loss=0.1943, simple_loss=0.2829, pruned_loss=0.05286, over 3136591.08 frames. ], batch size: 248, lr: 2.23e-03, grad_scale: 8.0 2023-05-02 23:02:24,888 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.2072, 2.8161, 3.1228, 1.7898, 3.2423, 3.2762, 2.7231, 2.5265], device='cuda:1'), covar=tensor([0.0838, 0.0339, 0.0236, 0.1244, 0.0115, 0.0237, 0.0518, 0.0533], device='cuda:1'), in_proj_covar=tensor([0.0150, 0.0111, 0.0103, 0.0139, 0.0088, 0.0134, 0.0132, 0.0132], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-05-02 23:03:13,163 INFO [train.py:904] (1/8) Epoch 30, batch 6250, loss[loss=0.213, simple_loss=0.287, pruned_loss=0.06948, over 11597.00 frames. ], tot_loss[loss=0.194, simple_loss=0.2824, pruned_loss=0.05277, over 3124103.50 frames. ], batch size: 247, lr: 2.23e-03, grad_scale: 8.0 2023-05-02 23:03:16,359 INFO [optim.py:368] (1/8) 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 23:03:21,841 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.6446, 2.5984, 1.9901, 2.7060, 2.1952, 2.7930, 2.1485, 2.3832], device='cuda:1'), covar=tensor([0.0318, 0.0388, 0.1260, 0.0274, 0.0641, 0.0505, 0.1211, 0.0616], device='cuda:1'), in_proj_covar=tensor([0.0178, 0.0184, 0.0197, 0.0175, 0.0181, 0.0223, 0.0206, 0.0185], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-05-02 23:03:34,844 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=300617.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 23:03:53,438 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([5.3502, 5.4748, 5.7421, 5.7234, 5.7823, 5.4946, 5.3824, 5.1958], device='cuda:1'), covar=tensor([0.0338, 0.0584, 0.0458, 0.0429, 0.0429, 0.0461, 0.0844, 0.0478], device='cuda:1'), in_proj_covar=tensor([0.0441, 0.0503, 0.0481, 0.0447, 0.0527, 0.0509, 0.0585, 0.0412], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:1') 2023-05-02 23:04:31,360 INFO [train.py:904] (1/8) Epoch 30, batch 6300, loss[loss=0.1936, simple_loss=0.2828, pruned_loss=0.05217, over 15303.00 frames. ], tot_loss[loss=0.1928, simple_loss=0.2818, pruned_loss=0.05185, over 3134415.22 frames. ], batch size: 190, lr: 2.23e-03, grad_scale: 8.0 2023-05-02 23:05:07,642 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.7505, 3.5870, 4.2259, 2.0056, 4.3862, 4.3635, 3.2454, 3.2095], device='cuda:1'), covar=tensor([0.0796, 0.0319, 0.0180, 0.1298, 0.0071, 0.0203, 0.0405, 0.0513], device='cuda:1'), in_proj_covar=tensor([0.0150, 0.0112, 0.0104, 0.0140, 0.0088, 0.0135, 0.0132, 0.0132], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-05-02 23:05:10,929 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 30, batch 6350, loss[loss=0.1775, simple_loss=0.2687, pruned_loss=0.04316, over 16537.00 frames. ], tot_loss[loss=0.1957, simple_loss=0.2839, pruned_loss=0.05376, over 3130722.70 frames. ], batch size: 75, lr: 2.23e-03, grad_scale: 8.0 2023-05-02 23:05:53,978 INFO [optim.py:368] (1/8) 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 23:06:21,477 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=2.01 vs. limit=2.0 2023-05-02 23:06:29,163 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=300729.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 23:06:39,023 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.7831, 3.7371, 3.8613, 3.6271, 3.8552, 4.2376, 3.9347, 3.6716], device='cuda:1'), covar=tensor([0.2365, 0.2587, 0.3432, 0.2522, 0.2674, 0.2038, 0.1700, 0.2419], device='cuda:1'), in_proj_covar=tensor([0.0435, 0.0645, 0.0720, 0.0525, 0.0697, 0.0737, 0.0555, 0.0699], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-05-02 23:06:44,746 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.78 vs. limit=2.0 2023-05-02 23:07:07,467 INFO [train.py:904] (1/8) Epoch 30, batch 6400, loss[loss=0.1755, simple_loss=0.2665, pruned_loss=0.04224, over 16419.00 frames. ], tot_loss[loss=0.1971, simple_loss=0.2842, pruned_loss=0.05495, over 3112806.17 frames. ], batch size: 146, lr: 2.23e-03, grad_scale: 8.0 2023-05-02 23:07:35,222 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.8967, 4.7467, 4.9076, 5.1383, 5.3248, 4.7579, 5.2957, 5.3311], device='cuda:1'), covar=tensor([0.2407, 0.1484, 0.2114, 0.0963, 0.0826, 0.1047, 0.0923, 0.0733], device='cuda:1'), in_proj_covar=tensor([0.0685, 0.0830, 0.0959, 0.0850, 0.0644, 0.0669, 0.0707, 0.0820], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-02 23:07:42,791 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=300777.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 23:07:44,566 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.9666, 4.1928, 4.0128, 4.0627, 3.7806, 3.8152, 3.8378, 4.1939], device='cuda:1'), covar=tensor([0.1162, 0.0889, 0.1070, 0.0839, 0.0803, 0.1768, 0.0959, 0.1059], device='cuda:1'), in_proj_covar=tensor([0.0726, 0.0876, 0.0721, 0.0682, 0.0560, 0.0558, 0.0734, 0.0687], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:1') 2023-05-02 23:08:01,151 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.1715, 3.4934, 3.4484, 2.2322, 3.2463, 3.5408, 3.3230, 1.9415], device='cuda:1'), covar=tensor([0.0613, 0.0089, 0.0090, 0.0510, 0.0134, 0.0139, 0.0112, 0.0551], device='cuda:1'), in_proj_covar=tensor([0.0140, 0.0093, 0.0094, 0.0137, 0.0105, 0.0118, 0.0101, 0.0133], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0003, 0.0004], device='cuda:1') 2023-05-02 23:08:04,614 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.26 vs. limit=2.0 2023-05-02 23:08:18,838 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 30, batch 6450, loss[loss=0.1628, simple_loss=0.255, pruned_loss=0.03529, over 16518.00 frames. ], tot_loss[loss=0.1964, simple_loss=0.2843, pruned_loss=0.05423, over 3105390.55 frames. ], batch size: 75, lr: 2.23e-03, grad_scale: 8.0 2023-05-02 23:08:24,002 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=2.71 vs. limit=5.0 2023-05-02 23:08:24,280 INFO [optim.py:368] (1/8) Clipping_scale=2.0, grad-norm quartiles 1.818e+02 2.836e+02 3.339e+02 4.246e+02 8.060e+02, threshold=6.678e+02, percent-clipped=2.0 2023-05-02 23:09:38,124 INFO [train.py:904] (1/8) Epoch 30, batch 6500, loss[loss=0.2022, simple_loss=0.2914, pruned_loss=0.05653, over 16940.00 frames. ], tot_loss[loss=0.1943, simple_loss=0.282, pruned_loss=0.05329, over 3110078.98 frames. ], batch size: 109, lr: 2.23e-03, grad_scale: 8.0 2023-05-02 23:09:52,586 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=300863.0, num_to_drop=1, layers_to_drop={3} 2023-05-02 23:10:58,874 INFO [train.py:904] (1/8) Epoch 30, batch 6550, loss[loss=0.2042, simple_loss=0.2947, pruned_loss=0.05683, over 16795.00 frames. ], tot_loss[loss=0.1972, simple_loss=0.2851, pruned_loss=0.05468, over 3081270.14 frames. ], batch size: 39, lr: 2.23e-03, grad_scale: 8.0 2023-05-02 23:11:01,651 INFO [optim.py:368] (1/8) 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] (1/8) Epoch 30, batch 6600, loss[loss=0.1947, simple_loss=0.2824, pruned_loss=0.05354, over 16670.00 frames. ], tot_loss[loss=0.1978, simple_loss=0.2865, pruned_loss=0.05462, over 3084508.52 frames. ], batch size: 57, lr: 2.23e-03, grad_scale: 8.0 2023-05-02 23:12:48,796 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=300973.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 23:13:07,226 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.2753, 3.7008, 3.6573, 2.3828, 3.3785, 3.6953, 3.3834, 2.2080], device='cuda:1'), covar=tensor([0.0601, 0.0065, 0.0081, 0.0479, 0.0124, 0.0135, 0.0136, 0.0479], device='cuda:1'), in_proj_covar=tensor([0.0140, 0.0092, 0.0094, 0.0137, 0.0104, 0.0118, 0.0101, 0.0132], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0003, 0.0004], device='cuda:1') 2023-05-02 23:13:38,993 INFO [train.py:904] (1/8) Epoch 30, batch 6650, loss[loss=0.1774, simple_loss=0.2763, pruned_loss=0.03929, over 16768.00 frames. ], tot_loss[loss=0.1985, simple_loss=0.2865, pruned_loss=0.05525, over 3080059.65 frames. ], batch size: 83, lr: 2.23e-03, grad_scale: 4.0 2023-05-02 23:13:43,931 INFO [optim.py:368] (1/8) 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 23:13:48,260 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.66 vs. limit=2.0 2023-05-02 23:13:59,853 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.7130, 3.7806, 2.3189, 4.4368, 2.9332, 4.3345, 2.5451, 3.0793], device='cuda:1'), covar=tensor([0.0363, 0.0454, 0.1968, 0.0213, 0.0920, 0.0592, 0.1632, 0.0937], device='cuda:1'), in_proj_covar=tensor([0.0178, 0.0183, 0.0197, 0.0175, 0.0181, 0.0223, 0.0205, 0.0185], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-05-02 23:14:08,904 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.7987, 4.8413, 5.1317, 5.1068, 5.1481, 4.8553, 4.8143, 4.6371], device='cuda:1'), covar=tensor([0.0323, 0.0536, 0.0400, 0.0407, 0.0469, 0.0369, 0.0832, 0.0498], device='cuda:1'), in_proj_covar=tensor([0.0443, 0.0507, 0.0484, 0.0449, 0.0530, 0.0513, 0.0588, 0.0412], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:1') 2023-05-02 23:14:55,112 INFO [train.py:904] (1/8) Epoch 30, batch 6700, loss[loss=0.1943, simple_loss=0.2935, pruned_loss=0.0476, over 16719.00 frames. ], tot_loss[loss=0.1965, simple_loss=0.2846, pruned_loss=0.05427, over 3122781.71 frames. ], batch size: 134, lr: 2.23e-03, grad_scale: 4.0 2023-05-02 23:16:11,421 INFO [train.py:904] (1/8) Epoch 30, batch 6750, loss[loss=0.1819, simple_loss=0.2607, pruned_loss=0.0515, over 17033.00 frames. ], tot_loss[loss=0.1969, simple_loss=0.284, pruned_loss=0.05489, over 3112760.83 frames. ], batch size: 55, lr: 2.23e-03, grad_scale: 4.0 2023-05-02 23:16:15,728 INFO [optim.py:368] (1/8) 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 23:16:16,820 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.8711, 3.9159, 4.1346, 4.1096, 4.1268, 3.9191, 3.9242, 3.9249], device='cuda:1'), covar=tensor([0.0368, 0.0612, 0.0419, 0.0427, 0.0530, 0.0461, 0.0896, 0.0530], device='cuda:1'), in_proj_covar=tensor([0.0444, 0.0508, 0.0485, 0.0449, 0.0531, 0.0514, 0.0591, 0.0413], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:1') 2023-05-02 23:17:07,616 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=301140.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 23:17:29,010 INFO [train.py:904] (1/8) Epoch 30, batch 6800, loss[loss=0.1892, simple_loss=0.2866, pruned_loss=0.04593, over 17111.00 frames. ], tot_loss[loss=0.1972, simple_loss=0.2843, pruned_loss=0.05507, over 3105622.67 frames. ], batch size: 49, lr: 2.23e-03, grad_scale: 8.0 2023-05-02 23:17:35,475 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=301158.0, num_to_drop=1, layers_to_drop={2} 2023-05-02 23:17:43,389 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.19 vs. limit=2.0 2023-05-02 23:18:43,934 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=301201.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 23:18:47,474 INFO [train.py:904] (1/8) Epoch 30, batch 6850, loss[loss=0.1718, simple_loss=0.2782, pruned_loss=0.03273, over 16800.00 frames. ], tot_loss[loss=0.1982, simple_loss=0.2852, pruned_loss=0.05555, over 3109214.07 frames. ], batch size: 102, lr: 2.23e-03, grad_scale: 8.0 2023-05-02 23:18:51,707 INFO [optim.py:368] (1/8) 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 23:19:11,190 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.4959, 1.7529, 2.1550, 2.4379, 2.4819, 2.6679, 1.9059, 2.6399], device='cuda:1'), covar=tensor([0.0232, 0.0609, 0.0351, 0.0386, 0.0372, 0.0252, 0.0626, 0.0186], device='cuda:1'), in_proj_covar=tensor([0.0198, 0.0198, 0.0187, 0.0193, 0.0211, 0.0168, 0.0204, 0.0168], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-02 23:20:04,219 INFO [train.py:904] (1/8) Epoch 30, batch 6900, loss[loss=0.1799, simple_loss=0.2825, pruned_loss=0.03866, over 16714.00 frames. ], tot_loss[loss=0.1989, simple_loss=0.2877, pruned_loss=0.05508, over 3114974.49 frames. ], batch size: 89, lr: 2.23e-03, grad_scale: 8.0 2023-05-02 23:20:04,749 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.5422, 4.5702, 4.8505, 4.8100, 4.8498, 4.5635, 4.5365, 4.4604], device='cuda:1'), covar=tensor([0.0396, 0.0702, 0.0468, 0.0478, 0.0556, 0.0504, 0.1038, 0.0563], device='cuda:1'), in_proj_covar=tensor([0.0445, 0.0508, 0.0485, 0.0450, 0.0531, 0.0515, 0.0591, 0.0414], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:1') 2023-05-02 23:20:30,224 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.1627, 2.4522, 2.0681, 2.2242, 2.7712, 2.4454, 2.7640, 3.0096], device='cuda:1'), covar=tensor([0.0211, 0.0525, 0.0646, 0.0559, 0.0319, 0.0449, 0.0254, 0.0338], device='cuda:1'), in_proj_covar=tensor([0.0234, 0.0244, 0.0235, 0.0235, 0.0246, 0.0242, 0.0241, 0.0244], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-02 23:20:31,928 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.9231, 5.3455, 5.4925, 5.2047, 5.3143, 5.8244, 5.2930, 5.1024], device='cuda:1'), covar=tensor([0.1048, 0.1852, 0.2568, 0.1994, 0.2159, 0.0896, 0.1712, 0.2478], device='cuda:1'), in_proj_covar=tensor([0.0434, 0.0644, 0.0719, 0.0525, 0.0695, 0.0734, 0.0555, 0.0700], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-05-02 23:20:35,557 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=301273.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 23:21:17,876 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.62 vs. limit=2.0 2023-05-02 23:21:20,933 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=2.78 vs. limit=5.0 2023-05-02 23:21:22,698 INFO [train.py:904] (1/8) Epoch 30, batch 6950, loss[loss=0.1949, simple_loss=0.2776, pruned_loss=0.05612, over 16252.00 frames. ], tot_loss[loss=0.2012, simple_loss=0.2892, pruned_loss=0.05663, over 3105659.13 frames. ], batch size: 165, lr: 2.23e-03, grad_scale: 8.0 2023-05-02 23:21:26,932 INFO [optim.py:368] (1/8) 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 23:21:47,103 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.5284, 1.7919, 2.1343, 2.4973, 2.5087, 2.7298, 1.8836, 2.6652], device='cuda:1'), covar=tensor([0.0269, 0.0612, 0.0374, 0.0396, 0.0375, 0.0253, 0.0681, 0.0195], device='cuda:1'), in_proj_covar=tensor([0.0199, 0.0200, 0.0188, 0.0194, 0.0212, 0.0169, 0.0205, 0.0169], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0002], device='cuda:1') 2023-05-02 23:21:50,652 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=301321.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 23:22:29,787 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.4587, 3.6945, 3.8341, 2.1845, 3.2931, 2.6474, 3.8073, 4.0374], device='cuda:1'), covar=tensor([0.0268, 0.0821, 0.0636, 0.2236, 0.0843, 0.1045, 0.0610, 0.0892], device='cuda:1'), in_proj_covar=tensor([0.0162, 0.0172, 0.0172, 0.0158, 0.0149, 0.0134, 0.0146, 0.0185], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0005], device='cuda:1') 2023-05-02 23:22:38,933 INFO [train.py:904] (1/8) Epoch 30, batch 7000, loss[loss=0.1796, simple_loss=0.2822, pruned_loss=0.03855, over 16786.00 frames. ], tot_loss[loss=0.2002, simple_loss=0.2888, pruned_loss=0.05578, over 3095692.25 frames. ], batch size: 83, lr: 2.23e-03, grad_scale: 8.0 2023-05-02 23:23:46,245 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.87 vs. limit=2.0 2023-05-02 23:23:55,570 INFO [train.py:904] (1/8) Epoch 30, batch 7050, loss[loss=0.2414, simple_loss=0.3069, pruned_loss=0.08797, over 11308.00 frames. ], tot_loss[loss=0.2009, simple_loss=0.2902, pruned_loss=0.05579, over 3099624.31 frames. ], batch size: 247, lr: 2.23e-03, grad_scale: 8.0 2023-05-02 23:24:00,750 INFO [optim.py:368] (1/8) 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 23:24:38,154 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 30, batch 7100, loss[loss=0.1752, simple_loss=0.2679, pruned_loss=0.04123, over 16878.00 frames. ], tot_loss[loss=0.1994, simple_loss=0.2882, pruned_loss=0.05528, over 3093193.70 frames. ], batch size: 42, lr: 2.23e-03, grad_scale: 8.0 2023-05-02 23:25:20,934 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=301458.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 23:26:14,395 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=301491.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 23:26:16,018 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=4.83 vs. limit=5.0 2023-05-02 23:26:21,065 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=301496.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 23:26:33,760 INFO [train.py:904] (1/8) Epoch 30, batch 7150, loss[loss=0.195, simple_loss=0.2831, pruned_loss=0.05345, over 16443.00 frames. ], tot_loss[loss=0.1981, simple_loss=0.2865, pruned_loss=0.0549, over 3116487.50 frames. ], batch size: 35, lr: 2.23e-03, grad_scale: 8.0 2023-05-02 23:26:36,531 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=301506.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 23:26:37,392 INFO [optim.py:368] (1/8) Clipping_scale=2.0, grad-norm quartiles 1.858e+02 2.724e+02 3.109e+02 4.036e+02 9.536e+02, threshold=6.218e+02, percent-clipped=1.0 2023-05-02 23:27:49,036 INFO [train.py:904] (1/8) Epoch 30, batch 7200, loss[loss=0.1658, simple_loss=0.2586, pruned_loss=0.03653, over 16616.00 frames. ], tot_loss[loss=0.1958, simple_loss=0.2842, pruned_loss=0.05373, over 3099949.29 frames. ], batch size: 57, lr: 2.23e-03, grad_scale: 8.0 2023-05-02 23:29:10,980 INFO [train.py:904] (1/8) Epoch 30, batch 7250, loss[loss=0.1724, simple_loss=0.2627, pruned_loss=0.0411, over 16529.00 frames. ], tot_loss[loss=0.1935, simple_loss=0.2819, pruned_loss=0.05261, over 3097072.28 frames. ], batch size: 68, lr: 2.23e-03, grad_scale: 8.0 2023-05-02 23:29:15,161 INFO [optim.py:368] (1/8) 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 23:29:39,695 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.5924, 4.6579, 4.9851, 4.9445, 4.9762, 4.6728, 4.6206, 4.5194], device='cuda:1'), covar=tensor([0.0400, 0.0591, 0.0397, 0.0424, 0.0466, 0.0425, 0.0988, 0.0542], device='cuda:1'), in_proj_covar=tensor([0.0443, 0.0507, 0.0484, 0.0448, 0.0528, 0.0513, 0.0589, 0.0412], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:1') 2023-05-02 23:30:27,540 INFO [train.py:904] (1/8) Epoch 30, batch 7300, loss[loss=0.2089, simple_loss=0.301, pruned_loss=0.05842, over 16102.00 frames. ], tot_loss[loss=0.1934, simple_loss=0.2818, pruned_loss=0.05253, over 3099003.34 frames. ], batch size: 165, lr: 2.23e-03, grad_scale: 8.0 2023-05-02 23:31:44,855 INFO [train.py:904] (1/8) Epoch 30, batch 7350, loss[loss=0.1855, simple_loss=0.2665, pruned_loss=0.05225, over 16991.00 frames. ], tot_loss[loss=0.1952, simple_loss=0.2831, pruned_loss=0.05366, over 3087871.73 frames. ], batch size: 53, lr: 2.23e-03, grad_scale: 4.0 2023-05-02 23:31:50,962 INFO [optim.py:368] (1/8) 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 23:32:44,259 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=301743.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 23:33:00,863 INFO [train.py:904] (1/8) Epoch 30, batch 7400, loss[loss=0.2104, simple_loss=0.297, pruned_loss=0.06189, over 16435.00 frames. ], tot_loss[loss=0.1969, simple_loss=0.2846, pruned_loss=0.05466, over 3079677.96 frames. ], batch size: 68, lr: 2.23e-03, grad_scale: 4.0 2023-05-02 23:33:18,374 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.8396, 4.9632, 5.2375, 5.1819, 5.2533, 4.9600, 4.9079, 4.7526], device='cuda:1'), covar=tensor([0.0378, 0.0500, 0.0431, 0.0462, 0.0467, 0.0421, 0.0966, 0.0490], device='cuda:1'), in_proj_covar=tensor([0.0441, 0.0503, 0.0481, 0.0446, 0.0525, 0.0509, 0.0585, 0.0410], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:1') 2023-05-02 23:33:53,663 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=301786.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 23:34:10,903 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=301796.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 23:34:23,456 INFO [train.py:904] (1/8) Epoch 30, batch 7450, loss[loss=0.1861, simple_loss=0.2791, pruned_loss=0.04658, over 16903.00 frames. ], tot_loss[loss=0.198, simple_loss=0.2852, pruned_loss=0.05536, over 3071087.31 frames. ], batch size: 116, lr: 2.23e-03, grad_scale: 4.0 2023-05-02 23:34:24,252 INFO [zipformer.py:625] (1/8) 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] (1/8) Clipping_scale=2.0, grad-norm quartiles 1.990e+02 2.877e+02 3.334e+02 4.200e+02 6.879e+02, threshold=6.668e+02, percent-clipped=3.0 2023-05-02 23:35:30,252 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=301844.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 23:35:45,205 INFO [train.py:904] (1/8) Epoch 30, batch 7500, loss[loss=0.2087, simple_loss=0.2944, pruned_loss=0.06147, over 15467.00 frames. ], tot_loss[loss=0.1978, simple_loss=0.2857, pruned_loss=0.05495, over 3057139.79 frames. ], batch size: 192, lr: 2.23e-03, grad_scale: 4.0 2023-05-02 23:35:50,368 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.8767, 2.0246, 2.1855, 3.5242, 1.9211, 2.3364, 2.1611, 2.1665], device='cuda:1'), covar=tensor([0.1949, 0.4287, 0.3429, 0.0787, 0.5448, 0.3024, 0.4138, 0.4130], device='cuda:1'), in_proj_covar=tensor([0.0426, 0.0479, 0.0390, 0.0340, 0.0449, 0.0550, 0.0453, 0.0562], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-02 23:36:57,866 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=301899.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 23:37:04,214 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.29 vs. limit=2.0 2023-05-02 23:37:05,270 INFO [train.py:904] (1/8) Epoch 30, batch 7550, loss[loss=0.1868, simple_loss=0.2869, pruned_loss=0.04335, over 16776.00 frames. ], tot_loss[loss=0.1967, simple_loss=0.2844, pruned_loss=0.05453, over 3059448.77 frames. ], batch size: 124, lr: 2.23e-03, grad_scale: 4.0 2023-05-02 23:37:07,750 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=301905.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 23:37:11,510 INFO [optim.py:368] (1/8) Clipping_scale=2.0, grad-norm quartiles 1.823e+02 2.767e+02 3.333e+02 4.100e+02 7.380e+02, threshold=6.666e+02, percent-clipped=1.0 2023-05-02 23:38:13,815 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([1.7654, 2.5367, 2.2567, 3.2764, 2.1867, 3.4870, 1.5974, 2.7307], device='cuda:1'), covar=tensor([0.1448, 0.0807, 0.1409, 0.0240, 0.0183, 0.0442, 0.1821, 0.0876], device='cuda:1'), in_proj_covar=tensor([0.0175, 0.0183, 0.0203, 0.0208, 0.0208, 0.0220, 0.0212, 0.0202], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-05-02 23:38:23,954 INFO [train.py:904] (1/8) Epoch 30, batch 7600, loss[loss=0.2004, simple_loss=0.2863, pruned_loss=0.05725, over 16138.00 frames. ], tot_loss[loss=0.1965, simple_loss=0.2835, pruned_loss=0.0548, over 3055695.70 frames. ], batch size: 165, lr: 2.23e-03, grad_scale: 8.0 2023-05-02 23:38:34,216 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=301960.0, num_to_drop=1, layers_to_drop={1} 2023-05-02 23:38:43,774 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=301966.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 23:39:47,150 INFO [train.py:904] (1/8) Epoch 30, batch 7650, loss[loss=0.1978, simple_loss=0.2905, pruned_loss=0.05251, over 16741.00 frames. ], tot_loss[loss=0.1978, simple_loss=0.2843, pruned_loss=0.05561, over 3066823.52 frames. ], batch size: 89, lr: 2.23e-03, grad_scale: 8.0 2023-05-02 23:39:53,135 INFO [optim.py:368] (1/8) Clipping_scale=2.0, grad-norm quartiles 2.050e+02 2.820e+02 3.410e+02 4.459e+02 8.175e+02, threshold=6.820e+02, percent-clipped=3.0 2023-05-02 23:41:00,847 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.5045, 3.4859, 3.4652, 2.6462, 3.3215, 2.1874, 3.1187, 2.7373], device='cuda:1'), covar=tensor([0.0205, 0.0163, 0.0220, 0.0222, 0.0129, 0.2278, 0.0164, 0.0271], device='cuda:1'), in_proj_covar=tensor([0.0184, 0.0174, 0.0213, 0.0185, 0.0190, 0.0217, 0.0202, 0.0178], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-02 23:41:06,721 INFO [train.py:904] (1/8) Epoch 30, batch 7700, loss[loss=0.2063, simple_loss=0.2905, pruned_loss=0.06102, over 16810.00 frames. ], tot_loss[loss=0.1973, simple_loss=0.2837, pruned_loss=0.05549, over 3070006.51 frames. ], batch size: 39, lr: 2.23e-03, grad_scale: 8.0 2023-05-02 23:41:57,467 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=302086.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 23:42:15,614 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.84 vs. limit=2.0 2023-05-02 23:42:18,129 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=302099.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 23:42:26,380 INFO [train.py:904] (1/8) Epoch 30, batch 7750, loss[loss=0.19, simple_loss=0.2877, pruned_loss=0.04616, over 16919.00 frames. ], tot_loss[loss=0.1972, simple_loss=0.284, pruned_loss=0.05521, over 3074629.49 frames. ], batch size: 96, lr: 2.23e-03, grad_scale: 4.0 2023-05-02 23:42:33,833 INFO [optim.py:368] (1/8) 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 23:42:35,639 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.6538, 3.6516, 2.2577, 4.2164, 2.8278, 4.1573, 2.4581, 3.0196], device='cuda:1'), covar=tensor([0.0311, 0.0419, 0.1782, 0.0249, 0.0869, 0.0563, 0.1500, 0.0816], device='cuda:1'), in_proj_covar=tensor([0.0177, 0.0182, 0.0196, 0.0174, 0.0180, 0.0221, 0.0204, 0.0184], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-05-02 23:43:06,657 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.1226, 2.0726, 2.6848, 3.1036, 2.9429, 3.5882, 2.2477, 3.5737], device='cuda:1'), covar=tensor([0.0273, 0.0673, 0.0430, 0.0382, 0.0401, 0.0206, 0.0678, 0.0181], device='cuda:1'), in_proj_covar=tensor([0.0200, 0.0201, 0.0189, 0.0195, 0.0212, 0.0169, 0.0206, 0.0170], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0002], device='cuda:1') 2023-05-02 23:43:12,802 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 30, batch 7800, loss[loss=0.2346, simple_loss=0.3008, pruned_loss=0.08425, over 11202.00 frames. ], tot_loss[loss=0.1985, simple_loss=0.285, pruned_loss=0.05595, over 3074489.25 frames. ], batch size: 247, lr: 2.23e-03, grad_scale: 4.0 2023-05-02 23:44:13,333 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.43 vs. limit=2.0 2023-05-02 23:44:14,840 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.5644, 3.6544, 2.2307, 4.1304, 2.8049, 4.0918, 2.3652, 2.9217], device='cuda:1'), covar=tensor([0.0321, 0.0423, 0.1878, 0.0330, 0.0884, 0.0619, 0.1654, 0.0941], device='cuda:1'), in_proj_covar=tensor([0.0178, 0.0183, 0.0197, 0.0174, 0.0180, 0.0222, 0.0205, 0.0185], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-05-02 23:44:16,269 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.70 vs. limit=2.0 2023-05-02 23:44:55,079 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=302200.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 23:45:00,018 INFO [train.py:904] (1/8) Epoch 30, batch 7850, loss[loss=0.2354, simple_loss=0.3022, pruned_loss=0.08431, over 11576.00 frames. ], tot_loss[loss=0.198, simple_loss=0.2855, pruned_loss=0.05524, over 3089558.40 frames. ], batch size: 247, lr: 2.23e-03, grad_scale: 4.0 2023-05-02 23:45:06,745 INFO [optim.py:368] (1/8) 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 23:46:15,232 INFO [train.py:904] (1/8) Epoch 30, batch 7900, loss[loss=0.1909, simple_loss=0.2908, pruned_loss=0.0455, over 16887.00 frames. ], tot_loss[loss=0.1967, simple_loss=0.2844, pruned_loss=0.05452, over 3099975.49 frames. ], batch size: 96, lr: 2.23e-03, grad_scale: 4.0 2023-05-02 23:46:16,927 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=302255.0, num_to_drop=1, layers_to_drop={2} 2023-05-02 23:46:25,269 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=302261.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 23:46:25,363 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=302261.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 23:47:22,419 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 30, batch 7950, loss[loss=0.2391, simple_loss=0.3089, pruned_loss=0.08463, over 11358.00 frames. ], tot_loss[loss=0.1979, simple_loss=0.2851, pruned_loss=0.05533, over 3084314.32 frames. ], batch size: 248, lr: 2.23e-03, grad_scale: 4.0 2023-05-02 23:47:41,024 INFO [optim.py:368] (1/8) Clipping_scale=2.0, grad-norm quartiles 1.902e+02 2.712e+02 3.186e+02 3.569e+02 8.017e+02, threshold=6.372e+02, percent-clipped=2.0 2023-05-02 23:48:48,632 INFO [train.py:904] (1/8) Epoch 30, batch 8000, loss[loss=0.2804, simple_loss=0.3471, pruned_loss=0.1068, over 11141.00 frames. ], tot_loss[loss=0.1997, simple_loss=0.2865, pruned_loss=0.05649, over 3071501.37 frames. ], batch size: 248, lr: 2.23e-03, grad_scale: 8.0 2023-05-02 23:48:54,530 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=302357.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 23:49:53,787 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=4.79 vs. limit=5.0 2023-05-02 23:49:57,476 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=302399.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 23:49:59,071 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.3942, 4.5039, 4.7002, 4.4545, 4.5694, 5.0562, 4.5107, 4.2496], device='cuda:1'), covar=tensor([0.1534, 0.1908, 0.2187, 0.1954, 0.2216, 0.1022, 0.1748, 0.2435], device='cuda:1'), in_proj_covar=tensor([0.0432, 0.0641, 0.0717, 0.0524, 0.0693, 0.0734, 0.0555, 0.0698], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-05-02 23:50:03,899 INFO [train.py:904] (1/8) Epoch 30, batch 8050, loss[loss=0.1963, simple_loss=0.2851, pruned_loss=0.05376, over 16739.00 frames. ], tot_loss[loss=0.199, simple_loss=0.286, pruned_loss=0.05597, over 3073129.35 frames. ], batch size: 62, lr: 2.23e-03, grad_scale: 8.0 2023-05-02 23:50:11,660 INFO [optim.py:368] (1/8) Clipping_scale=2.0, grad-norm quartiles 1.866e+02 2.982e+02 3.316e+02 3.990e+02 1.007e+03, threshold=6.633e+02, percent-clipped=4.0 2023-05-02 23:50:30,167 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.9636, 4.1815, 4.0166, 4.0355, 3.7800, 3.8172, 3.8629, 4.1794], device='cuda:1'), covar=tensor([0.1150, 0.0975, 0.1072, 0.0936, 0.0781, 0.1712, 0.1009, 0.1050], device='cuda:1'), in_proj_covar=tensor([0.0729, 0.0878, 0.0723, 0.0686, 0.0558, 0.0561, 0.0737, 0.0691], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:1') 2023-05-02 23:50:56,694 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.4609, 3.4100, 2.6833, 2.1920, 2.2168, 2.3112, 3.5403, 3.0262], device='cuda:1'), covar=tensor([0.3354, 0.0693, 0.2029, 0.3103, 0.2999, 0.2368, 0.0564, 0.1555], device='cuda:1'), in_proj_covar=tensor([0.0340, 0.0276, 0.0317, 0.0330, 0.0308, 0.0282, 0.0305, 0.0353], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:1') 2023-05-02 23:51:10,082 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=302447.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 23:51:19,818 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.9402, 2.1712, 2.2933, 3.3967, 2.1443, 2.4522, 2.2689, 2.3110], device='cuda:1'), covar=tensor([0.1636, 0.3672, 0.3180, 0.0777, 0.4419, 0.2617, 0.3716, 0.3453], device='cuda:1'), in_proj_covar=tensor([0.0426, 0.0480, 0.0390, 0.0340, 0.0448, 0.0550, 0.0452, 0.0561], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-02 23:51:21,038 INFO [train.py:904] (1/8) Epoch 30, batch 8100, loss[loss=0.2065, simple_loss=0.2995, pruned_loss=0.05674, over 15331.00 frames. ], tot_loss[loss=0.1981, simple_loss=0.2855, pruned_loss=0.05535, over 3070853.05 frames. ], batch size: 190, lr: 2.23e-03, grad_scale: 8.0 2023-05-02 23:51:56,163 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=302477.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 23:52:06,367 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.5399, 2.2736, 1.8816, 2.0458, 2.5602, 2.2551, 2.2430, 2.6845], device='cuda:1'), covar=tensor([0.0252, 0.0462, 0.0576, 0.0484, 0.0278, 0.0420, 0.0230, 0.0288], device='cuda:1'), in_proj_covar=tensor([0.0232, 0.0244, 0.0234, 0.0235, 0.0246, 0.0242, 0.0242, 0.0244], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-02 23:52:35,993 INFO [train.py:904] (1/8) Epoch 30, batch 8150, loss[loss=0.1659, simple_loss=0.2515, pruned_loss=0.0401, over 17240.00 frames. ], tot_loss[loss=0.1959, simple_loss=0.283, pruned_loss=0.05438, over 3096500.90 frames. ], batch size: 52, lr: 2.22e-03, grad_scale: 8.0 2023-05-02 23:52:43,486 INFO [optim.py:368] (1/8) 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 23:53:28,215 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=302538.0, num_to_drop=1, layers_to_drop={1} 2023-05-02 23:53:29,797 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.90 vs. limit=2.0 2023-05-02 23:53:52,239 INFO [train.py:904] (1/8) Epoch 30, batch 8200, loss[loss=0.1841, simple_loss=0.2667, pruned_loss=0.05078, over 16736.00 frames. ], tot_loss[loss=0.1944, simple_loss=0.2805, pruned_loss=0.05411, over 3108369.37 frames. ], batch size: 124, lr: 2.22e-03, grad_scale: 8.0 2023-05-02 23:53:53,939 INFO [zipformer.py:625] (1/8) 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] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=302556.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 23:54:03,193 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=302561.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 23:55:04,514 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.19 vs. limit=2.0 2023-05-02 23:55:12,501 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 30, batch 8250, loss[loss=0.1607, simple_loss=0.2638, pruned_loss=0.02884, over 16814.00 frames. ], tot_loss[loss=0.1909, simple_loss=0.2791, pruned_loss=0.05134, over 3095365.94 frames. ], batch size: 102, lr: 2.22e-03, grad_scale: 8.0 2023-05-02 23:55:22,071 INFO [optim.py:368] (1/8) 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 23:55:22,471 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=302609.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 23:56:01,939 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=302632.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 23:56:35,913 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 30, batch 8300, loss[loss=0.1735, simple_loss=0.2621, pruned_loss=0.04248, over 11803.00 frames. ], tot_loss[loss=0.1871, simple_loss=0.2769, pruned_loss=0.04869, over 3082381.59 frames. ], batch size: 248, lr: 2.22e-03, grad_scale: 8.0 2023-05-02 23:57:04,681 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=3.07 vs. limit=5.0 2023-05-02 23:57:40,407 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.4607, 4.4934, 4.8166, 4.7975, 4.7863, 4.5675, 4.4819, 4.4886], device='cuda:1'), covar=tensor([0.0496, 0.0878, 0.0609, 0.0543, 0.0597, 0.0661, 0.1164, 0.0606], device='cuda:1'), in_proj_covar=tensor([0.0440, 0.0502, 0.0480, 0.0445, 0.0525, 0.0509, 0.0585, 0.0410], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:1') 2023-05-02 23:57:43,527 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=302693.0, num_to_drop=1, layers_to_drop={0} 2023-05-02 23:58:01,686 INFO [train.py:904] (1/8) Epoch 30, batch 8350, loss[loss=0.169, simple_loss=0.2716, pruned_loss=0.03317, over 15255.00 frames. ], tot_loss[loss=0.1842, simple_loss=0.2758, pruned_loss=0.04628, over 3076741.83 frames. ], batch size: 190, lr: 2.22e-03, grad_scale: 8.0 2023-05-02 23:58:09,434 INFO [optim.py:368] (1/8) 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 23:58:20,151 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.1514, 4.1440, 4.4645, 4.4326, 4.4381, 4.2176, 4.1705, 4.2023], device='cuda:1'), covar=tensor([0.0408, 0.0858, 0.0511, 0.0496, 0.0528, 0.0573, 0.1005, 0.0548], device='cuda:1'), in_proj_covar=tensor([0.0439, 0.0500, 0.0478, 0.0443, 0.0522, 0.0507, 0.0583, 0.0408], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:1') 2023-05-02 23:59:22,770 INFO [train.py:904] (1/8) Epoch 30, batch 8400, loss[loss=0.1615, simple_loss=0.2638, pruned_loss=0.02957, over 16769.00 frames. ], tot_loss[loss=0.1809, simple_loss=0.2731, pruned_loss=0.0444, over 3048873.59 frames. ], batch size: 102, lr: 2.22e-03, grad_scale: 8.0 2023-05-02 23:59:45,791 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.3909, 2.8664, 3.2376, 2.0036, 2.8464, 2.1718, 3.0811, 3.1273], device='cuda:1'), covar=tensor([0.0309, 0.0995, 0.0550, 0.2284, 0.0873, 0.1104, 0.0674, 0.0923], device='cuda:1'), in_proj_covar=tensor([0.0159, 0.0170, 0.0169, 0.0156, 0.0148, 0.0132, 0.0145, 0.0183], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0005], device='cuda:1') 2023-05-03 00:00:37,350 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.2389, 4.3117, 4.5071, 4.2523, 4.3913, 4.8629, 4.4246, 4.1209], device='cuda:1'), covar=tensor([0.1673, 0.2045, 0.2105, 0.2040, 0.2291, 0.0966, 0.1602, 0.2444], device='cuda:1'), in_proj_covar=tensor([0.0426, 0.0634, 0.0707, 0.0517, 0.0685, 0.0725, 0.0549, 0.0690], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-05-03 00:00:44,411 INFO [train.py:904] (1/8) Epoch 30, batch 8450, loss[loss=0.1611, simple_loss=0.2628, pruned_loss=0.02968, over 16757.00 frames. ], tot_loss[loss=0.1785, simple_loss=0.2715, pruned_loss=0.0428, over 3052849.48 frames. ], batch size: 83, lr: 2.22e-03, grad_scale: 8.0 2023-05-03 00:00:52,080 INFO [optim.py:368] (1/8) 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 00:01:31,128 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=302833.0, num_to_drop=1, layers_to_drop={3} 2023-05-03 00:02:04,571 INFO [train.py:904] (1/8) Epoch 30, batch 8500, loss[loss=0.1556, simple_loss=0.2488, pruned_loss=0.03123, over 16249.00 frames. ], tot_loss[loss=0.175, simple_loss=0.2681, pruned_loss=0.04092, over 3059692.46 frames. ], batch size: 165, lr: 2.22e-03, grad_scale: 8.0 2023-05-03 00:02:08,477 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=302856.0, num_to_drop=0, layers_to_drop=set() 2023-05-03 00:02:19,445 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.6067, 3.6700, 3.4560, 3.1454, 3.2755, 3.5553, 3.3708, 3.4026], device='cuda:1'), covar=tensor([0.0608, 0.0689, 0.0321, 0.0288, 0.0452, 0.0512, 0.1325, 0.0550], device='cuda:1'), in_proj_covar=tensor([0.0312, 0.0474, 0.0364, 0.0365, 0.0358, 0.0421, 0.0252, 0.0435], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002], device='cuda:1') 2023-05-03 00:02:48,156 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 30, batch 8550, loss[loss=0.1698, simple_loss=0.2671, pruned_loss=0.03628, over 16714.00 frames. ], tot_loss[loss=0.1728, simple_loss=0.266, pruned_loss=0.03983, over 3045001.50 frames. ], batch size: 124, lr: 2.22e-03, grad_scale: 8.0 2023-05-03 00:03:27,651 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=302904.0, num_to_drop=0, layers_to_drop=set() 2023-05-03 00:03:38,833 INFO [optim.py:368] (1/8) 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 00:03:57,111 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=3.10 vs. limit=5.0 2023-05-03 00:04:20,363 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.7020, 4.0489, 4.1426, 2.9471, 3.5400, 4.1737, 3.7641, 2.5261], device='cuda:1'), covar=tensor([0.0490, 0.0075, 0.0053, 0.0372, 0.0135, 0.0094, 0.0097, 0.0447], device='cuda:1'), in_proj_covar=tensor([0.0139, 0.0091, 0.0093, 0.0135, 0.0103, 0.0116, 0.0099, 0.0131], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0003, 0.0004], device='cuda:1') 2023-05-03 00:04:30,224 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.8258, 4.8270, 4.5736, 3.8588, 4.6767, 1.7405, 4.4634, 4.2996], device='cuda:1'), covar=tensor([0.0084, 0.0078, 0.0217, 0.0357, 0.0111, 0.2899, 0.0133, 0.0299], device='cuda:1'), in_proj_covar=tensor([0.0182, 0.0173, 0.0211, 0.0183, 0.0188, 0.0216, 0.0199, 0.0176], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-03 00:04:40,744 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.25 vs. limit=2.0 2023-05-03 00:04:45,034 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=302942.0, num_to_drop=0, layers_to_drop=set() 2023-05-03 00:05:04,936 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 30, batch 8600, loss[loss=0.1814, simple_loss=0.2658, pruned_loss=0.04848, over 12354.00 frames. ], tot_loss[loss=0.1723, simple_loss=0.2664, pruned_loss=0.03912, over 3044346.21 frames. ], batch size: 248, lr: 2.22e-03, grad_scale: 8.0 2023-05-03 00:06:16,015 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=302988.0, num_to_drop=1, layers_to_drop={2} 2023-05-03 00:06:39,316 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=303000.0, num_to_drop=0, layers_to_drop=set() 2023-05-03 00:06:43,572 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([1.9279, 2.7457, 2.6597, 2.0052, 2.5183, 2.8067, 2.6806, 1.9468], device='cuda:1'), covar=tensor([0.0424, 0.0099, 0.0094, 0.0358, 0.0166, 0.0117, 0.0119, 0.0464], device='cuda:1'), in_proj_covar=tensor([0.0139, 0.0091, 0.0092, 0.0135, 0.0103, 0.0116, 0.0099, 0.0131], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0003, 0.0004], device='cuda:1') 2023-05-03 00:06:46,603 INFO [train.py:904] (1/8) Epoch 30, batch 8650, loss[loss=0.1622, simple_loss=0.2601, pruned_loss=0.03218, over 16220.00 frames. ], tot_loss[loss=0.1705, simple_loss=0.2649, pruned_loss=0.03804, over 3049446.40 frames. ], batch size: 165, lr: 2.22e-03, grad_scale: 8.0 2023-05-03 00:06:58,834 INFO [optim.py:368] (1/8) Clipping_scale=2.0, grad-norm quartiles 1.356e+02 2.161e+02 2.643e+02 3.389e+02 5.700e+02, threshold=5.286e+02, percent-clipped=3.0 2023-05-03 00:08:31,204 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 30, batch 8700, loss[loss=0.1637, simple_loss=0.2494, pruned_loss=0.03906, over 12496.00 frames. ], tot_loss[loss=0.168, simple_loss=0.2622, pruned_loss=0.03692, over 3051018.67 frames. ], batch size: 247, lr: 2.22e-03, grad_scale: 8.0 2023-05-03 00:10:14,424 INFO [train.py:904] (1/8) Epoch 30, batch 8750, loss[loss=0.1719, simple_loss=0.2759, pruned_loss=0.03396, over 16416.00 frames. ], tot_loss[loss=0.1678, simple_loss=0.2625, pruned_loss=0.0365, over 3067375.61 frames. ], batch size: 68, lr: 2.22e-03, grad_scale: 8.0 2023-05-03 00:10:25,189 INFO [optim.py:368] (1/8) Clipping_scale=2.0, grad-norm quartiles 1.363e+02 2.109e+02 2.507e+02 3.042e+02 7.094e+02, threshold=5.015e+02, percent-clipped=1.0 2023-05-03 00:10:38,716 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=303113.0, num_to_drop=0, layers_to_drop=set() 2023-05-03 00:11:23,570 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=303133.0, num_to_drop=1, layers_to_drop={0} 2023-05-03 00:11:46,444 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.74 vs. limit=2.0 2023-05-03 00:12:06,341 INFO [train.py:904] (1/8) Epoch 30, batch 8800, loss[loss=0.1757, simple_loss=0.2744, pruned_loss=0.03854, over 16685.00 frames. ], tot_loss[loss=0.1652, simple_loss=0.2604, pruned_loss=0.03497, over 3067784.66 frames. ], batch size: 134, lr: 2.22e-03, grad_scale: 8.0 2023-05-03 00:12:12,849 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.1858, 2.5557, 2.6394, 2.0030, 2.7644, 2.7992, 2.5366, 2.5383], device='cuda:1'), covar=tensor([0.0602, 0.0252, 0.0251, 0.0905, 0.0115, 0.0273, 0.0420, 0.0415], device='cuda:1'), in_proj_covar=tensor([0.0146, 0.0108, 0.0100, 0.0136, 0.0085, 0.0130, 0.0128, 0.0128], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0003], device='cuda:1') 2023-05-03 00:13:02,844 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=303181.0, num_to_drop=0, layers_to_drop=set() 2023-05-03 00:13:26,390 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.24 vs. limit=2.0 2023-05-03 00:13:50,920 INFO [train.py:904] (1/8) Epoch 30, batch 8850, loss[loss=0.1494, simple_loss=0.2429, pruned_loss=0.02799, over 12340.00 frames. ], tot_loss[loss=0.166, simple_loss=0.2626, pruned_loss=0.03473, over 3048308.09 frames. ], batch size: 248, lr: 2.22e-03, grad_scale: 8.0 2023-05-03 00:14:00,695 INFO [optim.py:368] (1/8) 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 00:14:49,593 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=192, metric=1.96 vs. limit=2.0 2023-05-03 00:15:03,334 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=303237.0, num_to_drop=0, layers_to_drop=set() 2023-05-03 00:15:38,682 INFO [train.py:904] (1/8) Epoch 30, batch 8900, loss[loss=0.1898, simple_loss=0.284, pruned_loss=0.0478, over 15470.00 frames. ], tot_loss[loss=0.1657, simple_loss=0.2628, pruned_loss=0.0343, over 3025205.54 frames. ], batch size: 191, lr: 2.22e-03, grad_scale: 8.0 2023-05-03 00:15:49,490 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=303258.0, num_to_drop=0, layers_to_drop=set() 2023-05-03 00:16:15,055 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([1.9722, 2.3821, 2.3342, 3.1678, 1.8196, 3.2851, 1.8413, 2.7716], device='cuda:1'), covar=tensor([0.1203, 0.0693, 0.1097, 0.0201, 0.0078, 0.0346, 0.1551, 0.0693], device='cuda:1'), in_proj_covar=tensor([0.0173, 0.0180, 0.0199, 0.0203, 0.0203, 0.0215, 0.0209, 0.0198], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-05-03 00:17:03,858 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=303288.0, num_to_drop=1, layers_to_drop={1} 2023-05-03 00:17:43,146 INFO [train.py:904] (1/8) Epoch 30, batch 8950, loss[loss=0.1736, simple_loss=0.2624, pruned_loss=0.04243, over 12762.00 frames. ], tot_loss[loss=0.1658, simple_loss=0.2623, pruned_loss=0.03467, over 3037407.90 frames. ], batch size: 248, lr: 2.22e-03, grad_scale: 8.0 2023-05-03 00:17:53,311 INFO [optim.py:368] (1/8) 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 00:18:16,176 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=303319.0, num_to_drop=0, layers_to_drop=set() 2023-05-03 00:18:23,944 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.6588, 3.7017, 4.2328, 2.2723, 3.5489, 2.7606, 3.9846, 4.0608], device='cuda:1'), covar=tensor([0.0200, 0.0841, 0.0455, 0.2103, 0.0678, 0.0906, 0.0580, 0.0875], device='cuda:1'), in_proj_covar=tensor([0.0157, 0.0168, 0.0167, 0.0155, 0.0146, 0.0131, 0.0143, 0.0180], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:1') 2023-05-03 00:18:53,963 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 30, batch 9000, loss[loss=0.1481, simple_loss=0.2404, pruned_loss=0.0279, over 16308.00 frames. ], tot_loss[loss=0.1631, simple_loss=0.2592, pruned_loss=0.03354, over 3060058.29 frames. ], batch size: 146, lr: 2.22e-03, grad_scale: 8.0 2023-05-03 00:19:32,193 INFO [train.py:929] (1/8) Computing validation loss 2023-05-03 00:19:42,082 INFO [train.py:938] (1/8) Epoch 30, validation: loss=0.1431, simple_loss=0.2465, pruned_loss=0.01984, over 944034.00 frames. 2023-05-03 00:19:42,083 INFO [train.py:939] (1/8) Maximum memory allocated so far is 18145MB 2023-05-03 00:20:36,978 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.9455, 4.7603, 4.9720, 5.1195, 5.3289, 4.7844, 5.3536, 5.3337], device='cuda:1'), covar=tensor([0.2126, 0.1516, 0.1960, 0.0976, 0.0614, 0.0912, 0.0633, 0.0856], device='cuda:1'), in_proj_covar=tensor([0.0660, 0.0800, 0.0923, 0.0827, 0.0623, 0.0647, 0.0683, 0.0790], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-03 00:20:37,032 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.1735, 3.7803, 4.3931, 2.1041, 4.4936, 4.6366, 3.4546, 3.5435], device='cuda:1'), covar=tensor([0.0620, 0.0274, 0.0176, 0.1171, 0.0078, 0.0129, 0.0350, 0.0383], device='cuda:1'), in_proj_covar=tensor([0.0145, 0.0108, 0.0099, 0.0135, 0.0085, 0.0129, 0.0127, 0.0127], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0003], device='cuda:1') 2023-05-03 00:20:44,502 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([5.3136, 5.7138, 5.4651, 5.5191, 5.2370, 5.2286, 5.0750, 5.8016], device='cuda:1'), covar=tensor([0.1253, 0.0828, 0.0905, 0.0766, 0.0767, 0.0719, 0.1215, 0.0761], device='cuda:1'), in_proj_covar=tensor([0.0712, 0.0854, 0.0704, 0.0666, 0.0544, 0.0545, 0.0715, 0.0672], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:1') 2023-05-03 00:20:44,633 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.7526, 3.9032, 2.4051, 4.3469, 3.0156, 4.2316, 2.6288, 3.1468], device='cuda:1'), covar=tensor([0.0283, 0.0318, 0.1547, 0.0250, 0.0762, 0.0539, 0.1373, 0.0745], device='cuda:1'), in_proj_covar=tensor([0.0173, 0.0177, 0.0191, 0.0168, 0.0176, 0.0215, 0.0200, 0.0179], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-05-03 00:21:25,798 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.1001, 3.2029, 3.5838, 2.1598, 3.0755, 2.3543, 3.4992, 3.4769], device='cuda:1'), covar=tensor([0.0266, 0.1018, 0.0568, 0.2240, 0.0835, 0.1017, 0.0705, 0.1119], device='cuda:1'), in_proj_covar=tensor([0.0157, 0.0167, 0.0167, 0.0155, 0.0146, 0.0131, 0.0143, 0.0180], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:1') 2023-05-03 00:21:28,458 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 30, batch 9050, loss[loss=0.1752, simple_loss=0.2635, pruned_loss=0.04349, over 13215.00 frames. ], tot_loss[loss=0.1635, simple_loss=0.2598, pruned_loss=0.0336, over 3072690.83 frames. ], batch size: 250, lr: 2.22e-03, grad_scale: 8.0 2023-05-03 00:21:38,010 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=303408.0, num_to_drop=0, layers_to_drop=set() 2023-05-03 00:21:39,556 INFO [optim.py:368] (1/8) 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 00:23:13,931 INFO [train.py:904] (1/8) Epoch 30, batch 9100, loss[loss=0.1619, simple_loss=0.2583, pruned_loss=0.03271, over 16754.00 frames. ], tot_loss[loss=0.164, simple_loss=0.2595, pruned_loss=0.03427, over 3067750.73 frames. ], batch size: 76, lr: 2.22e-03, grad_scale: 8.0 2023-05-03 00:23:33,685 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=303464.0, num_to_drop=0, layers_to_drop=set() 2023-05-03 00:24:19,399 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.1441, 2.2231, 2.1872, 3.8452, 2.1353, 2.5142, 2.3396, 2.3589], device='cuda:1'), covar=tensor([0.1491, 0.4113, 0.3564, 0.0637, 0.4691, 0.2851, 0.4121, 0.3725], device='cuda:1'), in_proj_covar=tensor([0.0419, 0.0471, 0.0385, 0.0332, 0.0441, 0.0540, 0.0446, 0.0551], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-03 00:24:41,588 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.6205, 3.6751, 3.5017, 3.2000, 3.3118, 3.5586, 3.3557, 3.4407], device='cuda:1'), covar=tensor([0.0570, 0.0877, 0.0331, 0.0279, 0.0497, 0.0584, 0.1429, 0.0543], device='cuda:1'), in_proj_covar=tensor([0.0308, 0.0466, 0.0361, 0.0360, 0.0354, 0.0415, 0.0248, 0.0428], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-03 00:24:58,771 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.4177, 4.7228, 4.5272, 4.5790, 4.2800, 4.2485, 4.1890, 4.7498], device='cuda:1'), covar=tensor([0.1235, 0.0852, 0.1012, 0.0753, 0.0792, 0.1480, 0.1203, 0.0846], device='cuda:1'), in_proj_covar=tensor([0.0711, 0.0854, 0.0703, 0.0666, 0.0544, 0.0545, 0.0715, 0.0671], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:1') 2023-05-03 00:25:12,863 INFO [train.py:904] (1/8) Epoch 30, batch 9150, loss[loss=0.1489, simple_loss=0.2455, pruned_loss=0.02616, over 15453.00 frames. ], tot_loss[loss=0.1642, simple_loss=0.2602, pruned_loss=0.0341, over 3064816.57 frames. ], batch size: 190, lr: 2.22e-03, grad_scale: 4.0 2023-05-03 00:25:17,669 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.5877, 2.5278, 1.8751, 2.7194, 2.0633, 2.7489, 2.1296, 2.3511], device='cuda:1'), covar=tensor([0.0373, 0.0376, 0.1318, 0.0245, 0.0725, 0.0497, 0.1308, 0.0646], device='cuda:1'), in_proj_covar=tensor([0.0174, 0.0177, 0.0191, 0.0168, 0.0176, 0.0215, 0.0201, 0.0179], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-05-03 00:25:25,335 INFO [optim.py:368] (1/8) 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 00:25:32,350 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.45 vs. limit=2.0 2023-05-03 00:26:22,512 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=303537.0, num_to_drop=0, layers_to_drop=set() 2023-05-03 00:26:54,951 INFO [train.py:904] (1/8) Epoch 30, batch 9200, loss[loss=0.1657, simple_loss=0.2582, pruned_loss=0.03665, over 12233.00 frames. ], tot_loss[loss=0.1615, simple_loss=0.2562, pruned_loss=0.03343, over 3045026.03 frames. ], batch size: 248, lr: 2.22e-03, grad_scale: 8.0 2023-05-03 00:27:31,978 INFO [zipformer.py:625] (1/8) 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] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=303585.0, num_to_drop=0, layers_to_drop=set() 2023-05-03 00:28:29,158 INFO [train.py:904] (1/8) Epoch 30, batch 9250, loss[loss=0.1531, simple_loss=0.2331, pruned_loss=0.03657, over 11959.00 frames. ], tot_loss[loss=0.1614, simple_loss=0.2558, pruned_loss=0.03347, over 3029610.65 frames. ], batch size: 248, lr: 2.22e-03, grad_scale: 8.0 2023-05-03 00:28:41,901 INFO [optim.py:368] (1/8) 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 00:28:49,603 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=303614.0, num_to_drop=0, layers_to_drop=set() 2023-05-03 00:29:39,007 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 30, batch 9300, loss[loss=0.1562, simple_loss=0.2471, pruned_loss=0.03265, over 16413.00 frames. ], tot_loss[loss=0.1602, simple_loss=0.2543, pruned_loss=0.03307, over 3017501.80 frames. ], batch size: 146, lr: 2.22e-03, grad_scale: 8.0 2023-05-03 00:31:23,164 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.0562, 2.2320, 2.2329, 3.6926, 2.1350, 2.5422, 2.3028, 2.3186], device='cuda:1'), covar=tensor([0.1583, 0.3825, 0.3559, 0.0656, 0.4627, 0.2758, 0.4145, 0.3753], device='cuda:1'), in_proj_covar=tensor([0.0419, 0.0471, 0.0386, 0.0332, 0.0441, 0.0540, 0.0446, 0.0551], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-03 00:32:05,575 INFO [train.py:904] (1/8) Epoch 30, batch 9350, loss[loss=0.1724, simple_loss=0.2682, pruned_loss=0.03836, over 15438.00 frames. ], tot_loss[loss=0.1601, simple_loss=0.2545, pruned_loss=0.0329, over 3039902.33 frames. ], batch size: 191, lr: 2.22e-03, grad_scale: 8.0 2023-05-03 00:32:13,809 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=303708.0, num_to_drop=0, layers_to_drop=set() 2023-05-03 00:32:16,690 INFO [optim.py:368] (1/8) 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 00:33:46,246 INFO [train.py:904] (1/8) Epoch 30, batch 9400, loss[loss=0.1715, simple_loss=0.2746, pruned_loss=0.03417, over 16403.00 frames. ], tot_loss[loss=0.1599, simple_loss=0.2544, pruned_loss=0.03269, over 3048169.81 frames. ], batch size: 146, lr: 2.22e-03, grad_scale: 8.0 2023-05-03 00:33:50,141 INFO [zipformer.py:625] (1/8) 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] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=303759.0, num_to_drop=0, layers_to_drop=set() 2023-05-03 00:35:26,339 INFO [train.py:904] (1/8) Epoch 30, batch 9450, loss[loss=0.1641, simple_loss=0.2531, pruned_loss=0.03758, over 12638.00 frames. ], tot_loss[loss=0.1611, simple_loss=0.2564, pruned_loss=0.03289, over 3048850.26 frames. ], batch size: 248, lr: 2.22e-03, grad_scale: 8.0 2023-05-03 00:35:36,995 INFO [optim.py:368] (1/8) 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 00:35:37,894 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=303810.0, num_to_drop=0, layers_to_drop=set() 2023-05-03 00:36:50,316 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.4384, 4.5601, 4.7104, 4.4806, 4.5276, 5.0663, 4.5921, 4.2872], device='cuda:1'), covar=tensor([0.1449, 0.1905, 0.2000, 0.2168, 0.2580, 0.1028, 0.1648, 0.2453], device='cuda:1'), in_proj_covar=tensor([0.0418, 0.0622, 0.0698, 0.0507, 0.0673, 0.0718, 0.0540, 0.0676], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-05-03 00:37:08,345 INFO [train.py:904] (1/8) Epoch 30, batch 9500, loss[loss=0.1575, simple_loss=0.2551, pruned_loss=0.0299, over 16641.00 frames. ], tot_loss[loss=0.1605, simple_loss=0.2557, pruned_loss=0.0326, over 3062531.48 frames. ], batch size: 57, lr: 2.22e-03, grad_scale: 8.0 2023-05-03 00:37:43,354 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=303871.0, num_to_drop=0, layers_to_drop=set() 2023-05-03 00:37:56,946 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.8570, 3.1626, 3.5272, 2.0192, 3.0619, 2.2419, 3.3827, 3.3071], device='cuda:1'), covar=tensor([0.0286, 0.0886, 0.0539, 0.2338, 0.0777, 0.1066, 0.0653, 0.1054], device='cuda:1'), in_proj_covar=tensor([0.0157, 0.0166, 0.0166, 0.0154, 0.0146, 0.0130, 0.0142, 0.0179], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:1') 2023-05-03 00:38:36,340 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=303896.0, num_to_drop=0, layers_to_drop=set() 2023-05-03 00:38:44,714 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.15 vs. limit=2.0 2023-05-03 00:38:54,494 INFO [train.py:904] (1/8) Epoch 30, batch 9550, loss[loss=0.1788, simple_loss=0.2797, pruned_loss=0.03891, over 16175.00 frames. ], tot_loss[loss=0.1609, simple_loss=0.2558, pruned_loss=0.03299, over 3071475.56 frames. ], batch size: 165, lr: 2.22e-03, grad_scale: 8.0 2023-05-03 00:39:08,517 INFO [optim.py:368] (1/8) 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 00:39:18,190 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=303914.0, num_to_drop=0, layers_to_drop=set() 2023-05-03 00:39:49,634 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 30, batch 9600, loss[loss=0.1829, simple_loss=0.2805, pruned_loss=0.04267, over 16905.00 frames. ], tot_loss[loss=0.1622, simple_loss=0.2574, pruned_loss=0.03348, over 3071996.43 frames. ], batch size: 109, lr: 2.22e-03, grad_scale: 8.0 2023-05-03 00:40:43,425 INFO [zipformer.py:625] (1/8) 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] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=303962.0, num_to_drop=0, layers_to_drop=set() 2023-05-03 00:41:15,690 INFO [zipformer.py:625] (1/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=303974.0, num_to_drop=1, layers_to_drop={0} 2023-05-03 00:42:24,325 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.7006, 2.0623, 1.7640, 1.9247, 2.3704, 2.0104, 1.9663, 2.5029], device='cuda:1'), covar=tensor([0.0237, 0.0598, 0.0727, 0.0652, 0.0392, 0.0574, 0.0258, 0.0341], device='cuda:1'), in_proj_covar=tensor([0.0225, 0.0238, 0.0229, 0.0230, 0.0240, 0.0237, 0.0233, 0.0236], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-03 00:42:31,192 INFO [train.py:904] (1/8) Epoch 30, batch 9650, loss[loss=0.1647, simple_loss=0.2573, pruned_loss=0.03603, over 16234.00 frames. ], tot_loss[loss=0.1633, simple_loss=0.2593, pruned_loss=0.03368, over 3072824.45 frames. ], batch size: 165, lr: 2.22e-03, grad_scale: 4.0 2023-05-03 00:42:48,100 INFO [optim.py:368] (1/8) 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 00:43:30,043 INFO [scaling.py:679] (1/8) Whitening: num_groups=1, num_channels=384, metric=4.95 vs. limit=5.0 2023-05-03 00:43:40,607 INFO [zipformer.py:625] (1/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=304035.0, num_to_drop=1, layers_to_drop={1} 2023-05-03 00:44:20,846 INFO [train.py:904] (1/8) Epoch 30, batch 9700, loss[loss=0.1625, simple_loss=0.2509, pruned_loss=0.03706, over 12296.00 frames. ], tot_loss[loss=0.1624, simple_loss=0.2581, pruned_loss=0.03331, over 3061151.18 frames. ], batch size: 248, lr: 2.22e-03, grad_scale: 4.0 2023-05-03 00:44:30,054 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=304059.0, num_to_drop=0, layers_to_drop=set() 2023-05-03 00:45:15,232 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.38 vs. limit=2.0 2023-05-03 00:45:26,601 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.25 vs. limit=2.0 2023-05-03 00:45:48,825 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.4339, 3.8509, 3.9706, 2.6799, 3.5147, 3.9724, 3.6697, 2.3024], device='cuda:1'), covar=tensor([0.0600, 0.0076, 0.0056, 0.0432, 0.0125, 0.0110, 0.0084, 0.0531], device='cuda:1'), in_proj_covar=tensor([0.0138, 0.0090, 0.0091, 0.0134, 0.0103, 0.0114, 0.0097, 0.0129], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:1') 2023-05-03 00:46:04,097 INFO [train.py:904] (1/8) Epoch 30, batch 9750, loss[loss=0.1729, simple_loss=0.2747, pruned_loss=0.03557, over 16801.00 frames. ], tot_loss[loss=0.1622, simple_loss=0.2573, pruned_loss=0.03356, over 3060669.72 frames. ], batch size: 124, lr: 2.22e-03, grad_scale: 4.0 2023-05-03 00:46:09,338 INFO [zipformer.py:625] (1/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=304107.0, num_to_drop=0, layers_to_drop=set() 2023-05-03 00:46:17,046 INFO [optim.py:368] (1/8) 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] (1/8) Epoch 30, batch 9800, loss[loss=0.162, simple_loss=0.2628, pruned_loss=0.03056, over 16453.00 frames. ], tot_loss[loss=0.1618, simple_loss=0.2577, pruned_loss=0.03293, over 3068196.96 frames. ], batch size: 68, lr: 2.22e-03, grad_scale: 4.0 2023-05-03 00:48:04,643 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=304166.0, num_to_drop=0, layers_to_drop=set() 2023-05-03 00:49:24,990 INFO [train.py:904] (1/8) Epoch 30, batch 9850, loss[loss=0.1685, simple_loss=0.2602, pruned_loss=0.03841, over 16914.00 frames. ], tot_loss[loss=0.1624, simple_loss=0.2589, pruned_loss=0.03291, over 3067507.39 frames. ], batch size: 109, lr: 2.22e-03, grad_scale: 4.0 2023-05-03 00:49:39,349 INFO [optim.py:368] (1/8) 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 00:50:19,499 INFO [zipformer.py:625] (1/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=304230.0, num_to_drop=0, layers_to_drop=set() 2023-05-03 00:50:27,222 INFO [scaling.py:679] (1/8) Whitening: num_groups=8, num_channels=96, metric=1.17 vs. limit=2.0 2023-05-03 00:51:13,314 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=304252.0, num_to_drop=0, layers_to_drop=set() 2023-05-03 00:51:16,010 INFO [train.py:904] (1/8) Epoch 30, batch 9900, loss[loss=0.1623, simple_loss=0.2643, pruned_loss=0.03018, over 15304.00 frames. ], tot_loss[loss=0.1619, simple_loss=0.2587, pruned_loss=0.03252, over 3059182.61 frames. ], batch size: 190, lr: 2.22e-03, grad_scale: 4.0 2023-05-03 00:52:03,609 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.0247, 2.0963, 2.0945, 3.6131, 1.9940, 2.3641, 2.2186, 2.1855], device='cuda:1'), covar=tensor([0.1538, 0.4070, 0.3628, 0.0730, 0.4975, 0.3012, 0.3984, 0.4205], device='cuda:1'), in_proj_covar=tensor([0.0419, 0.0471, 0.0385, 0.0332, 0.0441, 0.0539, 0.0446, 0.0550], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-03 00:52:13,218 INFO [zipformer.py:625] (1/8) 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] (1/8) Epoch 30, batch 9950, loss[loss=0.1557, simple_loss=0.2538, pruned_loss=0.0288, over 16704.00 frames. ], tot_loss[loss=0.1631, simple_loss=0.2607, pruned_loss=0.03282, over 3064344.61 frames. ], batch size: 76, lr: 2.22e-03, grad_scale: 4.0 2023-05-03 00:53:31,713 INFO [optim.py:368] (1/8) 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 00:54:20,738 INFO [zipformer.py:625] (1/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=304330.0, num_to_drop=1, layers_to_drop={3} 2023-05-03 00:55:15,632 INFO [train.py:904] (1/8) Epoch 30, batch 10000, loss[loss=0.1726, simple_loss=0.2618, pruned_loss=0.04173, over 17050.00 frames. ], tot_loss[loss=0.1629, simple_loss=0.2602, pruned_loss=0.03282, over 3074697.80 frames. ], batch size: 55, lr: 2.22e-03, grad_scale: 8.0 2023-05-03 00:56:57,638 INFO [train.py:904] (1/8) Epoch 30, batch 10050, loss[loss=0.156, simple_loss=0.2542, pruned_loss=0.02889, over 16868.00 frames. ], tot_loss[loss=0.1633, simple_loss=0.2607, pruned_loss=0.03292, over 3077440.09 frames. ], batch size: 96, lr: 2.22e-03, grad_scale: 8.0 2023-05-03 00:57:10,559 INFO [optim.py:368] (1/8) 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 00:57:42,866 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([2.0533, 3.1979, 3.2244, 2.2167, 2.9117, 3.2885, 3.1082, 1.9132], device='cuda:1'), covar=tensor([0.0587, 0.0085, 0.0079, 0.0427, 0.0158, 0.0094, 0.0111, 0.0532], device='cuda:1'), in_proj_covar=tensor([0.0136, 0.0089, 0.0090, 0.0133, 0.0102, 0.0112, 0.0096, 0.0128], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:1') 2023-05-03 00:57:52,074 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([3.0869, 2.2005, 2.1954, 3.7242, 2.1783, 2.5412, 2.3041, 2.3396], device='cuda:1'), covar=tensor([0.1514, 0.4010, 0.3516, 0.0629, 0.4456, 0.2611, 0.3962, 0.3474], device='cuda:1'), in_proj_covar=tensor([0.0417, 0.0469, 0.0384, 0.0330, 0.0439, 0.0536, 0.0443, 0.0547], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-03 00:58:34,282 INFO [train.py:904] (1/8) Epoch 30, batch 10100, loss[loss=0.1578, simple_loss=0.2508, pruned_loss=0.03244, over 16855.00 frames. ], tot_loss[loss=0.1632, simple_loss=0.2607, pruned_loss=0.03287, over 3085313.53 frames. ], batch size: 116, lr: 2.22e-03, grad_scale: 8.0 2023-05-03 00:58:40,331 INFO [zipformer.py:1454] (1/8) attn_weights_entropy = tensor([4.0730, 4.0436, 3.9666, 3.2704, 3.9567, 1.7969, 3.7444, 3.5129], device='cuda:1'), covar=tensor([0.0165, 0.0196, 0.0243, 0.0355, 0.0151, 0.3102, 0.0184, 0.0379], device='cuda:1'), in_proj_covar=tensor([0.0181, 0.0172, 0.0209, 0.0179, 0.0187, 0.0216, 0.0198, 0.0175], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-05-03 00:58:57,144 INFO [zipformer.py:625] (1/8) 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] (1/8) Done!